Download 1 In Press, Proceedings of the 4th Workshop on Discourse

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

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

Document related concepts

Junction Grammar wikipedia , lookup

Ukrainian grammar wikipedia , lookup

Polish grammar wikipedia , lookup

Proto-Indo-European verbs wikipedia , lookup

Ancient Greek grammar wikipedia , lookup

Old Irish grammar wikipedia , lookup

Inflection wikipedia , lookup

Preposition and postposition wikipedia , lookup

Modern Hebrew grammar wikipedia , lookup

Chinese grammar wikipedia , lookup

Macedonian grammar wikipedia , lookup

Swedish grammar wikipedia , lookup

Germanic weak verb wikipedia , lookup

Antisymmetry wikipedia , lookup

Scottish Gaelic grammar wikipedia , lookup

Portuguese grammar wikipedia , lookup

Japanese grammar wikipedia , lookup

Latin syntax wikipedia , lookup

Germanic strong verb wikipedia , lookup

Georgian grammar wikipedia , lookup

Sotho verbs wikipedia , lookup

Italian grammar wikipedia , lookup

Lexical semantics wikipedia , lookup

Icelandic grammar wikipedia , lookup

Yiddish grammar wikipedia , lookup

English clause syntax wikipedia , lookup

Hungarian verbs wikipedia , lookup

Kagoshima verb conjugations wikipedia , lookup

Russian grammar wikipedia , lookup

Serbo-Croatian grammar wikipedia , lookup

Spanish grammar wikipedia , lookup

Old English grammar wikipedia , lookup

Pipil grammar wikipedia , lookup

Navajo grammar wikipedia , lookup

Transcript
In Press, Proceedings of the 4th Workshop on Discourse
University of Texas, March 2006
Discourse Modes Across Languages
Carlota S. Smith
University of Texas
This article presents a linguistic study of discourse at the local level. There is a
good deal of research on relatively small stretches of discourse, notably information
structure, anaphora, discourse markers, and discourse relations. The work reported here
considers another aspect of local text structure, the discourse mode. Global discourse
structure is pragmatically based in context and expectation, I take it: discourse of a given
genre represents an activity with its own purpose and understood structure (Levinson
1979/91).1 Therefore the global study of discourse is not primarily linguistic.
The term ‘discourse mode’ refers to local stretches of text that are intuitively
recognizeable as having a particular force and function. I posit five modes: Narrative,
Report, Description, Informative, and Argument. Each makes a different contribution to
a discourse. The list is not exhaustive - it omits conversation and procedural discourse,
for instance - but includes the modes that commonly appear in written texts, the main
focus of the inquiry so far. The discourse modes each have linguistic correlates in
English: a characteristic cluster of distributional patterns and interpretations (Smith
2003). The main distributional factor is the type of situation entity a text introduces into
the universe of discourse.
1
I would like to thank the audience at the workshop for helpful discussion, questions, and
comments.
1
The modes seem quite general: narrative and description, for instance , are surely
universal. Thus we might expect them to be found across languages, mutatis mutandis.
But the key to this notion of mode is linguistic form, so that it can be generalized only
through detailed investigation of each language.
Here I consider evidence for the discourse modes in Mandarin Chinese, Navajo,
and French. I am looking for linguistic correlates, that is, distributional patterns for
situation entities in these languages. Since they are of different families, this study will
give some indication of whether and how the notion of discourse mode generalizes across
languages. I will be particularly interested in the argument mode, which depends on
clausal complement structures that vary widely across languages. The facts of Mandarin
raise an interesting point about linguistic evidence: if a language does not offer different
linguistic forms with distinct possible distributional patterns, frequency of a given pattern
can provide another type of evidence (see §2 below). I will suggest that we recognize
both ‘distributional possibilities’ and ‘actual distributional patterns’ as linguistic
evidence.
Section 1 introduces the notion of discourse modes and, briefly, the relevant
linguistic correlates for English. §2-4 discuss the three languages in turn, with comments
on the notion of distributional correlate; §5 concludes.
§1. Background
As introduction, I give two examples of text passages which shift from one
discourse mode to another. The shifts make salient the difference between one discourse
modes. The passages are excerpted from full texts. Passage (a) shifts from Argument to
2
Narrative at sentence 2, a new paragraph; passage (b) shifts from Information to
Narrative at sentence 3, in the middle of the paragraph.
(a) Argument to Narrative
...1 I feel reasonably certain of the final verdict on the current impeachment affair
