Download Introduction to Specific Language Impairment/SLI

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Document related concepts

Navajo grammar wikipedia, lookup

Modern Hebrew grammar wikipedia, lookup

Arabic grammar wikipedia, lookup

English clause syntax wikipedia, lookup

Georgian grammar wikipedia, lookup

Chichewa tenses wikipedia, lookup

Malay grammar wikipedia, lookup

Latin syntax wikipedia, lookup

Germanic strong verb wikipedia, lookup

Sanskrit grammar wikipedia, lookup

Udmurt grammar wikipedia, lookup

Esperanto grammar wikipedia, lookup

Portuguese grammar wikipedia, lookup

Kannada grammar wikipedia, lookup

Ukrainian grammar wikipedia, lookup

Modern Greek grammar wikipedia, lookup

Old Irish grammar wikipedia, lookup

Macedonian grammar wikipedia, lookup

Lithuanian grammar wikipedia, lookup

Ancient Greek verbs wikipedia, lookup

Old Norse morphology wikipedia, lookup

Ancient Greek grammar wikipedia, lookup

Inflection wikipedia, lookup

French grammar wikipedia, lookup

Italian grammar wikipedia, lookup

Spanish verbs wikipedia, lookup

Grammatical tense wikipedia, lookup

Hungarian verbs wikipedia, lookup

Spanish grammar wikipedia, lookup

Scottish Gaelic grammar wikipedia, lookup

Old English grammar wikipedia, lookup

Pipil grammar wikipedia, lookup

Yiddish grammar wikipedia, lookup

Swedish grammar wikipedia, lookup

Serbo-Croatian grammar wikipedia, lookup

English verbs wikipedia, lookup

Polish grammar wikipedia, lookup

Transcript
1
Grammatical Aspects of Specific Language Impairment:
A Linguistic Perspective
Andrew Radford, University of Essex, 17 March 20061
Contents
Page
Preface
2
1. Introduction to SLI
3
2. The Perceptual Deficit account of SLI
12
3. The Feature Blindness account of SLI
18
4. The Rule Deficit account of SLI
24
5. The Agreement Deficit account of SLI
31
6. The Agreement-and-Tense-Omission (ATOM) account of SLI
38
7. The Dependency Deficit account of SLI
51
8. Empirical Study I: The acquisition of past tense marking
60
9. Empirical Study II: The acquisition of s-forms
75
10. Empirical Study III: The acquisition of inflected main-verb forms
92
11. Empirical Study IV: The acquisition of subject case-marking
100
12. References
108
13. Appendix: Transcripts from a 4-year-old American boy with SLI
111
1
This coursebook should be cited in references in the following form:
Radford, A (2006) Grammatical Aspects of Specific Language Impairment: A Linguistic Perspective, unpublished
coursebook, University of Essex (http://courses.essex.ac.uk/lg505)
2
Preface
Overview
This book provides a linguistic perspective on the grammatical development of children with a language
disorder known as Specific Language Impairment/SLI. It is intended for intermediate or upper
undergraduates and beginning graduates. It presupposes no previous knowledge of SLI, and assumes only
the kind of background knowledge of traditional linguistic concepts (such as e.g. morpheme) which
students acquire from an introductory Linguistics textbook such as Radford et al. (1999). It is written in an
approachable style, avoiding unnecessary jargon and theoretical sophistication (or sophistry), with copious
footnotes designed to answer the kinds of questions which beginning students ask. The main focus
throughout is on English-speaking children with SLI.
Key features
There are three key features of the book which merit comment. The first is that (in recognition of the fact
that students have increasing demands on their time these days and so have less and less time to devote to
preparing for courses), I have kept the book short, so that you should be involved in doing not more than
an hour or two of preparatory reading for each of the core chapters. I have also built each core chapter
round a single key article – because past experience of asking students to read several articles before a
class is that virtually none of them manage to read all of the articles, and most read no more than one (and
yes, you’ve guessed it, each student reads a different article!).
A second key feature of the book is that it is designed to be interactive. Each of the core chapters contains
a brief summary (generally around 5 pages or so in length) of the relevant key article, followed by a
workbook section containing several pages of questions and exercises. The questions are designed to get
you to demonstrate that you have understood key points in the article, and to think critically and
independently about the relevant research. The exercise material is designed to get you to analyse small
sets of data and show how they can be used to test particular claims made by a particular theory. The
overall aim of the workbook section is to foster an interactive teaching style in the classroom.
A third key feature of the book is that it contains a substantial empirical component designed to get you to
undertake a small-scale piece of original research for yourself on a number of alternative topics. Detailed
step-by-step guidance is provided on how to undertake the research, and how to write it up.
Organisation
The book begins with an overview of SLI, followed by 6 core chapters outlining key theories of SLI, with
each chapter including several pages of material for class discussion. The remaining four chapters provide
step-by-step guidance on how to conduct a small-scale piece of empirical research into the nature of SLI
on your own, using a set of transcripts of the language production of a four-year-old boy with SLI which
you are provided with (in the Appendix).
3
1. Introduction to SLI
The Nature of SLI
Specific Language Impairment (conventionally abbreviated to SLI, and also referred to as developmental
dysphasia) is a developmental language disorder characterised by Gleason (2001, p. 504) as involving
‘delayed or deviant language development in a child who exhibits no cognitive, neurological or social
impairment’. Children with SLI show impaired language development from birth (with problems which
may either disappear during childhood or persist into adulthood) but are normal in other aspects of their
physical, mental and social development, and so have:
(1) • a score within the normal range on non-verbal intelligence tests2
• normal hearing3, and no recent ear infections
• no oral abnormalities4
• no neurological dysfunctions5 (e.g. brain lesions, brain injury, cerebral palsy, seizures)
• no behavioural, emotional, communicative or social problems (e.g. no autism)
See Bishop (1997), Leonard (1998), and the collection of papers in Bishop and Leonard (2000) and Levy
and Schaeffer (2003) for more detailed discussion of diagnostic criteria for SLI.
Prevalence and Persistence of SLI
Leonard (1989) estimates that around 6% of children suffer some form of language impairment (with
1.5% having a tested language age of less than two thirds of their tested mental age), and that SLI affects
around 3 times more males than females. Some SLI children6 seem to grow out of SLI during childhood,
while for others the symptoms persist into adulthood. Bishop & Edmundson (1987) report that of 68
children diagnosed as suffering from SLI at 4 years of age, 56% continued to show poor scores in
language tests at ages 5;6 and 8;6; Johnson et al (1999) report that 73% of the preschool children they
tested who were language-impaired continued to be language-impaired at age 19. However, van der Lely
and Battell (2003) suggest that children who ‘grow out of’ SLI may not be suffering from SLI at all, but
rather may simply be late developers’ (i.e. children who develop in the same way as normal children, but
more slowly). The issue amounts to whether SLI is seen as a disorder which makes it impossible for
children to acquire certain aspects of language, or as a disorder which makes them very slow at acquiring
the relevant aspects of language.
Comparing SLI children with TD (= typically developing) children
In order to determine the extent to which the language of SLI children is impaired, we need to compare
them with typically developing (= TD) children. However, this raises the question of what kind of TD
children we compare SLI children with. There are three main groups of TD children which SLI children
are typically compared with, namely:
(2) • chronological age controls (i.e. TD children of the same chronological age)
• mental age controls (i.e. TD children who have comparable scores on non-verbal intelligence tests)
• language age controls (i.e. TD children with comparable scores on tests of vocabulary or grammar)
Studies of SLI children’s grammatical development usually compare SLI children with TD children at the
same stage of grammatical development. A conventional way of quantifying a child’s grammatical
development is in terms of Mean Length of Utterance (conventionally abbreviated to MLU). A child’s
MLU is calculated by taking a sample of (say) 50 or 100 spontaneous speech utterances produced by the
E.g. a score of 85 or above on Arthur’s (1952) adaptation of the Leiter International Performance Scale or on the
British Ability Scales (Elliott et al. 1978).
3
Determined by a pure tone screening test.
4
As determined by a screening procedure like that in St. Louis and Ruscello (1981)
5
Determined by passing a screening test for neurological function such as that in Tallal, Curtiss and Kaplan (1980).
6
I shall use the term SLI children throughout – though you might note that some regard this term as ‘politically
incorrect’ and prefer the term children with SLI.
2
4
child, and calculating the mean number of morphemes7 per utterance in the child’s speech sample (by
adding up the total number of morphemes which the child produces and dividing it by the number of
utterances the child produces): see Brown (1973) and Miller et al. (1981) for procedures for computing
MLU. So, for example, if we were studying a group of 5-year old SLI children, we might compare them
both with 5-year old TD children (= chronological age controls), and with TD children who have
comparable MLU scores (= MLU controls). This would in effect mean that we’d compare 5-year old SLI
children with 3-year-old TD children, since 5-year old SLI children generally have much the same MLU
as 3-year-old TD children.
Differences between SLI and TD children
If we compare SLI children with typically developing age controls we find:
(3) • SLI children have scores significantly below TD children on language tests (e.g. SLI children’s MLU
is only around two thirds that of TD children, so that a 3-year old SLI child will have the language
age of a 2-year old TD child, and a 5-year old SLI child will have the language age of a 3-year-old)
• TD children produce their first word around 11 months of age, SLI children around 23 months
• TD children produce their first multiword combinations (e.g. Want chocolate) around 17 months,
SLI children around 37 months.
• SLI children show a higher proportion of errors then TD children
Overall characteristics of SLI
SLI children typically show some (or all) of the following types of impairment:
(4) • phonological (e.g. problems with consonant clusters and syllable-final consonants)
• grammatical (e.g. sporadic omission of affixes/inflections and articles/particles)
• lexical (delayed acquisition of words – e.g. first word appears around 23 months in SLI children, but
around 11 months in TD children; SLI children also have word-finding problems)
• comprehension problems (e.g. problems with understanding metaphors)
• reading problems
Note, however, that SLI children differ in the extent to which their comprehension and/or production
abilities are impaired: for more details, see the literature survey in Leonard (1998) chapter 3. Van der Lely
(2003) notes that SLI in some children is primarily characterised by a grammatical deficit, and refers to
the relevant type of SLI as GSLI. In this book, I focus on grammatical errors made by SLI children,
concentrating mainly on those around 4-5 years of age.
Overview of the grammatical errors made by SLI children
SLI children are traditionally said to make two main kinds of grammatical error, namely:
(5) • omission errors (sometimes omitting an item8 which would be obligatory in the adult grammar)
• commission errors (sometimes using an item in an inappropriate form or position).
For example, in a context where an adult would use a sentence like Where’ve they gone?, SLI children
around 4-5 years of age will sometimes produce the correct adult form, sometimes produce a sentence like
(6a) below, sometimes a sentence like (6b), and sometimes a sentence like (6c):
(6)(a)
7
Where they gone?
(b) Where’s them gone?
(c) Where they’ve gone?
Morphemes are grammatical units: e.g. an item like cats is a single word but two morphemes, comprising the noun
stem cat and the plural suffix -s. Although MLU is generally computed in terms of the mean (i.e. average) number of
morphemes per utterance for languages like English in which many words carry no overt inflections, for languages
like Italian or Spanish in which all words typically consist of a stem plus an affix or inflection, MLU is generally
computed simply in terms of words. When this is done, the abbreviation MLU(W) – i.e. mean length of utterance
measured in terms of words – is often used, for the sake of clarity.
8
The term item here is used informally to subsume both independent words and inflections/affixes/clitics.
5
(6a) involves an omission error, namely omitting the auxiliary HAVE. (6b) involves two commission
errors, one being the use of the wrong form of HAVE – using the third person singular form (ha)s instead
of the third person plural form (ha)ve – and the other being the use of the wrong form of the subject
pronoun (using the accusative form them in place of the nominative form they). (6c) involves a different
kind of commission error, namely placing the auxiliary HAVE after the subject they instead of before it (i.e.
failing to invert the auxiliary with the subject and so failing to position the auxiliary in front of the subject,
yielding the word order they have instead of the required adult order have they). In the various sections
below, I give a (superficial) classification of typical grammatical errors made by SLI children. Unless
otherwise stated, the illustrative data come from the Leonard files on the CHILDES data base
(MacWhinney 1993; http://childes.psy.cmu.edu). These are transcripts of spontaneous speech samples
from eleven SLI children aged between 3;8 and 5;7 collected by Larry Leonard; the children are identified
in the files only by letters (as child A, child B, child C, child D etc.). Accordingly, a letter like [D] in
square brackets after a particular example below indicates that the example was produced by child D.
Omission of affixes
A frequent type of grammatical error made by SLI children is omission of affixes. The term affix is used
to describe a grammatical morpheme which cannot stand on its own as an independent word, but which
must be attached to a host of an appropriate kind (i.e. to a suitable kind of word or expression). Affixes
found in English include:
(7) • genitive ’s9 on noun expressions like the president’s (as in The president’s car is bomb-proof)
• plural -s on nouns like cars (as in two cars)
• present-tense -s on verbs like knows (as in He knows her well)
• past-tense -d on verbs like lied (as in The president lied to Congress)
• progressive/gerund -ing on verbs10 (as in He was nodding off while listening to the talk).
Omission of an affix typically results in the production of a bare form (i.e. the production of an
uninflected word form which comprises a stem without any overt inflection). Illustrative examples of
affix-omission errors produced by SLI children (from the Leonard corpus) are given in the table below:
(8) Affix-omission made errors by children with SLI
Illustrative example
You know what my doctor name is? (= ‘my doctor’s name’)
Then I wanna put more sticker on (= ‘more stickers’)
My dad drink tea (= ‘My dad drinks tea’)
I drop him (= ‘I dropped him’)
We were do Superman (= ‘We were doing Superman’)
child
E
J
E
J
D
Item omitted
genitive ’s
plural -s
present-tense -s
past tense -d
progressive -ing
Omission of auxiliaries
Many auxiliaries in English have both a full form and a (contracted) clitic form. The term clitic denotes a
reduced form of another word, and has the property that it must cliticise (i.e. attach) to an appropriate kind
of word which can serve as its host. In its clitic form, a contracted auxiliary attaches to the word
immediately preceding it – as in the examples below:
Since children invariably use genitive ’s to mark possession, you can think of this informally as possessive ’s.
When used together with the auxiliary BE in sentences like He is sleeping an ing-form marks an action in progress
and is said to be a progressive participle in that use. In (most) other uses, ing-forms of verbs are said to be gerunds
(e.g. ‘I don’t want him upsetting her’).
9
10
6
(9) Full and clitic auxiliary forms in adult English
full form
clitic form
John is lying/John is lazy11
John’s lying/John’s lazy
What does he want?
What’s he want?
John has gone home
John’s gone home
They have done it
They’ve done it
You will regret it
You’ll regret it
He would like to see you
He’d like to see you
She had already left
She’d already left
Who did you see?
Who’d you see?
In spoken English (i.e. the form of English which constitutes the speech input to young children), clitic
forms are generally preferred to full forms wherever possible – though in some contexts a full form is
required (e.g. in John’s lying, I know he is we cannot replace the word is at the end of the sentence by its
clitic counterpart ’s). However, SLI children frequently omit auxiliaries in contexts where adults would
use a clitic form of the auxiliary in colloquial English, as we see from the examples below:
(10) Auxiliary-omission made errors by children with SLI
Illustrative example
child
How you get this out? (= ‘How d’you get this out?’)
A
Where this go? (= ‘Where’s this go?’)
A
It not wood (= ‘It’s not wood’)
D
Somebody coming (= ‘Somebody’s coming’)
D
He got no eyes (= He’s got no eyes’)
A
I been there (= ‘I’ve been there’)
D
Yeah, that be fun (= ‘Yeah, that’ll be fun’)
D
Item omitted
auxiliary d(o)
auxiliary (doe)s
copula (i)s
auxiliary (i)s
auxiliary (ha)s
auxiliary (ha)ve
auxiliary (wi)ll
Omission of pronouns
As well as omitting auxiliaries, SLI children also sometimes omit pronouns – more specifically, weak
pronouns (i.e. those which are unstressed and non-contrastive in use) like I/we/you/he/she/it/they. Such
weak pronouns often have a reduced form: for instance, he can be reduced to ’e in colloquial English, in a
sentence such as (H)e’s lying, I know (h)e is. Examples of pronoun-omission by children with SLI are
given in the table below; as some of the examples show, children may sometimes omit both a subject
pronoun and an auxiliary in the same clause (usually when one cliticises to the other in adult English, e.g.
as in d’you = do you):
(11) Omission of subject pronouns by children with SLI
Illustrative example
child Item(s) omitted
But can fit? (= ‘But can it fit?’)
A
subject pronoun it
Need no more (= ‘I need no more’)
B
subject pronoun I
How do skate?12 (= How does it skate?’)
B
subject pronoun it
If you chew on it, break (= ‘…it’ll break’)
A
it’ll (subject pronoun it and auxiliary will)
Got small knees (= ‘He’s got small knees’)
A
He’s (subject pronoun he and auxiliary has)
How pull this off? (= ‘How d’you pull this off?’)
A
d’you (auxiliary do and subject pronoun you)
11
In structures like John is lazy, John is in Paris and John is a doctor, the word BE is traditionally said to be a
copula (a Latin word meaning ‘link’) in that it serves to provide the sentence with a verb which links the subject
John with the non-verbal (predicative) expression following it – i.e. lazy/in Paris/a doctor. However, even in this
copular use, BE also functions as an auxiliary, and so (e.g.) undergoes inversion in questions like Is John lazy?
12
The child immediately afterwards asks ‘How do it skate?’ (omitting the -es inflection on do).
7
Omissions of articles and particles
A further characteristic of the speech production of SLI children is omission of articles and particles13.
The word the is termed a definite article in traditional grammars, and the word a(n) is traditionally
classified as an indefinite article. Articles are frequently omitted by SLI children in obligatory contexts
(i.e. in contexts where it would be obligatory for an adult speaker to use an article). Once such obligatory
context for the use of articles is with expressions containing a singular count noun14. For example, We
can’t say *Can I borrow computer? in English because computer is a singular count noun, and singular
count nouns require an article (or similar kind of word) in front of them (as we see from the fact that we
say Can I borrow a/the computer?).
The word to when followed by a verb in its (bare/uninflected) infinitive form is traditionally termed an
infinitive particle: so, in I want to go home, to is an infinitive particle and go is an infinitive verb form.
SLI children sometimes omit infinitival to in obligatory contexts (i.e. in contexts where adults require it)
A second kind of grammatical particle is the genitive case particle of. In expressions like the cover of the
book, the word of is traditionally said to be a Norman genitive case particle (because it corresponds to
French de ‘of’) – whereas the ’s affix in the book’s cover is said to be an Anglo-Saxon case particle. SLI
children frequently omit genitive particles. (I return to consider case errors below.)
A third type of grammatical particle are clause-introducing particles (in more recent work, termed
complementisers) like that/for/if in sentences like ‘He said that he was tired’, ‘She’s keen for him to be
there’, and ‘I doubt if he will come’. SLI children sometimes omit such particles in obligatory contexts15.
Examples of article and particle omission by SLI children are given below:
(12) Omission of articles and particles by children with SLI
Illustrative example
child
Can I finish game? (= ‘the game’)
A
Sue wanna take bath (= ‘a bath’)
C
Julie wanted eat his breakfast (= ‘to eat’)
C
What kind books? (= ‘What kind of books?’)
G
Time me to get another one? (= ‘Time for me to…’)
A
Item omitted
definite article the
indefinite article a
infinitive particle to
genitive case-particle of
clause-introducing particle for
Explaining omission errors
Why should it be that SLI children frequently omit affixes, clitics, articles and particles? There are two
opposing schools of thought relating to this question. One view known as the Perceptual Deficit
Hypothesis (and articulated by Leonard 1989) is that the items concerned are omitted because they are
difficult for SLI children to perceive because of their lack of phonetic substance. The perceptual
difficulties stem from the fact that the omitted items are typically unstressed, comprise only one or two
(consonant or vowel) segments, and often contain a short, weak schwa vowel (as in the articles the and a,
and particles like infinitival to and genitive of) or have no vowel at all (as in the clitic form of have found
in They’ve gone home). A diametrically opposed Grammatical Deficit view (articulated in an early form
by Gopnik 1990, and in more recent forms by Clahsen Bartke and Göllner 1997 and by Wexler, Schütze
and Rice 1998) is that children have problems in processing grammatical features and so (perhaps in order
to reduce their grammatical processing load) sometimes omit some of them. We can illustrate the
difference in the two approaches by the following set of examples.
Suppose that an SLI child omits plural -s in an expression like two car. The perceptual deficit account
claims that plural -s poses perceptual problems because it comprises only the consonant segment |z|. The
The term particle is used to denote a ‘short’ word which does not easily fit into conventional systems of
grammatical categories/parts of speech (e.g. isn’t a typical noun, verb, adjective, adverb, preposition etc.)
14
A count noun is a noun like chair which can be counted (in that we can say one chair, two chairs etc.). By
contrast, a noun like furniture which cannot be counted (cf. *one furniture, *two furnitures etc.) is called a
non-count noun or mass noun.
15
An obvious complication is posed by the fact that the particle/complementiser that can frequently be omitted in
adult English, so that alongside ‘He said that he was tired’ we can also say ‘He said he was tired’.
13
8
grammatical deficit account posits that omission of the plural -s ending stems from the fact that the child
omits the plural-number feature on the noun car (perhaps to reduce the feature-processing burden).
Now suppose that an SLI child omits past tense -d in ‘Daddy play with me yesterday’. The perceptual
deficit account claims that past -d poses perceptual problems because it comprises only the consonant
segment |d|. The grammatical deficit account posits that omission of the past-tense -d ending occurs
because the child does not specify the verb for past tense (i.e. omits the past tense feature on the verb).
Finally, suppose that a child omits the present-tense auxiliary ’s in a sentence like Daddy’s working and so
says simply Daddy working. The perceptual deficit account says that this is attributable to perceptual
problems caused by the fact that ’s is a clitic with little phonetic substance, comprising only the consonant
segment |z|. The grammatical deficit account (by contrast) says that children omit items which are lack one
or more of the grammatical features which they would carry in adult English: so (for example), if the child
fails to mark the auxiliary for tense, the auxiliary will be omitted16.
Commission errors: inflections
In addition to omission errors, SLI children frequently made commission errors. The commonest pattern
of commission error we find involves producing over-regularised forms of irregular words. For example,
the noun mouse in English has the irregular plural form mice: but SLI children sometimes add the regular
plural affix -s to the stem mouse and so produce the over-regularised form mouses. Similarly, in place of
the irregular past tense form bought SLI children sometimes use an over-regularised form like buyed
(formed by adding the regular past tense ending -ed to the stem buy). Indeed, sometimes we find multiply
inflected forms in which the regular past tense ending is attached to an adult irregular past tense form,
resulting in forms like wented (with regular -ed attached to the irregular past form went). In addition, in
structures in which a (present or past) tensed auxiliary or verb is followed by an infinitive form, SLI
children will sometimes use a tensed form in place of the infinitive (resulting in a tense-copying structure
containing two tensed verbs, e.g. He didn’t played with me instead of He didn’t play with me). Examples
of these errors are given below:
(13) Commission errors with inflections made by children with SLI
Illustrative example
child Nature of error
Firemans live over here (= ‘Firemen’)
G
over-regularised noun plural form
Something hurted you (= ‘hurt you’)
E
over-regularised past tense form
This broked (= ‘This broke’)
E
multiply inflected past form (past broke + past -ed)
It doesn’t really closes
A
tense-copying from does onto closes
Case-marking errors
A further type of commission error made by SLI children are case errors. Personal pronouns in English
have a number of distinct case forms. For example, the third person singular masculine personal pronoun
in English has three different forms – namely, the nominative form he, the genitive form his, and the
accusative (or objective) form him. Some personal pronouns have separate weak (= short) and strong
(= long) genitive forms. A weak genitive form like my is used when the genitive pronoun modifies an
immediately following noun expression of some kind, as an my new car (where genitive my modifies the
noun expression new car); a strong genitive form like mine is used when the genitive does not modify an
immediately following noun expression (as in This is mine or Mine is bigger than yours). The case carried
by a pronoun determines the spellout (i.e. superficial form) of the pronoun, as the table below illustrates:
This could be argued to be because although a child’s lexicon (i.e. mental dictionary) tells the child (e.g.) that BE
is spelled out as (i)s if it carries the features ‘third person singular present tense’, it provides the child with no way of
spelling out a tenseless form of BE which simply carries the features ‘third person singular’. Of course, BE would
also be omitted if it was not marked for agreement with the subject and so had no person and/or number features.
16
9
(14) Case Spellout Table: How case is spelled out on personal pronouns in adult English
Pronoun Spellout
1.SG
I if nominative, mine if a strong genitive, my if a weak genitive, me otherwise
1.PL
we if nominative, ours if a strong genitive, our if a weak genitive, us otherwise
2.SG/PL yours if a strong genitive, your if a weak genitive, you otherwise
3.M.SG
he if nominative, his if genitive, him otherwise
3.F.SG
she if nominative, hers if a strong genitive, her otherwise
3.N.SG
its if genitive, it otherwise
3.PL
they if nominative, theirs if a strong genitive, their if a weak genitive, them otherwise
NB: 1/2/3 denote 1st/2nd/3rd person; SG = singular; PL = plural; M = masculine; F = feminine; N = neuter
The otherwise form of the pronoun is termed its default form (the idea being that this is the form used by
default if the conditions for using other forms are not met).
Studies of SLI children report that they make case errors with personal pronouns: however, these case
errors are not random, but rather show a systematic pattern. The pattern we typically find is that SLI
children alternate between the correct adult forms and default forms. So, in a nominative context (i.e. a
structure in which adults would use a nominative pronoun like I/we/he/they etc.) SLI children will
typically alternate between correctly using the nominative form of the pronoun (so saying e.g. I want one)
and incorrectly using the default form (so saying e.g. Me want one). Likewise, in a genitive context (i.e. a
context in which adults produce a structure like his car where the genitive pronoun his marks possession)
SLI children will usually alternate between correctly producing the genitive form (saying his car) and
incorrectly producing the default form (saying him car). A further type of case error which SLI children
make is to confuse the two genitive forms – e.g. using a strong form like mine where an adult would use a
weak form like my (or conversely) – e.g. saying Mine car or This is my.
As we noted earlier, SLI children sometimes also omit the genitive case affix ’s in obligatory contexts, and
so alternate between saying Daddy’s car and Daddy car. If we assume that noun expressions have the
suffix ’s added to them when genitive in case but are otherwise bare (i.e. uninflected), then we could see
the use of a bare possessor like Daddy as being an instance of the use of a default form (= Daddy) in a
context where adults require the genitive form Daddy’s. If so, omission of genitive ’s would be a
grammatical error (use of a default form in place of a genitive form) rather than a perceptual error
(omission of a morpheme because of perceptual problems posed by its lack of phonetic substance).
Typical examples of case errors made by SLI children are given below:
(15) Case errors made by children with SLI
Illustrative example
child
Me don’t want those
D
Him eyes have water in
D
Mine baby fits
B
You know what my doctor name is?
E
Nature of error
use of default form me for nominative I
use of default form him for genitive his
use of strong genitive mine for weak genitive my
omission of genitive ’s
Commission errors: word order
A further type of commission error made by SLI children are word-order errors. The main type of word
error made by English SLI children involves failure to invert subject and auxiliary in (main-clause)
questions – as illustrated earlier by example (6c) Where they’ve gone? Rather more rare is the converse
error of incorrectly inverting an auxiliary with its subject in a complement-clause question. Examples of
both types of auxiliary inversion error are given below:
(16) Auxiliary inversion errors made by children with SLI
Illustrative example
child Nature of error
Which one I can do?
C
failing to invert can with I
Where this is?
C
failing to invert is with this
Know what one is it?
A
inverting is with it in a complement clause
10
Nature of commission errors
What’s the nature of the commission errors made by SLI children? It’s hard to see how errors which
involve using an inappropriate form of an item (e.g. using the default form them in place of the adult
genitive form their to mark the possessor in them house) could be the result of a perceptual deficit, since
the adult genitive form their should not pose perceptual problems (and would seem to have more or less
the same amount of phonetic substance as the default form them). Nor is it obvious how word order errors
like failure to apply auxiliary inversion could be the result of a perceptual deficit, since the problem is not
that the child doesn’t perceive the auxiliary and omits it, but rather that the child uses the auxiliary, but
fails to move it in front of the subject in a main clause, or (conversely) wrongly moves it in front of the
subject in a complement clause. It would seem more likely, then, that commission errors reflect a
grammatical (rather than a perceptual) deficit.
Overall nature of SLI
An important debate in the SLI literature revolves round the question of whether SLI is the result of
delayed development (with SLI children going through the same stages as TD children but more slowly,
and sometimes reaching a plateau – i.e. a stage beyond which they never progress), or deviant
development (with SLI children showing abnormal patterns of development not seen in TD children): this
can be succinctly referred to as the delay/deviance debate. A second debate relates to whether SLI is a
homogeneous disorder (in which all affected children show the same forms of language impairment) or a
heterogeneous disorder (in which different children show different forms of language impairment). A
related question is whether SLI represents a global impairment of grammar (with all aspects of grammar
being affected) or a selective impairment (with only some aspects of grammar being affected). Yet a
further controversy surrounds the question of whether SLI children show evidence of other primary (e.g.
cognitive) impairments or not: see Leonard (1998, chapter 5) for a summary of research arguing that SLI
children may show a cognitive deficit, and are less good at pretend play and mental imagery (e.g.
imagining objects in different orientations) than typically developing (= TD) children.
Genetic basis of SLI
A fair amount of evidence has been accumulated in support of the view that SLI is a genetically
transmitted disorder. Family studies show that between a quarter and two-thirds of immediate family
members (siblings and parents) of children with SLI have language problems (e.g. Gopnik and Crago
1991 report that around half the members of a 3-generation family they studied were affected by SLI; and
Rice, Wexler and Hanney 1998 report that the incidence of language disorders in families which have a
child with SLI is 22%, compared to only 7% in those which do not). Twin studies show a concordance
rate of 80-86% for identical twins (i.e. that in 80-86% of cases where one twin is diagnosed as having SLI,
the other has also been diagnosed as having SLI), and a concordance rate of 38-48% for non-identical
twins (see Tomblin and Buckwalter 1998 for discussion of the implications of twin studies). The fact that
SLI is three times more frequent in boys than girls also suggests a genetic basis for the disorder. Genetic
studies have shown that people affected with SLI have an anomaly in chromosome 7, as well as other
genetic anomalies (See Fisher et al. 1998, Tomblin et al 1998, Lai etc al. 2001, and a study by the SLI
consortium 2002).
11
Workbook section
§1.1 Sentences produced by children with SLI
Below are a number of sentences produced by the SLI children in the Leonard files on the CHILDES database (the child producing each sentence being identified by a letter: e.g. (B) denotes child B)17:
1 Maybe goes on this one (B) 2 What say? (B) 3 Can get us some them? (C) 4 Do this come out? (C)
5 Billy wanna has his blocks out (C) 6 The tree must broken off (C) 7 Superman have him hands up (D)
8 And they’re jump in water (D) 9 This is mine daddy’s (D) 10 I will be Chad brother (D)
11 Them is boys (D) 12 Me don’t know how do it (D) 13 How you knowed? (E) 14 It cames off (F)
15 I didn’t sawed you come in (G) 16 Think her too growed up (G) 17 What is this is? (H)
18 What next one is? (H) 19 Hope him gonna hit him butt (J) 20 Me no like him (K).
Discuss the nature of the errors found in each sentence.
The adult equivalents of these sentences are: 1 Maybe it goes on this one. 2 What d’you/did you say? 3 Can you
get us some of them? 4 Does this come out? 5 Billy wants to have his blocks out. 6 The tree must have broken off.
7 Superman has his hands up. 8 And they’re jumping in the water. 9 This is my daddy’s. 10 I will be Chad’s
brother. 11 They are boys. 12 I don’t know how to do it. 13 How did you know? 14 It came off. 15 I didn’t see
you come in. 16 I think she’s too grown up. 17 What is this? 18 What is the next one? 19 I hope he’s gonna hit
his butt. 20 I don’t like him.
17
12
2. The Perceptual Deficit account of SLI
Key article
Leonard LB (1989) ‘Language learnability and specific language impairment in children’, Applied
Psycholinguistics 10: 179-202
OUTLINE OF LEONARD’S PERCEPTUAL DEFICIT HYPOTHESIS
Larry Leonard (over the past two decades) has developed a particular theory of SLI known as the
Perceptual Deficit Hypothesis. He maintains (1989, p.181) that ‘SLI children have notable limitations in
the use of certain inflections, auxiliaries and closed-class morphemes18 such as articles’, noting (1989,
p.182) that:
(1) The most obvious error in the speech of older SLI children is the omission of obligatory grammatical
morphemes. The striking feature of this usage is that omissions occur in sentences containing
otherwise complex structures such as conjoining, complementation and relativization19.
The fact that SLI children can produce complex sentences like (2) below20 suggests to Leonard that the
errors they make are not due to processing overload problems caused by complex structures:
(2)(a)
(b)
(c)
Then he went home and tell mother – his mother – what he doing that day
Then about noontime those guy went in and eat and warm up
That boy climbing a rope to get to the top of the rope
Further evidence supporting this view comes from the fact that the same omission errors occur in shorter
sentences like (3) below (where processing complexity doesn’t come into play):
(3)(a)
He want play that violin
(b)
Those men sleeping
(c)
Can I play with violin?
Why SLI is not a grammatical deficit
On the basis of data from English children with SLI (= ESLI children) and Italian children with SLI
(= ISLI children), Leonard argues that the deficit which affects the language of SLI children is not an
intrinsically grammatical one (e.g. an inability to form plurals, or present/past forms etc.). In support of
this claim, he adduces a number of arguments, including the following:
(4)(i) ESLI children perform relatively well on irregular forms but relatively poorly on regulars (e.g. they
perform better on irregular past tense forms like gave than on regular past tense forms like played)
(ii) ESLI children perform better on uncontractible copula/auxiliary21 forms than on contractible forms
(iii) ISLI children perform better on marking morphological properties (e.g. plural number on nouns,
tense/agreement on verbs, concord between nouns and modifiers22) than ESLI children
Closed-class items are items belonging to a category with a very small number of members – e.g. auxiliaries,
determiners, pronouns etc. They are so-called functors/function words or what Brown (1973) called grammatical
morphemes. By articles Leonard means the so-called definite article the and the indefinite article a(n).
19
Conjoined structures are those like Daddy came home and he played with me where two structures are joined
together by a co-ordinating conjunction like and/or. A complementation structure is one containing a main clause
and a complement clause like Daddy says that we can go to the zoo where the clause [that we can go to the zoo] is
the complement of the verb say. A relativisation structure is one containing a relative clause which modifies a noun
or pronoun expression: in a sentence like He is someone who I like, the clause who I like is a relative clause in the
sense that it ‘relates to’ (i.e. refers back to and modifies) the pronoun someone.
20
The data in (1) and (2) are from a 16-year old child with SLI reported on in a study by Weiner (1974).
21
On the copula/auxiliary distinction, see footnote 19. Contractible forms are forms which can be contracted;
uncontractible forms are those which cannot be contracted. For example, in a sentence such as John is lying, I know
he is, the first occurrence of BE is contractible to ’s but the second is uncontractible.
22
The term agreement is traditionally used to denote matching the person/number features of an auxiliary/verb to
those of its subject; the term concord is used for matching the number features of a noun to those of a modifier (in
structures like this book/these books). In Italian, adjectives and determiners undergo number/gender concord with a
noun that they modify, as in questa bellissima macchina ‘this beautiful car’, where -a is a feminine singular suffix.
18
13
If the deficit in children with SLI were purely grammatical, Leonard reasons, we should expect them to
perform equally poorly on regular and irregular inflected forms, and on contractible and uncontractible
forms; likewise, we should expect Italian children with SLI to be just as bad at inflectional morphology as
English SLI children. The overall conclusion which Leonard draws from such observations is that SLI is
not an intrinsically grammatical deficit.
SLI as a perceptual deficit
Leonard (1989, p. 188) provides an Auditory Perceptual Deficit account of SLI under which the poor
performance of SLI children on certain grammatical morphemes is attributable to a perceptual deficit. The
relevant morphemes, he claims, cause perceptual difficulties because of their ‘low phonetic substance’23.
Key assumptions made by Leonard are that (because they have more ‘phonetic substance’):
(5)(i)
(ii)
(iii)
(iv)
Vowels and diphthongs are easier to perceive than consonants (and consonants are particularly
difficult to perceive when they occur at the end of syllables/words, or in a consonant cluster24)
Stressed vowels are easier to perceive than unstressed vowels
Long vowels and diphthongs are easier to perceive than short vowels
Full vowels are easier to perceive than reduced vowels25
Leonard, McGregor and Allen (1992) present experimental evidence in support of the claim that vowels
are more perceptually salient than consonants. They compared the performance of 8 monolingual SLI
children in the age range 4;6 to 5;7 with 8 TD children in the same age range on a perception test using
computer-synthesised syllables. They found that SLI children had ‘great difficulty’ in perceiving the
contrast between [das] and [da∫] (only 1 of the 8 SLI children passed the test, compared with 5/8 of the
TD children), but by contrast had no problems in perceiving the contrast between the vowels [i] and [u].
They conclude from this that there is evidence that consonants (particularly those at the end of a
syllable/word) pose more perceptual problems than vowels.
More specifically, Leonard claims that ‘SLI children are especially limited in their ability to perceive and
hypothesize low phonetic substance morphemes’26. What this implies is that the degree to which a child
will have problems in acquiring a particular morpheme will depend on whether it contains a stressed or
unstressed, long or short, full or reduced vowel – or no vowel at all. Leonard claims (1989, p.187) that the
following morphemes (by virtue of having little phonetic substance) pose problems for ESLI children:
(6)
23
past/perfect/passive -d, present tense -s, plural -s, genitive ’s, contracted forms (like the contracted
’s form of is/has, and the contracted ’ll form of will), the definite article the and the indefinite
article a, the infinitive particle to, and the complementiser that27
Note that phonetic substance is not a standard term in Phonetics; it is an informal term used by Leonard to indicate
the degree of perceptual salience that sounds have (i.e. how easy to perceive they are).
24
A consonant cluster is a set of two or more immediately adjacent consonants – e.g. str is a consonant cluster in the
word string.
25
To illustrate the difference: the ‘o’ vowel in the noun content (in a sentence like ‘Your claim has no empirical
content’) is a full vowel, whereas the ‘o’ vowel in the adjective content (e.g. in ‘He was feeling very content with
life’) is a reduced vowel, pronounced as schwa |ə|.
26
Note that the relevant children do not have a hearing deficit (i.e. they show no signs of deafness). Rather, the
deficit lies in their ability to process speech sounds.
27
The suffix -d is a past tense marker in a sentence like They arrested a group of England football fans yesterday, a
perfect participle marker when used with the auxiliary have in a sentence like They have just arrested a group of
England football fans and a passive participle marker when used with the auxiliary BE in a sentences such as A group
of England football fans were arrested. In a sentence such as Dan’s girlfriend likes roses, the suffix ’s on DAN marks
genitive case (and serves to indicate what we can loosely term ‘possession’), the suffix -s on the verb LIKE is a thirdperson-singular present-tense marker, and the suffix -s on the noun ROSE is a plural marker. The word that functions
as a complementiser (= complement-clause-introducing particle) in sentences like I said that I was tired (but is a
demonstrative determiner in That one is mine and a demonstrative pronoun in That is mine). Presumably, Leonard’s
reason for including the/a/to/that in his list of low-phonetic-substance morphemes is that they can have their vowel
reduced to schwa, as in |ðə, ə, tə, ðət|.
14
In terms of Leonard’s assumptions in (5) and (6), we can account for his observations in (4) as follows:
(4i): SLI children will find it easier to produce irregular past tense forms like gave than regular past tense
forms like played because the irregular past tense form gave involves marking past tense by the use of a
stressed diphthong |eI| in |geIv| which is phonetically substantial (and perceptually salient) by virtue of
being unreduced and relatively long, whereas the past tense of play is marked by the addition of the
(phonetically insubstantial) word-final consonant |d| in |pleId| (cf. Leonard 1989, p.187).
(4ii): Uncontractible forms (like is in ‘Yes it is’) are phonetically more substantial than contracted forms
(like ’s in It’s nice) by virtue of containing a vowel.
(4iii): In Italian, inflections are typically vocalic in nature (i.e. comprise or contain a vowel/diphthong),
whereas in English inflections are typically consonantal in nature. Hence, in an English sentence like
Mary sings, the third person singular inflection -s is phonetically insubstantial because it represents a
word-final consonant |z|, whereas in its Italian counterpart Maria canta, the corresponding third person
singular inflection -a has greater phonetic substance because it is a full (i.e. unreduced) vowel
The perceptual deficit which SLI children have means that they require many more exposures to a given
‘weak’ morpheme with little phonetic substance than a typically developing child in order to acquire it –
though the perceptual deficit will be much less evident for ‘strong’ morphemes with more phonetic
substance. This predicts (inter alia) that there should be a much greater difference between the
performance of SLI and TD children on (e.g.) contracted forms of the copula than on full (uncontracted)
forms. Moreover, as in typically developing children, grammatical processing (and hence the acquisition
of grammatical morphemes) will take longer for more complex morphemes (e.g. those carrying more
grammatical features) than for than less complex morphemes. This predicts that SLI children should
acquire grammatical morphemes in the same order as TDs – but more slowly.
Finally, a note on nomenclature. Leonard’s Perceptual Deficit account of SLI is frequently referred to
under the alternative name of the Surface Hypothesis, since it assumes that surface (i.e. superficial
perceptual) properties of morphemes determine how difficult they are to acquire.
WORKBOOK SECTION
Leonard’s core assumption is that perceptual factors play a key role in determining why SLI children have
problems in acquiring phonetically insubstantial items – though he also acknowledges that grammatical
processing plays an important part (with more complex items taking longer to process and acquire than
less complex one). The exercises below are intended to get you to explore the relative roles of perceptual
and grammatical factors in the acquisition of a range of different morphemes.
§2.1 Comparing SLI and TD children
Leonard (1992, 1995) reports on the results of a comparative survey of the grammatical performance of 10
SLI children (aged from 3;8 to 5;7) and 10 MLU-matched TD children (aged from 2;11 to 3;4), using data
based on spontaneous speech samples. He presents the following figures showing the mean percentage of
correct use in obligatory contexts for a number of grammatical phenomena by SLI and TD children:
15
Mean Percent Correct Use in Obligatory Contexts of Morphemes by SLI and TD children
Item
SLI children
MLU-matched TD children
regular past tense -d
32%28
65%
irregular past tense verb forms
65%
77%
regular 3Sg present tense -s
34%
59%
29
copula be
41%
71%
auxiliary be
21%
46%
auxiliary do
16%
27%
noun plurals
69%
96%
auxiliary inversion
66%
89%
nominative case
62%
71%
articles a/the
52%
62%
infinitival to
34%
46%
possessive ’s
32%
91%
genitive of
47%
88%
Discuss Leonard’s results, and how the Perceptual Deficit Hypothesis might account for them. Larry
Leonard (pc) points out that noun plural -s is on average about 40% longer than (third person singuar)
present -s, noting that this is due to the fact that the former appears in clause-final position more often than
the latter, and that fricatives are lengthened in clause-final position. What is the significance of this
observation, do you think?
§2.2 Production of word-final consonants
Paula Menyuk (1978) reported that SLI children have few problems producing a consonant at the end of a
word when it is part of the stem of the word (as with the final |z| of nose), but have considerable problems
producing a word-final consonant when it represents a grammatical inflection (as with the final |z| on a
plural noun like bees). More specifically, she remarks (1978, p.147) that (in sentence-repetition tasks,
where children are asked to repeat a word) ‘bees might be repeated as bee, but nose was never repeated as
no’. Myrna Gopnik (1990, p.143) makes a similar point, observing that ‘The child who is presumed to
have a perceptual problem with the final sound in boys has no problem with the word noise’. Leonard,
McGregor and Allen (1992, p.1077) likewise report that ‘higher percentages of use are seen for [d] in
braid than in played; similarly, percentages for [s] are higher in box than in rocks.’ What do you think are
the implications of these findings for the Perceptual Deficit Hypothesis?
§2.3 Sporadic omission and commission errors
Gopnik (1990, p.144) reports that SLI children typically show sporadic (and sometimes inappropriate)
production of grammatical morphemes like plural {s}. For example, she notes that an SLI child (PB) that
she studied produced utterances such as the following:
(a)
(b)
(c)
When they play, get points. When the girls lose, are sad
I was make 140 box. He only got two arena
You got a tape-recorders. You make one points
How far is PB’s performance on the italicised forms consistent with the Perceptual Deficit Hypothesis?
28
This figure indicates that the SLI children overall produced past tense -d in 32% of the contexts in which it would
be obligatory for an adult English speaker to use past tense -d.
29
The word BE is said to function as an auxiliary when followed by a verb (BE is a progressive (aspect) auxiliary in
He is waiting, and a passive (voice) auxiliary in He was arrested), and as a copula (i.e. linking verb) when followed
by a non-verbal expression – e.g. by an adjectival expression like lazy as in He is lazy, by a prepositional expression
like in Paris as in He is in Paris, or by a nominal (i.e. noun-containing) expression like a doctor as in He is a doctor.
16
§2.4 Performance on irregular pasts
In a study of a different group of children, Ullman and Gopnik report (1994, p.111) that the SLI and TD
subjects in their study perform just as well on regular pasts as on irregular pasts – and a parallel finding is
reported in a study by Vargha-Khadem et al. (1995). To what extent are these findings consistent with the
Perceptual Deficit Hypothesis?
§2.5 Italian verb inflections
Leonard (1989. p.197) reports that that Italian SLI children have greater problems in acquiring third
person plural verb inflections than third person singular inflections. He notes (ibid.) that none of the eight
Italian SLI children in his study used the third person plural inflection, and that all used the third person
singular form of the verb in plural contexts (i.e. with plural subjects). The relevant forms for a regular verb
like cantare ‘sing’ (whose stem form is cant-) are listed below (where |à| denotes a stressed vowel):
canta [kànta] ‘sings’ (third person singular present tense = ‘He/She sings’)
cantano [kàntano] ‘sing’ (third person plural present tense = ‘They sing’)
To what extent is this finding predicted by the Perceptual Deficit Hypothesis? Why do you think ISLI
children make this kind of error?
§2.6 Morphological uniformity
Jaeggli and Safir (1989, pp. 29-31) draw a distinction between those languages which are morphologically
uniform, and those which are non-uniform. A uniform language is one (like Italian) in which (regular)
words are of the form stem+affix and there are no regular words comprise simply a bare stem (without any
overt affix); a non-uniform language is one like English in which some regular word forms (e.g. working)
carry an overt affix and others (e.g. work do not). What role do you think morphological uniformity might
play in accounting for differences between ISLI and ESLI children?
§2.7 Verb-inflections and pro-drop
A major typological difference between Italian and English can be illustrated by the following dialogue:
SPEAKER A:
Dove abita Maria?
Where lives Maria?
(= ‘Where does Maria live?’)
SPEAKER B
Abita a Milano
Lives at Milan
(= ‘She lives in Milan’)
As these examples show, to refer to a person already mentioned, English requires the use of a pronoun like
she in a sentence like that produced by speaker B. However, Italian is said to be a pro-drop language (or
null-subject language), in that it allows a subject pronoun to be ‘dropped’ (i.e. omitted) or phonetically
null (i.e. ‘silent’) if it refers to someone already mentioned (like Maria above) – and in fact we find the
subject dropped in around two thirds of finite clauses in Italian. To what extent might this typological
difference between the two languages account for the observation made by Leonard that Italian children
perform better on producing inflected verbs than English children?
