Download Passive verb morphology: The effect of phonotactics on passive

This article was originally published in a journal published by
Elsevier, and the attached copy is provided by Elsevier for the
author’s benefit and for the benefit of the author’s institution, for
non-commercial research and educational use including without
limitation use in instruction at your institution, sending it to specific
colleagues that you know, and providing a copy to your institution’s
administrator.
All other uses, reproduction and distribution, including without
limitation commercial reprints, selling or licensing copies or access,
or posting on open internet sites, your personal or institution’s
website or repository, are prohibited. For exceptions, permission
may be sought for such use through Elsevier’s permissions site at:
http://www.elsevier.com/locate/permissionusematerial
Lingua 117 (2007) 1434–1447
www.elsevier.com/locate/lingua
co
py
Passive verb morphology: The effect of phonotactics on
passive comprehension in typically developing and
Grammatical-SLI children
Chloe Marshall, Theodoros Marinis 1, Heather van der Lely *
Centre for Developmental Language Disorders and Cognitive Neuroscience, Department of Human Communication
Science, University College London, 2 Wakefield Street, London WC1N 1PF, UK
on
al
Received 21 March 2006; received in revised form 23 July 2006; accepted 25 July 2006
Available online 24 October 2006
Abstract
r's
pe
rs
In this study we explore the impact of a morphological deficit on syntactic comprehension. A self-paced
listening task was designed to investigate passive sentence processing in typically developing (TD) children
and children with Grammatical-Specific Language Impairment (G-SLI). Participants had to judge whether the
sentence they heard matched a picture they were shown. Working within the framework of the Computational
Grammatical Complexity Hypothesis, which stresses how different components of the grammar interact, we
tested whether children were able to use phonotactic cues to parse reversible passive sentences of the form the X
was verbed by Y. We predicted that TD children would be able to use phonotactics to parse a form like touched
or hugged as a participle, and hence interpret passive sentences correctly. This cue is predicted not be used by
G-SLI children, because they have difficulty building complex morphological representations. We demonstrate that indeed TD, but not G-SLI, children are able to use phonotactics cues in parsing passive sentences.
# 2006 Elsevier B.V. All rights reserved.
th
o
Keywords: Specific Language Impairment; Syntax; Passives; Morphology; Phonotactics; Computational Grammatical
Complexity Hypothesis
Au
1. Introduction
Language processing by children, and the manner in which it breaks down in those who have
Specific Language Impairment (SLI), is currently a vibrant area of research. A number of studies
* Corresponding author. Tel. +44 20 7679 4049; fax: +44 20 7713 0861.
E-mail address: [email protected] (H. van der Lely).
1
Present address: School of Psychology and Clinical Language Sciences, University of Reading, UK.
0024-3841/$ – see front matter # 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.lingua.2006.07.001
C. Marshall et al. / Lingua 117 (2007) 1434–1447
1435
co
py
have investigated language processing in typically developing (TD) and/or SLI children (Clahsen
et al., 2004; Felser et al., 2003; Fonteneau and van der Lely, 2005; Marinis and van der Lely,
under revision; Montgomery and Leonard, 1998; Windsor, 2002; Montgomery, 2005). Such
investigations enable us to understand more fully the processing of particular linguistic forms in
typical development, and also to characterize how processing can break down.
The present study brings together two aspects of language that children with a particular form
of language impairment, Grammatical-SLI, find problematic—syntax and morphology. In this
paper we investigate the impact of phonotactics on the processing of passive sentences,
something that to our knowledge has not been studied in either typical or atypical development.
We begin this introduction with a discussion of the linguistic characteristics of G-SLI, and with
an outline of the model that we have proposed can account for G-SLI, the Computational
Grammatical Complexity Hypothesis. We then move on to discuss how passive sentences and
morphology are acquired in TD and G-SLI children, and outline the motivation for our study.
1.1. Linguistic characteristics of G-SLI
Au
th
o
r's
pe
rs
on
al
SLI is a deficit in the acquisition of language despite adequate hearing, non-verbal abilities
and opportunity, and the absence of any known neurological disorder (see Leonard, 1998, for a
review). The incidence of SLI in the preschool population is estimated at 7% (Tomblin et al.,
1997), but in some children the disorder persists until adulthood. Within the SLI population as a
whole, deficits have been diagnosed in syntax, morphology, phonology, pragmatics and the
lexicon. However, the population is also heterogeneous, with considerable variation in both the
severity and the linguistic pattern of impairment. Although this phenotypic heterogeneity is
widely acknowledged (Bishop, 1997; van der Lely, 2003), there is little consensus as to how it is
best described and accounted for, and consequently it is not accommodated within most theories
of SLI.
One pattern of impairment that has been characterized in detail is termed ‘Grammatical-SLI’
(for English: van der Lely, 1997, 2005; van der Lely et al., 1998; for Greek: Stavrakaki, 2002; for
Hebrew: Friedmann and Novogrodsky, 2004). This subgroup of SLI children is argued to have a
relatively homogenous pattern of linguistic deficits. G-SLI children’s language difficulties
encompass the core aspects of grammar—syntax, morphology and phonology. In syntax, there
are particular difficulties with the production and comprehension of complex syntactic forms,
specifically those involving syntactic dependencies characterized by ‘Movement’ (Chomsky,
1995), e.g. tense marking (production of bare stem forms, e.g. Yesterday he go to school),
WH-questions (gap-filling and lack of do-support, e.g. Who Mrs Scarlet saw someone in the
lounge?) and passives (incorrect assignment of thematic roles, in, for example, The elephant was
kicked by the mouse) (van der Lely, 1996a,b, 1998; van der Lely and Battell, 2003; van der Lely
and Stollwerck, 1997). In morphology, past tense and agreement suffixation are severely
affected, and children’s use of regular plurals inside compounds (e.g. *rats-eater) reveals that
their knowledge of plurality is unlike that of language-matched TD children (van der Lely, 1998;
van der Lely and Christian, 2000; van der Lely and Ullman, 2001). In phonology, children have
difficulties repeating non-words that contain complex syllabic and metrical structures, such as
consonant clusters and initial weak syllables (Gallon et al., submitted for publication; Marshall,
2004; Marshall et al., 2003). Lexical deficits are reported, but are less severe than the
grammatical deficits, and are argued to be secondary to them (van der Lely, 2005). Importantly,
the deficit is a persistent one—children are only included in the subgroup if they are at least 9
years old, and so are at an age when they are unlikely to ‘grow out of’ their impairment. Although
1436
C. Marshall et al. / Lingua 117 (2007) 1434–1447
co
py
the existence of this subgroup is controversial, the approach taken by van der Lely and her
colleagues to investigate SLI reduces the phenotypic heterogeneity that besets the majority of
studies of SLI, and makes the data more amenable to a coherent psycholinguistic explanation.