because I think history will see it as the climax of a six-year period marred by a
troubling and deepening failure of the Republican party to play within the
established constitutional rules.
2 It was on Election Night 1992, not very far into the evening, that the
Senate minority leader, Bob Dole, hinted at the way his party planned to conduct
itself in the months ahead: it would filibuster any significant legislation the new
Democratic President proposed.
(b) Information to Narrative
1 When a big whale dives, currents set in motion by the passage of so many
tons of flesh come eddying back up in a column that smooths the restless surface
of the sea. 2 Naturalists call this lingering spool of glassy water the whale's
footprint. 3 Out between the Hawaiian islandsof Maui and Lanai, Jim Darling
nosed his small boat into a fresh swirl. 4 The whale that had left it was visible 40
feet below, suspended head down in pure blueness with its 15-foot-long arms, or
flippers, flared out to either side like wings.
The examples come from The New York Times and the National Geographic;2 full
passages of each discourse mode are given in an appendix to this article.
Discourse Modes differ in the class of situation entity that predominates in a
given text passage, and the principle of temporal progression that holds for that passage.
The situation entities have distributional correlates; text progression involves principles
of interpretation. I uses the term ‘text passage’ for text segments long enough to establish
the linguistic features that distinguish a mode; two sentences is probably the minimum
length. This study discusses written texts.
2
Example (a) Alan Ehrenhalt, ‘Hijacking the Rulebook’: The NYTimes, Op-ed page,
December 20, 1998. Example (b): Douglas H. Chadwick, ‘Listening to Humpbacks’,
National Geographic July 1999.
3
A discourse introduces individuals, concepts, situations and times into the universe
of discourse. Each entity is licensed by information in a text. There is a well-known
distinction between event and state situation entities,3 expressed at clause level by the main
verb and its arguments (the verb constellation: Smith 1991/7. Events are dynamic, states
are not. For the discourse modes I have developed a new classification that recognizes
several classes of non-dynamic, stative situations. The extended classification of situation
entities distinguishes the five discourse modes and is perhaps their most important feature.
The three main classes of situation entity are Eventualities, specific events and
states; General Statives, generics and generalizing sentences that express a pattern or
regularity; and Abstract entities, complement clauses that express facts and propositions.
Examples of each appear below:
(1) Eventualities (specific)
a The lobster won the quadrille. Lee rehearsed.
b The cat is on the mat. The Colonel owns the farm.
(Events)
(States)
(2) General statives
Unicorns are mythical creatures; Lions eat meat.
(Generic)
John feeds the cat (every day). I drink coffee in the morning.
(Generalizing)
(3) Abstract entities
I know that Mary refused the offer.
(Fact complement)
I believe that Mary refused the offer.
(Proposition complemnt)
The complement clauses in (3) express a fact and a proposition; the sentences in which
they appear are stative. The three classes of situation entity are significant for discourse
modes because different entities predominate in passages of each mode, as in (4):
3
The analysis of English that I sketch in this article is developed more formally in the
framework of Discourse Representation Theory in Smith 2003.
4
(4) Predominant classes of entities and the discourse modes
Narrative:
Report:
Description:
Information:
Argument:
Eventualities
Eventualities, General statives
States, ongoing events, atelic events
General statives
Abstract entities, General statives
Different principles of text progression also distingushes the discourse modes. There are
three, each involving a different interpretation of tense.4 In narrative, situations are
related to each other, so that the function of tense is to indicate continuity.5 In
description, time is static and situations relate to an established time: tense is anaphoric.
In reports and other modes, situations relate to Speech Time and tense is deictic.
(5) Principles of temporal progression
Narrative:
Report:
Description:
Information:
Argument:
Eventualities related to each other, tense continuity
Situations related to speech time, tense deictic
Time a previous time, tense anaphoric
Atemporal, tense deictic
Atemporal, tense deictic
For discussion of tense and temporal progression, see Smith 2003, 2004. I concentrate in
this article on the situation entities.
The situation entities have distributional correlates in English, so that the
conceptual categories have linguistic expression. The linguistic correlates for events and
states are well known for English and some other languages as well (Vendler 1958,
4
For languages without tense such as Mandarin, the principles of text progression are the