§2.8 Grammatical complexity
Rice & Oetting (1993) report that SLI children perform significantly better on plural -s than on 3Sg
present -s: a group of SLI children they tested achieved 83% accuracy on noun plural -s but only 36% on
3Sg -s. How do you think the Perceptual Deficit Hypothesis might account for the relevant data?
§2.9 Comprehension data
Gopnik and Crago (1991) report that on a task testing comprehension of the noun plural -s morpheme,
the SLI children in their study achieved the same scores as language-matched normally developing
children. Children were given commands like ‘Please touch the book/the books’ to see whether they
would touch a single book or more than one book. Gopnik and Crago (1991, p.36) conclude that ‘These
results indicate that the subjects could perceive the difference between s-marked forms and unmarked
17
forms and could reliably associate this difference with number.’ What implications does this finding have
for the Perceptual Deficit Hypothesis?
§2.10 Production problems
To what extent might Leonard’s findings about the problems which SLI children have with English
inflections be attributable to production problems which SLI children are reported to have (e.g. by Fee
1994), particularly in producing syllable-final/word-final consonants?
§2.11 Case and word order errors
Loeb and Leonard (1991) report that children frequently use default (accusative) pronouns in contexts
where adults require nominative pronouns, and hence say e.g. Him did it rather than He did it. Leonard
(1995, 1281) reports that SLI children frequently fail to invert auxiliaries in questions, saying What
Daddy’s doing? rather than What’s Daddy doing? (or alternating between the two). To what extent can
such errors be attributed to a perceptual deficit?
§2.12 Overall verdict
What is your overall verdict on the strengths and weaknesses of Leonard’s Perceptual Deficit Hypothesis?
18
3. The Feature Blindness account of SLI
Key article
Gopnik, M. (1990) ‘Feature blindness: A case study’, Language Acquisition 1: 139-164
OUTLINE OF GOPNIK’S (1990) STUDY
In her 1990 paper on feature-blindness, Myrna Gopnik argues that SLI is the result of a genetic deficit
which prevents SLI children from acquiring grammatical features30 such as person, gender, number,
countability, commonness, tense and aspect31. Gopnik claims (1990, p.139) that ‘These features are
absent’ from SLI children’s grammars, and that consequently SLI children have ‘a grammar without
syntactico-semantic features’ (1990, p.145). Gopnik maintains that the feature deficit is global in the sense
that it affects all grammatical features.
Methodology of Gopnik’s 1990 study
Gopnik’s 1990 study is based primarily on data from a bilingual (French-English) Canadian boy from
Québec (identified simply as PB) aged 8-9 years (for details of PB’s history, see Gopnik 1990, p.163-4).
The data come from a mixture of spontaneous speech samples, prompted storytelling, repetition tasks and
grammaticality judgements. The main errors reported in Gopnik’s (1990) study are summarised below.
Number errors
Gopnik notes that PB frequently makes errors with number-marking on nouns – e.g. he uses plural noun
forms in singular contexts32 as in (1a), and conversely uses singular nouns in plural contexts as in (1b):
(1)(a)
You got a tape-recorders
(b)
He only got two arena
Gopnik concludes from examples like these that the grammatical feature (singular/plural) number is
absent from nouns in PB’s grammar, and that PB treats singular nouns like tree and plural nouns like trees
as forms which are intrinsically unspecified for number (hence which can be used in singular and plural
contexts alike). More specifically, Gopnik claims (1990, p.148):
The feature deficit hypothesis provides a single coherent explanation for the data. Without the feature
[plural], the s cannot be generated in the morphological component nor can feature-matching33
between the noun and the determiner operate. In this child’s grammar, s is regarded as a variant
phonological form with no associated meaning.’
In other words, Gopnik claims that SLI children assume that tree and trees are two alternative (entirely
equivalent) forms of the word TREE and hence can be used interchangeably (like e.g. dad and daddy).
Gopnik claims that a similar number blindness is revealed by PB’s treatment of nouns on grammaticality
judgements, repetition tasks, and writing tasks (see Gopnik 1990, pp. 148-9 for relevant data). She also
reports (1990, p.148) that ‘Errors in number also occur with pronouns. He, it and they occur in
spontaneous speech but are not reliably used to distinguish between singular and plural referents.’
30
Gopnik uses the more cumbersome term syntactico-semantic features: by this she seems to mean what Chomsky
(1995) calls formal features, and what I shall here refer to by Brown’s (1973) more traditional label grammatical
features.
31
Person is illustrated by the I/you/he contrast; gender by the he/she/it contrast; number by the dog/dogs contrast;
countability by the two chairs/*two furnitures contrast; commonness by the contrast between the chairman and *the
Richard; tense by the goes/went contrast; and aspect by the contrast between going/gone (going marking progressive
aspect and gone perfect aspect). Although she does not specifically mention case, presumably Gopnik would posit
that the case errors made by SLI children (e.g. saying Me want one rather than I want one) are further instances of an
overall grammatical feature deficit.
32
A singular/plural context is one in which an adult would require a singular/plural form.
33
By feature-matching, Gopnik means number concord between a determiner/quantifier and a noun which it
modifies (so that a noun is marked as plural if modified by a plural numeral like two and as singular if modified by a
singular numeral like one – cf. one dog/two dogs).
19
Commonness errors
In English (and other languages) there is a distinction between common nouns on the one hand and proper
nouns (also called proper names) on the other. Common nouns like man can generally be modified by a
determiner like the, but proper names like Andrew cannot (hence we say the man but not *the Andrew34).
Gopnik claims that PB shows no awareness of the distinctive properties of proper nouns and so treats them
like common nouns, incorrectly using the to modify proper names like Red Riding Hood in sentences like:
(2)
The Red Riding Hood arrive at his grandma’s house
Countability errors
The set of common nouns can be further differentiated into two different subtypes, namely count nouns on
the one hand, and mass nouns (or non-count nouns) on the other. Count nouns like chair have the property
that they can be counted (cf. one chair, two chairs, etc.), whereas mass nouns like furniture are so called
because they denote an undifferentiated mass which cannot be counted (cf. *one furniture, *two
furnitures). Gopnik reports that PB confuses count nouns with mass nouns: for example, he uses
(invariable) mass nouns like music as count nouns, pluralising them as in (3a) below; conversely he uses a
singular count noun like bicycle (which, like all singular count nouns in English, would normally have to
be modified by a determiner such as a/the/my/this etc.) as a mass noun (e.g. in contexts where adults
would use a mass noun like cycling) as in (3b):
(3)(a)
I play musics
(b)
I love bicycle
Gopnik concludes from examples such as (2) and (3) that PB has not mastered the commonness or
countability properties of nouns: this, she claims, supports her more general hypothesis that PB has a
global grammatical feature deficit (or, in her terms, is ‘blind’ to grammatical features)35.
Gender errors
Third person singular pronouns like he/she/it in English encode grammatical gender features, with it
encoding so-called neuter (or inanimate) gender, and he and she respectively denoting masculine and
feminine animate entities. PB produces a number of gender errors in pronouns, e.g. using he to refer back
to an inanimate antecedent, as in:
(4)
When the cup break he get repair
Here, PB fails to match the gender of the (animate) pronoun he to that of the inanimate expression the cup
which it refers to. This (in Gopnik’s view) shows that he is blind to features like gender.
Aspect errors
In adult sentences such as He is going home, the sequence is going marks progressive aspect (i.e.
indicates an action in progress): in this use, BE is a progressive auxiliary and the -ing form of the verb is a
progressive participle36. Gopnik observes that in contexts where adults would use progressive BE with a
verb in the progressive participle (-ing) form, PB may omit either BE as in (5a) below, or -ing as in (5b):
(5)(a)
This one is look
(b)
The dragon drying hisself
Gopnik (1990, p.157) hypothesises that in adult progressive sentences like He is sleeping, both the
auxiliary is and the verb sleeping carry an aspectual feature [Progressive-Aspect], and that there is a
feature-matching requirement for the aspectual feature of the auxiliary to match that of the verb (so that a
sentence denoting an action in progress requires use of both a progressive auxiliary and a progressive
34
Unless the proper name is modified by an expression such as that I once knew. Some nouns like daddy can
function both as common nouns (as in ‘You can be the daddy’) and as common nouns (as is ‘Daddy is naughty’).
35
If it is a characteristic of mass nouns that they have no intrinsic number features, PB’s confusion over the
countability properties of nouns will reduce to confusion over their number properties.
36
When not used in conjunction with BE, the ing-form of the verb generally functions as a gerund, e.g. in sentences
such as He enjoys playing tennis. In adult English sentences such as He has gone home, the auxiliary HAVE marks
perfect aspect (in that it marks the perfection – in the sense of ‘completion’ – of an action): however, Gopnik says
nothing about perfect aspect.
20
verb-form). She maintains that the progressive aspect feature is absent from PB’s grammar, with the
consequence that the feature-matching requirement does not hold – and so some sentences produced by
PB like (5a) contain a progressive auxiliary (is) without a matching progressive verb-form (looking), while
others like (5b) contain a progressive verb-form (drying) without a matching progressive auxiliary (is).
Tense errors
Gopnik claims that PB generally fails to mark tense in obligatory contexts37. For example, she notes
(1990, p.153) that for PB ‘Regular past tense forms never occur in spontaneous speech’. So, in a sentence
such as (6) below containing a past time expression like last time, the regular verb arrive appears as a bare
(i.e. uninflected) form which lacks the past tense suffix -d:
(6)
Last time we arrive
Gopnik administered an elicitation task to PB, asking him to convert present-tense sentences containing
the word today into past-tense sentences containing yesterday: in 27/29 cases where PB’s response was
unambiguous, it contained a bare verb form uninflected for past tense38. Similarly, Gopnik (1990, p.154)
observes that ‘PB virtually never added s to the verb to mark the third person singular present’ – hence the
absence of -s in a present-tense context in sentence (7) below:
(7)
He look at the other side of the tree
And in a sentence-repetition task, PB omitted the s-inflection from 16/19 (84%) s-inflected present tense
verb forms39.
Pro-drop errors
Gopnik notes that PB sometimes omits subject pronouns (a phenomenon widely referred to as pro-drop),
reporting (for example) that in a sentence repetition task, PB dropped the subject in 4/16 (25%) test
sentences. She argues that PB only omits subjects when he also omits tense-marking from verbs, her
hypothesis being that ‘subject pronouns can be dropped’ when ‘the feature [tense] is not present’ (1990,
p.158)40. For example, in a sentence repetition task, PB dropped the subject in 4/16 (25%) test sentences.
WORKBOOK SECTION
Gopnik’s key claim is that SLI children have a genetic deficit which makes them blind to (and hence
completely unable to acquire) any kind of grammatical features, with the resulting feature blindness
leading to a global grammatical deficit (i.e. an inability to acquire any grammatical feature of any kind).
However, there are a number of aspects of her feature blindness analysis which seem questionable. For
example, while it may well be the case that SLI children have problems in processing and acquiring
grammatical features, it could be argued that this results in them being slow to acquire grammatical
features and in marking them sporadically – rather than (as Gopnik claims) in being completely unable to
acquire grammatical features and hence never marking them. Moreover, subsequent research has
suggested that the feature deficit in SLI children may be selective rather than global, in the sense that SLI
children have more problems with some kinds of features (e.g. agreement features) than with other others.
The questions and exercises below are designed to get you to take a more critical look at Gopnik’s feature
37
I.e. in contexts where it would be obligatory for an adult to use a present-tense or past-tense verb/auxiliary.
The two correct responses were unspecified irregular verb forms. Gopnik also counts ambiguous forms like He cut
the grass as ‘correct’ past-tense responses: but this is inappropriate since it is impossible to be sure whether in such
cases cut is a bare stem/infinitive form or a past tense form. I have excluded the four such ambiguous cases from the
figures presented here.
39
The fact that PB correctly repeated 3/19 (16%) s-forms should not necessarily be taken as an indication that he is
beginning to acquire third person singular present tense -s. Gopnik makes the point that the ability to correctly repeat
a handful of model sentences may tell us very little about grammatical competence, particularly where the model
sentence is short, since ‘Short sentences can be repeated directly from memory’ (1990, p.157).
40
More specifically, following Guilfoyle (1984), Gopnik suggests that pronouns can only be dropped when they
cannot be case-marked: if the subject of a tensed clause (i.e. a clause containing a present or past tense
verb/auxiliary) is assigned nominative case, it follows that the subject of a tensed clause cannot be dropped.
38
21
blindness model, and at just how strong the evidence is which she adduces in support of the model. But
first, an exercise to test whether you understand the nature of the claims made by Gopnik.
§3.1 Analysing sentences produced by PB
The data below are examples of PB’s speech production (and grammaticality judgements) reported in
Gopnik (1990). In examples containing italicised pronouns, the expression which the pronoun refers back
to is bold-printed. (8-10) are examples of French (or mixed French/English) sentences produced by PB.
(Abbreviations used below are: M = masculine, F = feminine, Sg = Singular and Pl = Plural.)
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
You make one points
I was make 140 box
The dragon jumping
One machine clean all the two arena41
The Marie-Louise look at the bird
When the bus goes fast, he has an accident
The Red Riding Hood arrive at his grandma’s house.
Il prend un
gros
respiration
He takes a-M.Sg big-M.Sg breath-F.Sg
La
mère,
il
prend le
garçon
The-F.Sg mother-F.Sg, he-M.Sg takes the-M.Sg boy-M.Sg ‘The mother, he takes the boy’
Puis après
il cook le marshmallows
Then afterwards he cook the-M.Sg marshmallows-M.Pl
Discuss the nature of the errors made by PB in these sentences, and say how Gopnik’s feature-blindness
model would account for each of the errors. If you know French, discuss the extent to which some of the
relevant errors may be the result of transfer from French (given that PB is a French-English bilingual), and
discuss the methodological issue which this raises.
§3.2 A methodological issue
Although Gopnik claims that PB shows a deficit in respect of grammatical features, she presents no
systematic quantitative data42 relating to the frequency of the errors she discusses in naturalistic samples43
of PB’s speech (e.g. on the relative frequency of structures like two cars and two car, or on the relative
frequency of one car and one cars). To what extent do you think this is a methodological weakness – and
if so, why?
§3.3 Rice and Oetting’s study
Rice and Oetting (1993) report on a study they undertook of 81 SLI children 5 years of age and 92 MLUmatched normally developing (=ND) children (around 3 years of age). One of the phenomena they studied
was number marking on nouns. They report that only 4/152 (2.6%) of the plural nouns produced by the
SLI children in their study occurred in singular contexts (e.g. one dishes, it toys44), compared to 3/93
(3.2%) of the plural nouns produced by the TD children: by contrast, the SLI children in Rice and
Oetting’s study showed more frequent use of bare nouns45 in plural contexts (53/404 = 16%) than TDs
(38/473 = 7%), with almost all such errors involving quantifier+noun structures like two eye46. To what
extent are these findings compatible with the Feature Deficit Hypothesis?
41
This is reported by Gopnik to have been produced in a present-tense context.
More particularly, she presents no quantitative data on PB’s spontaneous speech output, even though she does
present limited quantitative data in relation to other types of speech output (e.g. from elicitation tasks).
43
A naturalistic speech sample is a sample of the spontaneous speech (or ‘free speech’) of one person in
conversation with another (the other person often being an interviewer/researcher).
44
Corresponding to the adult form ‘It’s a toy’.
45
I.e. uninflected nouns not carrying the plural -s affix.
46
This is arguably because number is marked by the use of the dual numeral two, and hence marking number on the
noun is (in some informal sense) redundant. More generally, it may be that (more frequently than their TD
42
22
§3.4 On the nature of the feature deficit in SLI children
SLI studies typically report that SLI children perform much better on some morphemes than on others. For
example, Rice and Oetting (1993) report that SLI subjects performed almost as well on plural -s as MLUmatched TDs (83%/93%), but performed much worse on 3Sg -s than the TDs (36%/83%). To what extent
do Rice and Oetting’s findings challenge the feature-blindness view of SLI? What if we modified
Gopnik’s theory in such a way as to claim that it’s not impossible for SLI children to acquire and process
grammatical features, but rather that they are very slow to process and acquire such features. Because of
the processing load which features pose for an SLI child with a slow processor, the more features a
morpheme marks, the more difficult it is to acquire and the longer it takes SLI children to acquire it. To
what extent would this revised slow processor model account for Rice and Oetting’s findings? And to
what extent would it account for the sentences produced by PB in §3.1?
§3.5 On the nature of PB’s tense errors
Gopnik’s Feature Blindness theory predicts that SLI children are blind to (and hence unable to acquire)
grammatical features, including tense features. Which of the following five types of italicised verb form
would her theory predict that SLI children produce in past tense contexts – and why?
1. Daddy went/go/goes/gone/going to Paris yesterday
Compare this prediction with the observations which Gopnik makes about the range of verb forms which
PB uses in past tense contexts. One observation she makes is that ‘Regular past tense forms never occur in
spontaneous speech’ (i.e. he uses bare forms like play rather than past forms like played in past contexts),
but elsewhere (1990, p.155, Table 8) she reports that PB produced 5 correctly inflected regular past tense
forms on a writing task, and that he also produced some present-tense s-forms on a writing task. She notes
that ‘frequent irregular past tense forms do’ get used by PB in past contexts47. On a tense transposition
task (in which PB was asked to transpose a present-tense sentence containing today into a past tense
sentence containing yesterday), PB produced two irregular past tense forms (though no regular past tense
forms). In addition, we find the following two (italicised) irregular past tense forms produced by PB cited
in Gopnik’s text:
(a)
I was make 140 box
(b)
Who did that?
She also notes (1990, p.154) that in his speech production PB ‘virtually never added s to the verb to mark
the third person singular present’ – though the text of her paper cites him producing the following
example:
(c)
When the bus goes fast, he has an accident
The text also shows PB producing a number of structures such as those below in which he uses
appropriate present-tense forms of BE in present-tense contexts:
(d)
The witch is coming
(e) I am riding a bicycle
(f) All the girls sing and they are dancing
To what extent are such data consistent with Gopnik’s feature-blindness view of SLI on which children
like PB are completely blind to grammatical features like tense?
§3.6 The nature of tense omission errors
Gopnik maintains that PB has a defective syntax, and claims that his omission of present tense -s and past
tense -d on regular verbs in obligatory contexts shows that verbs are not marked for tense in the syntax.
However, an alternative view of the inflectional errors made by SLI children is that their syntax is
relatively unimpaired (so that verbs are correctly marked as present/past tense in the syntax in appropriate
contexts) but that they have morphological problems which mean that they sometimes fail to spell out the
tense feature on a verb (or, informally, they sometimes ‘forget’ to add the appropriate inflection on the end
counterparts) SLI children tend to omit grammatical features when they are redundant (i.e. when the relevant features
make no essential contribution to determining meaning).
47
Presumably by this she means that high-frequency irregular past tense forms like went/came are used by PB.
23
of a verb to mark its tense). Gopnik assumes that the subject of a verb which is marked for present/past
tense in the syntax is assigned nominative case (and hence is spelled out as a nominative form like I/he
rather than as an accusative form like me/him). In the light of this assumption, consider the nature of the
tense errors made by PB on the italicised verbs in the following sentences:
(a)
(b)
When the cup break, he get repair
I wait in the Berri-de-Montigny48
§3.7 On the nature of pro-drop
Gopnik observes that PB sometimes omits subject pronouns in obligatory contexts (i.e. in contexts in
which adults would require the use of an overt subject pronoun like I/you/he etc.): this phenomenon is
traditionally referred to as pro-drop49. She suggests (1990, p.158) that ‘subject pronouns can be dropped’
when ‘the feature [tense] is not present’: what this implies is that if (in reply to a question like Where did
Jamie go?) an SLI child uses a tense-marked verb like (past-tense) went, the child will use a subject
pronoun and say ‘He went home’ (and will not omit he). But if the child uses a tenseless verb form like
go, he can omit the subject pronoun and simply say Go home. What problems are posed for this claim by
the fact that PB omits subject pronouns in English/French sentences such as those below?
(a)
(b)
Can watch them at the Forums
Ramène le feu
Returns the fire (‘He brings the fire back’)
More recent work by Liliane Haegeman and Luigi Rizzi (outlined in the collection of papers in Friedeman
and Rizzi 2000) maintains that in cryptic styles of spoken/written adult English, a weak (i.e. unstressed or
unemphatic) subject pronoun can be undergo a process of TRUNCATION whereby it is truncated (i.e.
dropped) when it is the first word in the main clause of a sentence50, resulting in subjectless sentences
such as Can’t find my wallet. Don’t know where I left it51. Must be at home. What bearing might this have
on sentences like (a/b)?
§3.8 PB’s aspectual errors
What kind of errors does Gopnik’s feature blindness model predict that PB will make in marking
progressive aspect? Is PB’s predicted behaviour consistent with his reported behaviour on an elicitation
task, in which he was asked to describe the actions of hand puppets? On the relevant task, he produced 19
structures classified by Gopnik as obligatory contexts for marking progressive aspect, 7 like (a) containing
both BE and -ing, 6 like (b) containing only -ing, and 6 like (c) containing only BE:
(a)
The queen is sleeping
(b)
The dragon jumping
(c)
The dragon is walk
Could we account for the relevant pattern of errors rather better if we made the traditional assumption that
SLI children make sporadic omission errors, but almost never make commission errors?
3.10 Overall verdict
What is your overall verdict on the strengths and weaknesses of Gopnik’s (1990) feature-blindness paper?
48
Berri-de-Montigny is a metro/underground train station. This sentence was used to describe an event which had
taken place the previous year.
49
An alternative terminology which is frequently employed is to say that children use null subjects in contexts in
which adults require weak overt subject pronouns (so that apparently subjectless sentences contain a ‘silent’ subject
pronoun which has no overt phonetic form, but has the same semantic/grammatical properties as weak pronouns).
50
In terms of the analysis in Rizzi (2000), a weak subject pronoun can have a null spellout when it is the specifier of
a root (i.e.) main clause. For clarity, I present his proposal in a simplified (hence somewhat inaccurate) form. In the
written language, subject-drop is found mainly in so-called diary styles, which is why the phenomenon is sometimes
referred to as diary-drop. Spoken French allows a more restricted form of subject-drop, in which a ‘meaningless’
(non-referential) expletive subject pronoun can be dropped – as in the title of the French TV programme Faut pas
rêver ‘Needs not dream’ = Il faut pas rêver ‘It needs not dream’ – i.e. ‘You don’t have to dream’
51
Note that because it is not the first word in the sentence, the I subject of left cannot be dropped.
24
4. The Rule-Deficit Account of SLI
Key article
Gopnik M & Crago MB (1991) ‘Familial aggregation of a developmental disorder’, Cognition 39: 1-50
OUTLINE OF GOPNIK AND CRAGO’S STUDY
Gopnik and Crago (henceforth abbreviated to GC) report on research into three generations of a family
living in the East End of London (known as the KE family), 16 out of the 30 members of which are
dysphasic (i.e. have SLI). Their research hypothesis (1991, p.12) is that dysphasics are ‘impaired in only
one subpart of the grammar and have other parts of the grammar intact’. More specifically, they argue that
dysphasics have essentially the same syntactic abilities as MLU-matched controls but have a genetic
deficit which makes them unable to construct regular morphological rules: consequently they can only
learn morphology by memorising individual word-forms. Accordingly, GC claim (1991, p.47) that
dysphasics ‘have a learning mechanism that sees each word as an independent item that must be learned
and entered into a lexicon that specifies its grammatical properties and meaning.’
Background: Dual mechanism model
GC’s paper assumes the dual mechanism model of the acquisition of morphology (outlined e.g. in Pinker
and Prince 1988 and Pinker 1991). Within this model, irregular forms are memorised forms which are
stored/listed in the mental lexicon52, whereas regular forms are computed or derived via application of
morphological rules like those sketched informally in (1) and (2) below which say how regular verb/noun
forms are derived by adding particular affixes (more precisely suffixes) to the stem form53 of the item:
(1)
A regular verb carries the suffix:
{s} if third person singular present
{d} if past/perfect/passive
{ing} if progressive/gerund
{ø} otherwise
(2)
A regular noun carries the suffix:
{s} if plural
{ø} otherwise
So, for example, the lexical entry (i.e. entry in the dictionary) for the regular noun BOY will contain simply
the stem form boy: its plural will be formed by adding the suffix -s to the stem form in accordance with
the first line of rule (2), so deriving boy-s; its singular form will be generated by adding the null suffix -ø
in accordance with the second line of rule (2), generating boy-ø 54. By contrast, the lexical entry for the
irregular noun MAN will contain not only the singular form man but also the irregular plural form men: the
presence of the irregular plural form men in the lexicon will then block application of the regular plural
formation rule (1i), so accounting for the ungrammaticality of *mans as the plural of man55. Similarly, the
lexical entry for a regular verb like PLAY will specify simply that it has the stem form play, and regular
morphological rules will determine that its third person singular present tense form is derived by adding -s
to the stem form in accordance with the first line of rule (1), its past tense and perfect/passive participle
forms are derived by adding -d56 in accordance with the second line of rule (1), its progressive participle
52
Using this terminology, we can draw a distinction between stored (i.e. memorised irregular) forms and derived or
computed forms (i.e. rule-generated regular forms).
53
What is here referred to as the stem form generally corresponds to the citation form found in dictionaries.
54
Underlying the last line of rule (1) is the traditional idea that seemingly uninflected forms (i.e. forms which have
no overt inflection) contain a null affix (traditionally called a zero morpheme). If we made rather different
assumptions, we could take the last line of rule (1) – and likewise the last line of rule (2) – to mean that the relevant
form is derived by adding no suffix to the stem form of the relevant item.
55
If man is listed in the lexicon as the stem form, the singular form man will be derived by the addition of a null
affix in accordance with the second line of rule (2).
56
The past tense form occurs in sentences such as He played tennis yesterday, the perfect participle form in
sentences such as I have never played squash, and the passive participle form in sentences such as The game was
played in a spirit of competitive camaraderie. Note in particular that the participle form used with the auxiliary have
is a perfect participle in English, and not (as in languages like French) a past participle. Because French does have
genuine past participles, it can use the relevant participial structures in genuine past tense contexts – e.g. J’ai vu
25
and gerund form is derived by adding -ing in accordance with the third line of rule (1), and its other
inflected forms are derived by the addition of a null affix -ø in accordance with the fourth line of rule
(1)57. By contrast, the lexical entry for an irregular verb like GO will specify that it has the stem form go,
the irregular perfect participle form gone and the irregular past tense form went; it will follow from this
that its (regular) third person singular present tense form is derived by adding -s to the stem form (giving
goes) in accordance with the second line of rule (1), its (regular) progressive participle form by adding
-ing to the stem form (giving going) in accordance with the fourth line of rule (1), and its other forms by
adding the null affix -ø to the stem form (giving go+ø) in accordance with the last line of rule (1)58.
Methodology of Gopnic and Crago’s (1991) study
GC provide comparative data on 6 dysphasic59 members of the KE family they studied (aged 16, 17, 40,
42, 45 and 74), and 6 unaffected members (aged 8, 12, 13, 14, 15 and 17). For 10 of the 12 subjects GC’s
data come from comprehension and production tasks administered to them, but for the remaining 2
(dysphasic) children the data come from school compositions written by the children once a week.
Test results
GC report that there was no significant difference between the performance of the dysphasic members of
the family and the unaffected members on tests involving grammatical comprehension or grammaticality
judgements relating to co-ordinate structures60, negative active and passive sentences61, possessive
constructions62, argument structure63, pronoun reference64, gender65, and number66. From this, they
conclude that there is no global impairment of grammar in the dysphasic members of the family.
However, on 5 other tests, the SLI family members performed significantly worse than the normals. One
Marie hier (literally I’ve seen Marie yesterday, idiomatically ‘I saw Marie yesterday’), whereas English perfect
participles (because they mark perfect aspect and not past tense) cannot be used in a past tense context (e.g. in a
sentence containing an adverb like yesterday).
57
Hence the form play-ø would occur e.g. as an infinitive form in I want to play, a subjunctive form in They insisted
that he play the national anthem, an imperative form in Play it again, Sam! and a present-tense indicative form in
They never play our tune. As pointed out in footnote 54, we could alternatively assume that bare forms like these
contain no affix rather than a null affix.
58
Note that an important implication of this assumption is that an irregular verb like GO is typically irregular only in
some (i.e. one or more) of its forms, but regular in others.
59
Recall that Developmental Dysphasia and Specific Language Impairment are equivalent terms: hence a
developmental dysphasic is an individual who has SLI.
60
Like ‘Here are 3 crayons. Drop the yellow one on the floor, give me the blue one and pick up the red one.’
61
Subjects were shown two pictures, one of a truck pulling a car and the other of a car pulling a truck, and were
tested on sentences like ‘The truck does not pull the car’ and ‘The truck is not pulled by the car’.
62
Subjects were shown a picture (e.g.) of a mother pushing a baby in a carriage and asked to ‘Show me the mother’s
baby’ or ‘Show me the baby’s mother’.
63
In a sentence like The prisoners escaped, the verb escaped is said (in traditional work in Predicate Logic) to
function as a predicate which has a single argument (its argument being its subject the prisoners). In a sentence like
The police arrested a suspect, the verb arrested has two arguments, namely its subject the police and its complement
a suspect. In John sent a parcel to Mary, the verb sent has three arguments, namely its subject John, and its two
complements a parcel and to Mary. In the argument structure sentences given to the family members, they were
asked to judge the grammaticality of 5 incorrect and 6 correct sentences: the incorrect sentences had argument
structure errors, as in He eats a cookie to the boy and The boy puts the book.
64
Subjects were tested for whether they treated two pronouns as coreferential (i.e. as referring to the same
individual) in sentences like ‘He washes himself’ on the one hand and ‘He washes him’ on the other.
65
Subjects were shown pictures of a man holding another man, a man holding a woman, a woman holding a man,
and a woman holding another woman and asked to point to the picture which can be described by saying e.g. ‘He
holds her’.
66
In one test, subjects were shown 16 objects (4 books, 4 crayons, 4 coins and 4 balloons) placed on a table in front
of them; the objects were randomly arranged in 8 piles, with each pile containing either a single item or 3 items of
the same kind. They were then given commands such as ‘Please touch the book’ or ‘Please touch the books’. In a
second test, they were asked to perform more complex tasks like ‘Put the crayon on the balloons’.
26
was on a test eliciting plurals of novel words67 (normals achieving a mean score of 83% correct and
dysphasics 47%). Another was on a grammaticality judgement test in which subjects were asked to judge
whether 9 correct and 21 incorrect test sentences (containing person/number/tense/aspect errors68) were
grammatical or not: the normals achieved a mean score of 92% correct judgements, and the dysphasics
57%. A third was on a test of their ability to produce verbs inflected for tense and aspect: normals
achieved a mean score of 72% correct, and dysphasics 30% correct69. A fourth was on a derivational
morphology task70 in which the normals achieved a mean score of 95% correct and the dysphasics 29%. A
fifth was on a narrative task in which the subjects were shown 6 cartoons and asked to tell the story they
depict: they were then scored in respect of the percentage of their referring expressions (i.e. the
expressions they used to refer to characters in the cartoons) which were pronouns (like he) or noun
expressions (like the man); normals showed 45% mean use of pronouns71, and dysphasics 9% (so that the
dysphasics used far fewer pronouns and far more noun expressions than the normals). In addition, GC
note that the dysphasics took significantly longer to respond to test stimuli than the normals: for example,
the mean response time for normals on the grammaticality judgement test for person/number/tense/aspect
was 52 seconds for the normals and 112 seconds for the dysphasics. GC also note that the dysphasics
reported that they found such judgement tasks very difficult, making remarks like ‘I’m not sure if that’s
right or wrong’. Gopnik and Crago conclude that the dysphasic members of the family show selective
impairment of specific grammatical abilities.
Results of the notebook study
Data for two of the dysphasic children (a 10-year-old boy T and an 11-year-old girl C) came from their
school notebooks. Every Monday throughout the school term, the children wrote a composition about
what had happened over the weekend. T’s notebook contained 25 entries with a mean length of around 35
words, C’s contained 21 entries with a mean length of around 50 words. Gopnik and Crago scored the use
of verb forms by the children, and report that they performed much better on irregular verbs than regular
verbs (T/C correctly inflecting 80%/90% of irregular verbs and 31%/39% of regular verbs respectively).
Gopnik and Crago’s conclusions
Noting that their dysphasic subjects had problems in pluralising nonsense words, GC conjecture that the
dysphasics lacked the regular plural formation rule (1i) which the normals had acquired. They account for
the good performance of the dysphasics on tests of their comprehension of the singular and plural forms of
real nouns by positing that dysphasics learn singular and plural forms of real words by memorising them
(and thereby are able to produce and comprehend memorised forms of real words). What this means in
more concrete terms is the following. For normal subjects, the lexical entry for a regular noun like BOY
will contain only the stem form boy, with the plural form boys derived by application of the pluralisation
rule (1i)72. By contrast, for dysphasics the lexical entry for the relevant noun will contain both the singular
form boy and the plural form boys73, and they will have no counterpart of the pluralisation rule (1i) in their
67
Also referred to as nonsense words. In this test, the subjects were shown pictures of imaginary animals. The
experimenter would point to a picture of an imaginary animal and say ‘This is a zoop’, and then point to a picture of
2 or 3 of the same animals and say ‘These are ...?’, inviting the subject to complete the sentence.
68
CG (1991, p.22) remark that their test sentences contained ‘errors in number The boy eats three cookie; person The
boy kiss a pretty girl; tense Yesterday the girl pet a dog; and aspect The little girl is play with her doll’.
69
Subjects were asked to complete the second sentence in sequences such as ‘Every day he walks 8 miles. Yesterday
he......’ ‘The boy always cries. Right now he.....’ ‘Yesterday the girl baked a cake. Tomorrow she......’ The percent
correct use figures given here are based on the more ‘generous’ of the two scoring procedures outlined by GC.
70
Subjects were asked to complete the second sentence in sequences such as ‘There is a lot of sun. It is very...’ (to
test whether they can form the adjectival derivative sunny from the noun sun). The figures given here are based on
the more generous of the two scoring procedures suggested by GC (under which both proud and prideful would be
taken as indicating that the subject can form an adjectival derivative of pride).
71
I.e. a mean of 45% of their referring expressions were pronominal, the other 55% being nominal.
72
Given the last line of the rule given earlier in (1) in the main text, the singular form boy will be derived by adding a
null affix -ø to the stem form boy, forming boy+ø.
73
In other words, dysphasics treat regular plural nouns in the same way as normal adults treat irregular nouns like
man/men.
27
grammar. More generally, GC hypothesise that dysphasics have a rule deficit which makes it impossible
for them to form morphological rules, and conjecture that the deficit is genetic in nature.
Predictions made by Gopnik and Crago’s analysis
The assumption that dysphasics have a genetic deficit which makes it impossible for them to construct
regular morphological rules makes a number of interesting predictions. One is that dysphasics will be
unable to inflect novel words (e.g. unable to inflect a nonsense word like plam for past tense): this is
because past tense formation in normal adults requires you to:
(3)(i)
(ii)
Go to your mental lexicon, see if you have an irregular past tense form listed for the verb in
question, and retrieve it if you have
If not, apply the regular past tense d-formation rule in (1)
Now, since your mental lexicon cannot in principle contain an entry for the past tense form of an invented
nonsense word like plam, the only way of forming its past tense is to fall back on the regular past tense
d-rule in (1): but since dysphasics (according to GC) cannot acquire regular morphological rules like (1),
the Rule Deficit Hypothesis predicts that they will be unable to produce the appropriate past tense form
plammed (and will instead just use the ‘bare’ uninflected infinitive plam).
A second prediction which the Rule Deficit Hypothesis makes is that dysphasics will be unable to form
over-regularised forms like (e.g.) past tense forms such as goed/comed/buyed/seed etc. This is for the
following reason. When normally developing children want to use a past tense form of an irregular verb
(say BUY), they will first (in accordance with (3i) above) go to their mental lexicon and see if it is a verb
for which they have an irregular past tense form stored in their lexicon; if they haven’t yet learned the
irregular past tense form bought (or have only just begun to learn it and still have problems retrieving it
because they haven’t heard or used it often enough to be able to retrieve it unfailingly), they will fail to
retrieve the adult irregular form bought and will instead fall back on (3ii), applying the regular past tense
d-rule in (1), so producing the over-regularised form buyed. But because the Rule Deficit Hypothesis
claims that dysphasics have a genetic deficit which makes them incapable of acquiring regular
morphological rules like (1), they cannot in principle fall back on (3ii) and so will instead simply use the
‘bare’ uninflected form buy e.g. in sentences like Daddy buy me one yesterday.
A third prediction which the Rule Deficit hypothesis makes is this. Since (by hypothesis) dysphasics have
to memorise all inflected forms (regular and irregular alike), and since we know that memorisation
improves with exposure and use (in the sense that the more often you are exposed to hearing other people
use a given word-form, the more quickly you memorise it and the more easily you are able to retrieve it
from your memory when you need it), this predicts that dysphasics will perform better on those forms
which they hear more frequently than other forms. Since many of the most frequent past tense verb forms
they hear and use tend to be irregular verbs like came/went, this might lead us to expect that dysphasics
will generally perform better on irregular than regular past tense verb forms.
Reasoning along similar lines, GC argue that the much better performance of the two dysphasics in the
notebook study on irregular than regular past tense forms can be accounted for in terms of their rule deficit
hypothesis. Irregular past forms are memorised, and listed as stored forms in the mental lexicon. Regular
forms cannot be rule-generated (because the dysphasics lack the capacity to form morphological rules)
and so can only be learned as memorised forms which are stored as such in the lexicon. Since
memorisation requires considerable exposure to the relevant forms74, dysphasics will only be able to
produce high-frequency inflected forms of regular verbs (i.e. verb forms which have a very high frequency
of occurrence in the adult speech input which they receive)75. In support of this claim, GC note that of the
74
Recent work by Maslen et al. (2003) estimates that normally developing children need to hear a given verb-form
(e.g. an irregular past tense form like ate) around 10,000 times (and make around 1,000 attempts at producing it)
before they can consistently produce the relevant form without error. Presumably dysphasics require far greater
exposure (perhaps twice as much).
75
GC (1991, p.41) conjecture that to memorize inflected forms (both regular and irregular) would take 30 or 40 years
to accomplish, and even then is likely to be accomplished imperfectly.
28
11 regular verbs in the notebook compositions, only 4 were correct in their first occurrence (namely
showed, asked, called and picked) and that three of these four verbs occur more frequently in past tense
forms than in present tense forms76.
WORKBOOK SECTION
The key assumption in GC’s (1991) paper is that SLI children have a genetic deficit which makes it
impossible for them to acquire regular morphological rules. However, one criticism which could be made
of such a claim is that it is too extreme, and that rather than suppose that SLI children are unable to
acquire regular morphological rules, we should instead suppose that they are much slower than typically
developing children in acquiring morphological rules (in the same way as GC showed that their SLI
subjects were twice as slow as their unaffected subjects in giving grammaticality judgments). This would
allow for the possibility that a dysphasic (at a particular stage of development) may have acquired some
but not all the regular morphological rules in a language (e.g. the noun-plural s-rule, but not the presenttense s-rule). The questions and exercises below are designed to encourage you to think more critically
about a range of aspects of GC’s paper.
§4.1 A methodological issue
GC claim that close study of the weekly notebook compositions of two of the dysphasic children provides
further evidence that the children acquire regular past tense forms by memorisation rather than ruleformation. For example, the first entry of the year in T’s notebook contains the following (verb forms used
in past tense contexts being italicised):
(a)
On Saturday I watch TV and I watch plastic man and I watch football. On Sunday I had pork and
potato and cabbage.
The teacher corrects the entry by adding -ed to each occurrence of watch. T gets watched right the next
time he uses it (5 weeks later) in the entry in (b) below, but does not add -ed to other regular verbs like
wash and dress in contexts where -ed is required: cf.
(b)
On Saturday I got up and I wash my self and I get dress and I eat my breakfast and I watched TV
all day and I went to bed. On Sunday, I got up and...
Again, the teacher corrects the notebook, changing wash/dress/get/eat to washed/dressed/got/ate. Again T
gets the corrected forms right in later work, and two stories later he writes:
(c)
On Saturday I got up and I got dressed 77 and I watched Motormouth...and I ate my dinner
GC (1991, p.39) conclude that ‘The pattern throughout the year demonstrates clearly that the subject
learns the individual past tense forms as they are corrected by the teacher, but does not generalize these
corrections to new verbs’. What do you think are the potential methodological pitfalls of using evidence
from a notebook study – and to what extent are the conclusions GC draw from the relevant notebook
evidence justified? Can you offer an alternative accounts of T’s behaviour?
§4.2 Ullman and Gopnik’s (1994) follow-up study
In a (1994) follow-up study of the KE family, Ullman and Gopnik (henceforth UG) examined 10 of the
family members (7 of whom had been diagnosed as language-impaired), testing their ability to form the
past tense of 16 existing regular verbs (e.g. look), 14 existing irregulars (e.g. dig), 12 novel regulars (e.g.
plam), and 14 novel irregulars (e.g. crive)78. Some of the dysphasic family members produced no novel
They cite the following frequency data from Kucera and Francis’ (1957) frequency dictionary: ask = 128,
asked = 398; call = 188, called = 401; pick = 55, picked = 78; show = 287, showed = 141.
77
The word dressed is not a past tense verb form, but rather a passive participle, perhaps used adjectivally.
78
Novel irregulars are nonsense verbs which might be expected to have an irregular past tense form by analogy with
existing irregulars: e.g. we might expect the novel verb crive to have the past tense form crove by analogy with the
existing (rhyming) irregular drive which has the past tense form drove.
76
29
regular past tense forms (e.g. plammed) at all and no over-regularised past forms (e.g. digged)79, although
they did nonetheless produce past tense forms of some existing (regular and irregular) past tense verbs. To
what extent is the behaviour of these family members consistent with GC’s Rule Deficit model?
§4.3 Performance on novel words
As noted above, GC report that their subjects achieve a significantly lower score on novel plurals than the
normal family-members (47% compared to 83%). Ullman and Gopnik (1994) report that one member of
the KE family (AD) produced 1 novel past tense form and a number of over-regularised past forms on a
production task. In a similar vein, Vargha-Khadem et al (1995, p.931) report (in relation to a past tense
production task with irregular verbs given to affected members of the KE family) that ‘Overregularisations constituted 41% of all errors’. To what extent are these findings consistent with GC’s
hypothesis that dysphasics are unable to form morphological rules? What alternative possibility can you
suggest?
§4.4 Findings from other SLI studies
Rice and Oetting (1993) report that the SLI subjects in their own study produced noun plurals like mans
and pronoun plurals like herselves. Reporting on a case study of a 10-year old boy with SLI (identified as
AZ), Heather van der Lely (1998, p 164) notes AZ producing nominals such as two foots and two mens, as
well as sentences like This is what they ated. In a study of 12 SLI children ranging in age from 8;2 to
12;11, Dorothy Bishop (1994) reports that with existing irregular verbs in past tense contexts, the children
(identified by letters and numbers) typically alternated between producing the correct irregular past form
and producing a bare form: cf.
(b)
(c)
Took it off (D17, in reply to ‘What did they do with the top part of the pram?’)
It take me a long time (D17 in reply to ‘Did it take you a long time to get better?)
Bishop also reports her SLI subjects producing over-regularised past tense/perfect participle forms like
falled, losed, throwed, runned, and taked and the ‘doubly inflected’ forms sawed and broked: cf.
(d)
(e)
(f)
(g)
And then Mummy taked to the garage to xxx (J03)
He falled in (J03, reply to ‘What did Andrew do when the ice gave way?’)
He sawed mine brother (J03, reply to ‘Has the doctor ever been to see you?’)
The car has broked down (J03)
In addition, Ramos and Roeper (1995) report on a 4-year old boy with SLI who shows 0% use of 3Sg
present tense {s} and genitive {s}, while nonetheless using (and over-generalising) plural {s}, past tense
{d} and progressive {ing}. Consider the implications of these various findings for the Rule Deficit Model
of SLI. Could we provide a better account of the relevant data if we assumed that SLI children have a
Generalised Morphological Deficit which causes them problems with acquiring and retrieving irregular
forms, and likewise with acquiring and applying regular morphological rules (so that e.g. they sometimes
under-apply the regular past tense d-rule it and fail to apply it to a regular verb, and conversely sometimes
over-apply it and so wrongly apply it to an irregular verb)? How would such a model account for SLI
children alternating (in past tense contexts) between forms like took, tooked, taked and take?
§4.5 Performance of dysphasics on existing irregulars
Ullman and Gopnik (1994) note that dysphasic members of the KE family had problems not only in
producing past tense forms of existing regular verbs but also in producing past tense forms of existing
irregular verbs: e.g. VA scored 0% correct on existing irregulars, ST 7%, KA 21% and AD 36%. They
also report (1994, p. 111) that ‘Irregular pasts were not produced more successfully than regular pasts’. In
much the same way, Vargha-Khadem et al. (1995) claim in relation to a past tense production task given
to dysphasic members of the KE family (pp.930-931) that ‘Their impairment was equally evident on
irregular forms…and regular forms’, further commenting in relation to irregulars that ‘The affected
79
Over-regularisation involves treating an irregular verb like dig (which has the past dug) as if it were regular and
adding the regular past tense ending -ed to its stem, forming digged.
30
members were impaired to about the same degree on these as on the regular forms’. Discuss the
implications of these findings for the Rule Deficit model of SLI.
§4.6 Differential performance on inflections
A number of studies report that SLI children perform much better on some inflections than on others. For
example, Rice & Oetting (1993) report that a group of SLI children they tested achieved 83% accuracy on
noun plural -s but only 36% accuracy on 3Sg -s. Likewise, Leonard (1989, p.181) notes that for the SLI
children he studied, the morpheme -ing was ‘only slightly below expectations based on MLU’ (i.e. SLI
children scored almost as highly on -ing as MLU-matched TD children); whereas Leonard (1995) reports
only 32% and 34% correct use for past tense -d and present tense -s respectively. Likewise, Dorothy
Bishop (1994) in a study of 12 SLI children ranging in age from 8;2 to 12;11 reports that her subjects
achieved between 92% and 100% correct use of plural -s, but 67% correct use of 3Sg present tense -s.
Discuss the implications of such studies for the Rule Deficit model of SLI.
§4.7 Progressive structures
The pattern of errors which GC (1991, p.41) report for progressive structures involves not only omission
of the progressive morpheme -ing in structures like (a) below but also omission of the progressive
auxiliary BE in structures like (b):
(a)
Carol is cry in the church
(b)
I walking down the road
To what extent can the Rule Deficit model of SLI account for either or both of these errors?
§4.8 Pronoun avoidance
One of the findings of Gopnik and Crago’s study was that SLI-affected members of the KE family showed
a much lower use of pronouns on the narrative task than the unaffected members, the affected members
showing 9% mean use of pronouns, and unaffected members 45%. To what extent does their Rule Deficit
hypothesis account for this finding – and (if not) how else do you think it might be accounted for?
§4.9 On the nature of the impairment in members of the KE family
Vargha-Khadem et al (1995, p.930) report that dysphasic members of the KE family showed ‘impaired
processing and expression of other areas of grammar’ (they cite problems with processing relative clauses)
‘grossly defective articulation of speech sounds, and, further, a severe extralinguistic orofacial dyspraxia’
(the latter resulting in difficulty in performing actions such as clicking their tongue, humming a tune,