The model proposed to account for the linguistic deficits in children with G-SLI is the
Computational Grammatical Complexity (CGC) Hypothesis (Marshall, 2004; Marshall and van
der Lely, in press; Marshall and van der Lely, submitted for publication; van der Lely, 2005). The
CGC Hypothesis claims that children with G-SLI have difficulty comprehending and producing
complex linguistic structures in three components of the computational grammar: syntax,
morphology and phonology. Furthermore, it stresses the cumulative and interactive effects that
these deficits have on linguistic constructions such as the English regular past tense, where
linguistic complexity in all three components of grammar is required (Marshall, 2004; Marshall
and van der Lely, submitted for publication; van der Lely, 2005). In this paper we investigate the
interaction between syntactic and morphological complexity in passive constructions.
1.2. Acquisition of the passive
th
o
r's
pe
rs
on
al
Full passives include a by-phrase, e.g. The elephant was kicked by the mouse. Short passives
lack the by-phrase, e.g. The elephant was kicked. TD children acquire short passives before full
passives (Horgan, 1978; Mills, 1986). Early acquired passives represent after-the-fact
observations about states, i.e. they describe a state, not an event (e.g. The tree is broken can
be interpreted similarly to The tree is green). Such passives are hence adjectival rather than
verbal. Furthermore, Maratsos et al. (1985) found that 4–5 year old children have more
difficulties in the comprehension and production of non-actional passives (e.g. with verbs such as
hear and see) than actional passives (e.g. with verbs such as comb and touch).
Verbal passives involve a more complex syntactic structure than adjectival passives. Whereas
adjectival passives are considered to be base-generated and not to involve any type of movement,
in verbal passives the object of the active verb (The mouse was kicking the elephant) becomes the
subject of the passive verb (The elephant was kicked by the mouse). This happens through
movement of the noun phrase from the object to the subject position, as illustrated in (1) below.
Au
This movement leaves a silent copy (‘trace’) behind, that is co-indexed with the moved constituent
and is assigned a thematic role by the verb. The moved constituent (the elephant) and the trace (ti)
form an (A)rgument-chain, i.e. they are both in argument positions of the same verb and they share
some syntactic features. According to Borer and Wexler (1987), young children cannot comprehend and produce verbal passives because they are unable to form A-chains, and thus they cannot
assign the thematic role to the moved constituent. In contrast, they do not make errors with
adjectival passives because these do not involve movement and A-chains. Borer and Wexler
suggested that the formation of A-chains is subject to maturation, and the mechanism responsible
for the formation of A-chains does not mature until at least the age of 5 or 6.
It has long been known that children with SLI have difficulty interpreting reversible passives
(Bishop, 1979). A series of studies by van der Lely and colleagues have characterised this deficit
in G-SLI children. G-SLI children have problems interpreting passive sentences when general
semantic and world knowledge cannot be used to facilitate comprehension, and where the child
C. Marshall et al. / Lingua 117 (2007) 1434–1447
1437
py
has to rely instead on abstract syntactic knowledge. This is the case when the event is improbable,
reversible, and when a novel verb is used (van der Lely, 1994, 1996a,b; van der Lely and Dewart,
1986). Moreover, in short but unambiguously verbal passive sentences (e.g. The fish is being
eaten), G-SLI children choose the syntactically simpler adjectival (stative) interpretation on a
substantial number of occasions. This pattern is emphasized for short passives that are ambiguous
between a stative and a verbal reading (e.g. The fish is eaten)—G-SLI children are more likely to
choose the stative reading than the verbal reading.
1.3. Morphology and phonotactics
Au
th
o
r's
pe
rs
on
al
co
One of the most severe and widely-reported deficits in SLI affects verb morphology, and in
particular the use of suffixes that mark tense and agreement (see review in Leonard, 1998). In
English the pattern is one of variable suffix omission, e.g. Yesterday I play_/played football, He
always watch_/watches television. One topic of very lively debate concerns the underlying cause
of the morphological deficit. According to some researchers, it reflects an impairment in auditory
processing, which affects the perception of suffixes of low perceptual salience and rapid duration
at the end of verbs (Joanisse and Seidenberg, 1998; Karmiloff-Smith, 1998; Marchman et al.,
1999). For others, however, the morphological deficit reflects an impairment in a grammatical
component of the language, although the nature of this impairment is likewise subject to vigorous
debate. It is variously proposed to be due to the child persisting in an extended optional infinitive
stage (Rice and Wexler, 1996), an inability to construct grammatical rules (Ullman and Gopnik,
1999), an inability to set up agreement relations (Clahsen et al., 1997) or an inability to construct
hierarchically branching complex morphological forms (Marshall and van der Lely, 2006; van
der Lely, 2005). This debate is theoretically relevant in that it relates to whether suffixation is
carried out using a single route, analogical mechanism, or whether it is instead a rule-based
process carried out by a separate component of the grammar (dual mechanism model). Those
who claim that morphological impairments stem from an auditory processing deficit interpret the
SLI data as supporting a single mechanism model (Joanisse and Seidenberg, 1998), whereas
those who claim that they stem from a linguistic deficit interpret the data as supporting a dual
mechanism model (Pinker and Ullman, 2002).