same, but expressed differently; see Smith & Erbaugh 2004.
5
Relative tenses also give temporal information: perfect tenses convey that a situation
precedes Reference Time, embedded futures that a situation follows Reference Time.
5
Dowty 1979, Smith 1991/7); I will assume them here. Some of these distributional
correlates are essentially semantic and hold across languages. Dynamism, for instance, is
closely associated with the notion of agency, so that adverbs of intention and control are
natural with event clauses, odd with statives: Mary carefully fed the cat, ?*John carefully
knew Greek. Another correlate of dynamism requires that a complement clause have a
volitional main verb, feed in the first exmple above. Others are syntactic, depending on
the particulars of a given language, for instance the English progressive, an imperfective,
is limited to non-statives in unmarked cases.
Generalizing statives involves classes and patterns of situations, rather than
particular individuals or situations (Krifka et al 1995). Generic sentences refer to classes,
e.g. unicorns and lions, as in (2); generalizing, or habitual sentences express a pattern of
situations, e.g. cat-feeding and coffee-drinking in (3). One distributional correlate in
English for generic sentences is a bare noun in subject position or the possibility of a bare
noun. Generalizing sentences have a frequency adverb or the possibility of a frequency
adverb. Both types of generalizing stative have a strong tense-aspect correlate in English:
if the clause has present tense, the verb form is simple rather than progressive. The verb
constellation is often dynamic - Lions eat meat, John feeds the cat. Stative verbs also
appear in generalizing sentences - Mary is often in love, Kate usually knows the answer.6
6
Generalizing statives pattern distributinally with states rather than events - for instance, in
appearing with simple non-progressive verb forms. However they have some of the
distributional properties of dynamism (Smith 1991/7). They can appear with forms
associated with agency and control, and with pseudo-cleft do. These distributional facts
reflect the hybrid nature of generalizing sentences. Although stative, they often have
dynamic verb constellations (e.g., play tennis) and they involve a pattern of dynamic
events.
6
Now consider the abstract entities expressed by Fact and Proposition complements.
They are situation entities introduced by verb constellations in clausal complements of certain
predicates. Clausal complements referring to facts and propositions have characteristic
distributional and other linguistic features, and thus - like the other classes of situation entities
- function as covert linguistic categories (Vendler 1967, Asher 1993, Peterson 1997).
Abstract entities differ from the other types of situations in how they relate to the world.
Eventualities and general statives are located spatially and temporally in the world; abstract
entities are not.
Facts are the objects of knowledge. Although not spatiotemporally located, they are
contingent for truth on situations being a certain way and arguably have causal powers. We
point out a fact, regret or rejoice in a fact. The class of Factive predicates includes such
verbs as matter, amuse, surprise, explain, realize, remember, discover, know, hear; be
tragic, important, significant, crazy, odd, mysterious, result, fact, explanation, consequence,
reason, upshot, etc.7 Questions of fact are empirical questions although facts are not part of
the furniture of the world.
Propositional complements are the objects of belief, the contents of mental states
like beliefs, expectations, decisions, intentions. Propositions are unlocated, are not
contingent, and do not have causal powers. Propositions are referentially opaque; Vendler
points out that this property reveals their subjectivity. This notion of proposition should
not be confused with the more general use of the term, in which a proposition is the
content that a sentence expresses, the sense of the sentence. Typically propositions appear
7
Kiparsky & Kiparsky 1970 explored the syntax and semantics of factive verbs, verbs
that take fact clauses as complements.
7
as clausal arguments of such predicates as believe, doubt, fear, hope, want, think, affirm,
deny; be unlikely, consistent.
English has three complement clause structures: that-clause (that the Germans
collapsed), gerundives (the Germans’ collapsing), complex nominals (the Germans’
collapse) - as well as infinitives, not treated here. Gerundives and nominals appear with
verb classes of different types, but that-clauses have a limited distribution. Event
predicates do not allow that-clauses, whereas Fact and Poposition predicates do allow
them, as in these examples from Vendler 1967.
(6) Clausal complements
a)*That Mary refused the offer was followed by silence.
b) I know that the German war effort collapsed.
c) That Mary refused the offer is unlikely.
(Event)
(Fact)
(Proposition)