biting their lip, licking their upper lip, smacking their lips, closing their left eye, sticking out their tongue
etc.). In addition, they note that 6 of the affected members of the KE family have both verbal and
performance Intelligence Quotient (IQ) scores below 85. What are the implications of these observations
for Gopnik and Crago’s Rule Deficit account of SLI?
§4.10 Findings of the Hurst study
Hurst et al (1990) tested 4 language-impaired members of the KE family and found that they ‘took a long
time to name pictures of objects with which they were familiar’ and sometimes produced inappropriate
responses ‘for example glass or tea for cup, and sky for star.’ What implications (if any) do you think their
observations might have for (e.g.) the poor performance of affected KE family members on the use of
inflected forms (such as past tense verb forms)?
§4.12 Overall Evaluation
In the light of the questions raised above, what’s your overall evaluation of the strong and weak points of
GC’s rule-deficit theory of SLI?
31
5. The Agreement Deficit account of SLI
Key article
Clahsen H, Bartke S and Göllner S (1997) ‘Formal features in impaired grammars: a comparison of
English and German SLI children’, Journal of Neurolinguistics 10: 151-171
OUTINE OF CLAHSEN, BARTKE AND GÖLLNER’S STUDY
Clahsen, Bartke and Göllner (henceforth CBG) use data from English and German children with SLI to
argue that SLI involves a selective grammatical deficit. They claim that both the English and German SLI
children in their study had far more problems in marking agreement than tense on (auxiliary and main)
verbs. In the light of the claim in Chomsky (1995) that tense features on verbs are interpretable (i.e. they
contribute to determining the meaning of sentences) whereas agreement features on verbs are
uninterpretable (i.e. make no contribution to meaning), CBG conclude that SLI children have problems
with acquiring uninterpretable features. Since understanding CBG’s work requires an understanding of the
difference between interpretable and uninterpretable features, let’s begin by looking at this.
Interpretable and uninterpretable features
Consider the grammatical features80 encoded by she and is in sentence (1a) below, which include those
given in (1b/c):
(1)(a)
(b)
(c)
He is in Paris
he = [third-person, singular-number, masculine-gender, nominative-case]
is = [third-person, singular-number, present-tense]
In his (1995) book The Minimalist Program, Chomsky draws a distinction between interpretable features
(i.e. those features which have semantic content and play a role in determining the meaning of sentences),
and uninterpretable features (i.e. those features which have no semantic content and play no role in
determining meaning, but rather simply serve some purely grammatical function). Chomsky maintains
that the person, number and gender properties of personal pronouns in English (which he terms their
φ-features ‘phi features’) are interpretable features, since a third person pronoun like he has a different
meaning from a second person pronoun like you, a singular pronoun like he has a different meaning from
a plural pronoun like they, and a masculine pronoun like he has a different meaning from a feminine
pronoun like she. By contrast, Chomsky claims that (nominative/accusative/genitive) case features are
uninterpretable81 – as we can illustrate in terms of the sentences in (2) below:
(2)(a)
(b)
(c)
I hadn’t expected [that he was announcing a takeover]
I hadn’t expected [him to announce a takeover]
I hadn’t expected [his announcing a takeover]
The case properties of each italicised pronoun in (2) are determined by the position which it occupies in
the syntactic structure containing it: the pronoun is nominative (he) in (2a) by virtue of being the subject
of the agreeing (auxiliary) verb was; it is accusative (him) in (2b) because it falls within the domain82 of
the transitive verb expect; it is genitive in (2c) because it is the subject of the gerund deciding (and the
80
Chomsky refers to grammatical features as formal features (since they play a role in determining the superficial
form of words), and CBG use this term throughout their paper. I shall keep to the traditional (and more familiar) term
grammatical features here.
81
More accurately, Chomsky claims that structural case features are uninterpretable. He draws a distinction between
structural and inherent case, structural case being assigned to an expression by virtue of the position it occupies in
the syntactic structure containing it, and inherent case being assigned to an expression by virtue of the semantic role
it fulfils. Case in English is generally taken to be structural.
82
Informally, domain can be characterised as ‘sphere of influence’. More formally, it can be defined in terms of the
relation c-command in ways which I will not explore here.
32
subject of a gerund in literary styles of English carries genitive case)83. In all three occurrences in (2), the
pronoun he/him/his has the same meaning (i.e. it is a third person singular masculine pronoun), suggesting
that case-features are uninterpretable and so play no role in determining the meaning of the pronoun.
Now consider the features carried by the copular (i.e. ‘linking’) verb is in (1a) He is in Paris – shown in
(1c) above. The present-tense feature carried by is must be an interpretable feature, since the meaning of
the sentence changes if we replace the present-tense form is by its past tense counterpart was. By contrast,
the person-number features of verbs are uninterpretable, since they are simply assigned to the verb via
agreement with the subject: i.e. the [third-person, singular-number] features carried by the verb is in (1)
He is in Paris are simply a copy of those carried by the subject he. As far as semantic interpretation (i.e.
meaning) is concerned, it is the person/number/gender features of the subject he which are interpretable
and so play a role in determining meaning (since they tell us that he refers to a single male entity which is
neither speaker nor addressee), not the person/number features of the verb is (which are uninterpretable84).
In the context of CBG’s paper, the important generalisations which come out of the discussion above are:
(3)(a)
(b)
The (person/number/gender) phi-features of noun/pronoun expressions are interpretable, but
their case-features are uninterpretable
The tense features of verbs are interpretable, but their person/number (agreement) features are
uninterpretable
The English children studied by CBG
The English data for CBG’s study come from elicitation tasks performed on a group of 9 SLI children in
the age range 10;00-13;0185. On one task, the children were prompted to produce 3Sg present tense forms
(using the prompt Every morning, my mum...), and on another to produce past tense forms (on a storytelling task using the prompt Once upon a time...). Both elicitation tasks were carried out twice for each
child, at intervals of a year. On the basis of the responses the children produced, they were scored for their
percent correct marking of case on subjects and tense/agreement on verbs. CBG report that the English
children in their study achieved relatively high scores on the past tense elicitation task (overall86 correctly
inflecting 76% of main verbs and 89% of auxiliaries for past tense in obligatory contexts), but much lower
scores on 3Sg present tense forms (overall correctly inflecting 49%87 of main verbs and 35% of auxiliaries
in obligatory contexts). They also report that the children achieved a 100% correct score on nominative
case-marking in obligatory contexts (all 217 of their subjects being assigned nominative case, according to
CBG).
Conclusions drawn from CBG’s English SLI study
The fact that the English SLI children achieved relatively high scores on past tense forms but much lower
scores on 3Sg present tense forms leads CBG to conclude that the reason for their poorer performance on
3Sg present tense forms is the fact that the latter encode not only tense but also agreement in (person and
number) phi-features with the subject. More specifically, they posit that SLI children have far greater
problems in marking agreement (viz. between subject and verb) than in marking tense: since past tense
83
Simplifying somewhat, we can say that verbs carrying the inflection -ing are traditionally classified as progressive
participles when used in conjunction with BE in sentences like He is lying, but as gerunds in other uses. The use of
genitive subjects with gerunds is restricted to formal (e.g. literary) styles, colloquial English using accusative
subjects with gerunds – as in ‘It was really upsetting, him not turning up to the party.’
84
If we replace the third person singular form is by the first person singular form am, we don’t change the meaning
of the sentence at all, but simply produce an ungrammatical sentence *He am in Paris.
85
The data originally came from a study conducted by Heather van der Lely, currently a professor in the Psychology
Department at University College London.
86
I.e. if we aggregate the figures from the 1993 and 1994 elicitation tasks.
87
There seems to be inconsistency in the figures given by CBG: on p.155, they report that the children correctly
inflected 40% of main verbs for 3Sg present -s in 1993 and 48% in 1994; yet over the page, they report that their
aggregate score was 49%. There would seem to be a scoring error of some kind (perhaps in the aggregate figure).
33
forms generally encode only tense88, the SLI children achieve high scores on this (76% for main verbs and
89% for auxiliaries); since 3Sg present forms encode not only present tense but also agreement, the SLI
children achieve much lower scores on this (49% on main verbs and 35% on auxiliaries).
Why should SLI children have greater problems in marking agreement on verbs than in marking tense on
verbs? CBG argue that this is a consequence of the tense features of verbs being interpretable and their
(person/number) agreement features being uninterpretable, and of the particular problems which
uninterpretable features pose for SLI children. However, CBG argue that it cannot be that all
uninterpretable features pose problems for SLI children, since the case features of pronouns are
uninterpretable, yet the English children in their study make no case-marking errors of any kind. They
therefore conclude that it is specifically uninterpretable phi-features which pose problems for SLI
children: since the uninterpretable person/number phi-features on verbs are agreement features, this
amounts to claiming that SLI children have an agreement deficit.
The German children studied by CBG
CBG’s study also includes one-hour spontaneous speech samples from 6 German SLI children aged 5;87;11. They report that the German SLI children they studied showed 99% correct tense marking89, but
only 64% correct marking of subject-verb agreement. They conclude (1997, p.157) that the relevant data
support their hypothesis that ‘Agreement phenomena cause major problems for English and German SLI
children’. They also report that the German SLI children generally show correct marking of number,
gender and person on noun and pronoun expressions: since these are interpretable features90, this is
consistent with their hypothesis that interpretable features pose few problems for SLI children. They also
note that subjects are correctly assigned nominative case (which would appear to be consistent with their
hypothesis that it is specifically uninterpretable phi features – e.g. agreement features of verbs – which are
impaired in SLI, not uninterpretable case features)91.
Tsimpli and Stavrakaki’s Uninterpretable Feature Deficit account
In an (1999) paper, Tsimpli and Stavrakaki (= TS), propose the more general hypothesis that SLIs have
problems in acquiring all uninterpretable features92: for succinctness, let’s refer to this as the
Uninterpretale Feature Deficit Model/UFDM. TS adduce empirical evidence in support of UFDM from a
study of a 5-year-old Greek girl with SLI called Eva. They claim that Eva has problems with agreement,
88
The one exception to this are past tense forms of BE, which inflect for both tense and agreement in standard
varieties of English (were being used with second person or plural subjects, and was with other subjects). In some
varieties of English, was is used with all choices of subject (cf. You was wrong and we was right).
89
This score is determined from their use of German preterite forms in obligatory contexts.
90
However, it is far from clear that gender features are always interpretable in German. For example, the noun
Mädchen ‘girl’ is neuter in gender in German, but denotes a female entity.
91
However, CBG concede that the strength of this finding is weakened if (as they assume) nominative is the default
case in German. Default case is the case assigned to an expression which could not otherwise be assigned case (i.e.
which is not within the domain of any case assigner). For example, in a sentence such as the following
(i)
Me lie to you? Never!
the subject me cannot be assigned case by the infinitive form lie, since infinitives do not case-mark their subjects in
English: hence, the subject is assigned accusative case by default (as a way of ensuring that it gets case). CBG’s
German SLI children produce infinitival main clauses such as:
(ii)
Der essen gänse ‘He eat.infinitive geese’ [= He eat geese]
and the subject is assigned default case (which is nominative in German).
CBG also discuss word order phenomena relating to the position of finite and nonfinite verbs in German, but their
discussion raises a number of complex technical issues which it would take us too far astray to delve into here.
92
This possibility was envisaged in Clahsen, Bartkle and Göllner (1997, p. 153), but rejected by them in favour of
their alternative view which sees SLI involving a selective deficit affecting only a subset of uninterpretable features –
namely agreement features.
34
case-marking, definite articles and clitic pronouns93. For example, they report that Eva correctly marked
subject-verb agreement in only 5/169 (3%) contexts with second person plural subjects; that she correctly
marked (unambiguous) case on nouns in only 3/13 (23%) obligatory contexts; that she omitted 3Sg object
clitics in 83/86 (97%) obligatory contexts; and that she omitted the definite article in 199/208 obligatory
contexts (96%). Since case and agreement features are uninterpretable (and since TS take object clitics
and definite articles to comprise sets of uninterpretable case and agreement features), these results are
consistent with an inability to acquire all uninterpretable features. By contrast, they report (p.52) that
‘Eva’s performance on the Greek tense system shows mastery of the present-past distinction’ – and since
tense is an interpretable feature, this observation is consistent with the view that interpretable features do
not pose particular problems for SLIs. The overall conclusion which Tsimpli and Stavrakaki draw is that
‘Non-interpretable features are not part of the SLI grammar’ (op.cit p.78).
WORKBOOK SECTION
CBG’s paper argues that SLI involves a selective grammatical deficit which affects the ability of affected
children to mark agreement. Moreover, they attempt to find a principled basis for this deficit in linguistic
theory (in terms of Chomsky’s distinctions between interpretable and uninterpretable features, and
between phi- and case-features). TS propose the more general hypothesis that all uninterpretable features
are impaired in SLI grammars. But is the evidence in support of their claims persuasive? The questions
and exercises below are designed to help you arrive at a reasoned answer to this question.
§5.1 On description and explanation
Any attempt to account for a set of findings about what SLI children do and don’t say must not only
describe the relevant data adequately, but must also provide a principled explanation of the relevant data
in terms of some relevant theory (whether of language, language acquisition, or language impairment).
To what extent are CBG and TS successful in providing a principled theoretical explanation for their
finding?
§5.2 Nominative case assignment
Finite verbs and auxiliaries in adult English require a nominative subject (e.g. the finite auxiliary has in
He has left requires a nominative subject like he and cannot have an accusative subject like him). CBG
(1997, p.158) account for this by supposing that a finite (auxiliary or non-auxiliary) verb94 has an
uninterpretable case feature which they refer to as ‘Assign NOM’ (meaning ‘Assign the feature
nominative-case to my subject’). Given their assumption about nominative case-marking, to what extent
are CBG correct in saying that their model predicts that SLI children will make no case-marking errors in
nominative contexts (i.e. in structures where adults would use a nominative subject)?
§5.3 On SLI children’s pronoun errors
In earlier work, Loeb and Leonard (1991) reported on a study of the acquisition personal pronouns by 8
English SLI children (in the age-range 4;0-5;0). Data were elicited from the children by using play
activities, by asking them to tell stories about using picture cards and to describe photos, and (where
necessary) by elicitation (e.g. trying to elicit the form is by saying ‘This shoe is dirty but...’ and pointing
to a picture of a clean shoe). The children’s correct/incorrect use of the 3rd person singular animate
93
Clitic pronouns are weak/unstressed pronouns which have the property that they must cliticise (i.e. attach) to an
appropriate kind of host (e.g. a finite verb).
94
A complication overlooked here is that within Chomsky’s CP/TP/VP analysis of clause structure, it is T which
carries the relevant case feature rather than the verb itself – but this does not affect the point of principle raised here.
35
pronouns (he/she) was scored in relation to the number of errors the children made in respect of the case,
number, person and gender of the pronoun95. The table below shows the raw number of pronoun errors
made by the children in Loeb and Leonard’s study96:
Raw number of pronoun errors made by the SLI children in Loeb and Leonard’s study
Child
Age
MLU
Case
Gender Number Person Ca+Ge Ca+Nu
SLI1
4;0
2.6
259
0
0
0
0
0
SLI2
4;3
2.8
30
0
0
0
1
4
SLI3
4;5
3.2
16
74
0
0
1
0
SLI4
4;4
3.5
180
0
1
0
0
0
SLI5
5;0
3.8
4
41
1
0
0
0
SLI6
4;3
4.3
70
0
0
0
0
0
SLI7
4;8
4.3
4
2
2
0
0
0
SLI8
5;0
4.5
5
16
0
0
0
0
Which aspects of Loeb and Leonard’s findings seem to prove problematic for CBG’s view that the core
deficit in SLI children lies in their ability to process uninterpretable agreement features of verbs? Does
TS’s view that SLI is characterised by an impaired ability to acquire all uninterpretable features fare any
better in accounting for the data? Are there any findings which neither model can account for? Why do
you think the children in Loeb and Leonard’s study made a large number of case errors, whereas the
English children in CBG’s study did not?
§5.4 Implications of other studies of inflectional morphology
A number of other studies of SLI (e.g. Leonard 1989, 1995; Rice and Oetting 1993) have reported that SLI
children (around the age of 5) tend to perform very well on noun plural forms (e.g. doggies) and on
progressive -ing verb forms (e.g. playing), but less (than half as) well on past tense marking (for regular
and irregular verbs alike). Are these findings consistent with the Agreement Deficit and/or Uninterpretable
Feature Deficit accounts of SLI?
§5.5 On interpreting CBG’s scores
CBG report that the ESLIs in their study (in obligatory contexts) showed 76% suppliance of past tense
marking on main verbs, but only 49% suppliance of third person singular present tense -s on main verbs,
concluding from this that ESLIs have more problems with marking uninterpretable agreement features on
verbs than with marking interpretable tense features. However, if we suppose that correct use of 3Sg
Present -s requires three separate features (person, number and tense) to be marked and that these three
features are independent of each other, and if we further suppose that the ESLIs in CBG’s study have a
0.76 probability of marking tense in obligatory contexts (i.e. they mark it in 76% of obligatory contexts),
then it is necessarily the case that the ESLIs must be better at marking either person or number (or both)
than at marking tense in order to account for their 49% suppliance of 3Sg Present -s. For example, if they
perform at the same level on marking number as on marking person, we would need to assume that they
have an 80% suppliance rate for person and an 80% suppliance rate for number in order to account for
their 49% suppliance of 3SgPres -s (because the probability of them marking person, number and tense
simultaneously in obligatory contexts for third person singular present tense -s would be 0.76 x 0.8 x 0.8
– i.e. 0.49). What are the implications of this observation for the Agreement Deficit model proposed by
CBG (and for the Uninterpretable Feature Deficit model of Tsimpli and Stavrakaki)?
95
The concentration on third person singular pronoun forms was motivated by the fact that Loeb and Leonard were
looking for a possible correlation between SLI children’s use of nominative he/she subjects and their use of s-marked
third person singular present tense verb forms like plays or is. I focus here on the children’s production of pronouns,
setting aside their production of s-marked verb forms. The gender errors made by the children typically involved
using he in contexts where adults would use she.
96
The case errors reported in column 4 almost all involve the use of an accusative subject (e.g. Him can’t swim) in a
nominative context. The final two columns show multiple errors: Ca+Ge = Case+Gender; Ca+Nu = Case+Number.
36
§5.6 A scoring issue: what’s right and what’s wrong
In a study like CBG’s in which the evidence presented is quantitative in nature, it is important to be sure
that scoring is accurate – i.e. in crude terms, to be sure that the right things are counted in the right way.
In this connection, what is the potential significance of the following quotation from their paper (p.155)?
‘Twenty obligatory contexts for 3rd sg. forms of auxiliaries were elicited, but only 7 were
correctly inflected (= 35%). The past tense task elicited 154 auxiliaries, 137 of which were
correctly inflected (= 89%). Here are some examples of agreement errors on auxiliaries:
(1) (a)
they was (JS 10;10) (b) he don’t know (RJ, 11;11)
(c) she do97 (AT, 13;01)’
§5.7 Verifying CBG’s figures
I have gone through the corpus used by CBG, and below list all the examples of contexts where adults
would require a 3.Sg present form of BE as a progressive auxiliary, HAVE as a perfect auxiliary, and DO as
a tense auxiliary98.
Contexts where adults would use a third person singular present form of the progressive auxiliary BE
1. Try get the cat, the man is99 (JW 11;03).
2. A horse is jumping over a gate (JS 11;10).
3. The boy’s picking it up for her (JS 11;10).
4. The boy’s trying to eat them (AZ11 12;03)
5. A boy is picking them up (RJ 12;11)
6. He’s jumping over a fence with a horse (AZ12 13;0)
7. A boy’s crying (AZ12 13;0)
8. The boy’s picking them up (AZ12 13;0)
9. Horse is jumping over a fence (CT 13;11)
10. The man climb…is climbing up the ladder to get a cat off the roof (CT 13;11)
11. He’s crying (CT 13;11)
12. She’s cuddling her teddy (SB 14;0)
13. He’s jumping over a…the gate (SB 14;0)
14. The boy’s posting it (SB 14;0)
15. The boy’s crying (SB 14;0)
16. She’s holding a doll (AT 14;01)
17. She’s going to pull her wellingtons off (AT 14;01)
18. The driver who drove it is fixing it (JS 9;10)
19. The bus is going down the road (AZ12 11;0)
20. And he’s trying to fix him (WL 11;05)
21. The apples are dropping out and the man picking them all up (AT 14;01)
22. And boy picking ’em up (JW 11;03)
23. He jumping over a gate (WL 11;05).
24. Apples fallen out on the boy who pinching them (JW 11;03)
25. The man taking the cat down on the ground (AZ11 12;03)
26. The dog taking slipper off him (AZ11 12;03)
27. Her hugging it (CT 13;11)
Contexts where adults would use a third person singular present form of the perfect auxiliary HAVE
28. The doggy’s got his shoes (JS 11;10).
29. A dog’s got his shoe (AZ12 13;0)
30. The dog has took the shoe off him (CT 13;11)
31. The girl has fallen down and she’s broken her glasses (SB 14;0)
What this child actually said was ‘She sometimes do some plays’.
Excluded from the list here are uses of HAVE and DO as main verbs, and uses of BE as a copular verb.
99
I assume the adult counterpart would be: Trying to get the cat, the man is
97
98
37
32. The dog has got his shoe (SB 14;0)
33. The lady has got a bag and it’s got a hole at the side (SB 14;0)
34. She hasn’t seen that she’s dropped…she’s dropped the apples (SB 14;0)
35. She sometimes buy stuff and then paint it what haven’t got coat of paint on it (RJ 11;11)
36. He been tied on (AZ12 13;0)
Contexts where adults would use a third person singular present form of the tense auxiliary DO
37. He don’t know (JW 10;3)
38. He don’t get hungry (JW 10;03)
39. What, when he don’t go to work? (JW 10;03)
40. And the bus don’t take no notice (AZ11 10;3)
41. And he don’t know how to (RJ 10;11)
42. He don’t know how to put his brakes on (AZ12 11;0)
To what extent do these examples data bear out CBG’s claim (in the quotation in question 5) about how
their subjects performed on marking agreement on auxiliaries in 3.Sg.present contexts? What is the
significance of the fact that in many varieties of spoken English, don’t is used with all types of subject
including third-person-singular subjects (as in the classic pop-song lament He don’t love me no more)?
§5.8 On the generality of CBG’s findings about German
Rice, Noll and Grimm (1997) report on two sets of spontaneous speech samples collected from 8 Germanspeaking SLI children in the age range 4;0-4;8 (the second sample collected a year after the first). They
claim that the German children in their study showed 88% correct use of the 3Sg agreement marker -t at
time TI and 100% at T2100, and 71% correct use of the 2Sg agreement suffix -st at time T1101 and 100% at
T2. They also report that the children overall102 also showed 97% correct use of overt forms of sein ‘be’103.
To what extent do Rice et al’s findings support the claims made by CBG about German SLI children?
§5.9 On scoring procedures
There are two different ways of scoring a child’s percent correct use of an agreement-marking item such
as the third person singular present tense -s affix in English. One is to ask: How often does the child use -s
in obligatory contexts (i.e. contexts where adults would use -s)? Another is to ask: When the child uses -s,
how often does the child use it with an appropriate kind of (third person singular) subject? Rice, Noll and
Grimm claim that the German children in their study scored 97% correct use of sein ‘be’, basing this
figure on the fact that when the children used sein ‘be’ (rather than omitting it) they correctly inflected it
in 97% (226/233) of the contexts in which it occurred. However, excluded from this score were 61 cases
where sein ‘be’ was omitted. What methodological issue is raised by their 97% correct score?
§5.10 Overall verdict
What is your overall verdict on the strengths and weaknesses of CBG’s Agreement Deficit hypothesis?
100
T1 is the time when the first spontaneous speech samples were collected, and T2 the time (a year later) when the
second were collected.
101
All their errors involved use of -t in contexts where -st is required. Since SLI children have phonological
production problems (e.g. with fricatives like |s| and with consonant clusters like |st|), the use of -t in place of -st may
well be a phonological rather than a grammatical error, as Rice, Noll and Grimm (1997, p.275) point out.
102
I.e. aggregating the figures from T1 and T2.
103
I.e. when the children used sein ‘be’ (rather than omitting it) they correctly inflected it in 98% (226/233) of the
contexts in which it occurred. Excluded from this score are 61 cases where sein ‘be’ was omitted.
38
6. The Agreement-and-Tense-Omission (ATOM) account of SLI
Key article
Wexler K, Schütze C & Rice M (1998) ‘Subject case in children with SLI and unaffected controls:
Evidence for the Agr/Tns Omission Model’, Language Acquisition 7: 317-344
OUTLINE OF WEXLER, SCHÜTZE AND RICE’S PAPER
In their 1998 paper, Ken Wexler, Carson Schütze and Mabel Rice (henceforth WSR) argue that SLI
involves a syntactic feature deficit which leads affected children to sometimes omit tense and agreement
features in obligatory contexts: they refer to the model they propose as ATOM (= Agreement & Tense
Omission Model). They argue that the case-marking of subjects as nominative or accusative by SLI
children directly correlates with whether or not they mark agreement on verbs104.
The Extended Optional Infinitives Model of SLI
In research into normally developing/ND children, Ken Wexler (1994) argued that TD children go through
a protracted stage (which generally lasts until around their 4th birthday) during which they alternate
between producing finite verbs105 and bare infinitives106 in contexts where adults require finite verb forms:
for obvious reasons, Wexler refers to this as the Optional Infinitives/OI stage. So, for example, in a past
tense context a (tearful) child at the OI stage (when asked why she is crying) might say either Daddy ate
my ice-cream or Daddy eat my ice-cream, alternating between the finite verb form ate and the bare
infinitive form eat. Wexler also observed that during the OI stage, children tend to omit auxiliaries and
copula BE in finite contexts, saying e.g. Daddy snoring rather than Daddy’s snoring, Daddy naughty rather
than Daddy’s naughty, Daddy gone rather than Daddy’s gone, and Daddy no like cabbage rather than
Daddy doesn’t like cabbage. The greater generalisation would appear to be that during the OI stage, TD
children alternate between producing finite and nonfinite clauses in finite contexts107.
In collaborative work, Ken Wexler and Mabel Rice (in conjunction with other researchers) have argued
that SLI children go through an Extended Optional Infinitives/EOI stage which typically lasts until they
are 7 or 8 years of age: see e.g. Rice, Wexler and Cleave (1995), Rice and Wexler (1996), and Rice,
Wexler and Herschberger (1998).
In early work, Wexler (1994) argued that Optional Infinitive structures (e.g. Daddy want one) result from
SLI children having a Tense Deficit in the sense that in contexts where adults mark (auxiliary or main)
verbs for present or past tense, SLI children sometimes mark tense but also sometimes leave verbs
underspecified for tense (so that the verb want in Daddy want one surfaces in the tenseless infinitive form
want rather than the present tense form wants or the past tense form wanted – the use of the infinitive form
want representing a tense-omission error under Wexler’s assumptions). In later collaborative work
between Ken Wexler and Carson Schütze (Schütze and Wexler 1996), it was argued that optional
infinitives can arise as a result of either tense or agreement features (or both) being underspecified (i.e.
omitted) – and this is the origin of the ATOM (Agreement/Tense Omission Model) outlined by WSR.
104
Throughout, I shall use the term verb to subsume both auxiliaries and main verbs. A technical complication which
I shall set aside is that within the theoretical framework which WSR adopt, tense and agreement are taken to be
properties of an abstract functor (traditionally labelled T/INFL). A slightly revised model appears in Wexler (2003) –
though I have not included this here since it is technical and says little new about case and agreement.
105
I.e. verbs inflected for tense/agreement, such as am/was/enjoyed/saw etc.
106
A bare infinitive form is an infinitival verb-form like say/find/help/play etc. used without the infinitive particle to.
The infinitival status of the verb-form is clearer in languages in which infinitives carry an overt inflection. For
example, French and German children at the OI stage produce verbs carrying an overt infinitive inflection in finite
contexts, giving rise to structures such as Fermer yeux Daniel ‘Close.Inf eyes Daniel’ or Thorsten Caesar haben
‘Thorsten Caesar have.Inf’ (where Inf is an infinitival suffix, and Caesar is the name of a doll).
107
I.e. They alternate between producing clauses which do or don’t contain a verb inflected for tense/agreement in
contexts where adults require a clause containing a verb inflected for tense/agreement.
39
Accounting for verb morphology within the ATOM model
WSR maintain that in finite contexts in adult English, verbs obligatorily carry person, number and tense
features (the value of their person and number features being determined by agreement with their subject).
So, in sentences like those in (1) below, the italicised verb forms will carry the parenthesised features:
(1)(a)
(b)
(c)
(d)
Daddy plays with me [plays = present-tense, third-person, singular-number]
Mummy and Daddy play with me [play = present-tense, third-person, plural-number]
Daddy played with me [played = past-tense, third-person, singular-number]
Mummy and Daddy played with me [played = past-tense, third-person, plural-number]
WSR posit that the relevant regular verb inflections in English108 encode the following features:
(2)(a)
(b)
(c)
-s = [present-tense, third-person, singular-number]
-d = [past-tense]
-ø = [any other set of features]
Informally, (2b) can be regarded as saying ‘If a regular verb carries the feature [past-tense], then
(irrespective of whether or not it also carries other features like person and number), it will have the
suffix -d added to its stem’. Let’s look at how we can account for the italicised verb forms in (1) in terms
of the assumptions made in (2).
In (1a), suppose that the syntax tells us we require the third person singular present tense form of the verb
PLAY. We then search through the list of affixes in (2) in order to find the affix which provides the best
match for the required set of features (and involves no mismatch). Since -s in (2a) is a perfect match, the
affix -s is added to the stem play- to form plays, deriving (1a) Daddy plays with me.
In (1b), suppose the syntax tells us that we require the third person plural present-tense form of PLAY. We
again search through the list of affixes in (2) in order to find the affix which provides the best match for
the required set of features (and involves no mismatch). The affix -s in (2a) will not do, since using it
would involve a mismatch between the plural number required by the syntax and the singular number
encoded by -s. Nor will the affix -d in (2b) do, since its use would involve a mismatch between the present
tense required by the syntax and the past tense encoded by -d. Accordingly, the inflection -ø in (2c) is used
by default (i.e. as a last resort, because the other two are not appropriate choices), since the use of -ø
involves no mismatch by virtue of being able to encode any other set of features. This results in the null
affix -ø being added to the stem play-, deriving Daddy and Mummy play-ø109 with me.
In (1c), suppose the syntax tells us that we require the third person singular past tense form of PLAY. We
search through the list of affixes in (2) in order to find the affix which best matches (and does not
mismatch) the set of features carried by the verb PLAY. The -s affix will not do, because of the mismatch
between the past tense feature specified in the syntax and the present tense feature specified in the entry
for -s in (2a). This time, the best match is -d, since the entry for -d in (2b) contains a [past-tense] feature
which matches the [past-tense] feature required by the syntax, but no further features (so that there is no
mismatch110 with the person/number features of the required form). Accordingly, the affix -(e)d is added
to the stem play-, deriving Daddy and Mummy play-ed with me. Essentially the same will be true in (1d),
where once again the [past-tense] feature of the affix -d in (2d) provides the best match for the features
required by the syntax in (1d).
108
They exclude (perfect/passive) participle uses of -d from their discussion, as well as the progressive participle and
gerund affix -ing. Note that -d and -s are preceded by e in the spelling (and hence surface as -ed/-es) when added to
certain types of stem, but this is a question of detail which I overlook here.
109
WSR assume that the verb in such cases carries a null affix. But (if you prefer) you can alternatively think of -ø in
(2c) as meaning ‘Don’t add any affix at all to the verb’.
110
WSR posit that all of the features of an affix must be present in the syntactic representation for that affix to be
used, but not all the features in a syntactic representation must be carried by an affix in order for the affix to be used.
40
Infinitives in finite contexts111 in the ATOM model
WSR argue that the bare infinitives which children sometimes produce in finite contexts come about when
the verb is underspecified for (i.e. lacks) tense and/or agreement features. Consider, for example, what
happens if the child omits one or more of the [bracketed] features carried by the italicised verb in a
structure like (3) below, repeated from (1a) above:
(3)
Daddy plays with me [plays = present-tense, third-person, singular-number]
For example, if the child leaves the [present-tense] feature unspecified (perhaps treating tense as
implicit112), the child’s syntactic structure will carry the features shown in (4) below:
(4)
Daddy play with me [play = third-person, singular-number]
If (as WSR assume) children’s affixes generally encode the same set of features as their adult counterparts
in (2) above, the child will search through the set of affixes in (2) to find the one which provides the best
match and no mismatch. However, the affix -s cannot be used here, since the entry for -s in (2a) says that
it can only be used where the syntax contains the feature [present-tense]; yet the (underspecified) syntactic
structure in (4) contains no tense feature of any kind on the verb. For the same reason, the affix -d cannot
be used either, since this encodes past tense and (4) contains no tense feature. Consequently, the null affix
-ø in (1c) is used by default, and since it can be used to spell out ‘any other set of features’ on a verb, there
is no feature-mismatch here. Accordingly the bare form play-ø is used in (4).
Now let’s make the alternative assumption that the child’s counterpart of (3) is specified for present tense,
but underspecified in respect of (person and number) agreement features, as shown informally below:
(5)
Daddy play with me [play = present-tense]
The child will then search through the affixes in (2) to find the best match for the [present-tense]
feature required by the syntax in (5). The affix -s will not, however, provide a good match, since the entry
for -s in (2a) says that it can only be used where the verb also carries the features [third-person, singularnumber], and yet the underspecified verb play in (5) carries only the feature [present-tense]. Nor is the
affix -d suitable, since this encodes past tense, and (5) tells us that we’re looking for a present-tense affix.
Consequently, the affix -ø is used by default, resulting in the bare verb form play-ø (which is bare in the
sense that it does not contain any overt affix)113.
What our discussion here highlights is that (within the ATOM model) a bare verb form like play in a finite
context in which adults require an inflected form like plays or played comes about by virtue of the verb
being underspecified for its tense and/or agreement features – e.g. if the verb is underspecified for tense as
in (4), or for agreement as in (5)114.
A finite context is a context where an adult would use a finite (auxiliary or main) verb – i.e. one marked for tense
and agreement.
112
E.g. perhaps assuming that it can be inferred from the context. So, for example, if a child replies to an adult
present-tense question like ‘What does Daddy do when he comes home?’ the child might treat tense as something
which can be inferred from the context (tense being seen, in some informal sense, as ‘given’ information – given by
the use of the present-tense auxiliary does by the person asking the question), and hence say ‘Daddy play with me’,
where play is not inflected for tense.
113
A further possibility which I won’t go into here is that Daddy play with me can also arise if the verb is
underspecified for both tense and agreement. A possibility which WSR do not discuss is that a verb may be
underspecified for one of its two (person and number) agreement features but not the other, resulting in partial
agreement (in person but not number, or conversely in number but not person) between subject and verb.
114
More generally, a verb in a finite context will surface as an uninflected bare (infinitive) form if the verb is
underspecified for any one or more of its person/number/tense features.
111
41
Auxiliary omission in finite contexts in the ATOM model
As noted earlier, both TD and SLI children frequently omit the auxiliaries BE/HAVE/DO in finite contexts,
hence producing auxilariless sentences such as:
(6)(a)
(b)
(c)
Daddy snoring/Daddy naughty (omission of is)
Daddy gone out (omission of has)
Daddy not like cabbage (omission of does)
WSR posit that the function of such auxiliaries is to encode tense and agreement properties. They further
assume that (just like main verbs) auxiliaries can optionally be underspecified for tense and/or agreement
in finite contexts in child grammars, and that when HAVE/BE/DO are underspecified for one or more of
their tense/agreement features, they are given a null phonetic spellout (i.e. so are ‘silent’115). So, in an
adult sentence such as Daddy is snoring, the auxiliary BE will be fully specified for tense and agreement
features, as in (7) below;
(7)
Daddy BE snoring [BE = present-tense, third-person, singular-number]
Accordingly, the auxiliary BE will be spelled out in the corresponding 3Sg present tense form is. But
suppose that a child marks the auxiliary BE for agreement but not for tense, as in (8) below:
(8)
Daddy BE snoring [BE = third-person, singular-number]
In such a case, the (tense-) underspecified auxiliary BE will be given a null spellout, so that the resulting
sentence is Daddy ø snoring, where ø is a null exponent of BE. Likewise, if the child specifies the
auxiliary for tense but not agreement as in (9) below:
(9)
Daddy BE snoring [BE = present-tense]
the (agreement-) underspecified auxiliary BE will again be given a null spellout, so deriving Daddy ø
snoring. In other words, an auxiliary-drop sentence like Daddy snoring will result whenever BE/HAVE/DO
is underspecified in respect of (i.e. lacks) one or more of its person/number/tense features116.
Although (overt forms of) the non-modal auxiliaries BE/HAVE/DO encode both tense and agreement
features, WSR posit that modal auxiliaries always encode tense, since they inflect for present/past tense
(cf. pairs such as will/would, shall/should, can/could, may/might), and never carry the third person
singular agreement inflection -s. So, the lexical entry for CAN will say that it is spelled out as follows117:
(10)
115
CAN
= can if [present-tense]
could if [past-tense]
A number of morphemes can have a null spellout in particular contexts in adult English. For example, possessive
’s has a null spellout when added to a noun ending in plural -s, as in the soldiers’ barracks. In colloquial English,
have and are can have a null spellout in questions like Where you been? or What you doing?, perhaps as a result of
more general phonological reduction processes (involving schwa-deletion). It may well be that auxiliaries are given a
null spellout by children (by default) when their lexicon contains no auxiliary form which can match the features
required in the syntax.
116
This is not intrinsically implausible, since BE can have a null spellout in certain structures in adult English. For
example, in the dialogue below, speaker B can either spell out BE overtly, or not:
A: You shouldn’t be rude about Queenie, like you were yesterday.
B: Me be rude about the old bag? No way!
Me rude about the old bag? No way!
117
(10) can be thought of as meaning ‘The word CAN is spelled out as the form can if it carries a present-tense
feature (irrespective of whether or not it also carries other features such as person and number), and is spelled out as
could if it carries a past tense feature (irrespective of whether or not it also carries other features like person and
number)’. Recall that WSR assume that auxiliaries in finite contexts in adult English always carry not only a present
or past tense feature, but also person and number agreement features. But if an SLI child has the same lexical entry
(10) for CAN as adults (in accordance with what WSR assume), the child will be able to use can not only when he
marks the auxiliary for both tense and agreement, but also when he marks it only for tense (omitting person/number
agreement features). This assumption turns out to have important consequences for how the child case-marks the
subject of can – as you will see shortly.
42
Case-marking in the ATOM model
Noun and pronoun expressions in adult English inflect for case. English is traditionally said to have three
cases, namely nominative, accusative (or objective), and genitive. Some pronouns have separate weak and
strong genitive forms, a weak genitive form like my being used when modifying a following noun
expression (as in my new car, where my modifies the noun expression new car), and a strong genitive
form like mine being used in other structures (e.g. in sentences like This is mine or Mine is bigger than
yours). The way in which the various case forms of personal pronouns are spelled out in English is shown
in the table below (repeated from chapter 1):
(11)
Pronoun
1.SG
1.PL
2.SG/PL
3.M.SG
3.F.SG
3.N.SG
3.PL
How case is spelled out on personal pronouns in adult English
Spellout
I if nominative, mine if strong genitive, my if weak genitive, me otherwise
we if nominative, ours if strong genitive, our if weak genitive, us otherwise
yours if strong genitive, your if weak genitive, you otherwise
he if nominative, his if genitive, him otherwise
she if nominative, hers if strong genitive, her otherwise
its if genitive, it otherwise
they if nominative, theirs if strong genitive, their if weak genitive, them otherwise
(1/2/3 denote 1st/2nd/3rd person; SG = singular; PL = plural; M = masculine; F = feminine; N = neuter.)
In contrast to personal pronouns, noun expressions have a common nominative-accusative form and a
distinct genitive ’s form (cf. nominative-accusative George/the president and genitive George’s/the
president’s).
The case carried by a pronoun or noun expression depends on its position within the structure containing
it. Below, I present a simplified version of the assumptions which WSR make about case assignment in
English:
(12)
Case Assignment (simplified)
A pronoun or noun expression is assigned
(i) nominative case if the subject of an agreeing verb118
(ii) genitive case if a possessor119
(iii) accusative case otherwise
Since otherwise forms are technically referred to as default forms, it can be said that accusative is the
default case in English120.
To see how (12) works, consider the case-marking of the italicised pronouns in the following sentence:
(13)
He has been fooling around, so I can see why she got upset
The auxiliary has here is a third person singular (present tense) form which agrees in person and number
with its (third person singular) subject he, so that he carries nominative case by virtue of agreeing with
has, in accordance with (12i). Since WSR hypothesise that all verbs marked for tense in adult English
agree in person and number with their subject, it follows from their assumptions that the present-tense
auxiliary can agrees in person and number with its (its first person singular subject I), so that can (as used
118
In the context of the discussion of agreement here, the term verb should be taken to include both auxiliary verbs
and main verbs: an agreeing verb is one which agrees with its subject in person and number. In more abstract
syntactic analyses of clause structure in which T/INFL is said to be the locus of tense/agreement features, an
expression is assigned nominative case via agreement in person and number with T/INFL, and the agreement
features attach to the auxiliary in T/INFL if there is one, and (if not) are otherwise lowered onto the main verb. The
idea that nominative case is assigned via person/number agreement with a verb dates back to work by Rouveret
(1980) and Raposo (1987) on tenseless agreement-inflected infinitives with nominative subjects in Portuguese.
119
WSR have nothing much to say about genitive case, or how SLI children case-mark possessors.
120
However, note that in some other languages (e.g. German or Arabic), nominative appears to be the default case.
See Schütze (2001) for a technical discussion of the notion of default case.
43
here) is a first-person-singular present-tense form: by virtue of agreeing in person and number with can,
the subject I is assigned nominative case in accordance with (12i)121. Given WSR’s assumption that all
present/past tense verbs in adult English agree in person and number with their subjects122, it follows that
the verb got (as used here) carries the features [past-tense, third-person, singular-number] via agreement
with its third person singular subject she, and conversely that she is assigned nominative case in
accordance with (12i) by virtue of agreeing in person and number with got. Given the impoverished nature
of verb morphology in present-day English, the agreement properties of past tense verbs are no longer
directly visible (except on the forms was/were).
Now consider the case-marking of the italicised pronouns in:
(14)
Please don’t report me to her!
Here, me is the object/complement of the verb report: by virtue of not being the subject of an agreeing
verb, and not being a possessor, me is assigned accusative case in accordance with (12iii). Likewise, her is
the object/complement of the preposition to, and so is also assigned accusative case in accordance with
(12iii). The Case Spellout List in (11) tells us how the various different case-forms of pronouns are spelled
out and includes the following entry:
(15)
3FSG = she if nominative, hers if strong genitive, her otherwise
What the ‘her otherwise’ condition in (11/15) means that a third person singular feminine pronoun gets
spelled out as her unless a nominative or strong genitive form is required: in other words, her is used
when we need a weak genitive form (as in her dress) or an accusative form (as in Help her!). More
generally, when accusative forms are required, they are always spelled out as the otherwise form.
Next, consider the case-marking of the italicised pronouns in:
(16)
My computer isn’t as powerful as yours
By virtue of being possessors (in the sense that my in the expression my computer denotes the person who
possesses the computer), the pronouns my and yours are assigned genitive case in accordance with (12ii).
Since my is immediately followed by a noun expression (namely the noun computer), it is spelled out as
the weak genitive form my (rather than the strong form mine), whereas since yours isn’t followed by a
noun expression, it is spelled out as the strong genitive form yours (rather than the weak genitive my).
Finally, consider the italicised pronouns in the (self-obsessed me-me-me) three-way conversation below:
(17)
SPEAKER A: Me, I reckon you’ll be a pop star one day
SPEAKER B: Me be a pop star? No way! Who’d wanna
SPEAKER C: Me!
be a pop star, anyway?
The pronoun me in (17A) serves the function of being the topic of the overall sentence, and the clause
which follows it (namely I reckon you’ll be a pop star one day) is referred to as the comment. In (17B),
ME is the subject of what is known as a ‘Mad magazine sentence’ (because it resembles a sentence which
appeared on the front cover of the American magazine Mad). In (17c), Me is a sentence fragment – i.e. an
expression which isn’t a complete sentence. An interesting question which arises in relation to the three
uses of the italicised pronoun in (17) is how it comes to be spelled out as the accusative form me. The
answer is that since me is neither the subject of an agreeing verb nor a possessor, it gets assigned
accusative case by default – i.e. via the otherwise condition in (12iii).
121
In earlier varieties of English, CAN overtly inflected for agreement in forms like thou canst. However, its
agreement inflections have been lost in the course of the historical evolution of the language. The counterpart of CAN
in languages like Italian overtly inflects for person/number agreement with its subject, as we see from forms like
posso/puoi/può/possiamo/potete/possono = ‘can.1SG/2SG/3SG/1PL/2PL/3PL’.
122
This agreement is more directly visible in languages like Spanish/Italian with a much richer overt subject-verb
agreement morphology than English.
44
Predictions of the ATOM model about case marking in SLI children’s grammars
The analysis of case-marking outlined in the previous section, taken together with the core assumption of
the ATOM model that (auxiliary and main) verbs can be underspecified for (i.e. lack) tense and/or
agreement features in finite contexts makes a number of specific predictions about the patterns of casemarking that we will find in the grammars of SLI children at the EOI stage (and indeed in TD children at
the OI stage) children. Predictions about the case-marking of subjects and objects are outlined below123.
Predictions about the case-marking of objects by SLI/TD children appear straightforward enough; the
object of a verb or preposition would be expected to be assigned accusative case by (12iii), and we should
therefore expect that SLI (and indeed TD) children would make no case errors with objects.
Predictions about the case-marking of subjects are rather more complex, because of the possibility that the
associated verb may (or may not) be underspecified for tense and/or agreement. There are three different
types of sentence which we need to consider here. A straightforward case are sentences containing a main
verb or a non-modal auxiliary like HAVE/BE/DO which is overtly inflected both for tense and for agreement
in person and number with its subject (e.g. as in Daddy snores, Daddy is snoring): such sentences will be
predicted always to have nominative subjects in accordance with (12i) because the verb agrees with the
subject (so that we expect to find He snores, He is snoring).
Less straightforward are modal auxilaries and past tense main verbs124. Both encode tense (e.g. can is a
present-tense modal auxiliary verb, could is a past tense modal auxiliary verb, and went is a past tense
main verb), but if SLI children at the Extended Optional Infinitives/EOI stage optionally mark agreement
on (auxiliary or main) verbs in finite contexts, modals and past tense verbs may either be specified for
agreement or not. If a modal or past tense verb is specified for agreement in person and number with its
subject, it is predicted to have a nominative subject by (12i) (as in He can swim, He could help and He
went out). But if a modal or past tense verb is underspecified for agreement, its subject will assigned
accusative case by default, in accordance with (12iii) (as in Him can swim, Him could help and Him went
out). Overall, then, SLI children at the EOI stage are predicted to alternate between using nominative and
accusative subjects for modals and past tense verbs.
Finally, let’s consider structures involving bare verbs (e.g. Daddy play with me) and ‘missing’ auxiliaries
(e.g. Daddy working, Daddy gone out, Daddy naughty). As we saw earlier, within the ATOM model there
are three ways in which such structures can arise, namely: (i) if the verb/auxiliary is specified for tense but