In order to probe whether the past tense difficulties demonstrated by English-speaking G-SLI
children best support a single or a dual mechanism model of morphology, Marshall and van der Lely
(2006) investigated the impact of verb final phonotactics on past tense production. Because regular
verbs stems frequently end in a consonant, suffixation often produces clusters. Two kinds of clusters
exist with regards to phonotactics—legal clusters, which also occur in monomorphemic words (e.g.
the /st/ at the end of past tense missed is also found in monomorphemic mist), and illegal clusters,
which only occur in suffixed forms (the /gd/ of hugged does not occur at the end of any
monomorphemic word of English). Illegal clusters, because they occur only in suffixed forms, are
less frequent than legal clusters, which occur in both suffixed and non-suffixed forms. This
frequency difference formed the basis of Marshall and van der Lely’s study.
Marshall and van der Lely (2006) hypothesized that if G-SLI children are impaired in forming
complex morphological forms, and are having to remember them by rote (van der Lely and
Ullman, 2001), then cluster frequency should impact on performance, with lower suffixation
rates for phonotactically illegal compared to legal inflected forms. In contrast, if typically
developing children are creating past tense forms using a suffixation rule, then the frequency of
the cluster formed by suffixation should not impact on performance in a past tense elicitation
task. Performance should therefore be equivalent on both phonotactically legal and illegal
1438
C. Marshall et al. / Lingua 117 (2007) 1434–1447
al
co
py
inflected forms (e.g. missed as compared to hugged), when past tense frequency and lemma
frequency (i.e. the sum of the frequencies of all morphological forms of the verb, e.g. miss,
missed, missing and misses) are controlled. This was indeed the case—whereas two groups of
G-SLI children were revealed to be less likely to mark tense on a verb when the cluster would
have been illegal, there was no impact of phonotactics on six groups of typically developing
children or on a group of individuals with Williams Syndrome.
Is phonotactics a phenomenon relevant only to tense or does it affect other morphemes with a
syntactic function, and if so, what can it reveal about linguistic processing in TD and languageimpaired children? For example, it might be the case that typically developing children are able to
use illegal cluster phonotactics to parse morphologically complex forms, but that this cue is not
available to children who have a morphological deficit, such as those with G-SLI. In this paper we
investigate the past participle, which for English regular verbs has same phonological form as the
past tense form of the verb. This enables us to investigate further the grammatical deficit in G-SLI:
compared to past tense forms, participles are syntactically different,2 but the phonology is the same
as for past tense forms. Past participle forms are lexically affixed, and so clearly involve
morphology. Hence we can investigate suffixed forms that are morphologically and phonologically
similar to tensed forms, but that participate in a different syntactic construction—the passive.
on
2. Method
rs
2.1. Participants
Au
th
o
r's
pe
Fourteen children and teenagers with G-SLI (11 boys) aged between 10;11 and 15;3 years
participated in this study. All have a professional diagnosis of SLI and are attending specialist
language schools or language units, but demonstrate non-verbal cognitive abilities in the normal
range (mean non-verbal IQ was 93.9 as measured by a composite of the Ravens Progressive
Matrices, Raven, 1998, and the Block Design subtest of the British Ability Scales, Elliott, 1996),
normal hearing, no articulation difficulties and appropriate emotional and social behaviour.
Furthermore, all participants meet the criteria for G-SLI (van der Lely and Batell, 2003; van
der Lely et al., 1998; van der Lely and Ullman, 2001), in that they have persistent problems with
grammatical aspects of language in production and comprehension, as evinced by their
performance on a range of standardised and non-standardised tests. Their mean z-score on a test
of sentence comprehension (Test for the Reception of Grammar, TROG, Bishop, 1989) is 1.8,
with an equivalent age of 6;5, while their mean z-score on a measure of single word
comprehension (British Picture Vocabulary Scales, BPVS, Dunn et al., 1997) is 1.5, with an
equivalent age of 7;8. In addition to these standardized tests, the G-SLI children were assessed on
a series of non-standardised tests that tap into the aspects of grammar that are particularly
problematic for them. These include the Verb and Agreement and Tense Test (VATT, van der
Lely, 2000) and the Test of Active and Passive Sentences (TAPS, van der Lely, 1996a,b). Each
participant made 20% or more errors on each of these tests (see van der Lely, 2005).
It is important to compare the performance of the G-SLI group to that of a wide age range of
TD children, in order to determine whether the pattern of performance in G-SLI children is
shown by TD children too. Three control groups of typically developing (TD) children
participated in this study (see Table 1). Two matched the G-SLI children on measures of sentence
comprehension or vocabulary. The youngest group, TD1, was a group of 14 children with a mean
2
Unlike the tensed form of the verb, past participles show no V to I movement (Grimshaw, 1990).
C. Marshall et al. / Lingua 117 (2007) 1434–1447
1439
Table 1
Summary of participants’ details
TD1 (N = 14)
TD2 (N = 19)
TD3 (N = 14)
G-SLI (N = 14)
Mean
S.D.
Mean
S.D.
Mean
S.D.