In clear cases like these, that-complement form and predicate classes correlate nicely.
However, the distributional pattern isn’t entirely reliable: certain classes of predicates are
flexible, counter to prediction, allowing that-clauses, gerundives, and nominals.8
Given this difficulty, Peterson 1997 proposed a substitution test to distinguish
clausal complements that express events, facts, and propositions. That-clauses and
8
Flexible verb classes include epistemic predicates (be unlikely) and psychological verbs
(surprise, horrify): they allow substantive, gerundive or a that-clause complements.
(i)
a The collapse of the Germans was unlikely.
b The German's collapsing was unlikely.
c That the Germans collapsed was unlikely.
(ii)
a The enemy's destruction of the city horrified us.
b The enemy's destroying the city horrified us.
c That the enemy destroyed the city horrified us.
Psychological predicates with a substantive are ambiguous between an event or fact
interpretation. For instance, the clausal subject of iiJohn's singing of the Marseillaise
surprised me may be interpreted either way.
8
indirect questions present three distinct patterns of grammaticality for predicates allowing
clausal complements, Peterson showed. The test substitutes a that-clause and an indirect
question for a complement, then checks for grammaticality. Fact predicates allow thatclauses and indirect questions; propositional predicates allow that-clauses, but not
indirect questions; event predicates allow neither. The examples illustrate:
(7) Substitution test patterns
Event
(a) *That Harold read the poem at school followed Mary’s singing.
(b) *What Harold did followed Mary’s singing
Fact
(c) I know that Harold read the letter too fast.
(d) I know how Harold read the letter.
Proposition
(e) ) It appears that Harold read the letter too fast.
(f) *It appears how Harold read the letter
The ill-formedness of indirect questions with verbs taking event and propositional
complements is semantic: indirect questions are semantically incompatible with such
verbs.9 The tests were applied successfully to Arabic, English, French, German, Hebrew,
Hindi, Kannada, and Marathi, using the closest equivalents to that-clauses and indirect
questions. Peterson claims that the categories are universally linguistic and conceptual.
9
The semantic incompatibility comes out clearly with an explicit question paraphrase,
for instance The answer to the question ‘What did Harold do?’ followed Mary’s singing
(Baker 1989). Compare to a similar paraphrase with know: I know the answer to the
question ‘What did Harold do’?
The indirect question complement must be differentiated from a ‘free relative’
such as the object nominal in Mary ate what John cooked. Free relatives are simply
nominals with reference, not indirect questions. The forms of the two overlap only
partially in English, but not necessarily in other languages. For instance, in English the
wh-form how many appears in indirect questions but not in free relatives.
9
Not all classes of verbs are well-behaved even with these tests, notably verbs of
communication and emotion.10
I add modals to the list of predicates that take abstract entities as complements:
they appear in English as auxiliary verbs (can, must etc.) and as adverbials (probably,
certainly, etc). Modals embed unrealized propositions, as Asher 1993 notes. This is an
important class of forms that widens the class of abstract entities that can be recognized
linguistically according to their distributional possibilities. Note that discourse also
expresses facts and propositions that are not clausal or modal complements. They cannot
be recognized linguistically.
The body of this article considers linguistic correlates of the main classes of
situation entity in Mandarin, Navajo, and French. The discourse modes can be recognized
if a language has linguistic corrrelates for these classes. In this article I concentrate on
abstract entities and the Argument mode in each language. For an application of the idea
of temporal progression to Mandarin see Smith & Erbaugh 2004.
§2 Mandarin Chinese
§2.1
Linguistic evidence
The evidence given for situation entities in English is distributional. When one class
of forms reliably co-occurs or fails to co-occur with another, we have a distributional
pattern, or correlation. Among eventualities, for instance, event verbs appear with the
10
Some exceptions can be explained, others cannot. Communication verbs allow indirect
questions in nonfactive uses, counter to his prediction, Peterson 1997 notes. He suggests
that in nonfactive uses with such the alleged indirect question is spurious, really a quotation.
Not all classes of counter-examples can be explained away, however.
10
progressive, whereas stative verbs do not; generalizing sentences allow frequency adverbs,
eventuality sentences do not. The correlates of abstract entities are clausal complements of
different forms. The forms vary in English, and the different forms pattern quite nicely with