not agreement; (ii) if it is specified for agreement but not tense; or (iii) if it is specified for neither tense
nor agreement. If the (auxiliary/main) verb is specified for (person and number) agreement but not tense,
the subject will be predicted to be nominative by (12i) (yielding e.g. He play with me, He working, He
gone out, He naughty). If the (auxiliary/main) verb is specified for tense but not agreement, the subject
will be accusative by (12iii): likewise, if the (auxiliary/main) verb is specified for neither tense nor
agreement, the subject is again predicted to be accusative by (12iii); in both cases, we expect to find
children producing structures such as Him play with me, Him working, Him gone out, and Him naughty
(where the subject him is a default form). Since two of the three possibilities discussed here yield
accusative subjects, we might expect accusative subjects to be more frequent than nominative subjects in
such structures (for any SLI child in the EOI period who leaves tense underspecified roughly as frequently
as agreement)125.
Overall, WSR classify the structures which SLI children at the EOI stage would be expected to produce in
contexts where adults use a finite clause (i.e. a clause containing an auxiliary/main verb marked for tense
and agreement) into the three main types indicated below:
WSR don’t discuss how SLI children at the EOI stage (and TD children at the OI stage) case-mark possessors.
It makes no difference for case-marking purposes whether the past tense verb is regular or irregular.
125
Note that WSR make no specific claims about whether children omit agreement features more or less frequently
than they omit tense features.
123
124
45
(18)(I)
Structures containing an (auxiliary/main) verb inflected for both tense and agreement (predicted
to have nominative subjects: e.g. He snores, He is snoring). Since such sentences overtly mark
agreement, WSR classify them as agreeing sentences.
(II)
Modal/past tense structures in which tense is marked, but agreement may or may not be marked
(predicted to have either nominative or default subjects: e.g. He/Him went out, He/Him can
swim). Since we can’t directly tell from the impoverished morphology of such verb forms
whether agreement is marked or not, WSR classify such sentences as ambiguous.
(III)
Bare verb/missing auxiliary structures which are ambiguous in respect of whether they are (i)
specified for agreement but not tense, (ii) specified for tense but not agreement, or (iii) specified
for neither tense nor agreement: (i) would yield nominative subjects (e.g. He snore, He snoring),
and (ii/iii) accusative subjects (e.g. Him snore, Him snoring). Since such sentences contain no
verb/auxiliary overtly inflected for agreement, WSR call them uninflected sentences. WSR
predict that accusative subjects will be more frequent in uninflected than ambiguous structures.
Thus, they write (1998, p.303): ‘We classified utterances into agreeing, ambiguous… and uninflected.
Agreeing forms include main verbs with -s (e.g. likes), and agreeing auxiliaries and copulas (e.g. is/are).
Ambiguous forms include past-tense verbs (e.g. liked) and modals (e.g. can). Uninflected forms include
main verbs missing -s (e.g. *Mary like126), omitted auxiliaries and copulas (e.g. *Mary going, *Mary
pretty), and the rare instances of uninflected auxiliaries (e.g. *Mary be). This category also included main
verbs that were intended as past tense semantically but were not marked as such’.
Methodology of WSR’s study
The data which WSR present in their paper are based on two sets of spontaneous speech samples and
experimental data (collected at 6 month intervals) from 23 SLI children with a mean age of 4;9 in round 1
and 5;5 in round 2127, and 20 MLU-matched TD children with a mean age of 3;0 at round 1 and 3;7 at
round 2. The experimental task was designed to elicit sentences with third person singular pronoun
subjects (to check both case and agreement marking).
Case-marking in WSR’s study
WSR report that (as expected under the ATOM model) both the SLI and the TD children virtually always
correctly case-mark pronouns in accusative contexts. Relevant data are given in (15) below:
Frequency of accusative pronouns128 in accusative contexts
Group
Round 1
SLI children
203/204 (99.5%)
MLU-matched TD children
166/166 (100%)
(19)
Round 2
244/245 (99.6%)
222/222 (100%)
By contrast, both the TD and SLI children make errors with the case-marking of subjects in nominative
contexts (i.e. in contexts where the corresponding adult sentence would have a nominative subject), as
shown by the data in the table below:
126
The asterisk indicates that the relevant structures are ungrammatical in adult English.
Rounds 1 and 2 denote the periods of time when the earlier/later sets of data were collected.
128
The data here relate only to pronouns which show an overt nominative/accusative contrast, namely I/me, we/us,
he/him, she/her and they/them. WSR exclude you and it from this count on the grounds that they have a common
nominative/accusative form (so it’s impossible to be sure whether you/it in a given sentence is nominative or
accusative).
127
46
(20) Mean percent correct use of nominative he/she subjects in nominative contexts129
Group/Round
Round 1
Round 2
Round 1
Round 2
Data type
spontaneous
spontaneous
elicited
elicited
SLI children
57%
77%
56%
59%
ND children
85%
83%
96%
86%
Given the assumptions of the ATOM model (namely that TD children optionally leave tense and/or
agreement underspecified until around their 4th birthday, and SLI children until around their 8th birthday,
and that agreement-underspecification leads to errors in the case-marking of subjects), we should expect to
find that both 5-year old SLI children and 3-year-old TD children will make case-marking errors with
subjects in nominative contexts, but that the SLI children will make more errors than the TD children
(since a 3-year old TD child is closer to the end of the EOI period than a 5-year old SLI child). These
predictions would seem to be borne out by the data in (23). All but one of the case errors on subjects
involved using an accusative subject in a context where adults would uses a nominative subject130.
Relation between nominative case and agreement
In principle, the ATOM model makes the prediction that sentences in which (person/number) agreement is
marked will have nominative subjects, whereas sentences in which agreement is not marked will have
accusative subjects. In terms of the classification in (21) above, WSR would predict that agreeing
sentences have nominative subjects, ambiguous sentences (i.e. sentences containing modals or past tense
verbs) should have either nominative or accusative subjects, and uninflected sentences (i.e. sentences with
a bare verb or missing auxiliary) should likewise have either nominative or accusative subjects – though
with accusative subjects occurring more frequently in uninflected than in ambiguous sentences, given their
assumptions. WSR summarise the predictions of the ATOM model in the following terms (1998, p.333):
‘Agreeing verbs should show the fewest non-Nom131 subjects, in principle none at all. Ambiguous verbs
should show some non-Nom subjects, more than agreeing verbs, and uninflected verbs should show the
highest proportion of non-Nom subjects.’
Adjusting for lexical gaps
WSR focus mainly on sentences with 3Sg M/F personal pronoun subjects, since regular verbs only inflect
for agreement in present-tense 3Sg forms, and since 3Sg M/F personal pronouns overtly inflect for the
nominative/accusative case contrast (cf. he/him and she/her132). However, the possibility that the children
may not have acquired the full set of pronouns found in adult English also needs to be controlled for. For
example, the pronoun she is often acquired relatively late133 by children: children who have not yet
acquired she have a lexical gap (i.e. have a gap in the set of pronoun forms listed in their lexicon relative
to the set of pronouns forms found in the adult lexicon). They often extend some other pronoun (typically
he or her) to take over the role played by she in the adult system: so, for example, a child who has not yet
acquired she might extend her from use as an accusative/genitive pronoun to additional use as a
nominative pronoun, hence using her in nominative contexts, as in Her’s naughty. For such a child, her
would therefore have a different status from its adult counterpart; whereas a 3FSg (third person feminine
129
This table includes only third person singular masculine/feminine pronouns (he/him/his/her).
WSR report one occurrence of a genitive his subject in their spontaneous speech data. However, as has often been
noted, it can sometimes to be hard for transcribers to determine whether a child says e.g. His sleep or He’s sleep (and
indeed to determine whether this is the child’s counterpart of He’s asleep or He’s sleeping). Because genitive his
subjects are virtually non-existent, WSR take her subjects to be accusative where they occur alongside she subjects,
and not genitive. Of course, if the children used genitive subjects, we’d also expect to find forms like Daddy’s can do
it where the subject Daddy surfaces in the genitive form Daddy’s: I assume these did not occur, since WSR do not
mention them.
131
Non-nominative (i.e. apparent accusative/genitive forms).
132
They exclude the inanimate (neuter) pronoun it, because it has the same form in nominative (cf. It is difficult) and
accusative (cf. Forget it!) contexts.
133
Perhaps because other 3Sg animate pronouns like he/him/his/her all begin with h-, making she irregular in respect
of its onset (i.e. its initial consonant).
130
47
singular) pronoun would be spelled out as in (21a) below in adult English, for a child with a defective
(she-less) pronoun paradigm, a 3FSg pronoun might be spelled out as in (21b):
(21)(a) ADULT: 3FSg = she if nominative, hers if a strong genitive, her otherwise
(b) CHILD: 3FSg = hers if a strong genitive, her otherwise
Because of this, for a child with the impoverished (she-less) pronoun system in (21b), we cannot tell what
case is carried by her subjects (in sentences like Her naughty, Her’s cheating, Her can do it etc.), since
her (for such a child) could in principle be a nominative, genitive or accusative/default form. To avoid
problems in determining the case properties of pronouns used by children with defective pronoun
paradigms134, WSR decided to exclude pronoun forms which are ambiguous in the child’s grammar from
their calculations.
WSR’s results
The figures which WSR arrive at for subject case-marking with third person singular animate (i.e.
masculine/feminine) pronouns in spontaneous speech samples are shown in the table below:
(22)
Frequency of he/she subjects in nominative contexts in spontaneous speech samples
Group
Clause type
Round 1
Round 2
SLI children
agreeing
30/43 (70%)
25/29 (86%)
SLI children
ambiguous
15/31 (48%)
27/38 (71%)
SLI children
uninflected
28/129 (22%)
19/38 (50%)
ND children
agreeing
51/55 (93%)
39/50 (78%)
ND children
ambiguous
19/24 (79%)
23/30 (77%)
ND children
uninflected
40/58 (69%)
13/22 (59%)
WSR’s conclusions
WSR draw 7 main conclusions about case-marking from their study. Firstly, it is specifically the casemarking of subjects which causes problems for SLI and TD children alike, not (e.g.) the case-marking of
objects. Secondly, SLI children make more frequent case-marking errors with subjects than MLUmatched TD children. Thirdly, the type of case-marking error SLI children make involves using an
accusative form in contexts where adults require a nominative form. Fourthly, the case-marking of
subjects (viz. whether subjects are spelled out as nominative or accusative forms) is determined by
whether clauses contain an auxiliary/main verb135 which agrees in person and number with the subject or
not. Fifthly, the overall characteristics of children’s verb morphology and pronoun case morphology
follow from the ATOM model, in which children leave verbs underspecified for tense and/or agreement
features in finite contexts, resulting in the production of Optional Infinitive/OI structures with accusative
subjects. Sixthly, the OI period is far more extended in SLI children (lasting until around their 8th
birthday) than in TD children (lasting until around their 4th birthday), so that a 3-year old TD child is
closer to the end of the OI period than a 5-year old SLI child (and hence a 5-year-old SLI child could be
expected to make more errors with verb morphology and subject case-marking than an MLU-matched
3-year old TD child). And finally, the ATOM model makes (limited) predictions about the relative
frequency of case and tense/agreement errors made by children in different types of clause structure (e.g.
with subject case-marking errors predicted to be most frequent with uninflected verbs, less frequent with
ambiguous verbs, and least frequent with agreeing verbs)136.
134
I.e. children who have not acquired the full range of pronouns found in adult English.
More precisely, an INFL/Agr constituent (a technical complication which I gloss over here). Recall that WSR
assume that an auxiliary like BE/HAVE/DO will be invisible (= have a null spellout) if underspecified for tense, or
agreement, or both.
136
However (as pointed out earlier), ATOM makes no specific claims about whether tense features are omitted more
frequently than agreement features – or conversely.
135
48
WORKBOOK SECTION
In many ways, WSR’s analysis replicates the findings of earlier work by Loeb and Leonard (1991), who
likewise showed that SLI children make case-marking errors with subjects rather than objects, and
hypothesised that these errors are the result of an agreement deficit (i.e. of verbs lacking subjectagreement features). However, WSR posit that SLI children additionally have a tense deficit in the sense
that they sometimes also fail to mark tense on verbs in obligatory contexts. WSR’s main contribution to
our understanding of SLI lies in their formulation of a specific developmental model – the ATOM model –
which incorporates a principled model of tense/agreement morphology, and which provides a principled
account of case-marking by SLI children in different types of clause structure. But does the ATOM model
account for the full range of sentences types produced by SLI children (in particular, the types of case
error they make), or are there some types or error not accounted for by ATOM – and if so, how can they
be accounted for? These are the questions which the exercise material below is designed to help you
answer.
§6.1 Data from child D in the Leonard corpus
Below are a selection of examples (from the Leonard files on the CHILDES data-base) of sentences with
produced by a male child with SLI (identified only as D) at age 4;4. Discuss the case-marking of subject
pronouns in these sentences, and say whether it conforms to the predictions made by ATOM or not.
Identify sentences which appear to be problematic for the ATOM account, and see if you can suggest
ways of dealing with them.
1 I’m strong
2 I’m trying kill him
3 I have Superman on tape
4 Where do I put this?
5 They’re jumping
6 They’re big
7 They’re shoes
8 Yeah, they do
9 I can’t do it
10 I don’t feel well
11 I don’t know what them are 12 Me don’t want those
13 Me don’t know how do it 14 I said I did all that 15 I gonna kill you
16 I been there
17 I in this
18 I not in that picture 19 Them mad at him
20 Them not plate 21 Them is big
22 Them is boys
23 Me’s right here 24 You’re wrong 25 You are gonna turn it?
The Standard English/SE equivalent of sentences which differ from their SE counterparts are as follows:
2 I’m trying to kill him 11 I don’t know what they are 12 I don’t want those 13 I don’t know how to do
it 15 I’m gonna kill you 16 I’ve been there 17 I’m in this 18 I’m not in that picture 19 They’re mad at
him 20 They’re not plates 21 They’re big 22 They’re boys 23 I’m right here. In relation to sentences
containing overt present tense forms of BE, assume that these are spelled out as follows:
BEPRESENT = am if first person and singular, are if second person or plural, is otherwise
Also consider (in addition to the possibilities envisaged by ATOM) the further possibility that SLI
children may sometimes mark a verb for partial agreement with its subject (in either person or number),
but that only a verb which agrees in both person and number with its subject has a nominative subject
(other verbs having default accusative subjects): show how these assumptions would help you account for
sentences containing I’m, me’s, they’re, them are and them is.
§6.2 Data from child J in the Leonard corpus
Below are all the examples (from the relevant file in the Leonard corpus on the CHILDES data-base) of
sentences with 3MSg (third person masculine singular his/him/his) pronouns produced by a female child
with SLI (identified only as J) at age 4;11. (Glosses or contextual information are provided in
parentheses). Discuss the case-marking of italicised pronouns, and say whether it conforms to the
predictions made by ATOM or not. Identify sentences which appear to be problematic for the ATOM
model, and see if you can suggest a way of dealing with them.
3MSg pronouns in accusative contexts
1. How about him? 2. Him got him. 3. I dropped him. 4. You stand him upside down.
5. I want him. 6. Dress him. 7. I can’t get him out. 8. Gonna put him right here. 9. Yeah, that him.
10. Him (Occurred immediately after he had asked himself ‘Now what do I need?’)
49
3MSg pronouns in nominative contexts
11. Him is. 12. Him can be nice. 13. Him can ride way, way in the back. 14. Maybe him don’t fit in here.
15. Him want to now (‘He wants…’). 16. Him want to (‘He wants...’) 17. Him catch (= ‘He caught’).
18. Him jump off the chair (‘He jumped…’). 19. Him walk off the table (‘He walked…’).
20. Him got him (‘He got…’). 21. Him got little pig barn (‘He’s got…’). 22. Except him got green and I
got red. (‘He’s got…I’ve got’) 23. Him walking like this (‘He’s walking…’). 24. Him driving (‘He’s
driving’). 25. And him riding (‘He’s riding’). 26. Him riding the car (‘He’s riding…’).
27. Him climbing the tree (‘He’s climbing…’).
28. Him running fast and him walking slow
(‘He’s…he’s’). 29. And him walking slow (‘He’s…’)
30. Him jumping on him car (‘He’s…. )
31. Him gonna go to bed (‘He’s…). 32. Hope him gonna hit him butt (‘…he’s…’)
33. Him gonna sit down in there (‘He’s). 34. Him gonna go in here (‘He’s…’)
35. Him bad guy (= ‘He’s a bad guy’). 36. Him the driver (‘He’s…). 37. Him on the chair (‘He’s…’).
38. Him in bed asleep (‘He’s…’). 39. Him under him bed turn (‘He’s…’).
3MSg pronouns in genitive contexts
40. Spilled his water.
43. Comb him hair.
46. Him under him bed turn.
49. Hope him gonna hit him butt.
41. The dog chewing him bone.
44. Hay to him donkey.
47. Him jumping on him car.
50. Him car go.
42. Zip him jacket up.
45. How pig get in him thing?
48. Walking on him car.
51. The baby drop him bottle.
§6.3 Looking at the behaviour of one of the children in WSR’s study
The following are extracts from the transcript of one of the recordings of one SLI child included in WSR’s
study. (The child was aged 4;6 at the time of the recording, and was subject number 19700128: I am
grateful to Mabel Rice for providing me with a copy of the transcript.) Below, I have listed all the
sentences with a third person masculine singular pronoun subject produced by the child in the transcript.
1. Him got away137
2. Him went up high stairs 3. Him did138
4. Him jump on light139
140
141
142
5. No, him jump
6. Him jump.
7. Him jump high
8. He eat it143
144
145
146
9. Him too big .
10. Him wild cat
11. Him a wild cat . 12. Him a wild cat
13. Now him high147 14. Now him back148
15. Him mad149
16. Him mad people150
151
152
17. Now him under table
18. Him right here my back
19. Him not hungry153. 20. Him up154
155
156
21. Him up there
22. Him upstairs
23. Now him up high157 24. Him walking street158
The interviewer responds ‘He got away?’
Produced after the interviewer said ‘Get off the table’.
139
This sentence was scored as an attempt at ‘He jumped on the light’.
140
Produced after the interviewer said ‘The cat didn’t jump up on there’.
141
The interviewer responds ‘He did. He jumped onto the ground’
142
The interviewer responds ‘He jumped high’.
143
The interviewer responds ‘He ate it’
144
Produced after the interviewer had said ‘Maybe Mommy can fit in there’.
145
The interviewer responds: ‘He’s a wild cat?’ In most of the sentences here, him/he refers to the cat being talked
about.
146
The interviewer responds: ‘It’s a wild cat’.
147
Produced after the interviewer says ‘The cat’s going upstairs’.
148
The cat has come back after going upstairs.
149
The interviewer then asks ‘Why’s he mad?’ Him mad was also produced on 4 other occasions.
150
The interviewer then says ‘He is mad at the people’.
151
The interviewer then says ‘Oh, he’s under the table’.
152
This sentence was scored as an attempt as ‘He is right here behind my back’
153
The interviewer responds ‘Oh, he’s not hungry’.
154
Produced after the child had previously said ‘This cat all tired’.
155
The interviewer responds ‘He’s upstairs’
156
Talking about a rat.
157
Interviewer replies ‘Oh, he’s up really high’
158
The interviewer then says: ‘He’s walking on the street?’
137
138
50
25. Now him eating159 26. Now him going160
27. Him hiding my back161
28. Him going
162
29. Him looking out there
30. Him sitting down 31. Him be up there163
32. Now he’s low
33. No, him don’t like peanut164
34. Him scratch165
35. Yeah, him go in166
36. Him want to lie down
37. Him bites people167
38. Him scratches people
The point of this exercise is to get you to look at which of the child’s utterances are in conformity with the
predictions of ATOM, and which are not. What you might do is the following.
(i) Look at the sentences 1-8 which the child appears to have produced in past tense contexts. Say what
range of different structures you find, and what range of structures ATOM would predict to occur, and try
to account for any discrepancy between predicted and occurring structures. More specifically, say whether
(and if so how) ATOM could account for sentences like 1, 6 and 8, discussing any problems which arise.
(ii) Look at the sentences 9-32 which occur in is contexts (i.e. in contexts where adults would use the verb
is or its contracted counterpart ’s – e.g. Him too big corresponds to the adult form He’s too big). Say what
range of different structures you find, and what range of structures ATOM would predict to occur, and try
to account for any discrepancy between predicted and occurring structures. In addition, say whether (and
if so how) ATOM could account for sentences like 30, 31 and 32, discussing any problems which arise.
(iii) Look at the sentences in 33-38, which all seem to occur in third-person-singular present-tense
contexts (the adult counterpart of 33 containing a 3Sg Present form of the auxiliary DO, and the adult
counterpart of 34-38 containing a 3Sg Present main verb). Say what range of different structures you find,
and what range of structures ATOM would predict to occur, and try to account for any discrepancy
between predicted and occurring structures. In particular, say whether (and if so how) ATOM could
account for sentences like 33, 34 and 37, discussing any problems which arise.
(iv) On the basis of the assumptions made in the ATOM model, try and calculate how frequently the child
marks tense and agreement. Compare the tense suppliance rate you arrive at with the agreement
suppliance rate, and try to account for any differences you find between the two.
§6.4 Overall verdict
What is your overall verdict on the strengths and weaknesses of the ATOM model? To what extent could
ATOM be said to be an explanatory model?
The interviewer responds ‘Now he’s eating, OK’
Produced after the interviewer says ‘He’s all done?’
161
The interviewer says ‘He’s hiding behind your back’.
162
The interviewer responds ‘Yeah, he’s looking out the window’
163
Glossed in the transcript as an attempt at ‘He is up there’.
164
Produced after the interviewer says ‘This elephant likes peanuts’. Bear in mind that in many varieties of
colloquial American English, don’t (like won’t) is used with any kind of subject, including a third person singular
subject like he/she/Daddy – hence the fact that soulful/doleful American pop songs typically lament that ‘He don’t
love me no more’ rather than that ‘He doesn’t love me any more’.
165
Produced just after the interviewer has said (about a cat) ‘Oh, he bites people’.
166
The interviewer responds by saying ‘Do wild cats go in houses?’
167
Produced just after the interviewer said ‘I want to play with the kitty’. The same sentence was also produced on
one other occasion.
159
160
51
7. The Dependency Deficit account of SLI
Key article
Van der Lely HKJ and Battell J (2003) ‘Wh-movement in children with grammatical SLI: A test of the
RDDR hypothesis’, Language 79: 153-181
OUTLINE OF THE DEPENDENCY DEFICIT MODEL
In work stretching back more than 10 years (See e.g. van der Lely, 1994, 1998, 2005; van der Lely &
Stollwerck, 1997; van der Lely and Battell 2003), Heather van der Lely has developed a model of SLI
which she terms the Representational Deficit for Dependency Relations Model (RDDR) – though it will
hereafter be referred to more succinctly as the Dependency Deficit Model (DDM). The core assumption of
DDM is that SLI children have problems in handling structure-dependent operations – i.e. grammatical
operations which involve a structural dependency between two different constituents, such as agreement
(e.g. of a verb with its subject), case-marking (e.g. of a subject by a verb) and movement (which involves
one constituent attracting another to move next to it). Van der Lely and Battell (2003) argue that DDM
gains empirical support from an elicited production study of the wh-questions produced by a group of
ESLIs (i.e. English subjects with SLI). Below is an outline of key claims made in their paper.
VAN DER LELY AND BATTEL’S STUDY
Van der Lely and Battell (henceforth VB) report on a study of the production of wh-questions by
a group of ESLI subjects with persistent problems in grammatical production and comprehension, but not
in other areas of language development168. They claim that the SLI subjects in their study fail to master the
syntax of the two types of movement operation involved in wh-questions (= preposing a wh-expression
and preposing an auxiliary), and argue that this is the result of difficulties they have in representing
syntactic dependencies between constituents. Since following the argumentation in their paper requires an
understanding of the assumptions they make about the syntax of adult wh-questions, let’s begin by taking
a look at the syntax of wh-questions in adult English.
Wh-questions in adult English
Before we look specifically at wh-questions, let’s first outline some basic assumptions made about the
structure of interrogative clauses (i.e. question structures). To make our discussion more concrete,
consider the structure of the sentence produced by speaker B in the following dialogue:
(1)
SPEAKER A:
What do you want to know?
SPEAKER B:
If Ralf can see her
In work in syntax over the past 30 years (dating back to Chomsky 1986), it would be assumed that speaker
B’s sentence is derived (i.e. formed) as follows. The verb see is merged (i.e. combined) with its (direct
object) complement her to form the VP/verb phrase see her. The (present-tense) T/tense-auxiliary can is
merged with the resulting VP see her and with its subject Ralf to form the TP (i.e. tense phrase – a term
used for an expression containing a present- or past-tense auxiliary) Ralf can see her. This TP is in turn
merged with the interrogative C/complementiser (a term used to denote a clause-introducing particle) if
to form the CP/complementiser phrase If Ralf can see her (where if serves the function of marking the
relevant clause as a yes-no question). Given these assumptions, speaker B’s sentence in (1) will have the
CP+TP+VP structure shown below:
(2)
[CP [C if] [TP Ralf [T can] [VP [V see] her]]]
But now suppose that instead of saying If Ralf can see her, speaker B had instead given the reply below:
(3)
SPEAKER A:
What do you want to know?
SPEAKER B:
Can Ralf see her?
They refer to the relevant form of SLI as G-SLI – i.e. a form of SLI which is characterised by a specifically
grammatical impairment.
168
52
Let’s suppose (in line with much thinking in current work in syntax) that all questions have a CP+TP+VP
structure, and that the derivation of Can Ralf see her? starts out in the same way as for If Ralf can see her.
If so, the verb see is merged with its complement her to form the VP/verb phrase see her, and the T/tenseauxiliary can is then merged with this VP and with its subject Ralf to form the TP/tense phrase Ralf can
see her. As in the case of (2), the next stage in the derivation of the sentence is for this TP to be merged
with an interrogative C/complementiser: but this time, instead of containing if, the relevant interrogative
complementiser is null (i.e. ‘silent’), because if is only used in complement clause questions, not in mainclause questions169. Merging the null complementiser with the TP we have already formed will derive the
structure shown in (4) below:
(4)
[CP [C ø] [TP Ralf [T can] [VP [V see] her]]]
However, the C constituent in a main-clause question (has a tense feature which) attracts the head T
constituent can of the TP Ralf can see her to move from the T position into the C position (an operation
traditionally called auxiliary inversion, but known in more recent work as T-to-C movement). The
relevant (T-to-C) movement operation is represented by the arrow in (5) below, with strikethrough on can
indicating the position occupied by can before it moved from T to C:
(5)
[CP [C Can] [TP Ralf [T can] [VP [V see] her]]]
In the light of our discussion of yes-no question structures like (5), let’s now turn to consider the structure
of wh-questions such as the following (adapted from VB p.157):
(6)
Who can Ralf see?
This is similar in part to Can Ralf see her – but with the exception that the (direct object) complement of
the verb see is the interrogative wh-pronoun who rather than her. More specifically, the derivation of the
sentence proceeds as follows. The V/verb see is merged with its complement who to form the VP/verb
phrase see who. The T/tense-auxiliary can is then merged with the VP see who and with its subject Ralf to
form the TP/tense phrase Ralf can see who. This is then merged with a null interrogative C constituent,
forming the CP shown in (7) below:
(7)
[CP [C ø] [TP Ralf [T can] [VP [V see] her]]]
As in main-clause yes-no questions, the null C constituent at the beginning of the clause attracts the
T-constituent (i.e. present-tense auxiliary) can to move from T into C. But since this is a wh-question, C
also attracts the wh-word who to move out of its original position into a new position at the front of the
overall CP (in front of the inverted auxiliary) – this second type of movement operation being known as
wh-movement. The effect of these two movement operations (T-to-C movement and wh-movement) is
indicated by the arrows below (where can and who indicate the original positions occupied by these two
words before they underwent movement to the italicised positions):
(8)
[CP Who [C can] [TP Ralf [T can] [VP [V see] who]]]
To introduce some technical terminology at this point, the inverted auxiliary can moves into the head C
position in CP, while the preposed wh-word who moves into the specifier position within CP (the
169
In traditional grammar, main-clause questions are termed direct questions, and complement-clause questions are
termed indirect questions. The structure if Ralf can see her produced by speaker B in (1) can be argued to be a
complement clause (= indirect) question if we consider speaker B’s reply to be an elliptical (i.e. abbreviated) form of
‘I want to know if Ralf can see her’, with all material other than the italicised complement clause undergoing ellipsis
(i.e. being ‘silent’).
53
specifier position in a phrase being a position in front of the head word of the phrase – so that Ralf can
likewise be said to be the specifier of the auxiliary can in a structure like (2) above).
As the brief outline of the syntax of main-clause wh-questions above illustrates, the derivation (i.e.
formation) of a main-clause wh-question like Who can Ralf see? involves two different structural
dependencies. One of these is a C-T dependency (i.e. a dependency between between C and T), in that C
attracts the auxiliary can to move from T into C. The other is a C-wh dependency (i.e. a dependency
between C and a wh-word), in that C attracts the wh-word who to become the specifier of C. If – as
claimed by DDM – people with SLI have problems in handling structural dependencies between
constituents, we’d expect to find that they have problems with the syntax of wh-questions: more
specifically (VB claim), we’d predict that they have problems with both auxiliary inversion (because this
involves a C-T dependency) and with wh-movement (because this involves in a C-wh dependency). VB’s
paper sets out to test this prediction.
The nature of VB’s study
VB’s study included 15 SLI subjects aged from 11;3 to 18;2, their performance being compared to that of
two groups of TD (typically developing) children, 12 grammar-matched children aged from 5;3 to 7;4,
and 12 vocabulary-matched children aged from 7;4 to 9;1. (For profiles of the children, see VB p.159.)
Data were collected on wh-questions170 containing who, what and which by getting the subjects to play a
version of Cluedo in which target wh-questions were elicited by prompts such as those in (9) below:
(9) Prompts and target responses for object questions
Prompt
Target response
Mrs Peacock saw someone in the lounge. Ask me who
Who did Mrs Peacock see in the lounge?
Mrs Brown placed something in the library. Ask me what What did Mrs Brown place in the library?
Professor Plum wore a coat. Ask me which one
Which coat did Professor Plum wear?
Results of VB’s study
The table in (10) below (adapted from VB’s table 4) reports on the types of wh-question errors made by
three groups of subjects in VB’s study.
(10)
Types of error produced by the subjects in the VB study
SLI subjects
Younger TD controls
0/15 (0%)
6/12 (50%)
No errors
3/15 (20%)
4/12(33%)
AUX errors
0/15 (0%)
1/12 (8%)
WH errors
12/15
(80%)
1/12 (8%)
WH and AUX errors
Older TD controls
12/12 (100%)
0/12 (0%)
0/12 (0%)
0/12 (0%)
The column headed SLI subjects tells you that none of the 15 SLI children in the study achieved errorfree performance on wh-questions, that 3 of them made errors only with auxiliary inversion, that none of
them made errors only with wh-movement, and that 12 of them made errors with both wh-movement and
auxiliary inversion. The column headed Younger TD controls tells you that of the 12 children in the
younger TD control group (aged 5-7 years), 6 produced no errors of any kind, 4 only made errors with
auxiliary inversion, 1 only made errors with wh-movement, and 1 made errors with both wh-movement
and auxiliary inversion. The column headed Older TD controls tells you that of the 12 children in the
older TD control group (aged 7-9 years), all produced completely error-free performance on
wh-questions, and hence made no errors with either wh-movement or auxiliary inversion.
170
VB collected data on both subject and object wh-questions. A subject wh-question is one in which the
wh-expression is the subject of the relevant verb (e.g. Who saw you? where who is the subject of saw); an object whquestion is one in which the wh-expression is the object of the relevant verb (e.g. Who did you see? where who is the
object of see). Here, I concentrate on the wh-object questions produced by the SLI subjects, partly because these are
the questions which posed greatest problems for the SLI subjects, and partly because the syntax of wh-subject
questions in adult English is a matter of controversy (e.g. about whether they involve wh-movement or not, and how
they can satisfy the WH-CRITERION without auxiliary inversion).
54
Errors made by VB’s subjects
Examples of sentences produced by the subjects in VB’s study which illustrate both the kind of auxiliary
errors made by their SLI subjects and the kind of wh-errors they made are given below:
(11)(a)
(b)
(c)
Who Miss Scarlett saw somebody?171
Which Reverend Green open a door?172
What did Colonel Mustard had something in his pocket?173
In order to understand what’s going in here, let’s take a closer look at the derivation of sentence (11a). The
V/verb see is merged with the pronoun somebody to form the VP/verb phrase see somebody. A T
constituent containing a past-tense affix (below denoted as Af) is then introduced into the derivation, and
merged with the VP already formed, and with the subject Miss Scarlett, forming the TP Miss Scarlett Af
see who. A null interrogative C is then merged with this TP, forming the CP shown in simplified form
below:
(12)
[CP [C ø] [TP Miss Scarlett [T Af] [VP [V see] somebody]]]
The specifier position in front of C is filled by directly merging the interrogative pronoun who in spec-C,
so deriving the structure:
(13)
[CP Who [C ø] [TP Miss Scarlett [T Af] [VP [V see] somebody]]]
When (as here) a tense affix is immediately adjacent to a following verb, it is lowered onto the verb in the
PF/morphophonological component of the grammar, via an operation traditionally referred to as Affix
Hopping, as shown by the arrow below:
(14)
[CP Who [C ø] [TP Miss Scarlett [T Af] [VP [V see ] somebody]]]
The resulting string see+Af is ultimately spelled out as saw, and consequently the overall structure in (14)
is ultimately spelled out as the sentence (11a) Who Miss Scarlett saw somebody?
The resulting sentence (11a) illustrates a double dependency failure. First of all, the SLI subject who
produces (11a) has failed to master the C-T dependency, in that C fails to attract the affix in T to move
into C. Moreover, the same individual has also failed to master the C-wh dependency in that the specifier
position in front of C is filled by merging the wh-pronoun who in spec-C, rather than by moving who from
(what should have been) its initial position as the (direct object) complement of the verb see into spec-C.
By contrast, a fluent native speaker of English would have derived the sentence as follows. The V/verb see
is merged with its complement who to form the VP/verb phrase see who. A past tense affix (Af) is then
merged with this VP and with the subject Miss Scarlett to form the TP Miss Scarlett see who. This TP is in
turn merged with a null interrogative complementiser to form the CP shown below:
(15)
[CP [C ø] [TP Miss Scarlett [T Af] [VP [V see] who]]]
The null complementiser then attracts the tense-affix in T to move into C, and at the same time also
attracts the wh-word who to move into spec-C, so deriving the following structure:
(16)
[CP Who [C Af] [TP Miss Scarlett [T Af] [VP [V see] who]]]
Because the tense affix in C is not immediately followed by a verb, it undergoes DO-SUPPORT – i.e. the
(meaningless) ‘expletive’ or ‘dummy’ stem DO is attached to the affix in order to support it (i.e. in order
to ensure that it has a verbal stem to attach to), and the resulting string DO+Af is eventually spelled out as
did – with the consequence that the overall structure in (16) is spelled out as the sentence Who did Miss
Response to ‘Miss Scarlet saw someone in the lounge. Ask me who’ (the target response being Who did Miss
Scarlet see in the lounge?)
172
Response to ‘Reverend Green opened a door. Ask me which one’ (the target response being Which door did Rev.
Green open?).
173
Response to ‘Something was in Colonel Mustard’s pocket. Ask me what’ (the target response being What was in
Colonel Mustard’s pocket?).
171
55
Scarlett see? However, as already noted, the corresponding sentence (11a) produced by the SLI subject
shows that neither the C-T dependency (involving movement from T to C) nor the C-wh dependency
(involving movement of a wh-word from within VP into spec-C) have been mastered. Sentences like (11a)
thus appear to provide empirical support for the Dependency Deficit model.
Now consider another sentence produced by one of VB’s SLI subjects, namely (11b) Which Reverend
Green open a door? This would appear to be derived as follows. The V/verb open is merged with its
(direct object) complement a door to form the VP open a door. A null T constituent (ø) is then merged
with this VP and with its subject Reverend Green to form the TP Reverend Green ø open a door. The
resulting TP is merged with a null interrogative complementiser to form the structure shown below:
(17)
[CP [C ø] [TP Reverend Green [T ø] [VP [V open] a door]]]
The specifier position in front of C is filled by directly merging the wh-word which in spec-C, so deriving
the structure:
(18)
[CP Which [C ø] [TP Reverend Green [T ø] [VP [V open] a door]]]
Since the T position appears not to contain a past tense affix, there is no affix to lower onto the verb open,
with the result that the structure in (18) is eventually spelled out as sentence (11b) Which Reverend Green
open a door? Like (11a), sentence (11b) exemplifies a double dependency failure, in that the SLI subject
who produces the sentence has not mastered either the C-T dependency (i.e. the need to fill the C position
by moving an affix or auxiliary in T into C), or the C-wh dependency (i.e. the need to fill the spec-C
position with a wh-word via movement rather than merger).
Finally, let’s see look at what goes wrong in (11c) What did Colonel Mustard had something in his
pocket? Let us suppose that this is derived as follows. The verb have is merged with the pronoun
something and with the prepositional phrase in his pocket to form the VP have something in his pocket. A
past tense affix (Af) then merged with this VP, and with its subject Colonel Mustard, forming the TP
Colonel Mustard Af have something in his pocket. The resulting TP is in turn merged with a null
interrogative complementiser, forming the structure shown below:
(19)
[CP [C ø] [TP Colonel Mustard [T Af] [VP [V have] something in his pocket]]]
Because an affix can only be attached to an immediately fllowing verb, and the past tense affix is
eventually attached to the verb have (with the resulting have+Af string eventually being spelled out as
had), it would appear that the tense-affix remains in T, and at no point moves into C. This being so, it
follows that the head C position of CP in (19) must be filled by directly merging the auxiliary did in C.
Likewise, the specifier position in front of C is filled by directly merging the wh-word what in spec-C.
These two merger operations result in the structure shown below:
(20)
[CP What [C did] [TP Colonel Mustard [T Af] [VP [V have] something in his pocket]]]
The past tense affix in T undergoes Affix Hopping in the PF/morphophonological component of the
grammar, forming the string have+Af which is eventually spelled out as had. Consequently, the overall
structure in (20) is eventualy spelled out as sentence (11c) What did Colonel Mustard had something in his
pocket? Once again, the structure involves a double-dependency error, with the SLI subject in question
failing to master the C-T dependency (i.e. failing to understand that C is filled by movement rather than
merger), and likewise failing to master the C-wh dependency (i.e. failing to understand that spec-C is
filled by movement rather than merger).
VB’s main claims
VB’s two main claims are the following. Firstly, SLI subjects have far more problems with the syntax of
wh-questions than language-matched TD controls. And secondly, the pattern of errors made by the SLI
subjects differs from the pattern of errors made by the TD subjects: most SLI subjects have problems with
both auxiliaries and wh-expressions, whereas most TD subjects have problems with neither (and if they
have any problems at all, they tend to have problems only with auxiliary inversion)174. This suggests to
174
VB suggest on p.167 that the one TD subject showing problems with both wh-movement and auxiliary inversion
may have ‘an undiagnosed language deficit’.
56
VB that SLI children do not understand the part played by movement operations in establishing syntactic
dependencies like the C-T and C-wh dependencies involved in main-clause wh-questions.
VB maintain that other theories of SLI are unable to explain their findings. For example, they maintain
that Leonard’s Perceptual Deficit hypothesis provides no account of errors such as that in (11c)
What did Colonel Mustard had something in his pocket? VB argue that such errors cannot plausibly be
taken to involve problems with ‘perception and processing of nonsalient morphemes’. Likewise, VB claim
that an approach which assumes a global deficit in all grammatical features (like Gopnik’s 1990 featureblindness account) could not account for the fact that the SLI subjects in their study sometimes produce
correct wh-questions, and sometimes produce incorrect forms. Similarly, they maintain (p.171) that
approaches which assume a selective feature deficit (e.g. in tense features as in Rice and Wexler 1996, or
in agreement features as in Clahsen, Bartke and Göllner 1987) ‘appear too narrow’ in that ‘neither of these
accounts would predict problems with WH-operator movement’, so that such accounts ‘do not provide
parsimonious explanations of the data for G-SLI subjects’. By contrast, they maintain that their own
RDDR model makes precisely the right prediction – namely that because people with SLI have problems
in handling structural dependencies between constituents, they will also have problems in handling the CT dependency involved in auxiliarty inversion, and the C-wh dependency involved in wh-movement, and
hence that they will show impaired performance on both auxiliary inversion and wh-movement in whquestions like Who did Ralf see? Their overall conclusion (p.170) is thus that ‘The findings of this study
confirm the predictions of the RDDR hypothesis and thus support this account of G-SLI’.
WORKBOOK SECTION
Below, I pose a number of questions designed to help you develop a critical evaluation of VB’s paper.
§7.1 Alternative interpretations of VB’s data
An interesting question which arises from VB’s research is whether the types of wh-question errors which
they report SLIs making doin fact provide strong empirical support for the dependency deficit which they
claim, or whether (at least some) of the structures produced by their SLI subjects can be accounted for in
ways which do not require us to assume any dependency deficit. For example, why is VB’s claim that an
SLI sentence such as:
1. Who Mrs Brown see?
shows a failure to understand the C-T dependency (and the need for the head C position of CP to be filled
by the auxiliary (di)d) weakened by the observation that adults (in informal speech) frequently give a null
spellout to a contracted inverted auxiliary, as in the following sentences (from the Bates files on the
CHILDES data-base) produced by mothers talking to children aged 1;8 and 2;4, where the mothers give a
null spellout to the parenthesised auxiliaries?
2. Where you going? (are). What she doing? (is). What they got in them? (have). How they sit? (do).
Then what she do? (did)
And why is the claim that SLI sentences such as
3. What did they drank?
show a similar C-T dependency deficit weakened by the claim made by Chomsky (in work since 1995)
that movement is a composite operation involving placing a copy of some constituent in a new position,
and then subsequently giving the copy in the original position a null spellout (i.e. a ‘silent’ pronunciation,
in effect deleting it)? In addition, evaluate the strength of VB’s claim that SLI sentences like
4. Which Reverend Green open a door?
involve the wh-word which being positioned in spec-C by merger rather than wh-movement. In this
connection, bear in mind that English has structures like what a goal! and how good a goal? in which the
indefinite article a can have a wh-expression modifying it (as its specifier). Bear in mind also that
Gavruseva and Thornton (2001) report that even typically developing children sometimes prepose a
wh-word on its own without pied-piping (i.e. ‘dragging’) along with it material modified by the wh-word,
57
- as illustrated by the structure in 5 below produced by a typically developing 6-year-old (where whose has
been preposed without the noun ball that it modifies):
5. Whose do you think ball went in the cage?
Finally, assess the strength of the claim made by VB that SLI sentences such as (6) below provide
empirical evidence that SLIs fill spec-C via merger rather than movement:
6. Who Miss Scarlett saw somebody?
In this connection, bear in mind that David Pesetsky (1997, 1998) has argued that wh-movement may
sometimes result in a partial copy of a moved wh-pronoun being left behind in the position out of which
the wh-pronoun initially moved, so resulting in a structure such as that in the bracketed relative clause
below:
7. He’s someone [who you’re never sure whether to trust him or not]
The partial copy that is left behind by movement in such cases (Pesetsky argues) is a pronoun which
carries essentially the same (e.g. person/number/gender/specificity) features as the wh-pronoun except for
its wh-feature (i.e. the partial copy is a non-wh counterpart of the wh-pronoun). What this could mean is
that if a non-specific wh-pronoun moves (e.g. who meaning ‘which person out of an unspecified set of
people’), it could leave behind a wh-less non-specific pronoun like somebody in the position out of which
it moves.
§7.2 Wh-movement in the Leonard corpus
Given that VB’s study is experimental in nature (in that they used a particular experimental technique to
elicit wh-questions from their SLI subjects), an interesting way of testing their theory empirically is to see
whether similar types of wh-question errors are found in a large naturalistic corpus (e.g. a publicly
accessible data-base containing transcripts of the spontaneous speech production of SLI children). One
such corpus is the Leonard SLI corpus on the CHILDES data-base (MacWhinney 1995), which contains
samples of the ‘free speech’ of 11 SLI children. Radford and Lin (2005) searched this corpus for examples
of main-clause non-subject175 wh-questions. They found that the SLI children in the Leonard corpus
produced a total of 294 questions containing an overt non-subject wh-expression, and that all of these
showed adult-like wh-movement. Why do you think it might be that there are no examples of wh-merger
structures like
(1)
Who Miss Scarlett saw somebody?
in the Leonard corpus – and what is the potential significance of this for the Dependency Deficit Model? In
addition, comment on the potential significance of the absence of wh-in-situ questions like (2) below in the
Leonard corpus (and indeed in van der Lely and Battell’s study):
(2)
They drank what?
Would the Dependency Deficit model lead us to expect SLI children to produce wh-in-situ questions like (2)?
§7.3 Wh-question errors in the Leonard corpus
Below are a representative sample of different types of non-adult-like wh-question structures (for mainclause non-subject wh-questions) produced by SLI children in the Leonard corpus referred to in §7.2 (the
child producing each example being identified by a parenthesised letter such as (B) after the example, and
the adult counterpart of each sentence being enclosed in inverted commas).
1 Which one I can do? (C ‘Which one can I do?)
2. What Kent’s gonna play with? (C ‘What’s Kent gonna play with?)
3. How you knowed? (E ‘How did you know?’)
175
I.e. questions where the wh-expression is not the subject of the clause containing it. The reason for excluding
subject questions is that wh-movement in them is seemingly ‘invisible’ and they do not involve auxiliary inversion
(for reasons discussed in Radford 2004).
58
4. What he did? (F ‘What did he do?’) )176
5. What you doing? (E ‘What are you doing?’)
6. What this for? (G ‘What is this for?’)
7. How much we got to do? (J ‘How much have we got to do?’)
8. How you get this out? (A ‘How d’you get this out?’)
9. What this do? (A ‘What’s this do?/What does this do’)
10. How open it up? (B ‘How d’you open it up?’)
11. What say? (B ‘What d’you say/What did you say?’)
12. Where go on? (B ‘Where’s it go on/Where does it go on?’)
13. How much long gonna be? (A ‘How much longer’s it gonna be?’)
14. These do? (C ‘What do these do?’)
15. What is this is? (H ‘What is this?’)
What different types of wh-question do we find in sentences 1-15 above, and to what extent do the various
question types show the kind of dependency deficit predicted by VB? What is the potential significance of
the fact that (as illustrated in the examples in 2 in §7.1) adults often give a null spellout to a contracted
inverted auxiliary in a wh-question? What is also the potential significance of the observation by Virginia
Valian (1991) that adults in rapid speech styles sometimes give a null spellout to a weak unstressed whword at the beginning of a sentence, hence (e.g.) giving a null spellout to what in What time is it? (More
generally, adults sometimes omit one or more unstessed syllables at the beginning of a sentence, so e.g.
giving a null spellout to the material marked by strikethrough in It’s a nice day, isn’t it?). You might also
care to note that (in a similar vein) Luigi Rizzi (2000) argues that even typically developing children often
give a null spellout to weak constituents at the edge (i.e. in the head or specifier position) of CP in a main
clause, so that a preposed auxiliary and/or a preposed wh-word are sometimes ‘silent’ if weak (i.e. if they
have little phonetic substance).
More specifically, prepare analyses of sentences 1, 2, 8, 11, 12 and 15. Assume that you have reached a
stage in the derivation of each sentence at which you have formed a CP containing a null interrogative
complementiser like that below (where ø marks a indicates that the relevant T constituent is null and so
contains no overt auxiliary, and strikethrough indicates that the relevant word ultimately receives a null
spellout and so is ‘silent’):
(1a)
(4a)
(5a)
(9a)
(11a)
(14a)
(15a)
[CP [C ø] [TP I [T can] do which one]]
[CP [C ø] [TP He [T Af] do what]]
[CP [C ø] [TP You [T are] doing what]]
[CP [C ø] [TP This [T Af] do what]]
[CP [C ø] [TP You [T Af] say what]]
[CP [C ø] [TP These [T Af] do what]]
[CP [C ø] [TP This [T is] what]]
Assume that T contains a past tense affix in (4a) and (11a), a (third person singular) present tense affix in
(9b), and a (third person plural) present tense affix in (14a). Assume also that an affix which remains in T
is eventually lowered onto an immediately following verb, but that an affix which moves to C has to
undergo DO-support (because there is no verb immediately following it for it to lower onto). Bear in mind
in relation to (15) that Chomsky has argued in work over the past decade that movement is a composite
operation involving placing a copy of some constituent in a new position, and then subsequently giving
the copy in the original position a null spellout (i.e. a ‘silent’ pronunciation, in effect deleting it).
176
The adult counterparts of child questions which differ in form from the corresponding adult questions are as
follows: 1. What is this? 2. What are you doing? 3. What is this for? 4. How much have we got to do?
5. How do you get this out? 6. What does this do? 7. How d’you open it up? 8. What d’you (= did you) say?
9. Where does it go on? 10. How much longer is it gonna be? 11. What do these do? 12. Which one can I do?
13. What’s Kent gonna play with? 14. How did you know? 15. What did he do?
59
§7.4 Two types of movement operation
Radford and Lin (2005) reported that (on non-subject wh-questions) the SLI children in the Leonard
corpus showed 100% correct suppliance of wh-movement, but only 35% correct suppliance of auxiliary
inversion. Would the Dependency Deficit Model predict this kind of difference? What if we were to posit
that wh-movement and auxiliary inversion differ from each other in that wh-movement is an interpretable