Mean
Chronological agea
Age range
7;0
6.6
8;7
5.2
10;1
8.0
TROGb
SS
z-Score
Equivalent age
13.5
100.6
0
6;9
2.6
13.3
0.9
1.8
16.4
108.2
0.5
9;4
1.9
13.0
0.9
1.9
16.7
100
0
9;7
BPVS II b
SS
z-Score
Equivalent age
66.7
103.8
0.3
6;5
12
11
0.7
1.2
87.1
109.4
0.6
8;7
9.2
8.2
0.5
1.1
91.6
99.9
0
9;3
8;9–10;9
1.5
11.8
0.8
1.5
13;5
18.9
10;11–15;3
12.5
73.6
1.8
6;5
3.0
7.4
0.5
1.6
78.3
77.6
1.5
7;8
18.0
11.4
0.8
2.0
py
7;9–8;11
co
6;0–7;8
12.5
11.6
0.8
1.7
S.D.
a
Mean = years;months, S.D. = standard deviation in months.
TROG: test of reception of grammar; BPVS II: British Picture Vocabulary Scale; SS = standard score, where
mean = 100 and standard deviation = 15; z-score, where mean = 0 and standard deviation = 1, and impaired is generally
considered to be 1.5 and below.
on
al
b
r's
pe
rs
age of 7;0 (range 6;0–7;8). This control group was matched to the G-SLI group (on raw score) on
the TROG (Bishop, 1989). There was no significant difference between the two groups’ scores on
the TROG (t(27) = 0.934, p = 0.358). The second group, TD2, comprised 19 children with a
mean age of 8;7 (range 7;9–8;11). This control group was matched to the G-SLI group on the
BPVS (Dunn et al., 1997). There was no significant difference between the two groups’ scores on
the BPVS (t(32) = 1.686, p = 0.101). The third group, TD3, consisted of 14 children with a mean
age of 10;1 (range 8;9–10;9). The sentence comprehension and vocabulary abilities of these
children were significantly higher than those of the G-SLI group (TROG: t(27) = 4.733,
p < 0.001; BPVS: t(27) = 2.09, p < 0.05), and so they will not be directly compared to the G-SLI
group, but they are included in this study so that we can investigate typical development over a
wide age range.
2.2. Design and material
Au
th
o
In this study we used a novel task that combines a picture-verification and a self-paced
listening or auditory moving window technique (Ferreira et al., 1996). In a picture verification
task participants typically see a picture, listen to a sentence, and have to decide whether the
sentence matches the picture. In a self-paced listening task, subjects listen to sentences in a wordby-word or phrase-by-phrase fashion, and press a button to receive successive words or phrasal
segments. This task provides a segment-by-segment measure of processing time. In our
experiment, participants first saw a picture on the computer screen. Then they listened to a
sentence segment-by-segment while the picture remained on the screen. We recorded how fast
participants processed each segment. After hearing each sentence, participants had to decide
whether the sentence they heard matched the picture. It is this judgment measure that concerns us
in this present paper—the reaction time data are presented separately for reasons of space
(Marinis and van der Lely, in preparation).
The material for our task comprised 135 sentences, including 15 practice sentences, 80
experimental sentences and 40 filler sentences. The experimental sentences were reversible
1440
C. Marshall et al. / Lingua 117 (2007) 1434–1447
active and passive sentences, whereby both the agent and the patient were animals, as illustrated
in (2) and (3) below:
(2)
py
(3)
I think that the squirrel with the gloves was bathing the tortoise at his house last
weekend.
I think that the squirrel with the gloves was bathed by the tortoise at his house last
weekend.
Experimental conditions
Active sentence – phonotactically legal verb – matching picture
Active sentence – phonotactically legal verb – non-matching picture
Active sentence – phonotactically illegal verb – matching picture
Active sentence – phonotactically illegal verb – non-matching picture
Passive sentence – phonotactically legal verb – matching picture
Passive sentence – phonotactically legal verb – non-matching picture
Passive sentence – phonotactically illegal verb – matching picture
Passive sentence – phonotactically illegal verb – non-matching picture
pe
rs
(4)
on
al
co
The pictures either matched the event in the sentence or they showed the event with the agent and
patient reversed.3
We included 20 different verbs, and each verb was used 4 times during the experiment. All
verbs were monosyllabic and had a regular past tense form. Half had an inflected form that ended
in a phonotactically legal cluster, whereas half had an inflected form that ended in an illegal
cluster (the full list of verbs is provided in Appendix A).
There are eight experimental conditions of interest, as shown in (4) below, which allow us to
explore the effects of sentence type (active versus passive), verb type (legal versus illegal cluster
phonotactics4) and picture type (match or mismatch to sentence):
Au
2.3. Procedure
th
o
r's
The auditory materials were spoken by female native speakers of English at a normal speaking
rate. We recorded the sentences in an anechoic room on a Sony digital tape recorder, and we used
CoolEdit software (Syntrillium) to edit the tapes. Four different experimental sets were created,
each containing one version of the experimental sentences and one of the two pictures. Each
participant encountered only one of those four sets containing 20 sentences of each condition.
Thus, each participant heard only one version of each experimental sentence, and saw only one of
the two pictures.
The experiment was carried out at the children’s home or at school, and consisted of two
20–30 min sessions. Participants were seated in front of a laptop and a push-button box, and were
3
The prepositional phrases are relevant only for the on-line effects. They were used to separate possible on-line effects
at the first NP from effects at the verb, and to avoid conflating an effect related to the assignment of theta roles and a wrap
up effect at the end of the sentence.
4
Note that cluster phonotactics, and the distinction between phonotactically legal and illegal clusters, is only relevant
in passive sentences, where the past participle form of the verb occurs.
C. Marshall et al. / Lingua 117 (2007) 1434–1447
1441
al
co
py
instructed to look at the picture on the computer screen and listen carefully to the pre-recorded
sentences over headphones. In the first session, each child received a fixed set of instructions, first
from the experimenter, and then through the headphones. They were taught how to press the
button to hear the sentences, and they were given 10 practice sentences in order to familiarise
themselves with the task. These practice sentences could be repeated if necessary. The second
session took place on a different day or after a sort break, and was preceded by a further 5 practice
sentences that could be repeated if necessary.