movement operation (in the sense that it has an effect on semantic interpretation/meaning), but auxiliary
inversion is an uninterpretable movement operation (i.e. one which has no effect on the meaning of a
sentence)177. What interpretive effect (i.e. effect on meaning) does wh-movement have? Well, if we follow
Radford et al (1999) in supposing that (for languages like English)
(i)
A clause is only interpreted as a non-echoic question if it is a CP with an interrogative specifier
then we can say that the interpretive effect of wh-movement in a question is to ensure that the relevant
structure begins with an interrogative wh-expression and so can be interpreted as a wh-question. At first
sight, it might seem that auxiliary inversion also has an interpretive effect, in that (we might suppose) a
clause is interpreted as a question only if it is a CP containing an auxiliary in the head C position of CP.
However, this claim seems untenable for a number of reasons. For one thing, a complement-clause
question like the CP bracketed below:
(ii)
She asked me [CP where [C ø] [TP he [T had] gone]]
contains an interrogative wh-word in spec-C but no auxilary in C – which means that it is wh-movement
and not auxiliary inversion which is crucial to ensuring that the bracketed clause is interpreted as
interrogative. And conversely, some structures containing an inverted auxiliary in C are not interrogative –
as can be illustrated in relation to the CP in (iii) below:
(iii)
[CP Never [C have] I heard such rubbish]
If it is indeed the case that wh-movement is an interpretable operation (and thus has an effect on meaning)
whereas auxiliary inversion is an uninterpretable operation (which therefore has no effect on meaning),
consider what the implications would be of generalising the Uninterpretable Feature Deficit Model of
Tsimpli and Stavrakaki (1999) and supposing that SLIs are good at handling interpretable features and
interpretable operations, but poor at handling uninterpretable features and uninterpretable operations.
Would this account fare better than the Dependency Deficit Model in accounting for why SLIs appear to
be much better at wh-movement than at auxiliary inversion?
§7.5 Overall verdict
In the light of the questions asked above, what is is your overall verdict on the strengths and weaknesses
of Heather van der Lely’s Dependency Deficit model?
177
Chomsky (1999) maintains that this is a consequence of wh-movement being a syntactic operation, but auxiliary
inversion being a PF (i.e. morphophonological) operation – but I will not go into relevant technical details here.
60
8: Empirical study I: Acquisition of past-tense marking
This chapter (and the next 3) aim to show you how to do a small-scale empirical178 research project
investigating particular aspects of the grammatical development of an SLI child, how to analyse the data,
how to use your research results to evaluate a number of different theories of SLI, and how to write up
your project. The data for your project come from a set of transcripts of the speech production of a
4-year-old American boy with SLI, known as JC. Brief details of JC and a copy of the transcripts can be
found in chapter 13. In this chapter, I show you how to use the JC corpus to undertake an empirical study
of The acquisition of past tense marking on main verbs by a 4-year-old American boy with SLI.
Your study might be organised into chapters/sections and subsections along the following lines.
1. Background (about 25% of your overall study)
1.1 Introduction
Start with a paragraph outlining the kind of developmental disorder which SLI involves (summarised on
p.4 of this book: see Bishop (1997), Leonard (1998), and the collection of papers in Bishop and Leonard
(2000) and Levy and Schaeffer (2003) for more detailed discussion of diagnostic criteria for SLI), and
point out that various aspects of grammar are reported to be impaired in SLI children, including their
ability to mark tense in obligatory contexts. There are a wide range of different models of SLI which seek
to account for the nature of the grammatical impairment in SLI children, and for the purposes of the
present study you are going to examine just a few of these (e.g. you might choose any three of the
Perceptual Deficit, Rule Deficit, Agreement Deficit and ATOM models). Your aim is to do a small-scale
empirical study designed to evaluate each of the models, by looking at the production of regular and
irregular past tense main verbs by a 4-year-old American boy with SLI.
1.2 Outline of the morphophonology of past tense marking in English
In this section, you should provide a brief account of the morphophonology of past tense marking in
English. The dual mechanism model of the acquisition of morphology (outlined e.g. in Pinker and Prince
1988, and Pinker 1991) maintains that irregular forms are memorised forms which are stored/listed in the
mental lexicon, whereas regular forms are computed or derived via application of morphological rules like
those sketched informally in (1) below (with morphemes enclosed in curly brackets):
(1)
A regular verb carries the suffix
{s} if third person singular present
{d} if past, perfect or passive
{ing} if progressive or gerund
{ø} otherwise
The precise phonological form which a morpheme is given in a particular structure is determined by
spellout rules like (4) below179:
(2)
{d} is spelled out as:
|Id| when attached to a stem ending in an alveolar stop – i.e. |t| or |d| (e.g. padded, batted)
|d| when attached to a stem ending in another voiced segment180 (e.g. moved, sneezed, paid)
|t| when attached to a stem ending in another voiceless segment181 (e.g. passed, packed, laughed)
So, for example, the lexical entry (i.e. entry in the dictionary) for the regular verb DIE will contain simply
the stem form die: its past tense form will be generated by adding the the suffix {d} to the stem form in
178
An empirical study is one which involves analysing a set of data in order to test one or more research hypotheses.
Since children are unable to read when they produce their earliest verb-forms, what is crucial is how such forms
are pronounced, not how they are written. Hence, I ignore the fact that e.g. past tense {d} is sometimes spelled -ed.
180
Since all vowels are voiced, this means ‘when attached to a stem ending in a vowel, or a voiced consonant other
than |d|’.
181
In other words ‘when attached to a stem ending in a voiceless consonant other than |t|’.
179
61
accordance with the second line of rule (1), so deriving die-d; the {d} suffix will be spelled out as |d| in
accordance with the second line of rule (2), so that the word is pronounced |da Id|. By contrast, the lexical
entry for the irregular verb CATCH will specify that it has the irregular past/perfect/passive form caught,
and the fact that the verb is listed as having an irregular past/perfect/passive form will block application of
the regular d-rule in (1) to irregular verbs like CATCH.
Empirical research is about testing research hypotheses, so in the next few sections you should provide a
short summary of the overall claims made by each of the models of SLI which you have decided to
evaluate, and then go on to say what specific prediction each makes about what SLI children do in past
tense contexts182 (these predictions being the research hypotheses which you are going to test). You
should provide evidence of having read the key research paper/s on each model you discuss (e.g. by
quoting a couple of key sentences from it/them).
1.3 Outline of the Perceptual Deficit model of SLI
In this section you should provide an outline of the Perceptual Deficit model183, noting that it claims that
consonantal inflections pose more perceptual problems for SLI children than vocalic inflections, and so
predicts that SLI children will perform worse on producing the past tense of regular verbs like played
(where past tense is marked by the consonantal inflection -d) than on irregulars like gave where past tense
is marked by vowel-change (replacing the |I| vowel of the citation form184 give by an [eI] diphthong). But,
as we saw in chapter 2, things are not quite as straightforward as that, for two reasons. Firstly – as we see
from (2) above – the regular past tense morpheme {d} has the three allomorphs |t|, |d| and |Id| and we
would therefore expect SLI children to perform better on the last of these three because it contains a
vowel: as Leonard (1989, p.187) notes, this is because ‘although all are low in phonetic substance, the
former (= |Id|) receives the benefit of greater acoustic energy from the vowel’. Moreover, if consonant
clusters cause additional perceptual problems, we’d expect the performance of SLI children on the purely
consonantal |t,d| allomorphs of past tense {d} to be worse when these are added to a verb stem ending in a
consonant (as in e.g. dumped which contains the cluster |mpt|) than to one ending in a vowel or diphthong
(as in tried, for example). An additional complicating factor is that not all irregular pasts are marked by
vowel-change alone (consonant change being used to mark the difference between the citation form send
and the irregular past tense form sent, for example). So, the Perceptual Deficit model would lead us to
expect that an SLI child should perform better on |Id| than on purely consonantal allomorphs of regular
past tense {d} and better on the latter when they don’t give rise to a consonant cluster than when they do,
and that they should perform better on irregulars which mark past tense by vowel change alone than on
other types of irregular.
1.4 Outline of the Rule Deficit model of SLI
In this section you should provide an outline of the Rule Deficit model185, which claims that SLI children
have a genetic rule-deficit that makes it impossible for them to acquire any regular morphological rules.
They are therefore unable to acquire the regular past tense d-rule, and this predicts that they will be unable
to create novel or over-regularised past tense forms186. Because of their genetic rule deficit, the only
learning mechanism which SLI children have is memorisation, with the consequence that they have to
182
I use this rather vague formulation here because some models of SLI make predictions not only about the
morphology of the verbs used by SLI children in past tense contexts, but also about the case properties of their
subjects – and you may want to test the latter as well as the former in your study.
183
Only if you have decided to include this model as one of the models that you are testing, naturally.
184
The citation form of a word is the form under which it is listed in a dictionary
185
Only if you have decided to include this as one of the models that you are testing.
186
However, it should be noted that Ullman and Gopnik (1994) maintain that in spite of their inability to form
implicit (i.e. subconsciously internalised) rules, SLI children who have undergone extensive therapy may learn an
explicit (i.e. consciously memorised) rule taught to them by therapists or teachers telling them to ‘Put -d on the end
of a verb in a past tense context’. This means that the production of over-regularised forms does not necessarily
‘prove’ that a child has developed a subconsciously internalised past tense formation rule. However, JC is only 4
years of age, and it is unlikely that he is old enough for therapists to attempt to teach him abstract grammatical rules.
62
memorise regular and irregular inflected forms alike. Since memorisation of a given form improves with
increased exposure to the relevant form, we therefore expect to find a frequency effect (viz. that the more
often a child hears a given past tense form, the more often the child is likely to produce it correctly). More
specifically, if Gopnik and Crago are right in claiming that SLI children have a rule-deficit, we should
expect to find a frequency effect with regular as well as irregular verbs (since, according to them, both
sets of verbs are acquired via the same mechanism of memorisation); but if Gopnik and Crago are wrong
and SLI children (like normally developing children) learn regular forms by rule and irregular forms by
memorisation, we should expect to find a frequency effect with (memorised) irregulars, but not with
(rule-derived) regulars.
1.5 Outline of the Agreement Deficit Model
In the next section, you should provide an outline of the Agreement Deficit model187, which claims that
children with SLI have considerable problems in acquiring uninterpretable agreement features, but few
problems in acquiring interpretable features like tense. Among the predictions made by the model are that
SLI children should fare much better on past tense forms (because these mark an interpretable past-tense
feature) than on third person singular present-tense/3SgPres forms (because these mark both an
interpretable present-tense feature and uninterpretable third-person-singular agreement features, and their
uninterpretable agreement features will therefore make 3SgPres forms harder to acquire).
1.6 Outline of the Agreement-and-Tense-Omission (ATOM) model of SLI
In the next section, you should provide an outline of the Agreement-and-Tense-Omission/ATOM
model188, which claims that SLI children go through an Extended Optional Infinitives/EOI stage during
which they sometimes fail to mark tense and/or agreement in finite contexts (i.e. in contexts where adults
use an auxiliary or main verb marked for tense and agreement). The presence or absence of agreement on
the verb will affect the case assigned to the subject (in that a nominative subject is used when agreement is
marked and a default accusative subject when it is not). ATOM makes the following prediction about how
an SLI child at the Extended Optional Infinitives/EOI stage will reply to a past tense question like ‘What
did Daddy do yesterday?’ One possibility is that the child will mark both tense and agreement: if so, the
verb play will carry the -d suffix (because it is marked for past tense) and the subject will be nominative
(because agreement is marked), giving He played with me. A second possibility is that the child may mark
tense on the verb, but not agreement: because past tense is being marked, the verb will then carry the past
tense suffix -d in accordance with line 2 of rule (1), but because agreement is not being marked, the
subject will have accusative case by default, resulting in Him played with me. A third possibility is that the
child may mark agreement on the verb, but not tense; because tense is not being marked, the verb will not
carry past tense -d and instead will be a bare infinitive form carrying a null affix in accordance with the
last line of rule (1); but because agreement is marked, the subject will be nominative so that we have He
play with me. The fourth and final possibility is that the child may use the verb PLAY without marking
either tense or agreement on it: absence of tense will mean that the verb does not carry -d, and absence of
agreement means that its subject will get accusative case by default, giving rise to Him play with me.
Because ATOM does not make any claims about whether agreement would be expected to be more or less
severely impaired than tense, it doesn’t make any precise predictions about whether overall there will be
(e.g.) more agreement-omission errors (with default accusative subjects) than tense-omission errors (or
conversely).
1.7 Summary of research hypotheses being tested
In this section, you should summarise the set of research hypotheses you have set out to test in your
project – e.g. along the lines shown below:
187
188
Only if you have decided to include this as one of the models that you are testing.
Only if you have decided to include this as one of the models that you are testing.
63
(3)
Perceptual Deficit model
SLI children perform better on (e.g. irregular) verbs where past tense is marked by a vowel than on
(e.g. regular) verbs where it is marked by a consonant
(4)
Rule Deficit model
SLI children will not produce spontaneous novel or over-regularised pasts189, and will perform
better on more frequent regular and irregular past-tense verb forms than on less frequent forms
(5)
Agreement Deficit model
SLI children have problems with uninterpretable agreement features, and so will perform better
on past tense forms (because these mark an interpretable past-tense feature) than on third person
singular present tense forms (because these mark not only an interpretable present-tense feature but
also uninterpretable third-person-singular agreement features)
(6)
Agreement-and-Tense-Omission/ATOM model
In past tense contexts, children alternate between producing past tense forms like played and bare
(infinitive) forms like play, and in both types of structure alternate between using nominative
subjects like I/we/he/she/they and default accusative subjects like me/us/him/her/them.
2. Research results (about 25% of the overall study)
2.1 Introduction
Begin with an Introduction (section 2.1) in which you note that you are setting out to test the predictions
made by your chosen theories of SLI by examining the types of structure produced by JC in contexts
where adults would use a past-tense main verb. In order to do this, you first need to draw up a list of
sentences containing main verbs which JC appears to use in past tense contexts (and you should include
this list as an Appendix to your eventual project). In order to help you with this task, below I have listed
sentences produced by JC containing (italicised) main verbs used in a past tense context (in the sense that
the italicised verbs forms arguably occur in a context where an adult would use a past tense main verb):
(7) List of (italicized) main-verb forms potentially used by JC in past tense contexts
5. He see snow on he chimney 16. Me used to have a dog, but somebody take it away
18. Now, going home cause it melted 23. Me got one from Michael 27. Me got chicken pox
32. Me fall asleep on the couch
33. Me said, me gotta hurry up and go up
35. Come up there, then me sleep up there
36. Me fall asleep up there
81. Like me see on the TV
97. No, somebody else taked it 101. Somebody else asked my mom to play outside with them
102. No one shoveled it
103. Me just jumpovered it
105. He shoveled someone else
112. He burn heself here
113. He eat it
114. No, took it off of...then he eat it
120. The teacher said when you eat a lot of food you got food in your mouth
125. I think this, he...umm, oh, he dump it 126. He shoveled him truck
127 And then he dump it 130. He shoveled him truck 131. Then he dump it
132. And then he drived away 133. Because, he want to put it 144. Then, he knocked him window
170. Long time ago you give me that first 174. He family, he lost he family
176/7. And he can't go and he family... die 179. He lost him duck 193. Oh, he caught me 198. Long time
ago I have a big eye 208. An owl did this with he eyes
210. And he, hooh, and someone look at there
and said ah! 211. And then, he scare them and said ha, ha, ha
213. Then, he talk to himself
214. Some wake up middle of night 215. Then, when someone go down, and the owl scare them
229. We see clown at...umm...
230. I see clown at a post office
231. Them have a party, and clown give me a balloon 232. He give a plane balloon
236. I never saw one of these stove
251. You lost one of these
189
As noted earlier, the only way SLI children could produce such over-regularised past-tense forms is if they have
consciously learned an explicit rule taught to them by a therapist or teacher.
64
259. Peter laughed you 267. Me teacher make cake 268. Her make it for Ms Peggy a long time ago
277. Her give me dad a lobster 278. Me mom put in here, cook them
279. Forgot to take them eyes out
280. And then, it give it to mom
281. He say, put it down
282. And then her say, ahhh!
283. And then her put on the floor, and we scare her 284. Her say, ahh it's moving
285. And then, them cook them up
286. And then it scared mom 290. And me dad eat it
291. Daddy cook it. 298. I know, he wanned to go and drop it
300. He jumped out
301. He bit me
303. Then, her drink some water 303 Water make cookie all gone 306. Maybe some lobster pinch him
311. Her have a snake in the sink
313. Then grandma get bitten by a snake
314. Then her got hurt
315. Then, me and xxx drop a xxx, then it hurt me
316. He try to eat me up
321. Long time ago, I go camp and hiking at the same time
323. I say, and I eat at hiking
324. And then at hiking, I xxx a picnic, so we eat 326. Then a bee eat a little bit my food
327. Then me said, oh!
328. Then me take it xxx 329. Then me go camping 330. Me eat all our food
337. Me mommy show me to make Easter eggs 339. I forgot
365. I got my hair from barber
366. But them cut my hair real tiny 379. He cut me
You should include the list in (7) as an Appendix to your project. Your first task is to use the list in (7) to
compile a frequency table like the (partial) table in (8) below, showing how often JC uses particular verbforms in past tense contexts:
(8) Raw frequency of verb-forms used by JC in past tense contexts
Type
Verb
A: correct form
B: bare form
1. regular
DUMP
dump = 3
1. regular
SCARE
scared = 1
scare = 3
1. regular
SHOVEL
shovelled = 4
2. novel
JUMPOVER
jumpovered = 1
3. mixed
BURN
burn = 1
4. Irregular
DRIVE
4. Irregular
EAT
eat = 7
4. Irregular
LOSE
lost = 3
4. Irregular
TAKE
took = 1
take = 2
5. Invariable irregular CUT
cut = 2
C: other form
drived = 1
taked = 1
In the first column of the table, I’ve classified verbs into five types in respect of how they form their past
tense. Regular verbs are those which have a regular adult past tense form ending in {d}. Novel verbs are
those like JUMPOVER (in this instance, created by fusing jump and over into a compound word) which
have no adult counterpart: such words typically have regular past tense forms (e.g. if we created a novel
verb GLURG, it would have the regular past tense form glurged). Mixed verbs are those like BURN which
can either have a regular past tense form ( = burned) or an irregular past tense form (= burnt). Irregular
verbs are those which have an irregular past tense form which is distinct from the bare/infinitive form of
the verb (e.g. DRIVE which has the irregular adult past tense form drove). Invariable irregulars are verbs
like CUT which have an irregular past tense form identical to their bare/infinitive form – as we see from
sentences like ‘He didn’t want to cutINFINITIVE the grass, but reluctantly he eventually cutPAST it’.
In the second column, I’ve listed verbs alphabetically by type. In the third column (labelled A: correct
form), the fourth column (labelled B: bare form), and the fifth column (labelled C: other form), I’ve
shown the raw frequency of individual verb-forms (a superscript number being used to indicate how often
each particular verb-form was produced by JC in a past tense context). So, for example, the entry for TAKE
says that (in past tense contexts) JC used the correct adult form took once, the bare form take twice, and
the over-regularised form taked once. Of course, the table I have produced in (8) is a partial table showing
only some of the verb-forms which JC uses in past tense contexts. You must use the data in (7) to produce
a complete table of your own which shows all the verb-forms produced by JC in past tense contexts.
65
Your own (complete version of) the frequency table in (8) should be included in an Appendix at the end of
your project. You can use this table as the basis for a number of simple frequency calculations in the next
few sections of your study, and use the research results you obtain to test the predictions made by the
particular theories of SLI that you are looking at. The particular calculations you need to make will
depend on the particular models of SLI that you are testing (so your particular project may not need to
include some of the sections below).
2.2 Calculating JC’s percent suppliance of correct regular and irregular past tense forms
Some models of SLI make specific claims about whether SLI children do or don’t perform better on
irregular verbs than on regular verbs. In order to test such claims, we therefore want to compare the
frequency with which JC correctly marks past tense on regular verbs with the frequency with which he
correctly marks past tense on irregular verbs. One way of doing this is in terms of the formulae shown
below (NB no. = number)
(9)
Percent suppliance of correct past =
tense form for regular verbs
no. of tokens of correct regular past forms x 100
no. of obligatory contexts for regular past forms
(10)
Percent suppliance of correct past =
tense form for irregular verbs
no. of tokens of correct irregular past forms x 100
no. of obligatory contexts for irregular past forms
What (10) means is as follows. For all regular (i.e. type 1) verbs in your own complete version of table
(8), count up the number of tokens (i.e. examples/occurrences) of correct forms listed in column A,
multiply this sum by 100, and then divide the resulting total by the number of times the relevant verbs
were used in past tense contexts (which you can arrive at by adding up the number of tokens of correct,
bare and other forms in columns A, B and C). So, if we did this calculation for the partial table in (8),
we’d end up with 5 tokens of 1A forms (1 occurrence of scared and 4 of shoveled). We’d then multiply
this figure by 100, giving a total of 500. We’d then divide this total by the number of tokens of verb-forms
in the cells in 1A, 1B and 1C together, i.e. 11 (comprising 4 occurrences of scared, 1 of shoveled, 3 of
dump and 3 of scare), giving us an overall percent score of 45% suppliance of correct past tense forms for
regular verbs (i.e. 500 ÷ 11).
What (10) means is this. For all type 4 verbs in your list (i.e. irregular verbs which have a past tense form
distinct from their bare form, and which do not have an alternative regular past tense form as well190),
count up the number of tokens of correct forms listed in column A, multiply this sum by 100, and then
divide the resulting total by the number of times the relevant verbs were used in past tense contexts (which
you can calculate by adding up the number of tokens of correct, bare and other irregular forms in columns
A, B and C). So, if we did this calculation for the partial table in (8), we’d end up with 4 tokens of 4A
forms (3 occurrences of lost and 1 of took). We’d then multiply this figure by 100, giving a total of 400.
We’d then divide this total by the number of tokens of regular verb-forms in the cells in 4A, 4B and 4C,
i.e. 15 (comprising 3 occurrences of lost, 1 of took, 7 of eat, 2 of take, 1 of drived and 1 of taked), giving
us an overall score of 27% suppliance of correct past tense forms for irregular verbs (i.e. 400 ÷ 15).
If these scores were representative of JC’s overall performance (and you obviously cannot assume that
they are, because they are computed on the basis of the partial table in 8, and your results need to be
computed in relation to a complete counterpart of the table in 8), they’d suggest that JC performs
substantially better on regular than irregular verbs. In chapter 3 of your project (where you analyse your
research results), you will eventually discuss whether your results are compatible with each of your
chosen theories of SLI.
190
We exclude verbs like CUT from this count, for the obvious reason that if a child produces the form cut in a past
tense context, you cannot be sure whether he is using a past tense form or a bare (infinitive) form. We also exclude
verbs like BURN, because alongside the irregular past tense form burnt they allow the alternative regular form
burned.
66
2.3 Testing the Perceptual Deficit model
Leonard’s Perceptual Deficit Hypothesis maintains that the problems which SLI children have with
certain grammatical morphemes are attributable to a perceptual deficit. The relevant morphemes, he
claims, cause perceptual difficulties because of their low perceptual salience; and he further maintains that
vowels are more perceptually salient (i.e. easier to perceive) than consonants. If this is so, we would
expect to find that SLI children perform better on past tense forms where the difference between the past
tense form and the bare/infinitive form is marked by a vowel (as in see/saw), than on those where it is
marked by a consonant (as in play/played). But the picture is clouded somewhat by the fact that some past
tense forms are marked by both a vowel and a consonant (e.g. as in wait/waited or sleep/slept) – and it is
not clear what Leonard’s theory would predict about how well SLI children fare on correctly marking past
tense on these forms (the vowel having a high perceptual salience, and the consonant having a low
perceptual salience, and the overall verb form having two markers of past tense rather than one). So, the
first thing we need to do in order to test Leonard’s Perceptual Deficit model is to draw up three lists of
verbs used in past tense contexts by JC: (i) those whose past tense form is marked by a consonant alone,
(ii) those whose past tense form is marked by a vowel alone, and (iii) other past tense forms (marked by
both a vowel and one or more consonants). For each verb form on each list, we need to say frequently he
supplied the correct form, and how frequently he supplied incorrect forms. I have drawn up three partial
tables below, to give you some idea of what your tables might look like. You need to exclude from your
tables any verb whose past tense form is identical to its bare/infinitive form (e.g. cut, for the obvious
reason that if a child says cut in a past tense context, you can’t tell whether he is using the past tense form
cut or the infinitive form cut).
(11) List of forms used by JC in contexts where adults use a consonant-marked past-tense form
Verb
A: Correct forms
B: Incorrect forms
ASK
asked = 1
HAVE
have = 3
SCARE
scared = 1
scare = 3
(12) List of forms used by JC in contexts where adults use a vowel-marked past-tense form
Verb
A: Correct forms
B: Incorrect forms
DRINK
drink = 1
SEE
saw = 1
see = 4
TAKE
took = 1
take = 2; taked =1
(13) List of forms used by JC where adults use a vowel- and-consonant-marked past-tense form
Verb
A: Correct forms
B: Incorrect forms
SAY
said = 5
say = 4
SLEEP
sleep = 1
WANNED
waned = 1
want = 1
Note that in classifying past-tense verb-forms into the three types listed above, you have to look at differences
in the spoken form of the past tense and the corresponding bare infinitive – not at differences in the written
language. It will help you if you do a phonetic transcription of each past tense and corresponding bare form.
Once you’ve drawn up complete versions of the partial lists in (11-13) above, you can then move on to
calculate JC’s mean percent correct score for each of these three groups of verbs, in terms of the following
calculation:
(14)
% suppliance of correct =
past-tense forms
no. of tokens of correct forms x 100
no. of tokens of correct forms + no. of tokens of incorrect forms
67
So, the percent suppliance of correct forms would be 25% for the consonant-marked forms in table (11) (i.e.
only 2 forms correctly marked for past tense out of a set of 8 forms occurring in past tense contexts), 20% for
the vowel-marked forms in table (12) (i.e. 2 correct forms in 10 attempts), and 50% for the consonant-andvowel-marked forms in table (13) (i.e. 6 correct forms in 12 attempts).
2.4 Testing the Rule Deficit model
Gopnik and Crago’s Rule Deficit Hypothesis claims that SLI children lack the ability to form regular rules,
and that they memorise all inflected verb forms (regular and irregular alike). If Gopnik and Crago are
right, we should expect to find a frequency effect with regular as well as irregular verbs (since both sets of
verbs are claimed by Gopnik and Crago to be acquired via the same mechanism of memorisation); but if
Gopnik and Crago are wrong and SLI children learn regular forms by rule and irregular forms by
memorisation, we should expect to find a frequency effect only with irregulars, not with regulars (because
all regular past tense forms will then be formed by application of the regular past tense d-rule).
In order to test for frequency effects, we need information on how frequently particular past tense forms
occurred in JC’s speech input. Unfortunately, we don’t have any information on this, so the best guess we
can make at input frequency is to use the figures from a frequency dictionary to estimate the relative
frequency of different past tense forms in his speech input. The only frequency dictionary currently
available for American English is Carroll (1971): it’s not ideal, because it’s based on written American
English (whereas the speech input to JC will largely have been a particular variety of spoken American
English) – but it’s the best we can do.
What you need to do is to compile two frequency tables like the partial tables shown in (12) and (13)
below – one for irregular verbs of type 4 (i.e. those which have a distinctive irregular past tense form and
which do not also have an alternative regular past tense form), and another for regular verbs of type 1.
Arrange the verbs in order of their adult frequency, with the most frequent verbs at the top of the list and
the least frequent at the bottom. List the correct (adult) past tense form of the verb in column 1, the raw
frequency per million words of the relevant form given in Carroll’s dictionary (i.e. the number of times it
occurred per million words in Carroll’s corpus) in column 2191, and JC’s suppliance of the correct form in
column 3 (where the figure 5/9 for said means that he used the verb SAY in past tense contexts on 9
occasions, and on 5 of these he supplied the correct past tense form said):
(15) Table of frequency of adult irregular past tense forms and JC’s suppliance of correct forms
adult past tense form adult frequency per million words
JC’s suppliance of correct form
had
20511
0/3
said
15309
5/9
did
7169
1/1
bit
600
1/1
ate
440
0/7
forgot
225
2/2
191
Carroll gives the following frequencies (per million words in adult English) for the past tense forms of irregular
verbs used in past tense contexts by JC: had = 20511; said = 15309; did = 7169; made = 7073; came = 4914; went =
4132; put = 3942; found = 3362; saw = 2900; got = 2626; took = 2490; cut = 1757; gave = 1534; brought = 1357;
lost = 820; caught = 793; fell = 790; bit = 600; hurt = 457; ate = 440; drove = 361; forgot = 225; slept = 200; woke =
133; drank = 106; burnt = 43; The corresponding frequency figures for regular verbs used by JC in past tense
contexts are as follows: used = 5607; looked = 3197; asked = 2924; wanted = 1637; tried = 1077; played = 810;
laughed = 639; showed = 478; jumped = 476; died = 459; dropped = 427; talked = 415; grabbed = 161; melted =
152; scared = 151; cooked = 147; knocked = 139; dumped = 38; pinched = 18; shovelled = 12.
68
(16) Table of frequency of adult regular past tense forms and JC’s suppliance of correct forms
adult past tense form adult frequency per million words
JC’s suppliance of correct form
looked
3197
0/1
asked
2924
1/1
wanned192
1637
1/2
jumped
476
1/1
died
459
0/1
dropped
427
0/1
Once you’ve compiled your own complete versions of the (partial) frequency tables in (15) and (16), you
then need to establish whether you have evidence for a frequency effect with irregular verbs on the one
hand, and with regulars on the other. You need to divide the verbs in each table into two groups – those
whose past tense form occurs relatively frequently in adult English (let’s call them frequent past-tense
forms), and those whose past tense forms occur relatively infrequently (let’s call them infrequent pasttense forms). It’s up to you to decide how you are going to classify verbs as frequent or infrequent – e.g.
maybe you will decide to take past-tense forms with an adult frequency of more than 2,000 (or 1,500? or
1,000? or 500? or 250?) to be frequent, and others to be infrequent. For the sake of argument, let’s
suppose that we take verbs with an adult frequency of more than 1,000 occurrences per million words to
be frequent: this would mean that the first three past-tense forms in each table above (viz. the irregulars
had/said/did and the regulars looked/asked/wanned) would be designated as frequent forms, and the last
three as infrequent forms.
What you do next is see whether JC performs better overall on frequent than infrequent irregulars, and
better overall on frequent than infrequent regulars. To do this for the irregulars in table (15), you add up
how often he produces the correct past tense form of frequent irregulars (= 6 times, comprising 5 correct
tokens of said and 1 correct token of did), multiply this by 100, and divide the resulting figure by the
number of times he uses frequent irregulars in past tense contexts (= 13, comprising 3 obligatory contexts
where adults would have used the past tense form had, 9 obligatory contexts for said and 1 for did). For
the tiny subset of data in table (15), this would give us a percent correct suppliance rate for frequent
irregulars of 46% (i.e. 6 x 100 ÷ 13). Doing a parallel calculation for the infrequent irregular pasts in table
(15) shows that JC produces 1 correct token of bit, 2 correct tokens of forgot and that there is 1 obligatory
context for the past tense form bit, 7 for ate and 2 for forgot: his percent suppliance rate for the correct
irregular pasts is thus 30% (3 x 100 ÷ 10).
To see whether JC performs better on the frequent regular pasts than on the infrequent ones, we do a
similar set of calculations in relation to table (16). He uses the 3 most frequent verbs 6 times in past tense
contexts, and produces only 2 correct past tense forms, so his percent suppliance of the correct past tense
form for frequent regular pasts is 33% (i.e. 2 x 100 ÷ 6). He uses the 3 least frequent verbs 3 times in past
tense contexts, and produces only 1 correct past tense form, so his percent suppliance of the correct past
tense form for frequent regular pasts is again 33% (i.e. 1 x 100 ÷ 3).
So, doing the relevant calculations on the verbs in the tables in (15) and (16) would lead us to the
conclusion that we have evidence of a frequency effect for irregular pasts (in that he performs
substantially better on frequent than infrequent irregular pasts), but not for regular pasts (where he
performs at the same level on both frequent and infrequent regular pasts). But bear in mind two things.
Firstly, these conclusions are based on partial tables which contain only a small number of the verb forms
used by JC in past tense contexts: your own conclusions must be based on complete tables which list all
the verb-forms used by JC in past tense contexts. Secondly, even complete tables may yield insufficient
data for you to draw any firm conclusions about whether there is (or isn’t) a frequency effect – especially
where any effect you find is marginal (e.g. a difference of only a handful of percentage points).
192
Wanned is the past/perfect/passive form of WANT in colloquial American English, the final t of the stem being
dropped in rapid speech in forms like wanna (= want to) and wanned (= wanted).
69
If you want to be particularly thorough, you could also try testing for a second kind of frequency effect
(mentioned in Gopnik and Crago 1991). You will note that JC performs particularly badly on some verbs:
e.g. he uses the bare form eat on 7 occasions in contexts where adults require the past form ate, and not
once does he produce the correct past tense form ate. A question which you might ask is whether this kind
of performance is associated with verbs whose bare form (e.g. eat) is more frequent than its past form (e.g.
ate) in adult English (using the raw frequency figures in Carroll’s dictionary193). To do this, you will need
to draw up a (complete version of) the (partial) table below for irregular verbs used in past-tense contexts
by JC):
(17) Table of frequency of adult irregular bare/past forms, and JC’s suppliance of past
Raw frequency of past form Raw frequency of bare form
JC’s suppliance of past
said = 15309
say = 3916
5/9
brought = 1357
bring = 1016
0/1
ate = 440
eat = 1616
0/7
saw = 2900
see = 8518
1/5
You can then see whether JC performs better at marking past tense on verbs on irregular verbs whose past
tense form is more frequent than its bare form. For example, if we take the first two verbs in the table in
(17), we see that both of them have past tense forms which are more frequent than their bare forms, and
that JC’s mean suppliance rate for the past tense forms of these two verbs is 5/10 (50%). By contrast, if we
take the last two verbs on the list, we see that their past forms are less frequent than their bare forms, and
that JC’s mean suppliance rate for these two items is only 1/12 (8%). If the results we obtain from the
partial table in (17) were replicated for the corresponding full table which you need to draw up for
yourself (if you decide to test for this kind of frequency effect), this would mean that we do find the kind
of frequency effect which might be expected if irregulars are learned by memorisation. Now, if Gopnik
and Crago are right in claiming that SLI children have a genetic rule deficit and so have to memorise
regulars as well, we might expect to find a parallel frequency effect with regulars – and to see whether this
is the case, you can draw up another table like that in (17) for regular bare/past forms. If we found such a
frequency effect with irregular pasts but not regular pasts, this would lead us to the conclusion that Gopnik
and Crago are wrong, and that irregulars pasts are memorised foms but regulars are rule-derived forms.
2.5 Testing the Agreement Deficit model
If one of the models of SLI which you have chosen to work on is the Agreement Deficit model, you will
need to compare how well JC peforms on past tense verbs (these not being overtly inflected for
agreement) with how well he performs on third person singular present tense forms (these being overtly
inflected for agreement). Hence, you will also need to calculate JC’s suppliance rate for regular and
irregular third person singular present tense forms. You can find detailed guidance on how to do this in
§2.4 and §2.5 of the next chapter.
2.6 Testing the ATOM model
The ATOM model claims that SLI children sporadically mark tense and agreement in obligatory contexts,
and that nominative subjects are found when agreement is marked, and accusative subjects when it is not
marked. Hence, if ATOM is one of the models you have chosen to work on, one of thin things you will
need to calculate is how frequently JC marks tense in past tense contexts. In order to do this, you can
193
Carroll gives the following figures for adult frequency (per million words) of bare forms of relevant irregular
verbs: have = 22337; do = 12695; make = 8333; come = 4676; go = 5388; find = 6916; see = 8518; get = 5700; take
= 4089; put = 3942; give = 3366; say = 3916; eat = 1616; cut = 1757; bring = 1016; fall = 824; sleep = 717; catch =
679; drive = 543; hurt = 457; drink = 347; forget = 314; lose = 268; bite = 172; wake = 113. Carroll also gives the
following adult frequency figures (per million words) for the bare forms of relevant regular verbs: use = 7009; look =
4933; want = 2655; show = 2734; play = 2113; try = 1958; talk = 1133; ask = 900; drop = 433; die = 360; jump =
356; laugh = 287; cook = 265; melt = 100; knock = 85; grab = 73; scare = 55; shovel = 60; burn = 43; dump = 33;
pinch = 30.
70
make use of the raw frequency figures in (your complete version of) the table in (8) above, repeated in
(18) below:
(15) Raw frequency of verb-forms used by JC in past tense contexts (repeated from (9) above)
Type
Verb
A: correct form
B: bare form
C: other form
1. regular
DUMP
dump = 3
1. regular
SCARE
scared = 1
scare = 3
1. regular
SHOVEL
shovelled = 4
2. novel
JUMPOVER
jumpovered = 1
3. mixed
BURN
burn = 1
4. Irregular
DRIVE
drived = 1
4. Irregular
EAT
eat = 7
4. Irregular
LOSE
lost = 3
4. Irregular
TAKE
took = 1
take = 2
taked = 1
5. Invariable irregular CUT
cut = 2
Which of the verb forms in this table are (and are not) marked for past tense? Well, the correct adult past
tense forms which appear in column A are clearly marked for past tense – though we have to exclude
invariable irregulars like CUT, since if a child uses the verb form cut in a past tense context, we have no
way of knowing whether he is using a past tense form or a bare infinitive form (since these two forms are
identical for verbs like CUT): hence, our calculations must be based on verbs with a distinctive past tense
form (i.e. one distinct from the corresponding bare/infinitive form). This means we include only verbs of
types 1-4, and exclude verbs of type 5. A second class of verb forms which are clearly marked for past
tense are over-regularised forms like drived and taked which appear in column C. By contrast, the
bare/infinitive forms in column B appear to be tenseless (i.e. not inflected for tense). Accordingly, we can
calculate how frequently JC marks tense in past tense contexts by doing the following simple calculation:
(19)
% marking of past tense on verbs =
with a distinctive past form
no. of tokens of verb-forms marked for past x 100
no. of obligatory contexts for past tense forms
Since verbs of types 1-4 have a distinctive past form (but not verbs of type 5), and since the forms in
columns A and C are marked for past tense, we can calculate the number of tokens of verb-forms
distinctively marked for past tense by adding up the number of tokens in columns 1A, 2A, 3A, 4A, 1C,
2C, 3C, and 4C: for the table in (15), the total comes to 12 (made up of 1 token of scared, 4 of shoveled, 1
of jumpovered, 3 of lost, 1 of took, 1 of drived, 1 of taked). Then we multiply this figure of 12 by 100,
giving us 1,200. We then need to calculate the number of obligatory contexts for past tense forms of verbs
with a distinctive past form, and we can do this by adding up the number of tokens in columns 1A, 2A,
3A, 4A, 1B, 2B, 3B, 4B, 1C, 2C, 3C and 4C together: for the table in (15), the relevant total comes to 28
(made up of 1 token of scared, 4 of shoveled, 1 of jumpovered, 3 of lost, 1 of took, 1 of drived, 1 of taked,
3 of dump, 3 of scare, 1 of burn, 7 of eat, and 2 of take). We then divide our top-line figure of 1,200 by
our bottom-line figure of 28, and arrive at a score of 43% marking of past tense (on verbs with a
distinctive past tense form) in obligatory contexts. Of course, I’ve computed this figure from the partial
table in (8/18): you need to compute your figure from a complete counterpart of the relevant table.
A second key claim made by ATOM is that SLI children have problems in marking on agreement in past
tense verbs, and that failure to mark agreement leads to the use of default accusative subjects. Because
past tense verbs don’t overtly inflect for agreement in English (with the exception of the irregular
auxiliary/copula forms was/were), you can’t tell directly whether a past tense form is marked for
agreement – e.g. by looking at whether it carries a third person singular -s inflection. But you can look at
the case assigned to its subject. On the ATOM assumption that a nominative subject occurs with an
agreeing (auxiliary or non-auxiliary) verb and a default accusative subjects with an agreementless verb,
we would expect to find that (in past tense contexts) children alternate between nominative and accusative
subjects (irrespective of whether they use past tense or bare forms – see the discussion in §1.6. Hence, we
71
need to look at the frequency with which JC uses nominative and accusative subjects in past tense
contexts. Of course, we can only do this for verbs which have subjects which are unambiguously marked
as either nominative or accusative. Hence, we have to exclude verbs with subjects like Daddy, you or it
because they have a common nominative-accusative form, and so there is no way of telling whether they
are nominative or accusative. In other words, we have to look at sentences with I/me, we/us, he/him,
she/her, or they/them subjects. The first step, then, is to go back to the list of sentences in (8) above, and
use it to create a sub-list like that below which shows the sentences produced by JC in past contexts with
unambiguously nominative subjects like I/we/he and those with unambiguously accusative subjects like
me/her194/them:
(20) List of sentences with nominative or accusative subjects used by JC in past-tense contexts
A: Nominative subject
B: Accusative subject
198. Long time ago I have a big eye
103. Me just jumpovered it
236. I never saw one of these stove
327. Then me said ‘Oh!’
321. Long time ago, I go camp and hiking…
328. Then me take it xxx
229. We see clown at...umm...
283. And we scare her
268. Her make it for Ms Peggy a long time ago
277. Her give me dad a lobster
314. Then her got hurt
126. He shoveled him truck
127 And then he dump it
132. And then he drived away
300. He jumped out
301. He bit me
316. He try to eat me up
379. He cut me
231. Them have a party
366. But them cut my hair real tiny
Note that the list I have drawn up in (20) is only a partial list (for illustrative purposes) based on a
randomly chosen subset of the relevant example sentences from the list in (7). Your own list has to contain
all the sentences produced by JC in past tense contexts which have unambiguously nominative or
accusative subjects195. Include your own complete version of the list in (20) as an Appendix to your
project – don’t include it in the main text (because it interrupts the flow of the text).
Once you’ve drawn up your own counterpart of the list in (20), you can use it to calculate the percentage
of nominative subjects used by JC in past tense contexts, in the manner shown below:
(21)
% of nominative subjects in
past tense contexts
=
no. of tokens of nominative subjects x 100
no. of tokens of nominative + accusative subjects
If we did this calculation for the (partial) list in (20), we first need to count the number of tokens of
nominative subjects in column A: this is 12 (made up of 3 tokens of I, 2 of we and 7 of he). We then
multiply this total of 12 by 100, giving 1,200. We then add up the number of tokens of nominative and
accusative subjects in columns A and B: this is 20 (made up of 3 tokens of I, 2 of we, 7 of he, 3 of me, 3 of
her and 2 of them). So, we divide our top-line figure of 1,200 by our bottom-line figure of 20, and arrive at
the conclusion that 60% of the subjects used by JC in past tense contexts are nominative. If the use of a
nominative subject is an indication of agreement being marked on the associated verb (as ATOM claims),
then this means that JC marks agreement on 60% of verbs used in past tense contexts.
Her could in principle be genitive, but given that JC doesn’t otherwise use genitive subjects (but does use a lot of
accusative subjects), it is far more likely to be accusative.
195
Only look at cases where the subject is a pronoun on its own, not at cases where the subject comprises a pronoun
modifying a noun – hence not at me mommy in utterance 337 Me mommy show me to make Easter eggs. This is
because the pronoun me here is not the subject of the sentence: rather, mommy is the subject, and me is a possessive
modifier of mommy.
194
72
A prediction of the ATOM model which you can test in relation to the list of sentences in your complete
version of table (20) is that SLI children will alternate between producing nominative and accusative
subjects both with bare verbs used in past tense contexts and with verbs overtly marked for past tense.
To give you an idea of what to do, in the table below, I have listed the raw frequency with which JC uses
unambiguously nominative and accusative subjects in past tense contexts in utterances 5-132:
(22) Frequency of JC’s unambiguously case-marked pronoun subjects in past tense contexts
A: nominative+past B: nominative+infinitive C: accusative+past
D: accusative+infinitive
he = 4
he = 7
me = 5
me = 4
2.6 Summary of research results
You should end your chapter with a brief section recapitulating your main research results.
3. Analysis of research results (around 35% of your overall study)
3.1 Introduction
Begin by saying that what you are aiming to do in this section/chapter is evaluate the extent to which the
research findings which you obtained in the previous chapter are consistent (or inconsistent) with the
research hypotheses which you outlined in chapter 1, and (more generally) with the particular models of
SLI which you are aiming to test. Since you are ultimately concerned with evaluating a range of theories
of SLI, perhaps the best way of organising the material in this chapter into sections is by theory – e.g.
discussing the extent to which your research findings are compatible with the Perceptual Deficit model in
section 3.2, with the Rule Deficit model in section 3.3, with the Agreement Deficit model in section 3.4
and with the ATOM model in section 3.5.
3.2 Evaluation of the Perceptual Deficit model
Begin by reminding the reader briefly (in one or two sentences) of the general assumptions underlying the
model, and the key prediction it makes about past tense – viz. that SLI children should perform better on
past tense forms marked by the use of a vowel than on those marked by use of a consonant. Discuss the
findings you yourself arrived at for the three types of past tense verb-form in (your complete version of)
the lists in (11), (12) and (13), and whether these are compatible with Leonard’s claim. If you found
evidence of some perceptual effect, was the effect strong (e.g. JC performed twice as well on some forms
than others), or weak (e.g. JC performed marginally better on some forms than other)196. If weak, could
they be attributable to simply not having enough data to make reliable calculations? And if you find no
evidence of perceptual effects of any kind, could this be because such effects do not exist (and the
Perceptual Deficit model is thereby falsified), or because you simply don’t have enough data to test
properly for such effects?
3.3 Evaluation of the Rule Deficit model
In section 3.4, you might evaluate Gopnik and Crago’s Rule Deficit model. Begin by providing a brief
outline of the general assumptions made in the model, and the specific predictions it makes about past
tense marking. The main research findings which are going to be relevant here are whether JC uses
overregularised forms (though pay heed to the footnote in this chapter about such forms), and whether you
have found the expected frequency effect for regular as well as irregular pasts. If you don’t find frequency
effects, do you conclude (i) that no such effects exist, or (ii) that they may well exist but you probably
don’t have enough data to test for them (or indeed that a frequency dictionary based on written English is
a poor reflection of the speech input a child receives)? What if we modified the Rule Deficit model so as
to claim not that SLI children fail to acquire any regular morphological rules, but rather that (at a given
196
Technically, what we should be looking for here is whether any effects you find are statistically significant or not:
but since this course is not designed to provide you with training in statistics, you need not be concerned with this.
73
stage of development) they may have acquired some but not all regular rules, and that even the rules they
have acquired may be applied sporadically (with better established rules being applied more frequently
than less well established rules)?
3.5 Evaluation of the Agreement Deficit model
Begin by providing a brief outline of its key prediction, namely that SLI children will perform well on past
tense forms because they only inflect for an interpretable tense feature, but badly on third person singular
present tense forms (because these also inflect for uninterpretable agreement features). Look first of all at
whether (as predicted) JC does well on regular past tense {d}, but badly on regular third-person-singular
present-tense {s}. Then look at whether he does well on irregular past tense forms, and badly on irregular
present-tense forms. If your findings don’t quite match the predictions of the model, consider why this
might be: e.g. maybe tense is a cognitively complex concept (Reichenbach arguing that tense involves a
complex relation between three points in time, namely speech time, event time and reference time); maybe
uninterpretable features slow down (rather than prevent) acquisition of particular morphemes, so that
high-frequency items will be acquired even if encoding uninterpretable features; maybe when children use
a bare irregular form like give in the past tense, they wrongly assume that it is the past tense form of GIVE,
and have learned that GIVE is irregular in respect of not taking the regular past tense d-suffix (unlike TELL
whose past tense is told), but not that the vowel in the stem changes from |I| to |eI| in the past tense.
3.6 Evaluation of the ATOM model
Begin by providing a brief outline of its core assumptions and what it predicts will happen in past tense
contexts – namely that we will find four types of structure: nominative+past forms like He played with me
yesterday; accusative+past forms like Him played with me yesterday; nominative+infinitive forms like He
play with me yesterday, and accusative+infinitive forms like Him play with me yesterday. The obvious
question for you to answer in relation to the data in your version of the table in (20) is whether you find
evidence that JC produces all four types of structure or not. You might want to consider possible problems
with ATOM – e.g. its abstractness and circularity, its failure to make precise predictions about how often
tense and agreement are marked, its failure to account for why children might be better at marking
nominative case on some pronouns than others (e.g. you might want to compare how well JC fares on e.g.
he subjects compared with I subjects).
Chapter 4. Summary and Conclusions (about 15% of your overall study)
4.1 Summary of research hypotheses
Begin this chapter with a section in which you provide a brief summary of the models of SLI and research
hypotheses which you outlined in chapter 1.
4.2 Summary of research findings
In this section, briefly summarise the main research findings you obtained in chapter 2.
4.3 Summary of evaluations
In this section, provide a brief summary of your evaluation (in chapter 3) of the relative strengths and
weaknesses of each of the theories of SLI you are testing, viewed from the perspective of how well they
account for your own research findings.
4.4 Overall conclusions
In this (the most important) section of your chapter, you need to give your considered final verdict on
which of the models you have looked at best accounts for your own research findings. If a particular
model doesn’t account for (some of) your findings, can it be modified in some way which would make it
compatible with your findings? If not, and if there is no one model of SLI which accounts for all your
research findings, does this take you towards a composite model which sees SLI as involving multiple
impairments, rather than a single type of impairment? (See my suggestions for what such a composite
model might look like in the section on chapter 4 in the next chapter.) To what extent would such a
74
composite model be able to account for your research findings about JC’s past tense marking? Be
particularly circumspect with your conclusions: these are usually the weakest point in student work, and a
weak conclusion will devalue your overall project. Avoid excessively general, unsubstantiated claims.
5. Appendices
After your four main chapters, include as an Appendix any lists of sentences which provide the raw data
from which you computed the figures for various tables included in the main body of your project, and
any tables from which you have computed suppliance rates (and any other figures you use).
6. List of references
After any Appendices, include a list of works (arranged alphabetically by author and date) which you have
cited in your project, using the standard author-date system described in the departmental handbook.
75
9: Empirical study II: Acquisition of English s-forms
In this chapter, we look at how to use the JC corpus to undertake an empirical study of The acquisition of
English s-forms by a 4-year-old child with SLI. Your study might be organised into sections and
subsections as follows.
1. Background (about 25% of your overall study)
1.1 Introduction
Give brief details (in a paragraph) on the kind of developmental disorder which SLI involves (summarised
on p.4 of this book: see Bishop (1997), Leonard (1998), and the collection of papers in Bishop and
Leonard (2000) and Levy and Schaeffer (2003) for more detailed discussion of diagnostic criteria for
SLI), pointing out that various aspects of grammar are reported to be impaired in SLI children, including
the ability to produce inflected forms. Then go on to say that there are a wide range of different models of
SLI which seek to account for the nature of the grammatical impairment in SLI children, and for the
purposes of the present study you are going to examine a subset of these (e.g. the Perceptual Deficit, Rule
Deficit, Agreement Deficit and ATOM models). Your aim is to do a small-scale empirical study designed
to evaluate each of the models, by looking at the production of three types of s-form in obligatory contexts
by a four-year-old American boy with SLI, known as JC. Provide brief details of JC’s background (which
you can find at the beginning of chapter 13). The three types of s-form you are going to look at are (i)
noun plural -s, (ii) genitive ’s, and (iii) third person singular present-tense -s. You’re going to begin by
outlining the morphophonology of s-forms in English in the next section
1.2 Outline of the morphophonology of s-forms in English
In this section, you need to provide a brief account of the morphophonology of each of the three s-forms
you are studying. The dual mechanism model of the acquisition of morphology (outlined e.g. in Pinker
and Prince 1988 and Pinker 1991) maintains that irregular forms are memorised forms which are
stored/listed in the mental lexicon, whereas regular forms are computed or derived via application of
morphological rules like those sketched informally in (1-3) below (with morphemes enclosed in curly
brackets, by convention):
(1)
A regular verb carries the suffix
{s} if third person singular present
{d} if past, perfect or passive
{ing} if progressive or gerund
{ø} otherwise
(2)
A regular noun carries the suffix:
{s} if plural
{ø} otherwise
(3)
A noun expression197 carries the suffix
{s} if genitive
{ø} otherwise
The precise phonological form which a morpheme is given in a particular structure is determined by
(morphophonological) spellout rules like (4) below:
The informal term noun expression is used to denote an expression which comprises or contains a noun – e.g. an
expression such as John or the girl next door. Note that genitive ’s is attached to the end of the last word in the
relevant noun expression, as in the girl next door’s.
197
76
(4)
{s} is spelled out as:
|Iz| when attached to a form ending in a sibilant consonant (e.g. passes, buzzes, catches, cages,
pushes, camouflages)
|z| when attached to a form ending in another voiced segment198 (e.g. calls, robs, bags, grows,
pays)
|s| when attached to a form ending in another voiceless segment199 (e.g. cuts, locks, breaks,
rocks)
|ø| when attached to a form already ending in {s} (e.g. the boys’ school, where genitive {s}
has a null spellout by virtue of being attached to a noun boys carrying plural {s})
So, for example, the lexical entry (i.e. entry in the dictionary) for the regular noun CAT will contain simply
the stem form cat: its plural form will be generated by adding the suffix {s} to the stem form in
accordance with the first line of rule (2), so deriving cats; this {s} morpheme will be spelled out as |s| in
accordance with the third line of rule (4), so that the word is pronounced |kats|. The singular form of CAT
will be generated by adding the null suffix {ø} in accordance with the second line of rule (2), generating
cat-ø. By contrast, the lexical entry for the irregular noun MOUSE will contain not only the singular form
mouse but also the irregular plural form mice: the presence of the irregular plural form mice in the lexicon
will then block application of the regular plural formation rule in (2), so accounting for why we don’t find
*mouses as the plural of mouse. Similarly, the lexical entry for a regular verb like PLAY will specify
simply that it has the stem form play, and the morphological rules in (1) will determine that its third
person singular present tense form is derived by adding {s} to the stem form, its past tense and
perfect/passive participle forms by adding {d}, its progressive participle and gerund forms by adding
{ing}, and all other forms by adding a null affix {ø}. By contrast, the lexical entry for an irregular verb
like GO will specify that it has the stem form go, the irregular perfect participle form gone and the
irregular past tense form went; it will follow from this that its (regular) third person singular present tense
form is derived by adding {s} to the stem form (giving goes) in accordance with the rule (1), its (regular)
progressive participle form is derived by adding {ing} to the stem form (giving going), and its other forms
are derived by adding the null affix -ø to the stem form (giving go+ø) in accordance with (1v).
Given that the dual mechanism account maintains that regular forms are rule-generated and irregulars are
stored, it is important to be able to differentiate between regular and irregular forms. This is not an issue
which arises with genitive ’s, since there are no irregular genitive s-forms for noun expressions in English
(though of course there are irregular genitive pronouns like my/our/your/her/their etc.). In the case of
plural nouns, the difference between regular plural forms like cats and irregulars like mice is relatively
clearcut. But the difference between regular and irregular forms is less clearcut in the case of third person
singular present-tense (3SgPres) s-forms. While the 3SgPres forms of the vast majority of verbs is entirely
regular in English (as with e.g. helps, works, sleeps, cries etc.), there are a handful of 3SgPres forms
which are potentially irregular – such as says, has, does and is. The irregularity of the verb SAY lies in the
fact that although its stem form is |seI| in forms like say/saying, it has the irregular stem-form |se| in the
inflected forms says and said. The forms has and does can be used either as main verbs (e.g. in He has a
flat in town or He does a lot of work for charity) or as auxiliaries (e.g. in Has he finished the assigment? or
Does he take sugar?), whereas the form is always functions as an auxiliary (and hence has the contracted
negative form isn’t and can undergo inversion in questions like Is it raining?). What’s irregular about the
verb HAVE is that it has the stem form hav- in forms like have/having, but the contracted stem form ha- in
the forms has and had. When used as an auxiliary, HAVE is even more irregular, in that its vowel can be
reduced to schwa (so that in Has he left? the auxiliary has can be pronounced |əz|, for example), and it can
even lose its vowel entirely (e.g. in He’s left). The irregularity of the 3SgPres form does lies in the fact
that it is derived by adding {s} not to the regular stem do- |du:|, but rather to the irregular stem doe- |d|,
yielding does |dz|. In its use as an auxiliary, the vowel can be reduced to schwa, so that in Does Jim like
her? the auxiliary does can be pronounced |dəz|. Indeed, does can be contracted down to ’s in certain
Since all vowels are voiced, this means ‘when attached to a stem ending in a vowel, or a voiced non-sibilant
consonant’.
199
In other words ‘when attached to a voiceless consonant which is not a sibilant’.
198
77
contexts (as in What’s he like for breakfast? = What does he like for breakfast?). An additional
complication to note is that in many varieties of (esp. American) English, it has the irregular 3SgPres
negative form don’t (as in the typical pop-song lament She don’t love me no more). As for the 3SgPres
form is, this is a suppletive form which seems to be unrelated to other forms of BE, and hence completely
irregular. It too has a contracted form ’s (e.g. in He’s lazy). Given the irregularity of says/has/does/is, it
seems likely that these are stored/memorised forms rather than rule-derived forms.
Empirical research is about testing research hypotheses, so in sections 1.3-1.6, you should provide a short
summary of the overall claims made by each of the models of SLI which you have decided to evaluate,
and then go on to say what specific prediction each makes about how well SLI children would be expected
to perform on each of the four types of s-form which you are looking at.
1.3 Outline of the Perceptual Deficit model of SLI
In this section, you should provide a brief outline of the Perceptual Deficit model and note that its key
claim is that consonantal inflections pose more perceptual problems for SLI children than vocalic
inflections, and so (on the face of it) appears to predict that SLI children will perform badly on all s-forms
(because they are consonantal). But things are not quite as straightforward as that, for two reasons. Firstly,
all three s-forms that you are concerned with here have the three overt allomorphs |s|, |z| and |Iz| as seen in
section 1.2, and we might therefore expect SLI children to perform better on the last of these three because
it contains a vowel: in addition, genitive ’s has an additional null allomorph (used e.g. with nouns which
end in regular plural -s, as in boys’ toys).
1.4 Outline of the Rule Deficit model of SLI
In this section, you should outline the Rule Deficit model, and note that it claims that SLI children have a
genetic rule-deficit which makes it impossible for them to acquire any regular morphological rules. They
are therefore predicted to be unable to acquire the noun plural s-rule which specifies that a regular noun
like cat is pluralized by adding the suffix {s} to it, so forming cats: likewise, they are predicted to be
unable to acquire the genitive s-rule which specifies that a noun expression like Daddy or the girl next
door which is marked as carrying genitive case has the affix {s} added to the last word of the expression,
so forming Daddy’s or the girl next door’s; similarly, they are predicted to be unable to acquire the present
tense s-rule which specifies that a regular verb has the suffix {s} attached to it if it is a third person
singular present tense form. It might therefore seem at first sight that the prediction made by the Rule
Deficit model is that SLI children will always use bare forms in contexts where adults use s-forms (and
hence will say e.g. two car instead of two cars, Daddy car instead of Daddy’s car, and Daddy play with
me instead of Daddy plays with me). But things are not quite as simple as that, because Gopnik and Crago
maintain that although SLI children acquire regular morphological rules, they can memorise inflected
forms (both regular and irregular alike), though cannot produce novel or over-regularised s-forms. Since
memorisation of a given form requires extensive exposure to the relevant form, we therefore expect to find
a frequency effect (viz. that the more often a child hears a given s-form, the more often the child is likely
to produce it correctly). This leads to the following predictions about how well SLI children fare on
s-forms. In the case of plural s-forms like eyes and 3Sg present-tense verbs forms (both regular forms like
plays and irregulars like is/has/does), the prediction is that the child should perform well on highfrequency forms (whether regular or irregular), and poorly on low-frequency forms (whether regular or
irregular) – and not produce novel or over-regularised forms. For genitive s-forms, the prediction is rather
more complex. If we assume that a child can memorise inflected words but not inflected phrases (because
there are an infinite set of possible phrases – cf. the man’s, the tall man’s, the tall handsome man’s, the
talk dark handsome man’s etc. – and it is in principle impossible to memorise an infinite set), we might
expect to find that SLI children can memorise genitive forms of high frequency individual words like
Daddy’s but not genitive forms of phrases like the man’s: if so, we might expect an SLI child to say
Daddy’s car but the man car. A child who has acquired the genitive s-rule might also be expected to
produce over-regularised genitive pronouns like e.g. I’s or me’s in place of my/mine.
78
1.5 Outline of the Agreement Deficit model of SLI
In this section, you should outline the Agreement Deficit model, and note that it claims that SLI children
have particular problems with acquiring uninterpretable agreement features, but not with other features
(e.g. not with interpretable tense features, or uninterpretable case features). They should therefore be good
at producing those s-forms which encode interpretable features: if we assume that plural nouns like dogs
carry an interpretable number feature (marking the presence of more than one dog), we should expect that
SLI children will perform well on plural s-forms of nouns. By contrast, we should expect SLI children to
be bad at third person singular present tense s-forms of verbs, since these carry uninterpretable agreement
features. Given that Clahsen, Bartke and Göller (1997) claim that it is specifically uninterpretable
agreement features which cause problems for SLI children, and not e.g. uninterpretable case features200,
we should expect SLI children to perform well on genitive ’s.
1.6 Outline of the Agreement and Tense Omission (ATOM) model of SLI
In this section, outline the ATOM model, and note that it claims that in finite contexts (i.e. in contexts
where adults use an auxiliary or main verb marked for tense and agreement), SLI children who are at the
EOI/Extended Optional Infinitives stage (generally, children under 7-8 years of age) sometimes fail to
mark tense and/or agreement, and that this affects the case assigned to the subject (in that a subject of an
agreeing verb is nominative, whereas the subject of an agreementless verb is assigned default accusative
case). The predictions made by ATOM are clearcut in relation to contexts for s-inflected verb-forms. In a
context where an adult would use a present-tense main verb like plays (e.g. in reply to a question like
What does Daddy do when he comes home?), we’d expect an SLI child to produce four different kinds of
structure. One would be a structure marking both tense and agreement: if the child has acquired the set of
verb inflections in (1), we’d then expect such a child to use the s-form plays, and to assigns its subject
nominative case (because agreement is marked), so resulting in He plays with me. A second would be a
structure in which agreement but not tense is marked; since (1) specifies that a null affix -ø is used by
default when either tense or agreement (or both) are not marked, and since ATOM assumes that
nominative subjects are used when agreement is marked, this will result in structures like He play with me.
A third is a structure in which tense but not agreement is marked; since (1) tells us that a default null affix
-ø is used when either tense or agreement (or both) are not marked, and since ATOM maintains that
accusative subjects are used when agreement is not marked, this will result in a structure like Him play
with me. The fourth type of structure we expect to find is one in which neither tense nor agreement is
marked; since (1) says that a default null affix -ø is used when either tense or agreement (or both) are not
marked, and since ATOM claims that default accusative subjects are used when agreement is not marked,
this will again result in a structure like Him play with me201.
ATOM also makes predictions about what SLI children will do in contexts where adults would use an
irregular (non-modal) auxiliary like is/has/does (or their contracted variants). For example, consider what
an SLI child would be expected to reply to question like What’s Daddy doing?. If both tense and
agreement are marked, the child will reply He’s snoring. If agreement but not tense is marked, the child
will say He snoring (because ATOM claims that BE/HAVE/DO when used as auxiliaries are null whenever
underspecified for tense and/or agreement, but the presence of agreement will trigger nominative case
assignment). If tense is marked without agreement (or indeed if neither tense nor agreement is marked),
the child will say Him snoring.
The predictions made by ATOM in the case of plural nouns would seem to be that no problems are
expected. The implication of the ATOM claim that SLI children have problems in marking agreement
features is presumably that SLI children will not have problems in marking number on nouns, but will
200
Recall that CBG reported the English SLI children in their study having 100% correct case-marking.
The picture would become more complicated for an SLI child who acquired the tense feature associated with the
affix -s but not the agreement features, and so treated -s as marking present tense only. Such a child would (under
ATOM assumptions) be predicted to use present-tense s-forms with subjects of any person or number, and with both
nominative and accusative subjects – for reasons which should be clear.
201
79
have problems in marking number on verbs which agree with those nouns202. So, if we suppose that the
number feature carried by a plural noun like dogs is not an agreement feature203, then since ATOM
specifies that only tense and agreement features pose problems for SLI children, the implication is
presumably that SLI children like JC should fare well on plural nouns (both regular and irregular).
The predictions made by ATOM in relation to genitive s-forms are difficult to work out, because Wexler,
Schütze and Rice (1998) do not talk about genitive case-marking in any detail. However, if (as they claim:
1998, p.323) ‘Case is a reflection of structural grammatical relations’ and if agreement is the core
structural relation in syntax, then it would seem reasonable to suppose that genitive case is assigned to a
nominal via some form of agreement. Since we find a possessor agreeing with a possessum (= possessed
entity) in person and number in languages like French (e.g. in structures like ma maison ‘my house’ where
the possessor ma ‘my’ is feminine singular in form because it agrees with the feminine singular possessum
maison ‘house’), we could assume that a possessor is assigned genitive case when it agrees (invisibly, in
languages like English) with the possessum204. What ATOM would then predict is that if possessor
agreement is marked, the possessor will be assigned genitive case (so giving rise to genitive-possesor
structures such as Daddy’s car and my car etc.); but if possessor agreement is not marked, the possessor
will instead be spelled out in the default (accusative) form, so giving rise to accusative-possessor forms
such as Daddy car and me car. Of course, since possessor-agreement is not overtly visible in English, any
such analysis is of necessity abstract in character.
1.7 Summary of research hypotheses
In section 1.7, you should summarise the predictions made by each of your chosen accounts of SLI, and
outline the research hypotheses that you are setting out to test. An important observation to make is that
the predictions made by individual models can be skewed by lexical gaps. For example, if a child has not
acquired a particular affix or a particular inflected form, then it is clear that the child will not use the
relevant affix/form – even if particular theories of SLI predict that it would be used by children who have
acquired it in certain contexts.
2: Research results (about 25% of your overall study)
2.1 Introduction
Begin this section with an introduction, saying that you are setting out to test the research hypotheses that
you outlined in section 1.7 by looking at the types of structure produced by JC in contexts where adults
would use s-forms. You are going to look at plural s-forms in section 2.2, genitive s-forms in section 2.3,
and present-tense s-forms of main verbs in section 2.4, and present-tense s-forms of auxiliaries in section
2.5 (auxiliaries and main verbs being dealt with in separate sections because of the differences ATOM
predicts about in how SLI children treat s-inflected auxilaries and s-inflected main verbs).
2.2 Research results for plural s-forms of nouns
Begin this section by saying that you drew up a list of sentences from the JC corpus which represent
obligatory contexts for plural s-forms – and include this list in the Appendix to your project. In order to
help you with this task, below I have drawn up a list of sentences produced by JC containing (italicised)
nouns used in (what seem to me to be) plural contexts (i.e. contexts where adults would have used a plural
s-form). You will see that I have also included the plural N-pronoun (i.e. pronominal noun) ones, because
202
In more technical terms, this would amount to claiming that SLI children have problems in marking number when
it is an uninterpretable feature on verbs, but not when it is an interpretable feature on nouns.
203
This is far from obvious in relation to expressions like two dogs, where it might be argued that the noun dogs
acquires its plural number feature via a form of agreement (traditionally called Concord) with the plural numeral
quantifier two.
204
For a variety of technical reasons which it would not be appropriate to go into here, a more principled analysis
would be to suppose that genitive case is assigned to a possessor via agreement with an abstract functional head
(perhaps with a null determiner of some kind, as suggested in Abney 1987).
80
it forms its plural by use of the same s-affix as regular plural nouns like dogs. Where the structure of the
relevant utterance is unclear, I’ve added the presumed adult counterpart in parentheses.
(5) List of (italicized) s-forms of nouns used by JC in potential plural contexts
11 Why him don’t have eyes? 44 When him crack tiny pieces up, and then put xxx. 92 I make some peppers
109 And put buns on it, and ketchup. 156 He have he hats on. 165 (Th)ey are straps. 183 He can fly cause
he got these wings. 208 An owl did this with he eyes. 227 I wanna play some games. 236 I never saw one of
these stove. 237 Me making hot dogs. 250 I have tea at boy scout (= ‘at the boy scouts’). 273 And then, you
could get lots fish in there. 277 Her give me dad a lobster, a two lobster (?= ‘a pair of lobsters’). 279 Forgot
to take them eyes out. 288 No, you can’t eat eyes 297 Two lobsters. 299 Lobsters don’t go there. 307
Lobsters live in an ocean. 333 We making books. 336 Easter eggs. 337 Me mommy show me to make
Easter eggs. 346 My dad make eggs, but mushy eggs. 353 Me mom hold in him hands. 356 He put in a box
with lots different and jelly bean. 358 Then mom and kids find easter eggs. 362 I need ten stickers. 369 We
got some new pictures. 371 I know these ones. 373 Them both have pinchers205. 374 Why them both have
pinchers? 375 Both got legs
You can use the data in this list to draw up a frequency table which shows the raw frequency of individual
s-forms produced by JC in obligatory plural contexts. The partial table in (6) below gives you an idea of
what your frequency table might look like (with numerals indicating the number of times JC used the
relevant word in plural contexts):
(6) Raw frequency of noun-forms used by JC in plural contexts
Noun
A: correct form
B: bare form
EGG
eggs = 5
LOBSTER
lobsters = 3
lobster = 1
C: other form
Having drawn up (your own complete version of) a frequency table like that in (6), you can then use the
data in the table to calculate JC’s percent suppliance of plural s-forms in obligatory contexts, in the
manner shown below:
(7)
% suppliance of plural s-forms of
nouns in obligatory contexts
= no. of tokens of correct plural s-forms of nouns x 100
no. of obligatory contexts for plural s-forms of nouns
If we do this calculation for the (small subset) of data in table (6), what we do is first of all add up the
number of tokens of correct s-forms in column A: this comes to 8 (viz. 5 tokens of eggs and 3 of lobsters).
We then multiply this total of 8 by 100, giving us a top-line figure of 800. We calculate the number of
obligatory contexts for plural s-forms of nouns by adding up the number of tokens of correct and incorrect
noun-forms in columns A, B and C: this comes to 9 (viz. 5 tokens of eggs, 3 of lobsters, and 1 of lobster).
We then divide our top-line total of 800 by our bottom-line total of 9, giving us a figure of 89% supplinace
of plural -s on nouns in obligatory contexts206. Of course, this figure is only based on a partial set of data:
your own score must be based on the complete set of data in (5), and may be somewhat different from this.
2.3 Research results for genitive s-forms
Begin this section by saying that you drew up a list of sentences from the JC corpus which represent
obligatory contexts for genitive ’s (in noun expressions like Daddy’s or the man next door’s) and that the
list is included in your Appendix. Below, I have drawn up a list of (italicised) noun expressions which JC
seems to use in genitive contexts (i.e. which he seems to use as possessors).
205
One possibility which you might want to consider is whether pincher is a novel noun formed by adding the
nominalising (i.e. noun-forming) suffix –er to the verb stem pinch to form the noun pincher. It seems to be used in
contexts where adults would use the word pincer.
206
If we excluded a pair lobster as indeterminate in status, this figure would rise from 89% to 100% suppliance.
81
(8) List of (italicized) noun expressions used by JC in genitive contexts (as possessors)
62 Me brother name Jack (= ‘My brother’s name is Jack’).
72 That can be Giovanni (= ‘Giovanni's’).
105 He shoveled someone else [?= ‘He shoveled someone else’s (snow)’].
157. A girl hat (‘A girl’s hat’)
159 This is somebody else fishing (= ‘somebody else’s fishing game’)
161 Now it’s gonna be Matthew (= ‘Matthew’s turn’) cause, cause...
263. Where Giovanni one? (‘Where’s Giovanni’s one?’)
264 Where Giovanni sticker? (‘Where’s Giovanni’s sticker?’)
276 Me sister name Dawn (‘My sister’s name is Dawn’).
You can then use the list in (8) to draw up a frequency table (which you include in the main body of your
text) showing the relative frequency with which JC uses s-forms and bare forms in genitive contexts.
Below, I’ve drawn up a partial table in order to give you an idea of what such a table might look like:
(9) Raw frequency of noun-expressions used by JC in genitive contexts (as possessors)
A: correct s-form
B: bare form
Giovanni = 3
me sister = 1
someone else = 2
Once you’ve drawn up your own complete version of the table in (9), you then need to calculate the
frequency with which JC marks noun expressions with genitive ’s in obligatory contexts, using the
procedure shown below:
(10)
% suppliance of genitive ’s on noun =
expressions in obligatory contexts
no. of tokens of genitive s-forms x 100
no. of obligatory contexts for genitive s-forms
This amounts to adding up the number of correct s-forms you find in all the A-cells in (your complete
version of) table (10), multiplying this by 100, and then dividing the resulting figure by the number of
s-forms and bare forms you find in all the A- and B-cells together. If we do this calculation for the small
fragment of data in table (10), we find that JC produces 0 (i.e. zero) s-forms in 6 genitive contexts, so his
percent suppliance of genitive s-forms in obligatory contexts is 0% (i.e. 0/6 x 100).
There is clearly little point in looking for possible perceptual effects (e.g. seeing if he performs better on
genitive forms which involve the use of the vowel-containing |Iz| allomorph than on those which involve
use of a purely consonantal allomorph like |z|), for the obvious reason that he performs equally badly (at a
0% suppliance level) on each. For the same reason, there is no point in trying to look for possible
frequency effects.
2.4 Research results for present-tense s-forms of main verbs
Begin this section by saying that you drew up a list of sentences from the JC corpus which represent
obligatory contexts for third person singular present tense (3SgPres) s-forms of main verbs – and include
this list in the Appendix to your project. In (11) below is a list of (italicised) main verbs which occur in
potential 3SgPres contexts in the JC corpus (with glosses in parentheses giving the adult English
counterpart)207:
I have excluded utterance 348. Her’s use with a green thing from this (and other) lists, since its precise status is
indeterminate. For example, ’s here may represent a contracted form of does (which would mean that he is using
DO-support inappropriately). Alternatively, JC may have been trying to say She uses a green thing, and because he
has not acquired the nominative pronoun she, uses the genitive form hers instead (though this doesn’t seem likely
either, since he generally does not use genitive pronouns as subjects).
207
82
(11) List of (italicized) main verbs used by JC in third person singular present-tense contexts
44. When him crack tiny pieces up, and then put xxx (‘When he cracks tiny pieces up and then puts...’)
45. Then dump into a truck again (‘Then dumps them into a truck again’)
55. He like Danny talking like that (‘He likes Danny talking like that’)
64. He only have a coat (‘He only has a coat’)
65. Now he have them two (‘Now he has those two’)
67. He have a doctor (‘He has a doctor’)
99. But, me mom let (‘But my mum lets (me)’)
116. When he hold here (‘When he holds (it) here’)
117. Me daddy like mustard (‘My daddy likes mustard’)
121. And it stick to you (‘And it sticks to you’)
156. He have he hats on (‘?= He has his hats on’)
195. Her standing and her see herself (‘She’s standing and she sees herself’ – describing picture)
346. My dad make eggs, but mushy eggs (‘My dad makes eggs...’)
349. Her use with a green cutter thing (‘She uses a green cutter thing’)
352. He hafta (‘He has to’)208
353. Me mom hold in him hands (‘My mum holds it in her hands’)
356. He put in a box with lots different and jelly bean (‘He puts (it) in a box with lots of different things and
jelly beans’)
357. Then he bring it (‘Then he brings it’)
360. Mom say, no, after lunch (‘Mom says ‘No, after lunch’)
370. Her need the camera to put in a locker (‘She needs the camera…’)
372. It look like a lobster (‘It looks like a lobster’)
376. He...him go first (‘He goes first’)
You can use the list in (11) to draw up a table (which you include in the main body of your text) showing
the relative frequency with which JC uses s-forms and bare forms of main verbs in 3SgPres contexts.
Below, I’ve drawn up a partial table (with items listed in alphabetical order) in order to give you an idea
of what such a table might look like:
(12) Raw frequency of main verbs used by JC in third person singular present contexts
Verb
A: correct s-form
B: incorrect bare form
HOLD
hold = 2
LIKE
like = 2
LOOK
look = 1
Once you’ve drawn up your own complete version of the table in (12), you then need to calculate the
frequency with which JC marks main verbs with {s} in third-person-singular present-tense contexts, using
the procedure shown below:
(13)
% suppliance of 3SgPres {s} on =
main verbs in obligatory contexts
no. of tokens of 3SgPres main-verb s-forms x 100
no. of obligatory contexts for 3SgPres main-verb s-forms
This amounts to adding up the number of correct s-forms you find in all the A-cells in (your complete
version of) table (12), multiplying this by 100, and then dividing the resulting figure by the number of
s-forms and bare forms you find in all the A- and B-cells together. If we do this calculation for the small
fragment of data in table (12), we find that JC produces zero s-forms of main verbs in five 3SgPres
contexts, so his percent suppliance of {s} on main verbs in 3SgPres contexts is 0% (i.e. 0/5 x 100).
There is clearly little point in looking for possible perceptual effects (e.g. seeing if he performs better on
3SgPres forms which involve the use of the vowel-containing |Iz| allomorph of {s} than on those which
involve use of a purely consonantal allomorph like |s| or |z|), for the obvious reason that he performs
equally badly (at a 0% suppliance level) on each. For analogous reasons, there is no point in trying to look
208
Hafta is a contracted form of have to, widely used in spoken American English.
83
for possible frequency effects (e.g. seeing whether he performs better on high- than low-frequency
3SgPres forms), since he scores equally badly (with a 0% suppliance rate) on all types of verb.
2.5 Research results for present-tense s-forms of auxiliaries
Begin this section by saying that you drew up a list of sentences from the JC corpus which represent
obligatory contexts for third person singular present tense auxiliary s-forms (both the full is/has/does
forms and their contracted variant ’s) – and include this list in the Appendix to your project. In (15) below,
I have drawn up such a list: in almost all cases, the relevant context is one for is209, the few contexts for
does or has being shown in parentheses.
(14) List of sentences produced by JC in contexts where adults use auxiliaries (i)s/(doe)s/(ha)s
A: Sentences in which JC uses the full forms is/has/does
57. This girl is. 58. But he bus is over here. 75. Is a dog under a table210 160. This is a girl and this is a boy.
161. This is somebody else fishing. 162. This is you? (= This is yours?) 191. Dirt is falling all over him.
198. This black shadow is her and her. 208. Ooh, this is a scary one. 211. Is green. 224. I’n see that one,
cause that one is very good. 248. This is gonna be hot dog. 249. This is gonna be coffee. 268. That
because is shamrock today. 334. My another one is lost. 340. That mean is already cook. 386. This is a
good one.
B: Sentences in which JC uses the contracted form ’s
34. It’s a little bit night time. 56. He’s sick. 59. But, he’s not have no money (= does). 78. It’s a cat.
82. And that’s not for a baby. 126. And it’s gonna be all gone. 130. It’s hard to find this. 156. It’s a plane
157. He’s funny. 163. Now it’s gonna be Matthew211 cause, cause... 164. It’s gonna be my, cause I'm
bigger than him. 173. It’s swinging. 175. He’s sad cause he can't go. 189. It’s a farm 190. He’s digging
up dirt. 199. It’s her hair. 201. I think it’s a ghost. 202. It’s not, imagination. 207. It’s a... 209. It’s a owl.
220. Oh, that’s different Barney. 226. It’s a boat. 262. Who’s that? 271. That’s my speech teacher, a
speech one. 276. It’s got eye (= has). 277. It’s a seagull. 284. Her say, ahh it’s moving.
C: Sentences in which JC omits the auxiliary (from the underlined position)
4. That_me friend. 17. He_happy. 29. How long_the song gonna be on? 30. Why_it not being on?
39. But it_a truck you dump dirt, and snow… 51. It_very sharp. 54. But he_not there. 57. Her_not.
60. Her_pretending to being a doctor. 62. Me brother name_Jack. 76. He_under the table 77. Her_laying.
88. He_making a mess. 108. Who_that? 110. This one, he_cooking up a hot dog. 113. Ooh, that_gross.
147. Why_her need this? (= does) 151. Her_sad. 177. And he_sad cause he_crying. 180. He_crying,
he_crying. 185. He can fly cause he_got these wings. (= has). 197. Her_standing. 206. He_flying.
218. It_just pretend. 219. He_not real. 223. He_got old one (= has). 230. _This a new game?
243. Daddy_got train (= has). 245. I don’t know what he_saying. 246. That_my brother.
258. That_not pepper. 265.Where_Giovanni one? 266. Where_Giovanni sticker? 268.That_because is
shamrock today. 278. Me sister name_Dawn. 301.Her_eating a cookie. 316. How_he gonna eat me up?
317. He_grabbing a lollipop. 383. He_taking a lollipop. 384.He_gonna bite you.
Using the data in table (14), you can then calculate how frequently JC supplies the auxiliaries
(i)s/(doe)s/(ha)s in obligatory contexts using the following procedure:
(15)
209
% suppliance of s-forms of auxiliaries =
no. of tokens of s-forms of auxiliaries
in obligatory 3gPres contexts
no. of obligatory contexts for s-forms of auxiliaries
The word is serves as a progressive auxiliary when followed by a verb in the progessive ing-form (as in He is
working) and as a copular (i.e. linking) verb when followed by an adjectival, prepositional or nominal expression (as
in He is happy/in Paris/a doctor). In both uses, is has the morphosyntax of an auxiliary, in the sense that e.g. it forms
negatives and questions without the use of DO-support (cf. He isn’t happy/Is he happy?).
210
In utterances 75, 211, 268 and 340, is would appear to be a contracted form of it’s – though this is not clear.
211
Apparently meaning ‘It’s gonna be Matthew’s turn’
84
The first thing you need to do is add up the number of times JC uses is/has/does/’s in the A- and B-rows in
table (14), and multiply this figure by 100, to give you your top-line figure. You calculate your bottomline figure (i.e. the number of obligatory contexts for s-forms of auxiliaries) by adding up the total number
of times JC uses and omits is/has/does/’s in A-, B- and C-rows columns in table (14). You then divide
your top-line figure by your bottom-line figure to give you JC’s percent suppliance of s-forms of
auxiliaries in obligatory contexts. If we were to do this calculation just for the (very small sample of)
utterances numbered between 50 and 60 in the corpus, we’d find that JC uses s-forms 4 times (in
sentences 56, 57, 58 and 59), and omits s-forms 4 times as well (in sentences 51, 54, 57, and 60). His
suppliance rate for auxiliary s-forms in these 8 sentences is therefore 50%. Of course, you have to
compute your own suppliance rate using all the data in table (14), not just the data from the 8 sentences I
have chosen here for illustrative purposes.
An interesting additional question to look at is whether JC fares better on uncontractible s-forms of
auxiliaries than on their contractible counterparts (as the Perceptual Deficit model would lead us to
expect). Auxiliary s-forms are uncontractible when (i) the last last word in a sentence (e.g. I wonder where
he is), (ii) followed by a weak pronoun at the end of a sentence (e.g. What is it?) and (ii) occurring after a
word ending in a sibilant (e.g. This is nice). If we exclude utterances 75, 211, 268 and 340 (because of the
indeterminate status of is in the relevant sentences), the only utterances in the JC corpus which constitute
obligatory contexts for uncontractible s-forms of auxiliaries are 57 (where is occurs at the end of a
sentence), and 58/160/161/162/208/248/249/386 where the auxiliary occurs after a word ending in the
sibilant |s|. If we do this, we can calculate JC’s suppliance rate for uncontractible s-forms of auxiliaries in
terms of the following formula:
(16)
% suppliance of uncontractible s-forms = no. of tokens of s-forms in uncontactible contexts
of auxiliaries in obligatory contexts
no. of obligatory contexts for contractible s-forms
For the reasons given above, utterances which could be taken to constitute obligatory contexts for
contractible s-forms of auxiliaries are 57/58/160/161/162/208/248/249/386.
You will then need to compare JC’s suppliance rate for uncontractible s-forms of auxiliaries with his
suppliance rate for contractible forms. Obligatory contexts for such forms are all the utterances in list (15)
except utterances which are indeterminate (75/211/268/340), which may involve truncation (230), or
which are obligatory contexts for uncontractible s-forms (57/58/160/161/162/208/248/249/386). You can
then compute his suppliance rate for contractible s-forms of auxiliaries using the following formula:
(17)
% suppliance of contractible s-forms =
of auxiliaries in obligatory contexts
no. of tokens of s-forms in contractible contexts
no. of obligatory contexts for contractible s-forms
The figures you arrive at in relation to (16) and (17) will enable you to copare how well JC fares on
supplying contractible and uncotractible s-forms of auxiliaries in obligatory contexts.
2.6 Summary of research results
In this section, briefly summarise your research results on how frequently JC supplies each of the different
types of s-form covered in your study.
3. Analysis of research results (around 35% of your overall study)
3.1 Introduction
Begin with an introduction saying that what you are aiming to do in this section/chapter is evaluate the
extent to which the research findings that you obtained in the previous chapter are consistent (or
inconsistent) with the research hypotheses which you outlined in chapter 1, and (more generally) with the
particular models of SLI which you are aiming to test. Since the main aim of your research is to use your
research results to evaluate a number of different theories of SLI, you are going to organise the material in
this chapter into sections by theory – e.g. discussing the extent to which your research findings are
85
compatible with the Perceptual Deficit model in section 3.2, with the Rule Deficit model in section 3.3,
with the Agreement Deficit model in section 3.4, and with the ATOM model in section 3.5. Below, I’ll
assume the results reported in relation to partial sets of data in §2.2 (= 89% suppliance of {s} on regular
nouns in plural contexts), §2.3 (= 0% suppliance of {s} on possessor in genitive contexts), 2.4
(= 0% suppliance of {s} on main verbs in 3SgPres contexts) and §2.5 (= 50% suppliance of s-forms of
auxiliaries in 3SgPres contexts) – though your figures (based on complete sets of data) may be somewhat
different from these, and your discussion must be based on your complete results, not my partial results.
3.2 Evaluation of the Perceptual Deficit model
On the face of it, the Perceptual Deficit model would appear to predict that SLI children should have
problems in acquiring all s-forms – and JC’s 0% suppliance of genitive {s} on noun expressions and
3SgPres {s} on main verbs would appear to be consistent with this claim. But the overall picture is not
quite so straightforward, because JC also achieves 89% suppliance of plural {s} on nouns and 50%
suppliance of s-forms of auxiliaries. It would appear that we need to assume that a perceptual deficit is
only part of a more complex picture. For example, suppose that learning to use a morpheme in appropriate
contexts involves a three-stage process – first perceiving the relevant form, then parsing it (i.e. working
out what grammatical features it carries), then practising using it; and suppose that all three stages take
considerably longer (e.g. twice as long) in SLI children as in normally developing children. We would
then expect to find three types of effect. One is a perceptual effect (of the type that Leonard discusses), so
that (all things being equal) we’d expect children to be better at vowel-containing forms than at purely
consonantal forms. A second effect we’d expect to find is a frequency effect, in that children at an
immature stage of development would be expected to perform better on forms which occur more
frequently in their speech input than those which occur less frequently (because all three stages in learning
how to use a morpheme require extensive exposure to the morpheme, and a child at a given stage of
development will have had more exposure to high-frequency than low-frequency forms). The third effect
we’d expect to find is a complexity effect, in that analysing the grammatical properties of a morpheme is
going to take longer if the morpheme is a complex one encoding several features than if it is a simple one
encoding a single feature (or if it encodes an uninterpretable feature, or an interpretable feature denoting a
cognitively complex concept).
You can in principle test for possible perceptual effects by looking at whether JC performs better on
vowel-containing allomorphs of s-forms than on purely consonant allomorphs. However, given that he has
a near-perfect score for plural s-forms (which would indeed have been perfect had we excluded the
indeterminate phrase a pair lobster), and a zero score for genitive s-forms and 3SgPres s-forms of main
verbs, there seems little point in doing so. But one place where you can look for possible perceptual
effects is in JC’s production of auxiliary s-forms, where the Perceptual Deficit model would lead us to
expect that he should perform much better on supplying uncontractible than contractible forms – and the
results you obtained in §2.5 will help you determine whether this is so. But even if JC is much better at
supplying uncontractible than contractible auxiliary s-forms, you might want to ask whether this
necessarily represents a perceptual deficit, or whether (e.g.) it could be the result of phonological
production problems caused by consonant clusters formed when ’s follows a word ending in one or more
consonants, and/or precedes a word ending in one or more consonants.
What about possible frequency effects? For example, could it be that one reason why JC performs well on
plural {s} but badly on genitive {s} and on 3SgPresent {s} is that plural s-forms occur far more frequently
than genitive s-forms or 3SgPresent s-forms? Unfortunately, although a frequency dictionary like Carroll
(1971) will tell you the relative frequency (per million words) of individual plural noun-forms and
3SgPresent verb-forms in adult English212, it won’t give you the frequency of genitive s-forms, since these
212
The frequency figures (per million words) for relevant 3SgPres s-forms in adult English given by Carroll are as
follows (with the frequency of the corresponding bare form in parentheses): is = ???? (be = 60852); has = 10369
(have = 22337); means = 1962 (mean = 1266); makes = 1311 (make = 8333); says = 1180 (say = 3916); goes = 930
(go = 5388); looks = 756 (look = 4933); needs = 616 (need = 2281); uses = 603 (use = 7009); sticks = 242 (stick =
501); likes = 229 (like = 9696); holds = 228 (hold = 1192); sees = 180 (see = 8515); brings = 169 (bring = 1016);
86
often attach to phrases (and there are an infinite set of possible phrases, so a frequency dictionary cannot
list them). To help you get round this problem (and because a frequency dictionary like Carroll’s isn’t
based on parental speech input to children), I calculated the frequency with which various s-forms
occurred in parental speech in 20 files from the Abe corpus on the CHILDES data-base (each file
representing about 10 minutes of conversation between a two-year-old normally developing boy called
Abe and his mother and/or father). During these files, Abe’s parents produced 119 tokens of regular noun
plural {s}, 91 of regular 3SgPres {s}, and 5 of genitive {s} on noun expressions; they also produced 343
tokens of the irregular 3SgPres auxiliary (i)s, 22 of the auxiliary (doe)s and 7 of the auxiliary (ha)s. Of
course, it has to be said that this is only a small sample of (3 hours or so) of the speech production of the
parents of one child – but at least they give you some kind of indication of likely relative frequencies.
In order to test for a possible complexity effect, you need to ask yourself how many grammatical features
each of the various s-forms produced by JC encodes, and whether this accounts for why he performs better
on some than others. But you need to bear in mind that you can only test for a complexity effect by
comparing simple and complex forms which have roughly the same frequency of occurrence, since
otherwise, there may be a frequency effect which skews the complexity effect. So, for example, if you
wanted to compare s-forms on nouns and main verbs, you’d need to compare sets of forms with roughly
the same frequency in adult English. Since your project is designed to test your intelligence and initiative
and not mine, I leave you to think out how to do this!
If you find evidence of the three effects suggested above, you could conclude this section by discussing
the extent to which the interaction of these three effects might account for your overall research findings.
How do you do that? By using your initiative and intelligence and demonstrating that you can think a
problem through (and find possible solutions) for yourself - after all, problem-solving is one of the key
skills which you’re supposed to demonstrate evidence of in your academic work!
3.3 Evaluation of the Rule Deficit model
The Rule Deficit model claims that SLI children are unable to acquire regular morphological rules like the
present-tense s-rule in (1), the plural s-rule in (2) and the genitive s-rule in (3): one prediction which this
makes is that children with SLI will not produce novel or over-regularised s-forms. The model maintains
that the only way they can acquire inflected forms (including s-forms) is by memorisation. Since
memorisation (and the ability to retrieve a memorised form) improves with increased exposure, we’d
expect to find a correlation between the suppliance rate for individual inflected word-forms and the
frequency of the relevant form (e.g. that an SLI child achieves 80% suppliance of a high-frequency form,
50% suppliance of a medium-frequency form, and 20% suppliance of a low-frequency form). So, the first
thing you need to look at is the extent to which you find any evidence of frequency effects. Since how to
do this is discussed in the previous section, I won’t repeat the relevant material here. But bear in mind the
possibility of a secondary frequency effect mentioned by Gopnik and Crago – namely that SLI children
tend to perform poorly on inflected forms which are less frequent than the corresponding bare form: for
example, could it be that one reason why JC uses the bare form like in contexts where adults use the
inflected (third person singular present tense) s-form likes is that like is 33 times more frequent than likes?
bear in mind that any frequency effect may in turn be skewed by a complexity effect (e.g. if two inflected
forms have the same frequency but differ in complexity, a child would be expected to be better at the
simpler one).
Suppose that you don’t find any clear evidence of any frequency effect. What then? You might feel
tempted to conclude that the Rule Deficit theory is thereby falsified. But before throwing out a theory, you
need to ask whether it can be modified in some way which will salvage it. Suppose, for example, that
rather than following Gopnik and Crago (1991) in hypothesising that SLI children (because of a genetic
deficit) are unable to acquire regular morphological rules, we instead hypothesise that they show a delay
puts = 156 (put = 3942); cracks = 81 (crack = 144); lets = 70 (let = 2176); dumps = 8 (dump = 33). The frequency
figures Carroll gives for relevant plural s-forms of nouns (and of their bare counterparts) are:
87
in acquiring morphological rules (e.g. perhaps they require twice as much exposure as normally
developing children, and hence take twice as long to acquire regular rules). If SLI children are indeed
considerably slower than normally developing children in acquiring regular morphological rules, we might
expect that a child (like JC) who is at an immature stage of development will show evidence of having
acquired some morphological rules, but not others. For example, suppose that JC has acquired the
defective (i.e. incomplete) set of regular morphological rules shown below:
(18)
A regular verb carries the suffix {d} if past213, {ing} if progressive or gerund, {ø} otherwise
(19)
A regular noun carries the suffix {s} if plural, {ø} otherwise
I leave you to think through whether this would account for (some or all of) your research results.
3.4 Evaluation of the Agreement Deficit model
The core assumption of the Agreement Deficit model is that SLI children have problems in acquiring
uninterpretable agreement features (but not interpretable features or other uniterpretable features). My
preliminary findings (based on a very small subsample of the JC corpus) seem to lend some support to this