In each trial, the child saw a picture on the computer screen. After 1000 ms, the presentation of
the sentence began. After listening to each segment, the child had to press a button on a pushbutton box as quickly as possible in order to receive the next segment. The end of the sentence
was indicated by a tone. The computer recorded the time between the onset of each segment and
the button press. Presentation of the stimuli and the recording of the reaction times were
controlled by the software package E-Prime (Psychology Software Tools). After hearing the
entire sentence, the child had to then decide whether it matched the picture. This was recorded by
the experimenter on a separate form. Participants did not receive feedback as to whether they had
answered accurately.
on
2.4. Predictions
r's
pe
rs
Our hypothesis concerns parsing: we predict that TD children will be able to use illegal cluster
phonotactics as a parsing cue, to aid them in interpreting passive sentences. In contrast, G-SLI
children will be unable to use phonotactics as a cue, because of their morphological deficit—they
have difficulty building up an abstract representation of stem + suffix, and are therefore not able
to use cues in speech stream that indicate the presence of a stem + suffix structure. We therefore
predict that TD children will be more successful at interpreting passive sentences when the
participle ends in a phonotactically illegal cluster. For G-SLI children, however, whether the
participle ends in a phonotactically legal or illegal cluster will have no effect. Hence we predict
that G-SLI children’s deficit in interpreting passive sentences, which has hitherto been
interpreted as a syntactic deficit, may have a partly morphological cause as well.
3. Results
Au
th
o
The scores from one child in the TD1 group and one in the TD2 group were more than 2
standard deviations below the mean for their own group, and so these data were excluded from
further analyses. Table 2 shows the accuracy of the remaining participants in each group.
In our analysis, we first compare data from the three control groups. We performed a 3
(participant group: TD1, TD2, TD3) 2 (sentence type: active, passive) 2 (verb type:
phonotactically legal, illegal) 2 (picture type: match, mismatch) ANOVA. There are
significant main effects of sentence type and picture type. Passive sentences are significantly
harder than active sentences, F(1,42) = 4.643, p = 0.037, and performance is significantly lower
when the picture and sentence do not match, F(1,42) = 30.884, p < 0.001. The lack of significant
interactions with group indicates that all groups behave in the same way with respect to the
factors of interest. However, there is a significant main effect of group, F(2,42) = 6.189,
p = 0.004. Post hoc t-tests (Bonferroni-corrected) indicate that the only significant difference is
between the TD1 and TD3 groups, p = 0.004.
Note that there is no significant main effect of verb type. This is important because it indicates
that neither set of verbs (phonotactically legal or illegal in the past participle form) is intrinsically
1442
C. Marshall et al. / Lingua 117 (2007) 1434–1447
Table 2
Mean % correct responses (standard deviation), for all groups
TD1
legal match
legal mismatch
illegal match
illegal mismatch
Total
Total TD
G-SLI
95.38
77.61
90.00
74.62
(5.19)
(14.73)
(10.00)
(17.13)
96.67
86.67
94.63
86.67
(4.85)
(15.72)
(6.68)
(21.14)
96.43
89.29
97.14
90.71
(4.97)
(7.30)
(6.11)
(9.97)
96.22
84.86
94.27
84.44
(4.90)
(13.91)
(7.97)
(18.04)
93.16
74.59
89.47
75.26
(9.46)
(24.29)
(13.94)
(25.68)
93.85
62.69
93.85
77.35
(8.70)
(25.55)
(9.61)
(23.46)
91.67
80.00
90.56
88.89
(11.50)
(16.80)
(10.56)
(13.67)
97.86
84.29
95.71
89.29
(4.26)
(11.58)
(6.46)
(12.69)
94.22
76.33
93.11
85.68
(9.17)
(20.18)
(9.25)
(17.29)
91.20
71.05
90.25
68.95
(12.13)
(23.78)
(9.56)
(24.24)
83.17 (14.30)
89.47 (12.62)
92.59 (7.92)
py
Passive
Passive
Passive
Passive
legal match
legal mismatch
illegal match
illegal mismatch
TD3
co
Active
Active
Active
Active
TD2
88.41 (11.61)
81.74 (13.67)
Au
th
o
r's
pe
rs
on
al
more difficult than the other, due to differences in properties unrelated to phonotactics, such as
imageability. However, verb type enters the picture when we consider higher order interactions.
There is a significant three-way interaction between sentence type, verb type and picture type,
F(1,42) = 6.072, p = 0.018, and there are significant two-way interactions between sentence type
and verb type, F(1,42) = 6.628, p = 0.014, and between verb type and picture type,
F(1,42) = 8.733, p = 0.005.
Our hypothesis concerned a particular comparison—we predicted better performance on
passive verbs ending in a phonotactically illegal cluster, although we did not make a prediction
according to picture type. Therefore, in order to reduce the number of pairwise comparisons
carried out and to maintain sensitivity, we compare passive legal match with passive illegal
match, and passive legal mismatch with passive illegal mismatch (we are not justified in
collapsing across picture type because of the three-way interaction). We reduce the alpha level to
0.025 to compensate for multiple comparisons. There is no significant difference between passive
illegal match and passive legal match, t(44) = 0.514, but performance is significantly better for
passive illegal mismatch compared to passive legal mismatch, p = 0.001. In other words,
performance is better for passive sentences containing a phonotactically illegal cluster, but only
in the mismatch condition.
Now we consider the performance of the G-SLI children compared to that of the TD1 controls
only, as these two groups were matched on general sentence comprehension and performed at
similar levels on this task.5 A 2 (participant group: TD1, G-SLI) 2 (sentence type: active,
passive) 2 (verb type: phonotactically legal, illegal) 2 (picture type: match, mismatch)
ANOVA reveals a significant main effect of match, F(1,25) = 29.879, p < 0.001, but no other
significant main effects. The lack of main effect of group indicates that we are justified in
comparing the relative pattern of performance of these two groups: qualitatively different
performance by the G-SLI children cannot be attributed to their being at a different level of
active/passive comprehension.
Again, our hypothesis concerns a particular comparison—we predicted that while the
typically developing children would perform better on passive verbs ending in a phonotactically
illegal cluster, G-SLI children’s impairment in morphology would render them insensitive to
phonotactics, and they would show no difference between passive sentences with illegal cluster
phonotactics compared to passive sentences with legal cluster phonotactics (we did not make a
5
Note that the results are very similar when we compare the G-SLI group to the vocabulary-matched LA2 controls.
C. Marshall et al. / Lingua 117 (2007) 1434–1447
1443
co
py
prediction according to picture type). The four-way interaction is not significant, but there is a
significant three way interaction between group, sentence type and verb type, F(1,25) = 4.261,
p = 0.050. To unpack this interaction we compare passive sentences with a phonotactically legal
verb to passive sentences with an illegal verb, first for the TD1 group and then for the G-SLI
group. For the TD1 group, t(12) = 2.084, p = 0.059, and for the G-SLI group, t(13) = 0.723,
p = 0.482.
The results of the t-test within the TD1 group are approaching significance, and together
with the highly significant difference between passive legal and illegal sentences when the
three control groups are considered together, we conclude that the G-SLI group show a
qualitatively different pattern of performance to typically developing children. While
typically developing children show an advantage for passive sentences whose past participles
contain phonotactically illegal clusters, the G-SLI children show no such advantage. This
qualitatively different pattern of performance suggests that G-SLI children process the verb in
passive sentences in a different way to TD children. We explore possible reasons for this in the
discussion in section 4.
al
4. Discussion
Au
th
o
r's
pe
rs
on
In this study we investigated whether typically developing (TD) and Grammatical-SLI
(G-SLI) children are able to use phonotactic cues to parse passive sentences. The performance of
TD children shows an improvement up to the age of 10. All groups show the same pattern of
performance with regards to whether or not the sentence they hear matches the picture they see—
they are more accurate in the match conditions. They are also more accurate at interpreting active
sentences than passive ones. However, the TD and G-SLI groups differ with respect to the impact
of phonotactic legality at the end of the past participle. TD children are more accurate in
interpreting a passive sentence when the participle has illegal cluster phonotactics. For G-SLI
children, however, cluster phonotactics has no impact. In this section we first discuss the effects
of match and phonotactics, and what they reveal about the processing of passive sentences in TD
and G-SLI children. We then discuss how these results contribute to our understanding of the
language deficit in SLI.
We first tackle the finding that all children are less accurate in interpreting a sentence when
that sentence does not match the picture they see, be that sentence active or passive. We
hypothesise that children make an initial assignment of thematic roles very early in the sentence,
before they encounter the verb and the second argument of the verb. For example, when the
picture shows a squirrel bathing a tortoise, and they listen to the mismatch sentence I think / that /
the squirrel / with the gloves / was bathed / by the tortoise / at his house / last weekend, the actor in
the picture is the subject of the sentence. We hypothesise that after listening to the segment the
squirrel, children make an initial assignment of the thematic role Agent to the first noun phrase of
the sentence, and expect to hear an active sentence. Then, when they encounter the segment was
bathed, they have to re-analyse the sentence, and assign the thematic role Patient/Theme to that
noun phrase. This re-analysis phase is prone to error in all the populations we tested. In order to
give the correct judgement, i.e. that the sentence does not match the picture, the child has to pay
special attention to the morphology of the verb if he/she is not to interpret the sentence as being
consistent with the picture.
As for the effect of cluster phonotactics in passive sentences, our explanation is as follows. In
order to interpret the sentence as a passive, and to assign thematic roles correctly, the child needs
to parse the participle into stem + suffix and interpret the aspectual feature of the suffix in relation
1444
C. Marshall et al. / Lingua 117 (2007) 1434–1447
th
o
r's
pe
rs
on
al
co
py
to the tense marker (the auxiliary). We hypothesise that TD children are able to represent a
morphologically complex form as stem + suffix. Illegal phonotactics provide a cue as to the
existence of a stem + suffix boundary. This is because forms such as hugged and touched can
only be suffixed, because those final clusters cannot occur in monomorphemic forms. Participles
such as kissed and called are, however, potentially ambiguous between a suffixed and
monomorphemic form (as in the pairs missed/mist, billed/build, wrapped/rapt, tacked/tact etc.).
For TD children, the effect of phonotactic illegality is only significant in the mismatch
condition—presumably performance in the two match conditions is too close to ceiling for any
difference to be apparent. Why are G-SLI children unable to make use of phonotactic cues in the
mismatch condition? We hypothesise, consistent with a series of studies of morphology in G-SLI
children (Marshall and van der Lely, 2006; van der Lely and Christian, 2000; van der Lely and
Ullman, 2001), that G-SLI children have difficulties parsing and creating morphologically
complex forms, and therefore to build a representation of stem + suffix. Therefore, G-SLI
children are unable to use the clue to morphological structure provided by the illegal clusters to
parse past participles in passive sentences.
These results are theoretically important because they challenge single mechanism models of
morphology. Such models reject the existence of a separate morphological component of the
grammar and hold that frequency is critical, with high frequency patterns being acquired earlier
and being available for analogy more readily than low frequency patterns. And yet here we have
an example of TD children experiencing greater success with lower frequency patterns
(phonotactically illegal clusters) than higher frequency patterns (phonotactically legal clusters).
Furthermore, our results also challenge the hypothesis that G-SLI children have difficulties
perceiving the –ed suffix—these children do no worse than their language matched controls on
the match sentences, where their performance is actually near to ceiling. Hence we have further
evidence against the auditory processing theory of SLI, at least for this particular subgroup of
children (see also van der Lely et al., 2004).