claim, in that JC achieves a very high (89%) suppliance rate for plural -s on nouns (which marks an
interpretable number feature), but a 0% suppliance rate for regular 3SgPres -s (which marks
uninterpretable agreement features). However, my preliminary finding that JC has a 0% suppliance rate
for genitive ’s seems incompatible with the prediction made by Clahsen, Bartke and Göllner (1997) that
SLI children are very good at case-marking (their paper reporting 100% correct suppliance of case
marking by the English children in their study). However, work in other studies (e.g. Loeb and Leonard
1990, and Bishop 1994) has shown that younger SLI children do indeed have problems with casemarking. We might therefore want to follow Tsimpli and Stavrakaki (1999) in concluding that SLI
children have problems with all uninterpretable features: let’s call this the Uninterpretable Feature Deficit
Hypothesis/UFDH. This would then account for why JC scores 89% correct on plural -s (which marks an
interpretable number feature), but 0% correct on 3Sg -s (which marks uninterpretable agreement features)
0% correct on genitive ’s (which marks an uninterpretable case feature). Note, however, that UFDH
would also predict that SLI children like JC should have problems in acquiring irregular genitive pronouns
like my/our/your/their etc, and should use default (accusative) forms in genitive contexts. You can test this
by calculating JV’s suppliance rate for genitive pronouns in obligatory contexts. Because your main
concern is with s-forms and not with pronouns, you can limit yourself to looking at how he case-marks the
most frequent type of pronominal possessor he uses – namely first person singular possessors. Below I
have listed all the first person singular possessors used by JC in the corpus:
(20) List of first person singular personal pronouns used in genitive contexts by JC
A: Genitive my possessors
B: Default accusative me possessors
101. Somebody else asked my mom to play outside
4. That me friend
with them
62. Me brother name Jack
162. It’s gonna be my, cause I’m bigger than him.
71. All of these can be me
166. I got on my shirt and have trouble doing my back 99. But, me mom let
203. I can make see my shadow
242. Me daddy
244. That my brother
267. Me teacher make cake
269. That’s my speech teacher…
276. Me sister name Dawn
326. Then a bee eat a little bit my food
277. Her give me dad a lobster, a two lobster
335. My another one is lost
278. Me mom put in here, cook them
346. My dad make eggs, but mushy eggs
309. Me grandma did long time ago from a snake
359. Then, I grab my one
337. Me mommy show me to make Easter eggs
365. I got my hair from barber
347. Me mom don’t use paint brush
366. But them cut my hair real tiny
353. Me mom hold in him hands
213
There are not enough relevant examples in the corpus to be sure whether JC makes productive use of regular
perfect and passive participle d-forms, so I have not included these here.
88
From this list, using prodecures which by now should be yawningly familiar, you can calculate JC’s
percent suppliance rate for first person singular genitive pronouns in obligatory contexts (multiplying the
number of italicised forms in column A by 100, and dividing this sum by the total number of italicised
forms in columns A and B together). If JC’s suppliance rate for my is much higher than for ’s, why do you
think this might be – e.g. could frequency be a factor?
3.5 Evaluation of the ATOM model
The ATOM model claims that SLI children have particular problems in marking tense and agreement, and
sporadically omit either or both of these features on verbs in finite contexts (i.e. in contexts where adults
would use a verb which is marked for both tense and agreement). ATOM also assumes that case is tied up
with agreement, and that SLI children use correctly case-marked forms when they mark agreement, and
default (accusative) forms when they fail to mark agreement.
If (in line with the discussion in §1.6) we assume that plural number on nouns is not an agreement feature,
we’d expect JC to show a very high suppliance rate for marking plural number on nouns in obligatory
contexts (the relevant rate being determined by the procedure in (7) above).
If (again in line with the discussion in §1.6) we take genitive case to be a reflex of possessor-agreement,
then one possible ATOM account of why JC never uses genitive {s} would be that he has not acquired
possessor-agreement, with the result that possessors are always assigned default accusative case (and so
surface as bare s-less nominals like Giovanni in Giovanni sticker). But an immediate problem with this
analysis is that ATOM claims that SLI children only sometimes fail to mark agreement in obligatory
contexts – a claim which would seem to predict that JC should alternate between structures containing an
agreementless accusative possessor (as in Daddy car where Daddy would be a bare accusative possessor)
and structures containing an agreeing possessor carrying the genitive {s} suffix (as in Daddy’s car, where
Daddy’s would be an s-marked genitive possessor). Still, maybe JC has acquired subject agreement
(between subject and verb) at this stage, but not possessor-agreement – perhaps because the former is far
more frequent than the latter (in that almost every sentence contains a subject agreeing with a verb, but by
no means every sentence contains a possessor). You can test whether JC does or doesn’t sporadically mark
possessor agreement by calcxulating his suppliance rate for first person singular genitive pronouns in the
contexts lised in table (20), uswing the procedure outlined in the previous section. If he sporadically marks
possessor agreement in genitive contexts, we’d expect him to alternate between (agreeing) genitive
possessors like my and (non-agreeing) default accusative possessors like me. Is this what we find?
Suppose (for the sake of argument) that he assigns genitive case to around half the first person pronouns
pronoun possessors he uses. What we’d then expect would be that around half of his nominal possessors
should carry the genitive {s} suffix, and around half should be bare (s-less) expressions. But this is not the
picture we find. Why do you think that might be?
Let’s now consider the predictions ATOM makes about how SLI children treat main verbs in 3SgPres
contexts, and whether these are borne out. For the reasons set out in §1.6, ATOM predicts that in contexts
where adults would produce a sentence like He plays with me, an SLI child who has acquired the third
person singular present tense morpheme {s} will alternate between producing nominative-subject
sentences like He plays with me and He play with me, and accusative-subject sentences like Him play with
me (though not *Him plays with me). But what would ATOM predict about a child like JC who seems not
to have acquired the 3SgPres morpheme {s}? I leave you to answer this question for yourself. In order to
determine the extent to which the relevant predictions about which types of subject will (and won’t) occur
with which types of verb in 3SgPres contexts, you need to use the data in the list in (11) above to draw up
a separate list of sentences produced by JC with nominative and accusative personal-pronoun subjects in
contexts where adults would use a 3SgPres main verb (excluding it subjects because these show no
nominative-accusative contrast). I’ve drawn up a partial list for illustrative purposes below (the relevant
subjects being italicised and their associated verbs bold-printed):
89
(21) List of nominative and accusative subjects that JC uses with main verbs in 3SgPres contexts
A: Nominative subjects
B: Default accusative subjects
55. He like Danny talking like that
44. When him crack tiny pieces up…
64. He only have a coat
195. Her standing and her see herself
When you’ve drawn up a complete version of the list in (25), you need to ask yourself to what extent the
set of structures JC produces is consistent with the predictions made by ATOM.
The last set of structures to look at are those containing auxiliary s-forms. For the reasons set out in §1.6,
ATOM predicts that in contexts where adults would say He’s sleeping, SLI children would say He’s
sleeping, He sleeping, and Him sleeping – but not *Him’s sleeping. In order to determine the extent to
which these predictions are borne out, you need to use the data in the list in (14) above to compile a
separate list of sentences produced by JC with nominative and accusative subjects in contexts where adults
would use a 3SgPres main verb (excluding it subjects because these show no nominative-accusative
contrast). Again, I’ve drawn up a partial list for illustrative purposes below:
(22) List of nominative and accusative subjects that JC uses with auxiliaries in 3SgPres contexts
A: Nominative subjects
B: Default accusative subjects
56. He’s sick
60. Her_pretending to being a doctor
17. He_happy
147. Why_her need this?
59. But, he’s not have no money.
223. He_got old one
When you’ve drawn up a complete version of the list in (22), you need to ask yourself to what extent the
set of structures JC produces is consistent with the predictions made by ATOM. Are there any
discrepancies, and (if so) how do you think they can be accounted for, and to what extent do they
undermine ATOM (if at all)?
Chapter 4. Summary and Conclusions (about 15% of your overall study)
4.1 Summary of research hypotheses
Begin this chapter with a section in which you provide a brief summary of the models of SLI and research
hypotheses which you outlined in chapter 1.
4.2 Summary of research findings
In this section, briefly summarise the main research findings you obtained in chapter 2
4.3 Summary of evaluations
In this section, provide a brief summary of your evaluation (in chapter 3) of the relative strengths and
weaknesses of each of the theories of SLI you are testing, viewed from the perspective of how well they
account for your own research findings.
4.4 Overall conclusions
In this (the most important) section of your chapter, you need to give your considered final verdict on
which of the models you have looked at best accounts for your own research findings. If a particular
model doesn’t account for (some of) your findings, can it be modified in some way which would make it
compatible with your findings? If not, and if there is no one model of SLI which accounts for all your
research findings, does this take you towards a composite model which sees SLI as involving multiple
impairments, rather than a single type of impairment?
Such a composite model might say something along the following lines in relation to how an SLI child
acquires a plural noun like cats (cf. the remarks I made earlier in section 3.2). The first stage (I) is
phonological processing, and perceiving that the form comprises the segments |kats|. The second stage (II)
is grammatical processing – i.e. determining what grammatical features the word cats carries, and what its
90
internal structure is. Children who are at stage I in respect of the acquisition of a given morpheme may go
through a short stage like that described in Gopnik’s (1990) Feature Blindness model when they have
learned cat and cats as separate word-forms, but are not aware of the grammatical distinction between the
two forms, and so use both forms in singular and plural contexts alike (hence alternating between a cat
and a cats, and between two cat and two cats). The first phase of grammatical processing (stage IIA) is
feature decomposition – i.e. decomposing words into the set of features they carry (e.g. realising that cats
encodes a plural number feature): a child who is at this phase will be learning plural forms (regular and
irregular alike) by memorisation. The second phase of grammatical processing (IIB) is morphemic
segmentation – e.g. abstracting a common plural formation pattern from contrasts like cat/cats, dog/dogs,
car/cars and segmenting each regular plural noun into a stem and a plural affix {s}, and working out what
different allomorphs plural {s} has, and what conditions govern the use of each allomorph. The third stage
is practising the use of the relevant morpheme – in this case, plural {s}. Let’s suppose that all three stages
of this process of learning how to use a morpheme take considerably longer in SLI children as in normally
developing children – maybe because they process all aspects of language (say) twice as slowly as
normally developing children. Such a composite model would enable us to incorporate elements of a
number of different theories of SLI. For example, a perceptual deficit of the kind postulated by Leonard
would mean that phonological processing takes much longer than in normal children. And If SLI children
are also much slower to analyse words into their component features than normally developing children,
then we could follow Gopnik (1990) in saying that they have a feature deficit (though not in the sense that
they are ‘blind’ to grammatical features and hence unable ever to acquire such features, but rather in the
sense that they are slow to acquire grammatical features). We might also expect to find that certain types
of feature are more difficult to acquire than others – e.g. perhaps uninterpretable features are intrinsically
more difficult to acquire than interpretable ones (as suggested by Clahsen, Bartke and Göllner 1997), and
perhaps some types of interpretable feature are particularly difficult to acquire (e.g. tense – as in the
ATOM model) because of their conceptual complexity. All other things being equal, we might also expect
that a morpheme which encodes multiple grammatical features will be more difficult to acquire than one
encoding a single feature. And if SLI children are slower at grammatical processing than normally
developing children, we’d expect them to be slower at segmenting words into stems and affixes, and at
acquiring the morphophonological rules which govern the use of affixes, so that at a particular phase (IIA)
in their development, they may have a rule deficit in the sense that they have not yet acquired one or more
regular morphological rules and so learn relevant inflected forms by memorisation – though Gopnik and
Crago’s 1991 claim that SLI children are unable ever to form any regular rules may simply be too strong,
and we also need to bear in mind the possibility that at a given point of development, an SLI child may be
at a later stage/phase in the acquisition of (say) plural {s} than genitive {s}. Moreover, it would seem
reasonable to suppose that morphemes with multiple allomorphs (whose distribution is determined by
complex phonological factors) will take longer to acquire than those with a singe allomorph, for the
obvious reason that the child has to work out the morphophonological rules governing the use of the
relevant allomorphs. And once children have acquired a morpheme (and learned the rules which govern its
use), they need time to practice using it – perhaps considerably more time than normally developing
children. The kind of composite model of SLI which I have outlined here would also lead us to expect to
find frequency effects, since the first two stages of the process – phonological and grammatical processing
– require extensive exposure (to thousands of tokens of a given morpheme), and the threshold number of
exposures required in order to achieve a given level of performance will be reached much earlier for highfrequency morphemes than for low-frequency ones.
Of course, it’s one thing to sketch out a composite model like that which I have outlined above which sees
SLI as affecting a number of different components of the grammar, and quite another to show how such a
model could account for your specific research results. The question which you have to answer in your
conclusion is: how (if at all) could such a composite model of SLI account for your specific results in
relation to how well JC fared on each of the different s-forms you looked at. Be particularly circumspect
in what you say in this part of your project: the conclusion is usually the weakest point in student work,
and (if so) will devalue your overall project.
91
5. Appendices
After your four main chapters, include as an Appendix any lists of sentences which provide the raw data
from which you computed the figures for various tables included in the main body of your project, and
any tables from which you have computed suppliance rates (and any other figures you use).
6. List of references
After any Appendices, include a list of works (arranged alphabetically by author and date) which you have
cited in your project, using the standard author-date system described in the departmental handbook.
92
10. Study III: Acquisition of inflected main-verb forms in English
In this chapter, I look at how to undertake an empirical study of The acquisition of inflected main-verb
forms in English by a 4-year-old child with SLI. Your study might be organised as follows.
1. Background (about 25% of your overall study)
1.1 Introduction
Give brief details (in a paragraph or two) of the kind of developmental disorder which SLI involves
(summarised on p.4 of this book: see Bishop (1997), Leonard (1998), and the collection of papers in
Bishop and Leonard (2000) and Levy and Schaeffer (2003) for more detailed discussion of diagnostic
criteria for SLI), pointing out that various aspects of grammar are reported to be impaired in SLI children,
including the ability to produce inflected forms. Then go on to say that there are a wide range of different
models of SLI which seek to account for the nature of the grammatical impairment in SLI children, and
for the purposes of the present study you are going to examine a subset of these (e.g. the Perceptual
Deficit, Rule Deficit, Agreement Deficit and ATOM models). Your aim is to do a small-scale empirical
study designed to evaluate each of the models, by looking at the production of three types of inflected
forms of main verbs in obligatory contexts by a four-year-old American boy with SLI, known as JC.
Provide brief details of JC’s background (which you can find at the beginning of chapter 13). The three
types of inflected forms of main verbs which you are going to look at are (i) past tense forms, (ii) third
person singular present tense forms, and (iii) progressive participle forms. You’re going to exclude other
inflected forms of main verbs (in particular, perfect participles, passive participles and gerunds) from your
study because there are insufficient obligatory contexts for them in the JC corpus for you to draw any firm
conclusions about his production of them. You’ll begin by outlining the morphophonology of the three
main-verb forms that you are focusing on.
1.2 Outline of the morphophonology of inflected forms of main verbs in English
In this section, you need to provide a brief account of the morphophonology of the three types of inflected
main-verb forms which you are studying. The dual mechanism model of the acquisition of morphology
(outlined e.g. in Pinker and Prince 1988 and Pinker 1991) maintains that irregular forms are memorised
forms which are stored/listed in the mental lexicon, whereas regular forms are computed or derived via
application of morphological rules like those sketched informally in (1) below (with morphemes enclosed
in curly brackets, by convention):
(1)
A regular verb carries the suffix
{s} if third person singular present
{d} if past, perfect or passive
{ing} if progressive or gerund
{ø} otherwise
The morphemes {s} and {d} have a variety of different allomorphs (i.e. variant forms), and the use of
these is determined by by morphophonological spellout rules like those in (2) and (3) below:
(2)
{s} is spelled out as:
|Iz| when attached to a form ending in a sibilant consonant (e.g. passes, buzzes, catches, cages,
pushes, camouflages)
|z| when attached to a form ending in another voiced segment214 (e.g. calls, robs, bags, grows,
pays)
|s| when attached to a form ending in another voiceless segment215 (e.g. cuts, locks, breaks,
rocks)
|ø| when attached to a form already ending in {s} (e.g. the boys’ school, where genitive {s}
has a null spellout by virtue of being attached to a noun boys carrying plural {s})
Since all vowels are voiced, this means ‘when attached to a stem ending in a vowel, or a voiced non-sibilant
consonant’.
215
In other words ‘when attached to a voiceless consonant which is not a sibilant’.
214
93
(3)
{d} is spelled out as:
|Id| when attached to a stem ending in an alveolar stop – i.e. |t| or |d| (e.g. padded, batted)
|d| when attached to a stem ending in another voiced segment216 (e.g. moved, sneezed, paid)
|t| when attached to a stem ending in another voiceless segment217 (e.g. passed, packed, laughed)
By contrast, the morpheme {ing} is uniformly spelled out as |Iŋ| in standard varieties of English (though as
|In| in non-standard varieties).
How the system works for regular main verbs can be illustrated in the following terms. The lexical entry
(i.e. entry in the dictionary) for the regular verb HELP will contain the stem form help-. The third person
singular present-tense form of HELP will be derived by adding the suffix {s} to the stem in accordance
with the first line of rule (1) so deriving help-s: the {s} suffix will be spelled out as |s| in accordance with
the third clause of rule (2), so that the word is pronounced |helps|. The past tense form of HELP will be
generated by adding the the suffix {d} to the stem form in accordance with the second line of rule (1), so
deriving helped; the {d} suffix will be spelled out as |t| in accordance with the third line of rule (3), so that
the word is pronounced |helpt|. The progressive participle form of HELP is derived by adding the suffix
{ing} to the stem in accordance with the third line of rule (1), so forming helping; the {ing} suffix is
spelled out as |Iŋ| in standard varieties, so that the resulting progressive participle is pronounced |help Iŋ|
By contrast, the lexical entry for the irregular main verb CATCH will specify that it has the irregular
past/perfect/passive form caught, and the fact that the verb is listed in the lexicon as having this irregular
form will block application of the regular d-rule in (1) to irregular verbs like CATCH. Since no verbs have
irregular progressive participle forms in English, its progressive participle will be formed by adding {ing}
to the stem form catch-, so forming catching. Likewise, since catch is regular in respect of its present
tense formation, its 3SgPres form will be derived by adding {s} to the stem form catch-, deriving catches.
While the 3SgPres form of the vast majority of main verbs is entirely regular in English (as with e.g.
helps, works, sleeps, cries etc.), there are three 3SgPres main-verb forms which are irregular – namely
says, has and does (the last two of which function as main verbs in sentences such as He has a flat in town
or He does a lot of work for charity)218. For details of ways in which these forms are irregular, see section
1.2 of chapter 9.
Empirical research is about testing research hypotheses, so in sections 1.3-1.6, you should provide a short
summary of the overall claims made by each of the models of SLI which you have decided to evaluate,
and then go on to say what specific prediction each makes about how well SLI children would be expected
to perform on each of the three types of inflected forms of main verbs which you are looking at. Of
course, if you are looking at only 3 of the 4 models of SLI dealt with here, you will omit the section on the
model that you are not dealing with.
1.3 Outline of the Perceptual Deficit model of SLI
In this section you should provide an outline of the Perceptual Deficit model, noting that it claims that
consonantal inflections pose more perceptual problems for SLI children than vocalic inflections, and
saying what this predicts about how SLI children will perform on producing the three types of inflection
you are concerned with here. In relation to progressive {ing}, the Perceptual Deficit model predicts that
this should be relatively unproblematic for SLI children, in that the morpheme {ing} contains a vowel. For
the predictions made about past tense forms of main verbs, see section 1.3 of chapter 8. For the predictions
about 3SgPres forms of main verbs, see section 1.4 of chapter 9.
Since all vowels are voiced, this means ‘when attached to a stem ending in a vowel, or a voiced consonant other
than |d|’.
217
In other words ‘when attached to a stem ending in a voiceless consonant other than |t|’.
218
Has and does also have alternative uses as auxiliaries, but if you are focussing on main verbs, you should not
include these auxiliary uses in your project. Likewise the irregular 3SgPres form is should be excluded from your
project because in all uses it functions as an auxiliary (e.g. in that it has the contracted negative form isn’t and
undergoes inversion in questions like Is it raining again?).
216
94
1.4 Outline of the Rule Deficit model of SLI
In this section you should provide an outline of the Rule Deficit model, which claims that SLI children
have a genetic rule deficit that makes it impossible for them to acquire any regular morphological rules.
They are therefore predicted to be unable to acquire the regular present tense s-rule, the past-tense d-rule,
or the progressive ing-rule outlined in (1) above, and are likewise predicted to be unable to create novel or
over-regularised forms219. Because of this genetic rule deficit, the only learning mechanism which SLI
children have is memorisation, with the consequence that they have to memorise all verb-forms, regular
and irregular alike. Since memorisation of a given form requires extensive exposure to the relevant form,
we therefore expect to find a frequency effect (viz. that the more often a child hears a given present, past
or progressive main-verb form, the more often the child is likely to produce it correctly). More
specifically, if Gopnik and Crago are right in claiming that SLI children have a rule deficit, we should
expect to find a frequency effect with both regular and irregular main verbs alike, with childen being
better at high-frequency inflected forms than at low-frequency inflected forms. But if Gopnik and Crago
are wrong and SLI children (like normally developing children) learn regular forms by rule, we should
expect to find that they perform at the same level on all regular inflected forms of the same type (e.g.
equally well or badly on all regular past-tense d-forms of main verbs).
1.5 Outline of the Agreement Deficit model of SLI
In this section you should provide an outline of the Agreement Deficit model, which claims that SLI
children have particular problems with acquiring uninterpretable agreement features, but not with
acquiring interpretable features. In relation to past tense forms, the model predicts that these pose few
problems for SLI children, since tense is an interpretable feature (and past tense forms do not overtly
inflect for uninterpretable person/number agreement features). Likewise, since progressive {ing} encodes
an interpretable progressive-aspect feature (indicating that the event described by the relevant verb is in
progress), we should expect SLI children to have few problems in acquiring it. However, the picture is
different in relation to 3SgPres forms, in that the s-affix they carry encodes not only an interpretable
present-tense feature, but also uninterpretable (third-person, singular-number) agreement features, and we
would therefore expect that their uninterpretable agreement features will mean that s-forms are difficult
for SLI children to acquire.
1.6 Outline of the Agreement-and-Tense-Omission (ATOM) model of SLI
In this section, you should provide an outline of the Agreement-and-Tense-Omission/ATOM model,
which claims that SLI children go through an Extended Optional Infinitives/EOI stage during which they
sometimes fail to mark either tense and/or agreement in finite contexts (i.e. in contexts where adults use an
auxiliary or main verb marked for tense and agreement). The presence or absence of agreement on the
verb will affect the case assigned to the subject (in that a nominative subject is used when agreement is
marked and a default accusative subject when it is not). On what ATOM predicts will happen with main
verbs in past tense contexts, see section 1.7 of chapter 8. On ATOM’s predictions about what will happen
with main verbs in 3SgPres contexts, see section 1.6 of chapter 9. In relation to progressive {ing}, ATOM
seemingly predicts that SLI children should perform relatively well on progressive ing-forms, since {ing}
marks progressive aspect and not tense or agreement (and ATOM claims that it is tense and agreement
features which pose particular problems for children with SLI).
1.7 Summary of research hypotheses being tested
In this section, you should briefly summarise the set of research hypotheses you have set out to test in your
project in relation to each of the models of SLI which you have chosen to evaluate.
219
However, it should be noted that Ullman and Gopnik (1994) maintain that in spite of their inability to form
implicit (i.e. subconsciously internalised) rules, SLI children who have undergone extensive therapy may learn an
explicit (i.e. consciously memorised) rule taught to them by therapists or teachers to the effect that (e.g.) ‘A regular
verb used in a past tense context ends in -d’. This means that the production of over-regularised forms does not
necessarily ‘prove’ that a child has developed a subconsciously internalised past tense formation rule.
95
2. Research results (about 20% of the overall study)
2.1 Introduction
Begin by noting that in this chapter/section you are setting out to test the predictions made in your chosen
theories of SLI by examining the types of structure produced by JC in contexts where adults would use a
progressive ing-form, a past-tense form or a third-person-singular present-tense form of a main verb. You
will look at the range of verb-forms he uses in contexts where adults would use a progressive ing-form of
a main verb in §2.2, a past-tense main-verb form in §2.3, and a present-tense main-verb s-form §2.4. You
will then summarise your main research findings in §2.5.
2.2 Forms used by JC in contexts where adults use a progressive main verb
The first thing to do is to draw up a list of (like that in (5) below) of (italicised) verb-forms used by JC in
contexts where an adult would use a progressive-participle ing-form of a main verb:
(4) List of (italicised) verb-forms used by JC in contexts where adults require progressive {ing}
18. Now, going home cause it melted
28. Me talking (a)bout... 29. How long the song gonna220 be on?
221
30. Why it not being on?
60. Her pretending to being a doctor
77. Her laying
86. I’m cooking something for dinner
88. He making a mess
89. Me making nice and neat
96. Me watching about it
108. This one, he cooking up a hot dog
124. And it’s gonna be all gone
141. What beach you going?
150. Me making a car
161. Now it’s gonna be Matthew cause, cause...
162. It’s gonna be my, cause I’m bigger than him 171. It’s swinging
175. And he sad cause he crying
178. He crying, he crying
186. I’m gonna have to pick it up first
188. He’s digging up dirt
189. Dirt is falling all over him 195. Her standing and her see herself
204. He flying
246. This is gonna be hot dog 247. This is gonna be coffee
284. Her say, ahh it’s moving
302. Her eating a cookie
317. How he gonna eat me up?
318. He grabbing a lollipop
320. What’s I talking about?
322. Then, I eating food
325. But a bee eating it
333. We making books
377. He taking a lollipop
378. He gonna bite you
You can then use this list to calculate JC’s suppliance rate for progressive {ing} in accordance with the
following procedure:
(5)
Percent suppliance rate for = number of tokens of progressive ing-forms produced by JC
progressive {ing}
number of obligatory contexts for progressive {ing}222
If (for illustrative purposes) we were to do this calculation for the first 10 utterances in the above list, JC’s
suppliance rate for progressive {ing} would be 100% (based on the fact that he uses 10 progressive
ing-forms in 10 obligatory contexts).
2.3 Forms used by JC in contexts where adults use a past-tense main verb
In this section, you should follow the procedures set out in sections 2.1, 2.2 and the first part of 2.4
(relating to how frequently he marks past tense) of chapter 8.
2.4 Forms used by JC in contexts where adults use a third person singular present-tense main verb
In this section, you should follow the procedures set out in section 2.4 of chapter 9.
220
Note that gonna is a contracted form of going to (extremely common in colloquial American English), and so
should be treated as (containing) an ing-form.
221
This presumably corresponds to the adult form Why’s it not being on? What’s odd about this is that adults don’t
generally use the verb BE in a progressive ing-forms unless it denotes an activity (e.g. He’s being naughty again).
You could of course decide to exclude this utterance from your table, if you think it is of indeterminate status.
222
I.e. the number of times an adult would have used a progressive ing-form in the (adult counterpart of the) relevant
sentences.
96
2.5 Summary of your research results
In this section, you should present a summary of your main research results. For example, on the basis of
the small subset of data for which I did the relevant calculations in section 2.2 of this chapter, section 2.2
of chapter 8 and section 2.4 of chapter 9, I arrived at the conclusion that JC’s rate of supplying correct
forms of main verbs in obligatory contexts is 100% for progressive ing-forms, 0% for third person
singular present tense s-forms, 45% for regular past tense forms, and 27% for irregular past tense forms.
(Of course, since my calculations were done for illustrative purposes on only a small subset of the relevant
data, your results may be quite different from mine.)
3. Analysis of research results (around 40% of your overall study)
3.1 Introduction
Begin with an introduction saying that what you are aiming to do in this section/chapter is evaluate the
extent to which the research findings that you reported in section 2.5 are consistent (or inconsistent) with
the research hypotheses which you outlined in chapter 1, and (more generally) with the particular models
of SLI which you are aiming to test. Since the main aim of your research is to use your research results to
evaluate a number of different theories of SLI, you’re going to organise the material in this chapter into
sections is by theory, discussing the extent to which your research findings are compatible with the
Perceptual Deficit model in section 3.2, with the Rule Deficit model in section 3.3, with the Agreement
Deficit model in section 3.4, and with the ATOM model in section 3.5.
3.2 Evaluation of the Perceptual Deficit model
As noted in §1.3 in this chapter, the Perceptual Deficit model predicts that progressive {ing} should pose
comparatively few perceptual problems for an SLI child like JC, since it contains a vowel (albeit a short,
unstressed |I| vowel) – and my partial score223 of 100% correct use of progessive {ing} in obligatory
contexts by JC would be consistent with this prediction. By contrast, present-tense {s} would be expected
to pose far greater perceptual problems, for two reasons. Firstly, its two most frequent allomorphs |s| and
|z| are purely consonantal (the vowel-containing allomorph |Iz| being restricted to use with stems ending in
a sibilant consonant). And secondly, the fact that {s} has three distinct allomorphs (each subject to
complex phonological restrictions on its use) arguably increases the perceptual problems it poses – at
least, if we assume that (all other things being equal) a morpheme which has multiple allomorphs poses
greater perceptual problems than one which has a single allomorph. I leave you the task of thinking
through for yourself whether JC’s past tense marking is consistent with the Perceptual Deficit model: for
some relevant discussion, see section 3.2 of chapter 8. You might like to bear in mind the remarks made in
section 3.2 of chapter 9 – and in particular, the following paragraph:
Suppose that learning to use a morpheme in appropriate contexts involves a three-stage process –
first perceiving the relevant form, then parsing it (= working out the grammatical features it
carries), then practising using it: and suppose that all three stages take considerably longer (e.g.
twice as long) in SLI children as in normally developing children. We would then expect to find
three types of effect. One is a perceptual effect (of the type that Leonard discusses), so that (all
things being equal) we’d expect children to be better at vowel-containing forms than at purely
consonantal forms. A second effect we’d expect to find is a frequency effect, in that children at an
immature stage of development would be expected to perform better on forms which occur more
frequently in their speech input than those which occur less frequently (because all three stages in
learning how to use a morpheme require extensive exposure to the morpheme, and a child at a
given stage of development will have had more exposure to high-frequency than low-frequency
forms). The third effect we’d expect to find is a complexity effect, in that analysing the
grammatical properties of a morpheme is going to take longer if the morpheme is a complex one
encoding several features than if it is a simple one encoding a single feature.
I shall use the informal term partial score to mean ‘the score I arrived at on the basis of the partial set of data
which I looked at in the relevant section’.
223
97
I leave you to ponder on the possible significance of the above paragraph (and the wider discussion in
section 3.2 of chapter 9) for how you evaluate the extent to which your research results are compatible
with the Perceptual Deficit model. If you are looking for a possible frequency effect, you might like to
bear in mind the results I obtained from looking at child-directed parental speech on 20 of the Abe files on
the CHILDES data-base: in around 3 hours of conversation with him, Abe’s mother and father between
them produced 144 tokens of progressive {ing} on main verbs (including 15 tokens of gonna), 91 tokens
of 3SgPres {s} on main verbs, and 61 tokens of the regular past-tense affix {d} – along with 86 tokens of
irregular past-tense main verbs (including 13 tokens of fell, 10 of went, 8 of ate, 7 of made and 6 of said).
Now, while these figures are clearly based on a very small sample, they at least give us some idea of the
likely relative frequency of particular inflected forms in parental speech to children.
3.3 Evaluation of the Rule Deficit model
The Rule Deficit model claims that SLI children are unable to acquire regular morphological rules (like
the present-tense s-rule, the past tense d-rule and the progressive ing-rule in (1) in §1.2), and that the only
way they can acquire inflected forms (including s-forms) is by memorisation. Since memorisation (and the
ability to retrieve a memorised form) improves with increased exposure, we’d expect to find a correlation
between the suppliance rate for individual inflected word-forms and the frequency of the relevant form
(e.g. that an SLI child at a given stage of development might achieve 80% suppliance of a high-frequency
form, 50% suppliance of a medium-frequency form, and 20% suppliance of a low-frequency form). A
second frequency effect predicted by Gopnik and Crago is SLI children will have particular problems in
acquiring an individual inflected form that is less frequent than the corresponding bare form: for example,
could it be that one reason why JC uses the bare form like in contexts where adults use the inflected (third
person singular present tense) s-form likes is that like is 33 times more frequent than likes? So, the first
thing you need to do is to look at is the extent to which you find any evidence of frequency effects. For
example, could the reason why JC achieves 100% suppliance of progressive {ing} and 0% suppliance of
3SgPres {s} be that the ing-forms he uses have a much higher frequency of occurrence in adult English
than the s-forms of verbs he uses in 3SgPres contexts? To answer this question, you could go to an
American English frequency dictionary like Carroll (1971), and look at the frequency of occurrence of
each of the progressive ing-forms JC produces, and of the present-tense s-forms of each of the verb that JC
uses in a 3SgPres context224. If there is no such frequency effect (i.e. if it simply isn’t the case that the
relevant ing-forms are far more frequent than the relevant s-forms), you might want to consider whether
JC’s performance on ing-forms and s-forms of main verbs could be argued to be more consistent with an
alternative rule-based account which supposes that he has acquired the regular progressive ing-rule, but
has not acquired the regular 3SgPres s-rule. In other words, JC may have acquired the defective (i.e.
incomplete) set of regular morphological rules shown below:
(6)
A regular verb carries the suffix
{d} if past225
{ing} if progressive or gerund
{ø} otherwise a defective rule system such as the following:
If JC has indeed acquired the progressive ing-rule, this would falsify Gopnik and Crago’s claim that SLI
children are incapable of acquiring any regular morphological rules. But if JC has not acquired the
3SgPres s-rule, it would suggest that we should revise the Rule Deficit theory so as to hypothesise that SLI
224
Carroll gives the following frequency figures (per million words) for relevant -ing forms (and the corresponding
bare forms) in adult English: going = 2832 (go = 5388); being = 2092 (be = 60852); making = 1408 (make = 8333);
doing = 927 (do = 12695); moving = 871 (move = 1592); taking = 719 (take = 4089); talking = 613 (talk = 1133);
standing = 575 (stand = 1081); saying = 524 (say = 3916); flying = 508 (fly = 785); watching = 443 (watch = 969);
eating = 419 (eat = 1616); leaving = 301 (leave = ???); falling = 239 (fall = 824); cooking = 212 (cook = 265); crying
= 164 (cry = 327); digging = 124 (dig = 181); swinging = 108 (swing = 189); laying = 54 (lay = 54); pretending = 46
(pretend = 115); grabbing = 13 (grab = 73). The corresponding frequency figures (per million words) for s-forms of
verbs used by JC in adult English are given in the previous chapter.
225
There are insufficient relevant examples in the corpus to be sure whether JC makes productive use of regular
perfect and passive participle d-forms, so I have not included these here.
98
children are slow rather than unable to acquire regular morphological rules, and may go through a stage
where they have acquired some but not all regular morphological rules (so that e.g. JC is at a transitional
phase where he has acquired the ing-rule but not the s-rule).
You also need to discuss how to account for JC’s performance on main verbs in past tense contexts. The
partial results I reported earlier suggest that JC sporadically marks past tense in obligatory contexts
(around a third of the time). One possible account of this might be that he has not yet acquired the regular
past tense d-rule, but has memorised individual past tense forms, and sometimes manages to retrieve them
(but sometimes doesn’t, instead using a bare form). If so, we’d expect to find a frequency effect for both
regular and irregular past tense forms. But an alternative possibility is that he has indeed acquired the
regular past-tense d-rule, but is in the early stages of learning to use it and so (since learning to use a rule
requires extensive practice) only manages to apply it some of the time (using a bare verb-form by default
when he fails to apply the rule): irregular verbs would be learned by memorisation on this account. If he
has indeed acquired the regular past tense d-rule (and is in the early stages of learning to apply it), we’d
expect to find a frequency effect for irregular past tense verb forms, but not for regulars. You should look
at the discussion in section 2.3 of chapter 8 for how to test for frequency effects. You might also want to
look at whether JC performs better on regular than irregular pasts, and the possible implications of your
findings. You also need to discuss what these two alternative accounts of his performance on past-tense
main verbs (relating to whether he has or hasn’t acquired the past tense d-rule) would predict about
whether he will or won’t produce over-regularised or novel past tense forms, and whether you have any
evidence from the JC corpus which you can use to test this prediction.
3.4 Evaluation of the Agreement Deficit model
The core assumption of the Agreement Deficit model is that SLI children have problems in acquiring
uninterpretable agreement features (but not in acquiring interpretable features). The research hypothesis
which this gives rise to is that SLI children are expected to perform well on inflected forms which encode
only interpretable features (like progressive forms, and past-tense forms), but much worse on forms which
encode uninterpretable agreement features (like present-tense s-forms). My own partial results (to the
effect that JC shows 100% suppliance of progressive forms, 45%/27% suppliance of correct regular/
irregular past-tense forms, and 0% suppliance of 3SgPres s-forms) appear to be partly consistent with
what the model predicts, but raise the question of why JC should perform so much better on {ing} than on
past tense forms? Can you think of possible reasons for why this might be?
3.5 Evaluation of the ATOM model
The ATOM model claims that SLI children have particular problems in marking tense and agreement, and
sometimes omit either or both of these features on verbs in finite contexts (i.e. in contexts where adults
would use a verb which is marked for both tense and agreement). ATOM also assumes that case is tied up
with agreement, and that SLI children use correctly case-marked forms when they mark agreement, and
default (accusative) forms when they fail to mark agreement.
Since progressive {ing} marks neither tense nor agreement, ATOM seemingly predicts that SLI children
should perform well on this morpheme – and my partial score of 100% suppliance of progressive {ing} by
JC in obligatory contexts would seem to bear out this prediction (though leaves unanswered the important
theoretical question of why SLI children should have problems marking tense but not aspect, since both
are interpretable features. Can you think of reasons why this should be the case?)
In relation to testing the predictions which ATOM makes about how SLI children treat main verbs (and
their subjects) in 3SgPres contexts, see the discussion of main verbs in section 3.5 of chapter 9.
On the question of how to test the predictions made by ATOM about how SLI children treat main verbs
(and their subjects) in past-tense contexts, see the discussion in sections 2.4 and 3.6 of chapter 8.
99
Chapter 4. Summary and Conclusions (about 15% of your overall study)
4.1 Summary of research hypotheses
Begin this chapter with a section in which you provide a brief summary of the models of SLI and research
hypotheses which you outlined in chapter 1.
4.2 Summary of research findings
In this section, briefly summarise the main research findings you obtained in chapter 2
4.3 Summary of evaluations
In this section, provide a brief summary of your evaluation (in chapter 3) of the relative strengths and
weaknesses of each of the theories of SLI you are testing, viewed from the perspective of how well they
account for your own research findings.
4.4 Overall conclusions
In this (the most important) section of your chapter, you need to give your considered final verdict on
which of the models you have looked at best accounts for your own research findings. If a particular
model doesn’t account for (some of) your findings, can it be modified in some way which would make it
compatible with your findings? If not, and if there is no one model of SLI which accounts for all your
research findings, does this take you towards a composite model which sees SLI as involving multiple
impairments, rather than a single type of impairment? (For my suggestions about what such a composite
model might look like, since section 4.4 in the previous chapter.) To what extent (and precisely how)
would such a composite model account for your research findings? Be particularly circumspect with your
conclusions: these are usually the weakest point in student work, and could devalue your overall project.
5. Appendices
After your conclusions, include as an Appendix any lists of sentences which provide the raw data from
which you computed the figures for various tables included in the main body of your project
6. List of references
After any Appendices, include a list of works (arranged alphabetically by author and date) which you have
cited in your project. Use the standard author-date system to refer to relevant works in the main body of
your text (e.g. set out in the same way as the references in section 12 of this coursebook). See the
departmental handbook on how to set out references.
100
11: Empirical study III: Acquisition of subject case-marking in English
In this chapter, I look at how to undertake an empirical study of The acquisition of the case-marking of
subjects in English by a 4-year-old child with SLI. Your study might be organised as follows.
1. Background (about 25% of your overall study)
1.1 Introduction
Give brief details (in a paragraph or two) of the kind of developmental disorder which SLI involves
(summarised on p.4 of this book: see Bishop (1997), Leonard (1998), and the collection of papers in
Bishop and Leonard (2000) and Levy and Schaeffer (2003) for more detailed discussion of diagnostic
criteria for SLI), pointing out that various aspects of grammar are reported to be impaired in SLI children,
including case-marking (in children under 7-8 years of age, at least). But the pattern of case errors
reported in the SLI literature is far from random. For example, Loeb and Leonard (1991) and Bishop
(1994) reported that SLI children almost invariably assign correct (accusative) case to objects, but make
errors with the case-marking of subjects (using accusatives in contexts where adults require nominatives).
Work by Schütze and Wexler (e.g. Wexler 1994; Schütze and Wexler 1996; Schütze 1997; Wexler 1999)
on normally developing children and by Wexler, Schütze and Rice (1998) on SLI children has argued that
the subject case-errors made by normally developing and SLI children alike are the consequence of
children going through an Optional Infinitive/OI stage (or, in the case of SLI children, an Extended
Optional Infinitive/EOI stage) during which they sometimes fail to mark either tense or agreement on
verbs in finite contexts. In your project, you are aiming to test two different accounts of subject case errors
developed within the framework of the (E)OI model – one (outlined in Bromberg and Wexler 1995)
which sees subject case errors as the consequence of a tense deficit, and the other (outlined in Wexler,
Schütze and Rice 1998) which sees subject case errors as the consequence of an agreement deficit. Your
aim is to undertake a small-scale empirical study designed to evaluate these two accounts of subject case
errors, by looking at the case-marking of subjects in nominative contexts by a four-year-old American boy
with SLI, known as JC. Provide brief details of JC’s background and the transcripts you are using (which
you can find at the beginning of chapter 13). You’re going to begin by briefly looking at subject casemarking in unimpaired (adult) grammars in the next section.
1.2 Case-marking of subjects in unimpaired (adult) grammars
Sabine Iatridou (1993) argues that there is parametric variation between languages with respect to how
nominative case-marking works. In languages like Modern Greek (she argues), nominative case is
correlated with tense, in the sense that the subject of a tensed verb (i.e. one morphologically marked for
present or past etc. tense) is assigned nominative case. By contrast, in languages like Classical Greek,
nominative case is correlated with agreement, in the sense that the subject of an agreeing verb (i.e. one
which agrees in person and number with the subject) is assigned nominative case.
Iatridou’s research raises the question of whether tense or agreement is responsible for nominative case
assignment in English. In principle, either analysis would appear to work for English, because finite verbs
with nominative subjects in English typically inflect for both tense and agreement: for example, in He is
lying the auxiliary is has a nominative subject he, but because is carries both (present) tense and (third
person singular) agreement features, it could in principle be either tense or agreement (or a combination of
the two) which is responsible for assignment of nominative case to the subject he.
An additional complication is that when a subject does not fall within the domain of any case assigner, it
is assigned default case – this being accusative in English. Hence, in a so-called ‘Mad magazine’ sentence
like Me be disloyal to the party? Never! the subject pronoun me cannot be assigned nominative case
because the verb be is an infinitive form which carries neither tense nor agreement. Accordingly, the
subject is assigned accusative case by default (i.e. by virtue of being in a position where it would
otherwise not be assigned any case at all). For more detailed discussion, see chapter 6.
101
1.3 The Extended Optional Infinitives model of SLI
Wexler (1994) argued that normally developing children go through an Optional Infinitive/OI stage
(which they usually grow out of by the time they reach 4 years of age) during which they alternate
between producing inflected finite verbs and bare infinitives in finite contexts (i.e. in contexts where
adults would use a finite verb inflected for tense and agreement). Hence, in reply to a past tense question
like What did Daddy do yesterday? a child at the OI stage would alternate between using a finite verb like
played (as in Daddy played tennis) and an uninflected infinitive like play (as in Daddy play tennis). In
early version of the OI model, Wexler argued that children’s bare infinitives (in finite contexts) arise
because children sometimes fail to mark tense on verbs in obligatory contexts. Bromberg and Wexler
(1995, p.223) maintain that tense is responsible for nominative case assignment in English: if so, the
subject of a tense-underspecified verb will not be able to be assigned nominative case (because only a
tensed verb can assign nominative case to its subject), and so will be assigned accusative case by default
instead. This predicts that (in finite contexts) children at the OI stage will alternate between producing
structures containing a tensed verb with a nominative subject (as in He played tennis) and structures
containing a tenseless (bare, uninflected, infinitive) verb with a default accusative subject (as in Him play
tennis).
In a revision to the earlier version of the OI model, Schütze and Wexler (1996) argued that the bare
infinitive verb-forms produced by children at the OI stage can result from omission of either tense or
agreement (or both). They posited that agreement is responsible for nominative case assignment in
English, and that (in consequence) default accusative subjects arise when a verb is underspecified for
agreement (so that a child at the OI stage may produce a structure such as Him played tennis if the verb
played is specified for tense but not specified for agreement).
Wexler, Schütze and Rice (1998) argue that SLI children also go through an OI stage, but that this
typically lasts twice as long (until around 8 years of age) as for normally developing children: hence they
refer to it as the Extended Optional Infinitive/EOI stage. They follow Schütze and Wexler (1996) in
positing that bare infinitives in finite contexts arise via underspecification (= omission) of either tense or