Instead, our results provide further evidence that G-SLI children process certain aspects of
language differently to TD children (see also Fonteneau and van der Lely, 2005; Marshall and van
der Lely, 2006; van der Lely and Christian, 2000; van der Lely and Ullman, 2001). Most
parsimoniously for the results in this paper, is an explanation whereby G-SLI children have a deficit
in building morphologically complex forms, within a cognitive model of language in which TD
children have a dual route system for morphology. In G-SLI children the rule-based morphological
component of the grammar, responsible for building representations of stem + suffix, is impaired.
We have previously shown this for regular past tense and plural formation (van der Lely and
Christian, 2000; van der Lely and Ullman, 2001; Marshall and van der Lely, 2006), and now we
have shown that this deficit impacts on syntax too, in the parsing of passive sentences.
Au
5. Conclusions
The evidence we have presented from this study supports the Computational Grammatical
Complexity (CGC) account of G-SLI (Marshall, 2004; Marshall and van der Lely, in press; van
der Lely, 2005). The CGC Hypothesis provides a novel viewpoint for considering the deficit in
SLI. Rather than searching for a single locus for the deficit, e.g. in lower level sensory processing
or a deficit affecting tense, the CGC Hypothesis considers independent deficits in the different
components of grammar and their cumulative and interactive effects on different linguistic
structures. We have previously shown that for tense, syntactic, morphological and phonological
deficits all impact on tense, because regular past tense formation in English requires syntactic,
C. Marshall et al. / Lingua 117 (2007) 1434–1447
1445
morphological and phonological complexity. In this paper we have shown that for passive
comprehension, a linguistic structure that is already vulnerable in G-SLI because of its syntactic
complexity, the morphological deficit also has an impact. Furthermore, we suggest that studies of
typical language acquisition also need to consider the interaction between different levels of
linguistic representation.
py
Acknowledgements
Appendix A. List of verbs
Word
Legal
kicked
stroked
kissed
slapped
called
licked
smacked
chased
pulled
stopped
r's
pe
rs
on
Condition
al
co
We thank the children who participated in this study, and the speech and language therapists,
teachers and parents at Moor House, Dawn House, Whitefield School, Wortley High School, St.
Catherine’s School, St. John’s Upper Holloway C.E. Primary School, St. Andrews (Barnsbury)
C.E. Primary School, Radlett School and Southgate School for their help and cooperation. TM,
HvdL and this work were supported by the Wellcome Trust (no: 063713), and CM by the ESRC
(RES-000-23-0575) grants to HvdL. Finally, we would like to thank Ellie Thom, Hillit Hayan,
Sophie Tang and Nichola Gallon for helping with the data collection.
Au
th
o
Illegal
bathed
scrubbed
brushed
touched
warmed
splashed
washed
hugged
pushed
watched
References
Bishop, D.V.M., 1979. Comprehension in developmental language disorders. Developmental Medicine and Child
Neurology 21, 225–238.
Bishop, D.V.M., 1989. Test of Reception Grammar (TROG). Age & Cognitive Performance Research Centre,
Manchester: Manchester City.
1446
C. Marshall et al. / Lingua 117 (2007) 1434–1447
Au
th
o
r's
pe
rs
on
al
co
py
Bishop, D.V.M., 1997. Uncommon Understanding: Development and Disorders of Language Comprehension. Psychology Press, Hove, UK.
Borer, H., Wexler, K., 1987. The Maturation of Syntax. In: Roeper, T., Williams, E. (Eds.), Parameter-Setting and
Language Acquisition. Reidel, Dordrecht, The Netherlands.
Chomsky, N., 1995. The Minimalist Program. MIT Press, Cambridge, MA.
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.
Clahsen, H., Hadler, M., Weyerts, H., 2004. Speeded production of inflected words in children and adults. Journal of Child
Language 31, 683–712.
Dunn, L.M., Dunn, L.M., Whetton, C., Burley, J., 1997. The British Picture Vocabulary Scale, second ed. NFER-Nelson,
Windsor.
Elliott, C.D., 1996. British Ability Scales, second ed. NFER Nelson, Windsor.
Felser, C., Marinis, T., Clahsen, H., 2003. Children’s processing of ambiguous sentences: a study of relative clause
attachment. Language Acquisition 11, 127–163.
Ferreira, F., Henderson, J.M., Anes, M.D., Weeks Jr., P.A., McFarlane, D.K., 1996. Effects of lexical frequency and
syntactic complexity in spoken-language comprehension: Evidence from the auditory moving-window technique.
Journal of Experimental Psychology: Learning, Memory, and Cognition 22, 324–335.
Fonteneau, E., van der Lely, H.K.J., 2005. The neural correlates of syntactic and semantic processing in typically
developing children and children with G-SLI. In: Presentation at the 10th International Congress for the Study of Child
Language, Freie Universitat of Berlin, 25th–29th July 2005.
Friedmann, N., Novogrodsky, R., 2004. The acquisition of relative clause comprehension in Hebrew: a study of SLI and
normal development. Journal of Child Language 31, 661–681.
Gallon, N., Harris, J., van der Lely, H.K.J., submitted for publication. An investigation of phonological complexity in a
non-word repetition task in children with G-SLI.
Grimshaw, J., 1990. Argument Structure. MIT Press, Cambridge, MA.
Horgan, D., 1978. The development of the full passive. Journal of Child Language 5, 65–80.
Joanisse, M.F., Seidenberg, M.S., 1998. Specific language impairment: a deficit in grammar or processing? Trends in
Cognitive Sciences 2, 240–247.
Karmiloff-Smith, A., 1998. Development itself is the key to understanding developmental disorders. Trends in Cognitive
Sciences 2, 389–398.
Leonard, L., 1998. Children with Specific Language Impairment. MIT Press, Cambridge, MA.