agreement (or both), and for this reason, they refer to their model as the Agreement and Tense Omission
Model/ATOM. They too assume that accusative subjects arise from failure to mark agreement on verbs,
resulting in assignment of default accusative case to the subject.
You can conclude this section by saying that what you are setting out to do in your project is test the two
different (E)OI accounts of the subject case errors made by SLI children in finite contexts – one which
sees them as resulting from omission of tense-marking on verbs, and the other which sees them as
resulting from omission of agreement. In the next section, you’re going to look at the predictions made by
the tense-deficit account of the (E)OI stage about the case-marking of subjects in a range of different types
of finite context, and in the following section you’re going to look at the predictions made by the
agreement-and-tense deficit (ATOM) model.
1.4 Patterns of subject case-marking predicted by the Tense Deficit model
Under the tense-deficit account, we expect a child at the (E)OI stage to produce two different types of
sentence structure in contexts where adults use a past-tense main verb (e.g. Daddy played with me). One
possibility would be for the child to correctly mark the verb for past tense and (because tense triggers
nominative case assignment) assign nominative case to the subject (resulting in He played tennis); the
other would be for the child to fail to mark tense on the verb, so resulting in in structures like Him play
tennis (with the verb not carrying the past tense suffix {d} because it is tenseless, and the subject carrying
default accusative case because a tenseless verb cannot assign nominative case to its subject). Thus, the
child is predicted to produce sentences like He played tennis and Him play tennis (but not *Him played
tennis or *He play tennis) in such contexts.
Now consider what would be expected to happen in contexts where adults would produce a sentence
containing a third-person-singular present-tense (3SgPres) main verb (e.g. Daddy plays tennis). If tense
and agreement are marked (and the 3SgPres affix {s} has been acquired), the child will be predicted to use
a nominative subject (because tense in marked) and an s-inflected verb (e.g. He plays tennis). But if tense
102
is not marked, the verb will lack the {s} affix (because this can only be used when both tense and
agreement are marked) and the subject will be assigned accusative case by default (because only a tensed
verb has a nominative subject) – and so the child will say Him play tennis. Thus in 3SgPres contexts, the
child is predicted to say He plays tennis and Him play tennis, but not *Him plays tennis or *He play
tennis.
In contexts where adults would produce a bare present-tense main verb with a non-3Sg subject (i.e. a
subject which is not third person singular), children at the (E)OI stage are expected either to mark tense
and use a nominative subject (as in I play tennis), or fail to mark tense and use a default accusative subject
(as in Me play tennis). Of course, tense-marking in such cases would not be directly visible on the verb.
Now consider what is expected to happen in contexts where an SLI child produces a structure containing a
finite auxiliary (like am/are/is/was/were/does/do/did/has/have/had/will/would/can/could etc.). Since
auxiliaries are all tensed (e.g. can is a present-tense form and could a past-tense form), and tensed verbs
require a nominative subject, the prediction is that all auxiliaries used in finite contexts will have
nominative subjects (hence an SLI child is predcted to say I can swim and not *Me can swim).
Finally, consider what is expected to happen in auxiliary-omission structures (i.e. structures in which
children have omitted a finite auxiliary which would be obligatory in the corresponding adult sentence). If
(as claimed by Bromberg and Wexler 1995, p.223) auxiliary omission is the consequence of tense
omission, such sentences will be tenseless and so have accusative subjects by default – so that children are
predicted to say e.g. Him playing tennis but not *He playing tennis.
1.5 Patterns of subject case-marking in finite contexts predicted by the ATOM model
The agreement-and-tense deficit account assumed in the ATOM model makes subtly different predictions
(from those made by the tense deficit model, outlined in §1.4) about the case-marking of subjects by
children at the (E)OI stage. Since the ATOM predictions are outlined at length in chapter 6 of this
coursebook, I will not repeat the relevant material here.
1.6 Research hypotheses being tested
In this final section, you should summarise succinctly (perhaps in a neat tabular form) the range of
structures which children at the (E)OI stage are predicted to produce and not produce – and highlight the
differences in predictions made by the two accounts of the (E)OI stage which you are focussing on.
2. Research results (about 25% of the overall study)
2.1 Introduction
Begin by noting that in this chapter/section you are setting out to test the predictions made by the two
alternative (E)OI accounts of subject case-marking which you outlined in §1.4 and §1.5. You are going to
do this by looking at how JC case-marks subjects in a variety of different types of finite context. Because
you are concerned with his use of nominative and accusative subjects, you are only going to include
structures containing an unambiguously nominative pronoun like I/we/he/she/they or an unambiguously
accusative pronoun like me/use/him/her226/them (I’ll call such pronouns unambiguous). You are only
going to be concerned with overt subjects (not null subjects), for the obvious reason that you are looking
at unambiguous subject pronouns, and there is no way of knowing what case (if any) null subjects have. In
the various sections below, you are going to look at how JC case-marks (unambiguous) subjects in a range
of different types of structure which he produces in finite contexts.
226
In principle, her is ambiguous between being an accusative or genitive pronoun, but since JC makes extensive use
of accusative subjects (but not genitive subjects), we can assume that her subjects are accusative.
103
2.2 JCs case-marking of subjects with agreeing tense-marked auxiliaries
In (1) below, I have listed all the examples which occur in the JC corpus of an unambiguous pronoun
which occurs as the subject of an auxiliary227 (like am/are/is/was) which is overtly inflected for both tense
and agreement (in that e.g. (a)m is a first person singular present tense form of BE). The relevant pronoun
is italicised in each case.
(1)
I
he
they
List of unambiguous subject JC uses with auxiliaries inflected for tense and agreement
86. I’m cooking something for dinner 162. Cause I’m bigger than him 164. I’m bigger than him
186. I’m gonna have to pick it up first 201. I’m not scare of them 363. I’m gonna be so proud
320. What’s228 I talking about?
56. He’s sick 155. He’s funny 173. He’s sad cause he can’t go 188. He’s digging up dirt
59. But, he’s229 not have no money
165. (Th)ey are straps
On the basis of the data in table (1), you can then calculate the percentage of nominative and accusative
subjects which JC uses with agreeing auxiliaries – a straightforward task.
2.3 JCs case-marking of subjects with ambiguous auxiliaries
In (2) below, I’ve listed all examples of unambiguous pronouns used as subjects of what Wexler, Schütze
and Rice (1998) call ambiguous auxiliaries (e.g. can/can’t/could/don’t230/didn’t) – that is, auxiliaries
which are clearly marked for tense, but where it’s impossible to tell whether they are marked for
agreement (and hence agree invisibly with their subject) or not.
(2)
I
me
we
he
him
her
List of unambiguous subject pronouns used by JC with (underlined) ambiguous auxiliaries
1. After that I can do that 151. I can make a big big man 194. Ah, so happy I can get out of them
fish 203. I can make see my shadow 225. I can fall 167. I try to take it off, but I can’t put on
168. I can’t put this on 15. I don’t wanna play with that 19. I don’t know 61. I don’t have a
doctor set 129. I don’t know 240. I don’t know where her can cook 243. I don’t know what he
saying 248. I don’t, I like tea 222. I’n231 see that one, cause that one is very good
10. Me can’t back home to go shopping 70. Me can have this 22. Me don’t know
84. Me don’t have a cat on a bed
192. We can do all over again
85. He can play with that 181. He can fly 182. He can do it 183. He can fly cause he got these
wings 184. He can pretend fly 351. He can eat it 173. He’s sad cause he can't go 176. And he
can’t go and he family... die 115. No, he could burn heself 63. He don’t have this jacket 294. Why
he don’t have a nose?
11. Why him don’t have eyes?
147. Then her can make a pretty one 148. No, her can put up here 238. Her can cook something
240. I don’t know where her can cook
On the basis of the list in (2), you can calculate the percentage of nominative and accusative subjects
which JC uses with ambiguous auxiliaries.
227
This term (as used here) includes all uses of BE, since (in all uses) BE behaves like an auxiliary in e.g.
undergoing inversion in questions (cf. Are you ready?) and being negated without the use of DO (cf. He isn’t ready).
228
The form ’s here is arguably a contracted form of was (cf. What was I talking about?)
229
The form ’s here is arguably a contracted form of does.
230
Bear in mind that in colloquial American English, don’t is used with all types of subject, including third person
singular subjects (so that in numerous pop-songs we find the line He don’t love me no more, rather than the more
formal style He doesn’t love me any more). JC doesn’t produce doesn’t at all in the corpus.
231
Taken by the person who did the transcription to be a contracted form of didn’t.
104
2.4 JCs case-marking of subjects in null-auxiliary structures
In (3) below, I have listed all the unambiguous-subject structures produced by JC whose adult counterpart
would contain a finite auxiliary, but which JC produces without any overt auxiliary232:
List of unambiguous subject pronouns use by JC in null-auxiliary structures233
166. I got on my shirt and have trouble doing my back 220. I got a new Barney 223. I got a
different Barney 245. I got a train 334. But, I got another one 365. I got my hair from barber
322. Then, I eating food 95. I hear myself again? 201. When I be bigger234, I’m not scare of them
8. Me bigger than Gio and him 9. Me bigger him 46. Me too tired 28. Me talking bout...
me
89. Me making nice and neat 96. Me watching about it 150. Me making a car
237. Me making hot dogs 33. Me said, me gotta hurry up and go up
271. Me got Ms. Peggy speech teacher
286. And then it scared mom, so we gonna put him to trouble 333. We making books
we
252. But we got two more 272. We got a lotta fish for everyone 369. We got some new pictures
17. He happy 76. He under the table 175. And he sad cause he crying
he
88. He making a mess 108. This one, he cooking up a hot dog 178. He crying, he crying
204. He flying 243. I don’t know what he saying 287. And then he be trouble235 317. How he
gonna eat me up? 318. He grabbing a lollipop 377. He taking a lollipop 378. He gonna bite you
54. But he not there 217. He not real 183. He can fly cause he got these wings 221. He got old one
60. Her pretending to being a doctor 77. Her laying 195. Her standing and her see herself
her
302. Her eating a cookie 149. Her sad 57. Nuh uh, This girl is...her not 145. Why her need this?
them 374. Why them both have pinchers?
(3)
I
On the basis of the data in the list in (3), you can calculate the percentage of nominative and accusative
subjects which JC uses in null-auxiliary structures.
2.5 JCs case-marking of subjects in past-tense main-verb contexts
In (4) below, I have listed all the unambiguous-subject structures produced by JC in contexts where adults
would use an (underlined) verb in the past tense form.
(4)
I
me
we
he
232
List of unambiguous subject pronouns used by JC in past-tense main-verb contexts
198. Long time ago I have a big eye 230. I see clown at a post office 236. I never saw one of these
stove 321. Long time ago, I go camp and hiking at the same time 323. And I eat at hiking
339. I forgot
16. Me used to have a dog, but somebody take it away 23. Me got one from Michael
27. Me got chicken pox 32. Me fall asleep on the couch 33. Me said, me gotta hurry up and go up
81. Like me see on the TV 103. Me just jumpovered it 327. Then me said, oh!
328. Then me take it xxx 329. Then me go camping 330. Me eat all our food
229. We see clown at...umm... 283. And then her put on the floor, and we scare her
324. And then at hiking, I xxx a picnic, so we eat
5. He see snow on he chimney 105. He shoveled someone else 112. He burn heself here 113. He
eat it 114. No, took it off of...then he eat it 125. I think this, he...umm, oh, he dump it 126. He
shoveled him truck 132. And then he drived away 127. And then he dump it 130. He shoveled
him truck 131. Then he dump it 133. Because, he want to put it 144. Then, he knocked him
window 174. He family, he lost he family 179. He lost him duck 193. Oh, he caught me 211.
And then, he scare them and said ha, ha, ha 213. Then, he talk to himself 232. He give a plane
balloon 281. He say, put it down 298. I know, he wanned to go and drop it 300. He jumped out
301. He bit me 316. He try to eat me up 379. He cut me
Recall that it is being assumed here that BE is an auxiliary in all its uses, for reasons given earlier.
The idea behind this term is that in contexts where an adult would use an auxiliary structure like He’s working
(which contains the overt auxiliary (i)s), JC may produce a structure in which the auxiliary is null (i.e. ‘silent’)
234
There may be a missing will auxiliary here – though you may prefer to exclude the example as indeterminate.
235
The missing auxiliary here is presumably will.
233
105
her
them
268. Her make it for Ms Peggy a long time ago 277. Her give me dad a lobster, a two lobster
282. And then her say, Ahhh! 283. And then her put on the floor, and we scare her 284. Her say,
ahh it's moving 303. Then, her drink some water 311. Her have a snake in the sink 314. Then
her got hurt
231. Them have a party, and clown give me a balloon 285. And then, them cook them up 366. But
them cut my hair real tiny
On the basis of the data in the list in (4), you can calculate the percentage of nominative and accusative
subjects which JC uses in past-tense main-verb contexts (i) with tensed verbs, and (ii) with tenseless (bare)
infinitive verbs. (Of course, you will have to exclude verb-forms which are ambiguous between being past
tense forms and infinitive forms from these two counts.)
2.6 JCs case-marking of subjects in present-tense main-verb contexts
In (5) below, I have listed all the unambiguous-subject structures produced by JC in contexts where adults
would use an (underlined) main verb in an appropriate present-tense form.
(9)
I
JC’s use of unambiguous subject pronouns in present tense main verb structures
6. When I go iceskate me fall. 37. I just have a garbage truck, not a monster 92. I make some
peppers 93. I win 119. I eat a lot of ketchup 153. I see her and you 163. I have to do it all
by myself, cause, cause... 167. I try to take it off, but I can't put on 199. I think it's a ghost
202. I see he shadow 219. I have a new Barney 227. I wanna play some games 234. I wanna
play a stove 235. I like it, cause, cause, I wanna play 248. I don't, I like tea 249. I like warm tea
250. I have tea at boy scout 257. I put this and this
265. I remember 295. Ah I remember this 296. I remember that thing 298. I know, he wanned
to go and drop it 323. I say, and I eat at hiking 331. I mean this much
338. Now I do all by myself 359. Then, I grab my one 361. Then, I eat it after lunch
362. I need ten stickers 364. I know her 371. I know these ones
2. When me go outside to play me go like that 3. Me go like everywhere 6. When I go iceskate
me
me fall 7. Me cry 12. Me never have them real big 13. Me never take a shower
14. Then, me no have to go bath 21. Me have different puzzle 35. Come up there, then me sleep
up there 36. Me fall asleep up there 47. Me remember some of that 49. Me wanna hurt somebody
50. Me wanna point it and hurt somebody 74. Me know how to do it 91. Me put some pepper here
118. Me like ketchup 142. Me go beach not far away 143. Hey, me go faraway beach 152. Me
see her and you 270. Me know, but my...me get two cause...
295. Me remember 310. Me remember her house snake 319. Me remember at school...
350. We just give to Easter Bunny
we
355. We take one at a time
64. He only have a coat 65. Now he have them two 67. He have a doctor 156. He have he hats
he
on 352. He hafta 55. He like Danny talking like that 116. When he hold here
356. He put in a box with lots different and jelly bean 357. Then he bring it
44. When him crack tiny pieces up, and then put xxx 376. He...him go first
him
195. Her standing and her see herself 349. Her use with a green cutter thing
her
370. Her need the camera to put in a locker
137. That why them put a lot of sand 138. Them put all over 373. Them both have pinchers
them
On the basis of the data in the list in (4), you can calculate the percentage of nominative and accusative
subjects which JC uses in present-tense main-verb contexts. Although one would normally compute
separate figures for verbs with 3Sg and non-3Sg subjects, there seems to be little point in the case of JC
because (as discussed in chapter 9) he has a morphological deficit and has not acquired the 3SgPres
morpheme {s}, with the consequence that all his present tense verb-forms are bare – even those with 3Sg
subjects.
106
2.7 Summary of research results
In this section, you should succinctly summarise your research results – perhaps in tabular form (though if
you do this, make sure you explain what each part of the the table represents).
3. Analysis of research results (around 35% of your overall study)
3.1 Introduction
Begin with an introduction saying that what you are aiming to do in this section/chapter is evaluate the
extent to which the research findings that you reported in section 2.7 are consistent (or inconsistent) with
the tense-deficit account of the (E)OI stage on the one hand, and the ATOM (agreement-and-tense deficit)
account on the other. You’ll look at each of your major research findings in a different section below, and
evaluate the extent to which each of the two accounts is (or is not) compatible with your findings. In the
final section, you’ll provide an overall evaluation of which of the two accounts is best able to explain the
overall pattern of case-marking you find in the full range of structures produced by JC in finite contexts.
3.2 Evaluation of accounts of JCs case-marking of subjects with agreeing tensed-marked auxiliaries
Briefly summarise the predictions made by each of the two accounts, remind the reader of the research
results you obtained in §2.2, and provide an evaluation of the extent to which each account is compatible
with your results.
3.3 Evaluation of accounts of JCs case-marking of subjects with ambiguous auxiliaries
Briefly summarise the predictions made by each of the two accounts, remind the reader of the research
results you obtained in §2.3, and provide an evaluation of the extent to which each account is compatible
with your results.
3.4 Evaluation of accounts of JCs case-marking of subjects in null-auxiliary structures
Briefly summarise the predictions made by each of the two accounts, remind the reader of the research
results you obtained in §2.4, and provide an evaluation of the extent to which each account is compatible
with your results.
3.5 Evaluation of accounts of JCs case-marking of subjects in past-tense main-verb contexts
Briefly summarise the predictions made by each of the two accounts, remind the reader of the research
results you obtained in §2.5, and provide an evaluation of the extent to which each account is compatible
with your results.
3.6 Evaluation of accounts of JCs case-marking of subjects in present-tense main-verb contexts
Briefly summarise the predictions made by each of the two accounts, remind the reader of the research
results you obtained in §2.6, and provide an evaluation of the extent to which each account is compatible
with your results.
3.7 Overall evaluation of the two accounts of the (E)OI stage
In this final section, begin by providing an overall evaluation of the extent to which the tensed-based
account of nominative case assignment handles the results you reported in §3.2-§3.6. Then go on to
evaluate the extent to which the agreement-based account of nominative case assignment handles the same
results. Say which account seems to be superior to the other – and in what respects.
107
Chapter 4. Summary and Conclusions (about 15% of your overall study)
4.1 Summary of research hypotheses
Begin this chapter with a section in which you provide a brief summary of the two different versions of
the EOI model of SLI – the Tense Deficit and ATOM (Agreement-and-Tense Deficit) models, and what
each predicts about the case-marking of subjects.
4.2 Summary of research findings
In this section, briefly summarise the main research findings you obtained in chapter 2
4.3 Summary of evaluations
In this section, provide a brief summary of your evaluation (in chapter 3) of the relative strengths and
weaknesses of the Tense Deficit and ATOM accounts of SLI, viewed from the perspective of how well
they account for JC’s case-marking of subjects.
4.4 Overall conclusions
In this (the most important) section of your chapter, you need to give your considered final verdict on
which version of the EOI model best accounts for your own research findings. Does one of the two
accounts handle a much wider range of structures than the other? What possible drawbacks are there to
either or both of the accounts? For example, does JC perform better on some nominative pronouns than
others, and if so what might the sigificance of this be, and how might we account for it – e.g. is the
relative frequency of different nominative pronouns in adult English a possible factor?236 (You will find
some critical reflections on ATOM in my Children’s English coursebook (in chapter 6 on
Underspecification) on http://courses.essex.ac.uk/lg/lg620 (under acquisition coursebook). Be
particularly circumspect with your conclusions: these are usually the weakest point in student work, and
could devalue your overall project.
5. Appendices
After your conclusions, include as an Appendix any lists of sentences which provide the raw data from
which you computed the figures for various tables included in the main body of your project
6. List of references
After any Appendices, include a list of works (arranged alphabetically by author and date) which you have
cited in your project. Use the standard author-date system to refer to relevant works in the main body of
your text (e.g. set out in the same way as the references in section 12 of this coursebook). See the
departmental handbook on how to set out references.
Carroll’s (1971) frequency dictionary shows that (in adult American English), the frequency of unambiguously
nominative pronouns per million words is: he = 46249; they = 27620; I = 25932; we = 16452; she = 13653.
236
108
12: List of references
Abney, S.P. (1987) The English Noun Phrase in Its Sentential Aspect, PhD diss., MIT.
Arthur G (1952) The Arthur Adaptation of the Leiter International Performance Scale, Stoelting,
Chicago
Bernstein Ratner N (1984) ‘Patterns of vowel modification in mother-child speech’, Journal of Child
Language 11: 557-558
Bishop DVM (1994) ‘Grammatical errors in specific language impairment: Competence or performance
limitations?’, Applied Psycholinguistics 15: 507-550
Bishop, DVM (1997) Uncommon Understanding: Development and Disorders of Language
Comprehension in Children, Pychology Press, Hove
Bishop D & Edmundson A (1987) ‘Specific language impairment as a maturational lag: Evidence from
longitudinal data on language and motor development’, Developmental Medicine and Child Neurology
29: 442-459
Bishop DVM & Leonard LB (eds) Speech and Language Impairments in Children: Causes,
Characteristics, Intervention and Outcome, Psychology Press, Hove.
Brown R (1973) A First Language, Penguin Books, London
Carroll J B (1971) The American Heritage Word Frequency Book, Boston, Houghton Mifflin [library
classmark PE 1449; on restricted 3-day loan]
CHILDES (= Child Language Data Exchange System) http://childes.psy.cmu.edu
Chomsky N (1957) Syntactic Structures, Mouton, The Hague.
Chomsky N (1986) Barriers, MIT Press, Cambridge Mass.
Chomsky N (1995) The Minimalist Program, MIT Press, Cambridge Mass.
Chomsky N (1998) Minimalist Inquiries, MIT Working Papers in Linguistics, no.15
Chomsky N (1999) Derivation by Phase, MIT Working Papers in Linguistics, no. 18.
Chomsky N (2001) Beyond Explanatory Adequacy, unpublished manuscript, MIT.
Chomsky, N. (2004) ‘Three factors in language design’, unpublished paper, MIT.
Clahsen H, Bartke S & Göllner S (1997) ‘Formal features in impaired grammars: A comparison of English
and German SLI children’, Journal of Neurolinguistics, 10: 151-171
Elliott CD, Murray D and Pearson L (1988) British Ability Scales, NFER-Nelson, Windsor
Fee EJ (1995) ‘The phonological system of a specifically language-impaired population’, Clinical
Linguistics and Phonetics, 9: 189-209
Fisher S, Varga-Khadem F, Watkins K, Monaco A and Pembrey M (1998) ‘Localization of a gene
implicated in severe speech and language’, Nature Genetics 5: 11-16.
Freidemann M-A & Rizzi L (eds) The Acquisition of Syntax, Longman, London.
Gleason JB (2001, 5th edition) The Development of Children, Allyn and Bacon, Boston, MA.
Gopnik M (1990) ‘Feature blindness: A case study’, Language Acquisition 1: 139-164
Gopnik M & Crago MB (1991) ‘Familial aggregation of a developmental disorder’, Cognition 39:1-50
Guilfoyle E (1984) ‘The acquisition of tense and the emergence of lexical subjects in child grammars of
English, McGill Working Papers in Linguistics 2: 20-31
Hurst J, Baraitser M, Auger E, Graham F and Norell S ‘An extended family with an inherited speech
disorder’, Developmental Medicine and Child Neurology 32: 347-355.
Iatridou S (1993) ‘On nominative case assignment and a few related things’, MIT Working Papers in
Linguistics 19: 175-196
Johnson C et al (1999) ‘Fourteen-year follow-up of children with and without speech/language
impairments’ Journal of Speech and Hearing Research 42: 744-760
Lai CS, Fisher SE, Hurst JA, Vargha-Khadem JA & Monaco AP (2001) ‘A forkhead domain gene is
mutated in a severe speech and language disorder’, Nature 413: 519-523
Leonard LB (1989) ‘Language learnability and specific language impairment in children’, Applied
Psycholinguistics 10: 179-202
Leonard LB (1992) ‘Specific Language Impairment in three languages: Some cross-linguistic evidence’,
in P Fletcher and D Hall (eds) Specific Speech and Language Disorders in Children, Anetheium Press,
San Diego, chapter 9.
109
Leonard, LB (1995) ‘Functional categories in the grammars of children with Specific Language
Impairment’, Journal of Speech and Hearing Research 38: 1270-1283
Leonard LB, McGregor KK and Allen GD (1992) ‘Grammatical morphology and speech perception in
children with specific language impairment’, Journal of Speech and Hearing Research 35: 1076-1085
Leonard LB (1998) Children with Specific Language Impairment, MIT Press, Cambridge Mass.
Levy Y & Schaeffer J (eds) Language Competence Across Populations: Toward a Definition of Specific
Language Impairment, Erlbaum, Mahwah NJ.
Loeb D and Leonard LB (1991) ‘Subject case-marking and verb morphology in normally developing and
specifically-language-impaired children’, Journal of Speech and Hearing Research 34: 340-346.
MacWhinney B (1993) The CHILDES Database: Second Edition, Discovery Systems, Dublin, OH.
Marcus G, Pinker S, Ullman M, Hollander M. Rosen TJ & Xu F (1992) Overregularisation in Language
Acquisition, Monographs of the Society for Research in Child Development, 57.
Maslen RJC, Theakston AL, Lieven EVM and Tomasello M (2003) ‘A dense corpus study of past tense
and plural overregularization in English’, unpublished paper, University of Manchester
McDaniel D, McKee C & Smith Cairns H (eds) (1998) Methods for Assessing Children's Syntax, MIT
Press, Cambridge Mass ISBN 0-262-63190-3 (pb)
Menn L & Ratner NB (eds) (2000) Methods for Studying Language Production, Erlbaum, London
Menyuk P (1987) ‘Linguistic problems in children with developmental dysphasia’, in Wyke MA (ed)
Developmental Dysphasia, Academic Press, New York
Miller JF et al. (1981) Assessing Language Production in Children, Edward Arnold, London
Panagos, JM & Prelock PA (1982) ‘Phonological constraints on the sentence productions of languagedisordered children’, Journal of Speech and Hearing Research 25: 171-177
Pesetsky, D. (1997) ‘Optimality Theory and Syntax: Movement and Pronunciation’, in D. Archangeli &
D.T. Langendoen (eds) Optimality Theory: An Overview, Blackwell, Oxford, pp. 134-170.
Pesetsky, D. (1998) ‘Some optimality principles of sentence pronunciation’ in P. Barbosa, D. Fox,
P. Hagstrom, M. McGinnis & D. Pesetsky (eds) Is the Best Good Enough? MIT Press, Cambridge, pp.
337-383.
Pinker S (1991) ‘Rules of language’, Science 253: 530-535
Pinker S & Prince A (1988) ‘On language and connectionism: Analysis of a parallel distributed processing
model of language acquisition’, Cognition 8: 73-193
Radford A. (1990) Syntactic Theory and the Acquisition of English Syntax, Blackwell, Oxford
Radford A (2004) English Syntax, Cambridge University Press, Cambridge
Radford A et al (1999) Linguistics: An Introduction, Cambridge University Press, Cambridge
Radford A & Lin Y-A (2005) ‘On the nature of the grammatical deficit in English-speaking children with
Specific Language Impairment’, unpublished paper, University of Essex
Ramos E & Roeper T (1995) ‘Pronoun case assignment by a SLI child’, poster presentation presented to
convention organised by American Speech and Hearing Association
Raposo E (1987) ‘Case theory and Infl-to-Comp: the inflected infinitive in European Portuguese’,
Linguistic Inquiry 18: 85-109
Rice ML & Oetting JB (1993) ‘Morphological deficits of children with SLI: Evaluation of number
marking and agreement’, Journal of Speech and Hearing Research 36: 1249-1257
Rice ML, Noll KR & Grimm H (1997) ‘An Extended Optional Infinitive stage in German-speaking
children with Specific Language Impairment’, Language Acquisition 6: 255-295
Rice ML & Wexler K (1996) ‘Towards Tense as a clinical marker of Specific Language Impairment’,
Journal of Speech and Hearing Research 39: 1239-1257
Rice ML, Wexler K & Cleave P (1995) ‘Specific Language Impairment as a period of Extended Optional
Infinitive’, Journal of Speech and Hearing Research 38: 850-863
Rice ML, Wexler K & Herschberger S (1998) ‘Tense over time: The longitudinal course of Tense
acquisition in children with Specific Language Impairment’, Journal of Speech, Language and Hearing
Research 41: 1412-1431.
Rizzi L (1996) ‘Residual verb second and the WH-criterion’, in A. Belletti & L. Rizzi (eds) Parameters
and Functional Heads, Oxford University Press, Oxford.
Rizzi L (2000) ‘Remarks on early null subjects’ in Friedemann & Rizzi (eds), pp. 269-292.
110
Ross JR (1967) Constraints on Variables in Syntax, PhD diss., MIT (published as Infinite Syntax! by
Ablex Publishing Corporation, Norwood, New Jersey, 1986).
Rouveret A (1980) ‘Sur la notion de proposition finie: gouvernement et inversion’ Recherches
Linguistiques 9: 76-140, Vincennes
Schütze C and Wexler K (1996) ‘Subject case licensing and English root infinitives', in Stringfellow A,
Cahana-Amitay D, Hughes E and Zukowski A (eds) Proceedings of 20th Boston University Conference
on Language Development: Vol. 2, Cascadilla Press, Somerville Mass., pp. 670-681.
Schütze C (2001) ‘On the nature of default case’, Syntax 4: 205-238.
SLI consortium (2002) ‘A genomewide scan identifies two novel loci involved in specifc language
impairment’, American Journal of Human Genetics 70: 384-98.
St. Louis KO and Ruscello DM (1981) Oral Speech Mechanism Screening Examination – Revised,
Pro-Ed, Austin
Swanson L & Leonard LB (1994) ‘Duration of function-word vowels in mothers’ speech to young
children’, Journal of Speech and Hearing Research 37: 1394-1405
Tallal P, Curtiss S and Kaplan R (1980) Evaluation of the outcome of preschool impairments in language
disorders, Progress Report for US National Institutes of Neurological and Communicative Disorders and
Strokes
Tomblin JB & Buckwalter P (1998) ‘Hereditability of poor language achievement among twins’, Journal
of Speech, Language and Hearing Research 41: 188-199.
Tomblin JB, Nishimura C, Zhang X and Murray J (1998) ‘Association of developmental language
impairment with loci at 7q31’, American Journal of Human Genetics 63: A312.
Tsimpli I and Stavrakaki S (1999) ‘The effects of a morphosyntactic deficit in the determiner system;
The case of a Greek SLI child’, Lingua, 108: 31-85.
Ullman M & Gopnik M (1994) ‘The production of inflectional morphology in hereditary specific
language impairment’, McGill Working Papers in Linguistics 10: 81-118.
Van der Lely, H.K.J. (1994). Canonical linking rules: Forward vs. reverse linking in normally developing
and specifically language impaired children. Cognition, 51, 29-72.
Van der Lely, H.K.J. (1998). SLI in children: Movement, economy and deficits in the computationalsyntactic system. Language Acquisition 7: 161-192.
Van der Lely, H.K.J. (2005). Domain-specific cognitive systems: insight from Grammatical-SLI. Trends
in Cognitive Sciences 9: 53-59.
Van der Lely, H.K.J., & Battell, J. (2003). Wh-movement in children with grammatical SLI: A test of the
RDDR hypothesis. Language, 79, 153-181.
Van der Lely, H.K.J., & Stollwerck, L. (1997). Binding theory and specifically language impaired
children. Cognition, 52, 245-290.
Vargha-Khadem F, Watkins K, Alcock K, Fletcher P & Passingham R (1995) ‘Praxic and nonverbal
cognitive deficits in a large family with a genetically transmitted speech and language disorder’,
Proceedings of the National Academy of Sciences 92; 930-933.
Weiner P (1974) ‘A language-delayed child at adolescence’, Journal of Speech and Hearing Disorders
39: 202-212.
Wenzlaff M & Clahsen H (2002) ‘Tense and Agreement in German Agrammatism’, unpublished ms,
University of Essex.
Wexler K (1994) ‘Optional Infinitives, Head Movement and the Economy of Derivations’, in Lightfoot D
& Hornstein N (eds) Verb Movement, Cambridge University Press, pp. 305-350.
Wexler, K. (2003) ‘Lenneberg’s dream: Learning, normal language development, and specific language
impairment’, in Y. Levy and J.C Schaeffer (eds) Language Competence Across Populations, Lawrence
Erlbaum Associated, Mahwah, New Jersey.
Wexler K, Schütze C & Rice M (1998) ‘Subject case in children with SLI and unaffected controls:
Evidence for the Agr/Tns Omission Model, Language Acquisition 7: 317-344.
111
13: Appendix (Transcripts of samples of the language production of
a 4-year-old American boy with SLI)
PROFILE OF THE BOY (KNOWN AS JC)
JC is a white middle-class American boy from Western Massachusetts with white parents who speak
Standard American English. When tested he had an average score in non-verbal intelligence, a moderately
low score in verbal intelligence, moderately low scores in standard (TOLD-P237) tests of comprehension
and expression, average receptive and expressive vocabulary, age-appropriate articulation skills, normal
hearing acuity, and no known neurological or emotional problems. The JC corpus comprises 8 recordings
of his spontaneous speech (each of roughly 10 minutes in duration) at ages 4;3.15, 4;4.20, 4;5.3, 4;5.14,
4;5.16, 4;5.23; 4;6.3, and 4;6.12. Information is provided in parentheses on the contexts in which some of the
utterances produced by JC occurred, together with (in some cases) a gloss which represents the presumed
adult counterpart. xxx marks an unintelligible stretch of speech.
Recording 1: 10/1/94 age 4;3.15
1. After that I can do that
2. When me go outside to play me go like that
3. Me go like everywhere (‘I go like that everywhere’)
4. That me friend (‘That’s my friend’)
5. He see snow on he chimney (‘He saw snow on his chimney’)
6. When I go iceskate me fall (‘When I go ice-skating, I fall)
7. Me cry
8. Me bigger than Gio and him (‘I’m bigger that Gio and him’)
9. Me bigger him (‘I’m bigger than him’)
10. Me can't back home to go shopping
11. Why him don't have eyes? (‘Why don’t238 he have eyes?’)
12. Me never have them real big (‘I never have them really big’)
13. Me never take a shower
14. Then, me no have to go bath (‘Then I won’t have to go for a bath’)
15. I don't wanna239 play with that
16. Me used to have a dog, but somebody take it away
Recording 2: 16/2/94 age 4;4.20
17. He happy (‘He’s happy’)
18. Now, going home cause it melted [describing picture] (‘He’s going home...’)
19. I don't know (reply to ‘What did you tell your mom?’)
20. Not like this (reply to ‘Do you have puzzles at home?’)
21. Me have different puzzle (‘I have a different puzzle’)
22. Me don't know (reply to ‘Who did you give candy to?’)
23. Me got one from Michael
24. Chicken pox, chicken pox
25. No, but chicken pox, mmm, itchy (reply to ‘Did you have chicken pox?’)
26. No, it a long time ago (‘No, it was a long time ago’)
27. Me got chicken pox
28. Me talking bout... (‘I’m talking about...’)
237
Test of Language Development-Primary, taken by children aged 4-8 years of age. This tests understanding and
the use of spoken words, aspects of grammar and pronunciation.
238
In many (American and other) varieties of colloquial English, don’t is an invariable form like won’t, used with all
kinds of subjects including third person singular subjects. Hence, numerous American pop songs contain the lament
‘(S)he don’t love me no more’ (rather than the Standard English counterpart ‘He doesn’t love me any more’).
239
In colloquial (esp. American) English, want to is often pronounced wanna, and wanted as wanned. In effectm the
final –t of want in such forms is dropped.
112
29. How long the song gonna be on? (‘How long is the song gonna be on?’)
30. Why it not being on? (‘Why isn’t it on?’ – referring to tape recorder)
Recording 3: 1/3/94 age 4;5.3
31. Play Nintendo (reply to ‘What did you do?’)
32. Me fall asleep on the couch (‘I fell asleep on the couch’)
33. Me said, me gotta hurry up and go up
34. It's a little bit night time
35. Come up there, then me sleep up there (‘(I) went up there, then I slept up there’)
36. Me fall asleep up there (‘I fell asleep up there’)
37. I just have a garbage truck, not a monster (reply to ‘Do you have a monster truck?’)
38. It don't have a mouth
39. But it a truck you dump dirt, and snow, and other stuff (‘It’s a truck where you...’)
40. You can put everything in there
41. When you dump it
42. Some put sand and bring it to a different stop
43. And put garbage in the machine
44. When him crack tiny pieces up, and then put xxx (‘When he breaks tiny pieces up...’)
45. Then dump into a truck again (‘Then dumps them into a truck again’)
46. Me too tired (‘I’m too tired’)
47. Me remember some of that
48. Look at last time
49. Me wanna hurt somebody [talking about truck again]
50. Me wanna point it and hurt somebody
51. It very sharp (‘It’s very sharp’)
52. It can hurt somebody
53. It can poke somebody
54. But he not there (‘But he’s not there’)
55. He like Danny talking like that (‘He likes Danny talking like that’)
56. He's sick
57. Nuh uh, This girl is...her not (reply to ‘I think this girl is pretending she’s a bus driver’)
58. But, he bus (= ‘her bus’) is over here (reply to ‘Do you think she is a real bus driver?’)
59. But, he's not have no money (‘But she doesn’t have any money’)
60. Her pretending to being a doctor [describing picture]
61. I don't have a doctor set
62. Me brother name Jack (‘My brother’s name’s Jack’)
63. He don't have this jacket
64. He only have a coat (‘He only has a coat’)
65. Now he have them two
66. Not this kind
67. He have a doctor
68. Just a doctor
69. Not a doctor, but...
70. Me can have this
71. All of these can be me (= ‘mine’), and you can find another one for him
72. That can be Giovanni (= ‘Giovanni's’)
73. Got a girl again
74. Me know how to do it
75. Is a dog under a table (‘It’s240 a dog under a table’ – describing a picture)
76. He under the table (‘He’s under the table’)
It’s not clear what is represents here. For example, it may be an attempt as it’s in which the t is dropped in order
to simplify the consonant cluster ts
240
113
77. Her laying (‘She’s lying down’)
78. It's a cat
79. Nuh uh, Baby can just jump out (reply to ‘That’s a baby bed’)
80. You need a baby cage, so you can't get out
81. Like me see on the TV (‘Like I saw on the TV)
82. And that's not for a baby, cause only for a cage
83. Make a baby can go in it (?= ‘They make it so a baby can go in’)
84. Me don't have a cat on a bed [referring to picture]
85. He can play with that
86. I'm cooking something for dinner
87. You have to drink it when you take the coffee up
88. He making a mess (‘He’s making a mess’)
89. Me making nice and neat (‘I’m making it/them nice and neat’)
90. Pour it here
91. Me put some pepper here
92. I make some peppers
93. I win (‘I am the winner’)
Recording 4: 8/3/94 age 4;5.14
94. My hear myself again? (‘Can241 I hear myself again?’ – i.e. on the tape recorder)
95. I hear myself again? (reply to ‘What?’)
96. Me watching about it (reply to ‘Did you watch somebody brushing their teeth on TV?’)
97. No, somebody else taked it (reply to ‘Did your teacher take the TV into your room?’)
98. That because them mom don't let them (‘That’s because their mum doesn’t let them’ – reply to
‘Nobody could go outside because of the snow storm’)
99. But, me mom let (‘But my mum lets (me/us/them)’)
100. Them mom could let them play outside (‘Their mum could let them play outside’)
101. Somebody else asked my mom to play outside with them
102. No one shoveled it (reply to ‘Who shoveled the snow?’)
103. Me just jumpovered it (‘I just jumped over it’)
104. Me dad too (response to another child saying ‘Yeah, my dad’ in reply to ‘Nobody shoveled the
snow?’)
105. He shoveled someone else [?= ‘He shoveled someone else’s (snow)’]
106. Who that? (‘Who’s that?’, looking at a picture)
107. This one
108. This one, he cooking up a hot dog (‘This one, he’s cooking a hot dog’)
109. And put buns on it, and ketchup
110. Ketchup and mustard
111. Ooh, that gross (‘Ooh, that’s gross’)
112. He burn heself here (‘He burned himself there’)
113. He eat it (interviewer replies ‘He ate it?’)
114. No, took it off of...then he eat it (‘No, he took it off of... then he ate it’)
115. No, he could burn heself (reply to ‘Could he hold the hot dog like that?’)
116. When he hold here (‘When he holds it here’)
117. Me daddy like mustard (‘My daddy likes mustard’)
118. Me like ketchup
119. I eat a lot of ketchup
120. Know? The teacher said when you eat a lot of food you got food in your mouth (‘Know what? The
teacher said that when you eat a lot of food, you get food in your mouth’)
121. And it stick to you (‘And it sticks to you’)
122. You can't take it off
241
It may be that ‘Can I…’ got reduced to |kna I| and then |naI|, and this was then (mis-)transcribed as my.
114
123. Only you can brush your teeth very good (‘The only thing you can do is brush your teeth really well’)
124. And it's gonna be all gone (‘And then all the food will disappear’)
125. I think this, he...umm, oh, he dump it (‘I think this, he dumped it’)
126. He shoveled him truck (‘He shoveled it into his truck’)
127. And then he dump it (‘And then he dumped it’)
128. It's hard to find this
129. I don't know
130. He shoveled him truck (‘He shoveled it into his truck’)
131. Then he dump it (‘Then he dumped it’)
132. And then he drived away (‘And then he drove away’)
133. Because, he want to put it (‘Because he wanted to put it (somewhere)’ – reply to ‘Why did he do
that?’)
134. So you don't want no ice to fall down
135. When you go on the sand you trip xxx
136. No ice can go underneath sand
137. That why them put a lot of sand (‘That’s why they put a lot of sand there’)
138. Them put all over (‘They put it everywhere’)
139. Then it can go under it
140. Then step on it, no ice (‘Then when you step on it, there’s no ice’)
141. What beach you going? (interviewer replies ‘Oh, far away’)
142. Me go beach not far away (‘I go to a beach not far away’)
143. Hey, me go faraway beach
144. Then, he knocked him window (‘Then he knocked on his window’)
145. Why her need this? (‘Why does she need this?’)
146. Oh, so for to stick in here (‘In order to stick it in here’)
147. Then her can make a pretty one
148. No, her can put up here (‘No, she can put it up here’)
149. Her sad (‘She’s sad’)
150. Me making a car (‘I’m making a car’)
151. I can make a big big man
152. Me see her and you (interviewer replies ‘What?’)
153. I see her and you
154. It's a plane
155. He's funny
156. He have he hats on (‘?= He has his hat(s) on’)
157. A girl hat (‘A girl’s hat’)
158. This is a girl and this is a boy
Recording 5: 10/3/94 age 4;5.16
159. This is somebody else fishing (= ‘somebody else's fishing game’)
160. This is you? (‘Is this yours?’)
161. Now it's gonna be Matthew (= ‘Matthew’s turn’) cause, cause...
162. It's gonna be my (= ‘mine’, i.e. ‘my turn’), cause I'm bigger than him
163. I have to do it all by myself, cause, cause...
164. I'm bigger than him
165. (Th)ey are straps (reply to ‘What are those?’)
166. I got on my shirt and have trouble doing my back (‘I’ve got my shirt on and have trouble tucking the
back of the shirt in’)
167. I try to take it off, but I can't put on
168. I can't put this on
169. But a T-shirt, can take it off (‘But a T-shirt, I can take it off’)
170. Long time ago you give me that first (‘A long time ago, you gave me that beforehand’ – reply to ‘I’ll
give you the net after you catch a fish’)
115
171. It's swinging
172. Can't see it from here
173. He's sad cause he can't go
174. He family, he lost he family (‘His family, he lost his family’ – describing a picture)
175. And he sad cause he crying (‘And he’s sad because he’s crying’)
176. And he can't go and he family... (interviewer asks ‘What happened to his family?’)
177. Die
178. He crying, he crying
179. He lost him duck (‘He lost his duck’)
180. Why you put that toy way up there?
181. He can fly
182. He can do it
183. He can fly cause he got these wings
184. He can pretend fly (‘He can pretend to fly’)
185. All these fish can pretend fly
186. I'm gonna have to pick it up first
187. It's a farm
188. He's digging up dirt
189. Dirt is falling all over him
190. Can't do that Matthew!
191. Can do all over again (‘We can do it all over again’)
192. We can do all over again
193. Oh, he caught me
194. Ah, so happy I can get out of them fish
195. Her standing and her see herself (‘She is standing and she sees herself’ – describing picture)
196. This black shadow is her and her
197. It's her hair
198. Long time ago I have a big eye (‘A long time ago, I had a swollen eye’)
199. I think it's a ghost
200. It's not, imagination (‘It’s not, it’s imagination’)
201. When I be bigger, I'm not scare of them (‘When I’m bigger, I won’t be scared of them’)
202. I see he shadow (‘I can see his shadow’)
203. I can make see my shadow (‘I can make people see my shadow’)
204. He flying (‘He’s flying’)
205. It's a...
206. Ooh, this is a scary one
207. It's a owl
208. An owl did this with he eyes (= ‘his eyes’)
209. Is green (‘It’s green’)
210. And he, hooh, and someone look at there (= ‘looked up there’) and said ah!
211. And then, he scare (= ‘scared’) them and said ha, ha, ha
212. Very funny
213. Then, he talk (= ‘talked’) to himself
214. Some wake up middle of night (‘Some people woke up in the middle of the night’)
215. Then, when someone go down, and the owl scare them (‘When someone went downstairs, the
owl scared them’)
216. It just pretend
217. He not real
218. Oh, that's different Barney
219. I have a new Barney
220. I got a new Barney
221. He got old one
222. I'n (‘didn't’) see that one, cause that one is very good
116
223. I got a different Barney
224. It's a boat
225. I can fall
226. Jesse can do it
227. I wanna play some games
228. This a new game? (‘Is this a new game?’)
229. We see clown at...umm... (‘We saw a clown at...’)
230. I see clown (= ‘saw a clown’) at a post office
231. Them have a party, and clown give me a balloon (‘They had a party and the clown gave me a
balloon’)
232. He give a plane balloon (‘He gave me a balloon in the form of an airplane)
233. A funny one
234. I wanna play a stove (‘I want to play with a/the stove’)
235. I like it, cause, cause, I wanna play (reply to ‘Why do you like that stove?’)
236. I never saw one of these stove
237. Me making hot dogs
238. Her can cook something
239. Can cook at...somewhere
240. I don't know where her can cook
241. Daddy got train (Interviewer replies ‘Daddy got a train?’)
242. Me daddy (‘My daddy’)
243. I don’t know what he saying (Other child said something unintelligible)
244. That my brother (‘That’s my brother’)
245. I got a train (‘I’ve got a train’)
246. This is gonna be hot dog
247. This is gonna be coffee (Interviewer then asks ‘Do you like coffee?’)
248. I don't, I like tea (Interviewer then asks: ‘Do you like hot tea?’)
249. I like warm tea
250. I have tea at boy scout (= ‘at the boy scouts’')
251. You lost one of these
252. But we got two more
253. You take a cup and drink like this
254. You pour on here (Interviewer then asks ‘Do you have some ketchup?’)
255. How about pepper? (interviewer then says ‘I like pepper’)
256. That not pepper (‘That’s not pepper’)
257. I put this and this (Interviewer then says ‘Pepper makes me sneeze’)
258. It don't make me (‘It doesn’t make me (sneeze)’)
259. Peter laughed you (‘Peter laughed at you’)
260. Who's that? (hearing voice on the intercom. Interviewer replies ‘A teacher’)
261. You know her? (Interviewer replies ‘Yeah’)
262. Who?
263. Where Giovanni one? (‘Where’s Giovanni’s one?’)
264. Where Giovanni sticker? (‘Where’s Giovanni’s sticker?’)
265. I remember
Recording 6: 17/3/94 age 4;5.23
266. That because is shamrock today (reply to ‘Why is everyone wearing green today?’)
267. Know what? Me teacher make cake (‘My teacher made a cake’)
268. Her make it (= ‘She made it’) for Ms Peggy a long time ago (interviewer then asks ‘Who's Ms.
Peggy?’)
269. That's my speech teacher, a speech one (Interviewer then says ‘I thought Mrs. Holly was your
speech teacher?’)
270. Me know, but my...me get two cause...
117
271. Me got Ms. Peggy speech teacher (‘I’ve got Ms Peggy as my speech teacher’)
272. We got a lotta fish for everyone [pointing to toys]
273. And then, you could get lots fish in there
274. It's got eye (‘It’s got an eye’)
275. It's a seagull
276. Know what? Me sister name Dawn (‘My sister’s name is Dawn’)
277. Her give me dad a lobster, a two lobster (‘She gave my dad a lobster, a pair of lobsters’)
278. Me mom put in here, cook them (‘My mum put them in here to cook them’)
279. Forgot to take them eyes out (‘She forgot to take their eyes out’)
280. And then, it give it to mom (‘And then ?he gave it to Mum’)
281. He say, put it down (‘She said Put it down’)
282. And then her say, ahhh! (‘And then she said Ahhh’)
283. And then her put on the floor, and we scare her (‘And then she put it on the floor and we scared her’)
284. Her say, ahh it's moving (‘She said Ahh, it’s moving)
285. And then, them cook them up (‘And then they cooked them up’)
286. And then it scared mom, so we gonna put him to trouble
287. And then he be trouble (‘And then he’ll be in trouble’. Interviewer then asks ‘Did you eat the
lobster?’)
288. No, you can't eat eyes
289. Only you can eat skin (‘You can only eat the skin’)
290. And me dad eat it (‘And my dad ate it’)
291. Daddy cook it (‘Daddy cooked it’)
292. Hey, how do ... to pick them up? (referring to toys)
293. How do you do this?
294. Why he don't have a nose?
295. Me remember...ah I remember this
296. I remember that thing
297. Two lobsters
298. I know, he wanned242 to go and drop it
299. Lobsters don't go there
300. He jumped out
301. He bit me
302. Her eating a cookie (‘She was eating a cookie’ – describing a picture)
303. Then, her drink some water (‘Then she drank some water’)
304. Water make a cookie all gone (‘The water made the cookie melt’)
305. Lobster to eat for lunch
306. Maybe some lobster pinch him (‘Maybe some lobster pinched him’)
307. Lobsters live in an ocean (Interviewer then says ‘I think an ant bit him’)
308. You could have chicken pox (Interviewer then asks ‘Did you ever get bitten by an ant?’)
309. Me no, me grandma (= ‘my grandma’) did long time ago from a snake
310. Me remember her house snake
311. Her have a snake in the sink (‘She had a snake in the sink’. Interviewer then asks ‘In the sink?’)
312. No, amm, next to a xx
313. Then grandma get biten (= ‘got bitten’) by a snake
314. Then her got hurt (‘Then she got hurt’)
315. Then, me and xxx drop a xxx, then it hurt me
316. He try to eat me up (‘He tried to eat me up’)
317. How he gonna eat me up?
318. He grabbing a lollipop [describing a picture]
319. Me remember at school...ahm...
242
The t of want is dropped in colloquial American English in such forms, so that corresponding to Standard English
wanted we find colloquial wanned. This can be treated as a regular (but contracted) past form of want.
118
320. What's I talking about? (‘What was I talking about?’)
321. Long time ago, I go camp (= ‘went camping’) and hiking at the same time
322. Then, I eating food
323. I say, and I eat at hiking
324. And then at hiking, I xxx a picnic, so we eat
325. But a bee eating it (‘A bee was eating it/ate it’)
326. Then a bee eat a little bit my food (‘A bee ate a little bit of my food’)
327. Then me said, oh!
328. Then me take it xxx (‘Then I took it xxx’)
329. Then me go camping (‘Then I went camping’)
330. Me eat all our food (‘I ate all our food’)
Recording 7: 29/3/94 age 4;6.3
331. I mean this much (reply to ‘I don't want that many people in my room!’)
332. Not that again
333. We making books (‘We are making books’)
334. But, I got another one
335. My another one is lost
336. Easter eggs
337. Me mommy show me to make Easter eggs (‘My mummy showed me how to...’)
338. Now I do all by myself (‘Now I do them all by myself’. Interviewer then asks ‘How do you make
them?’)
339. I forgot
340. Only he mom can teach him make it good (‘Only his mum can teach him to make it properly’)
341. That mean is already cook (‘That means it’s already cooked’)
342. No, you don't give one
343. You hafta take it with a mitten
344. You hafta take it with a little holder thing
345. With a egg thing
346. My dad make eggs, but mushy eggs (‘My dad makes eggs...’)
347. Me mom don't use paint brush (‘My mum doesn’t use a paint brush’)
348. Hers use with a green thing (‘She uses a green thing’)
349. Her use with a green cutter thing (‘She uses a green cutter thing’)
350. We just give to Easter Bunny
351. He can eat it
352. He hafta
353. Me mom hold in him hands (‘My mum holds it in her hands’)
354. Very careful so you can't drop it
355. We take one at a time
356. He put in a box with lots different and jelly bean (‘He puts the egg in a box with lots of different
things and jelly beans’)
357. Then he bring it (‘Then he brings it’)
358. Then mom and kids find easter eggs
359. Then, I grab my one
360. Mom say, no, after lunch (‘Mom says ‘No, after lunch’)
361. Then, I eat it after lunch
362. I need ten stickers
363. I'm gonna be so proud
Recording 8: 7/4/94 age 4;6.12
364. I know her (Interviewer then says ‘I see you got a hair cut’)
365. I got my hair from barber
119
366. But them cut my hair real tiny (‘But they cut my hair really short’)
367. Who first?
368. Me or him?
369. We got some new pictures (Interviewer then asks ‘Why do you think she needs the camera?’)
370. Her need the camera to put in a locker
371. I know these ones
372. It look like a lobster (‘It looks like a lobster’)
373. Them both have pinchers (‘They both have pincers’)
374. Why them both have pinchers? (‘Why do they both have pincers?’)
375. Both got legs
376. He...him go first
377. He taking a lollipop
378. He gonna bite you
379. He cut me
380. This is a good one