Maratsos, M.P., Fox, D.E., Becker, J.A., Chalkley, M.A., 1985. Semantic restrictions on children’s passives. Cognition 19,
167–191.
Marchman, V.A., Wulfeck, B., Weismer, S.E., 1999. Morphological productivity in children with normal language and
SLI: a study of the English past tense. Journal of Speech Language and Hearing Research 42, 206–219.
Marinis, T., van der Lely, H.K.J., in press. On-line processing of wh-questions in children with G-SLI and typically
developing children. International Journal of Language and Communication Disorders.
Marinis, T, van der Lely, H.K.J., in preparation. Processing of passive sentences in English typically developing children
and children with Grammatical-SLI.
Marshall, C.R., 2004. The morpho-phonological interface in specific language impairment. Unpublished doctoral thesis,
University of London.
Marshall, C.R., van der Lely, H.K.J., 2006. A challenge to current models of past tense inflection: the impact of
phonotactics. Cognition 100, 302–320.
Marshall, C.R., van der Lely, H.K.J., in press. Derivational morphology in children with Grammatical-Specific Language
Impairment. Clinical Linguistics and Phonetics.
Marshall, C.R., Harris, J., van der Lely, H.K.J., 2003. The nature of phonological representations in children with
Grammatical-Specific Language Impairment (G-SLI). In: Hall, D., Markopoulos, T., Salamoura, A., Skoufaki, S.
(Eds.), Proceedings of the University of Cambridge First Postgraduate Conference in Language Research, Cambridge:
Cambridge Institute of Language Research, pp. 511–517.
Mills, A.E., 1986. The acquisition of German. In: Slobin, D. (Ed.), The Cross-Linguistic Study of Language Acquisition.
Lawrence Erlbaum Associates, Hillsdale, NJ.
Montgomery, J., 2005. Real-time language processing in school age children with specific language impairment.
International Journal of Language and Communication Disorders 40, 171–188.
Montgomery, J., Leonard, L., 1998. Real-time inflectional processing by children with specific language impairment:
effects of phonetic substance. Journal of Speech, Language, and Hearing Research 41, 1432–1443.
Pinker, S., Ullman, M., 2002. The past and future of the past tense. Trends in Cognitive Sciences 6, 456–463.
C. Marshall et al. / Lingua 117 (2007) 1434–1447
1447
Au
th
o
r's
pe
rs
on
al
co
py
Raven, J.C., 1998. Raven’s Progressive Matrices. Oxford Psychologists Press, Ltd., Oxford.
Rice, M.L., Wexler, K., 1996. Towards tense as a clinical marker of specific language impairment in English-speaking
children. Journal of Speech and Hearing Research 39, 1239–1257.
Stavrakaki, S., 2002. A-bar movement constructions in Greek children with SLI: evidence for deficits in the syntactic
component of language. In: Fava, E. (Ed.), Clinical Linguistics: Theory and Applications in Speech Pathology and
Therapy. Current Issues in Linguistic Theory. John Benjamins, Amsterdam & Philadelphia, pp. 131–153.
Tomblin, B., Records, N., Buckwater, P., Zhang, X., Smith, E., O’Brien, M., 1997. Prevalence of specific language
impairment in kindergarten children. Journal of Speech, Language and Hearing Research 40, 1245–1260.
Ullman, M.T., Gopnik, M., 1999. Inflectional morphology in a family with inherited SLI. Applied Psycholinguistics 20,
51–117.
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., 1996a. Specifically language impaired and normally developing children: verbal passive versus
adjectival passive sentence interpretation. Lingua 98, 243–272.
van der Lely, H.K.J., 1996b. The Test of Active and Passive Sentences (TAPS). Available from Author at the Centre for
Developmental Language Disorders and Cognitive Neuroscience. University College London, London, UK.
van der Lely, H.K.J., 1997. Language and cognitive development in a grammatical SLI boy: modularity and innateness.
Journal of Neurolinguistics 10, 75–107.
van der Lely, H.K.J., 1998. SLI in children: movement. Economy and deficits in the computational-syntactic system.
Language Acquisition 7, 161–192.
van der Lely, H.K.J., 2000. Verb Agreement and Tense Test (VATT). Available from Author at the Centre for
Developmental Language Disorders and Cognitive Neuroscience. University College London, London, UK.
van der Lely, H.K.J., 2003. Do heterogeneous SLI deficits need heterogeneous theories? SLI subgroups, G-SLI and the
RDDR hypothesis. In: Levy, Y., Schaeffer, J. (Eds.), Towards a Definition of Specific Language Impairment.
Lawrence Erlbaum, pp. 109–134.
van der Lely, H.K.J., 2005. Domain-specific cognitive systems: insight from grammatical specific language impairment.
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., Christian, V., 2000. Lexical word formation in grammatical SLI children: a grammar-specific or
input-processing deficit? Cognition 75, 33–63.
van der Lely, H.K.J., Dewart, M.H., 1986. Sentence comprehension strategies in specifically language impaired children.
The British Journal of Disorders of Communication 21, 291–306.
van der Lely, H.K.J., Stollwerck, L., 1997. Binding theory and specifically language impaired children. Cognition 62,
245–290.
van der Lely, H.K.J., Ullman, M., 2001. Past tense morphology in specifically language impaired children and normally
developing children. Language and Cognitive Processes 16, 177–217.
van der Lely, H.K.J., Rosen, S., McClelland, A., 1998. Evidence for a grammar specific deficit in children. Current
Biology 8, 1253–1258.
van der Lely, H.K.J., Rosen, S., Adlard, A., 2004. Grammatical language impairment and the specificity of cognitive
domains: relations between auditory and language abilities. Cognition 94, 167–183.
Windsor, J., 2002. Contrasting general and process-specific slowing in language impairment. Topics in Language
Disorders 22, 49–61.