certain verb classes, as noted above. The co-occurrence patterns are useful in distinguishing
fact and propositional clausal complements from event complements in a signficant number
of cases.
The approach cannot be applied directly to Mandarin clausal complements, because
there is one form of clausal complement, and one nominal form. Not surprisingly, many
classes of verbs appear with these forms. The distributional possibilities of such
complements do not correlate with particular classes of verbs.
In this case we must look for another type of linguistic evidence. I will show that
frequency provides it. In actual texts, we find that clausal complements tend to appear with
particular classes of verbs - although they can in principle appear with other classes.
Looking at actual texts, we find that the consistent choices people make honor the
distinctions between fact, propositional, and event complements. In an internet study, Fei
Ren found a striking difference in frequency of complements with certain verb classes: the
frequency pattern mimics the pattern of complement distribution in English (Ren 2005).
Complement clauses are much more frequent for verbs with fact and propositional
complements than for verbs with event complements. Thus pattern of use can provide
interesting linguistic evidence. I will return to this point below.
§2,2 Mandarin distributional correlates for situation entities
11
There are clear distributional correlates in Mandarin for two of the classes of
situation entities, Eventualities and General Statives.
Eventualities are events and particular states. Events are dynamic and
distinguishable in Mandarin. The semantic correlates for dynamism are like those of
English and other languages.11 One strong Mandarin-specific correlate is the verbal
suffix –le, a perfective aspectual viewpoint marker. Perfective –le appears only with
event clauses. Mandarin also has a linguistic correlate for ongoing events: the
progressive morpheme zai, similar to the English progressive. For examples see Smith
1991/78, Smith & Erbaugh 2004.
General statives are recognizable in Mandarin. Mandarin is a classifier language and
certain classifiers appear with generics: they are zho&ng or lei$ meaning ‘type’ or ‘sort’, as in
(8a). However a generic sentence need not have one of these classifiers, as in (8b) shows.12
Grammatical morphemes are glossed with small capitals.
(8) a zhe $ zho&ng yu@ ti& xi@ng xi$cha@ng e@r gua#ng hua@, cha@ng yue# li@mi&
This CL fish shape thin long and smooth, length about 15cm.
As for this type of fish, its shape is thin, long, and smooth, about 15 cm. long’.
b lu$ sha$ng de do$ngwu ru@ she@, bia&nfu@, qi#ngwa#, ci$wei$, xio@ng de&ng
Land above DE animal such as snake, bat, frog, hedgehog, bear etc
dou# you& do#ngmia@n de xi@xi$ng.
all have hibernate DE characteristic.13
11
For instance, dynamic predicates can be embedded under verbs of intention, whereas
statives cannot. See Smith 1991/7 for additional tests.
12
These and some of the other examples in this article are due to the internet corpus
search by Fei Ren; I thank her for her assistance.
13
is a grammatical morpheme which appears in several complex nominal
constriuctions.
DE
12
‘Land-dwelling animals such as snakes, bats, frogs, hedgehogs, and bears, etc., all
have the characteristic of hibernation / all hibernate’.
The actual appearance of a generic classifier is not a reliable clue for a generic sentence.
The possibility can be used as a test, however: a sentence is generic if it allows a generic
classifier.
Frequency adverbs are the main cue to generalizing sentences. They can but need
not appear. (9b) is both a generic and a generalizing sentence: it refers to tigers as a class
and also expresses a pattern of events
(9) a. dia$nhua$ li@ngshe#ng ta# cha@ng ti#ng da$o xi#nfa@n
telephone ring
he often listen until irritated
He often got irritated with the telephone ring.
b. lao&hu& ye& ji@shao& sha#nghai re@nlei$
tiger also rarely hurt
human beings
Tigers rarely hurt human beings either.
Again, the possibility of frequency adverbs can be used as a test. This is like
generalizing stative sentences in English.
Generalizing sentences are stative, so they do not appear with event correlates
such as the perfective suffix -le (Smith & Erbaugh 2004).
The pattern of linguistic correlates for eventualities and general statives enables
us can recognize the discourse modes of Narrative, Report, Decription, and Information
in Mandarin.
§2.3 Linguistic correlates for abstract entities
13
Recall that the English correlates for clauses expressing facts and propositions
depend on the distribution of complement clause structures. Mandarin has one clausal
complement structure and one complex nominal structure; both appear with a wide range
of predicates. Nominals appear freely in many distributional environments, and are not
discussed further. Modals have the form of verbal auxiliaries and are followed by a verb
and complement, as in several of the examples below.
Clausal complements appear in topic, subject or object position; there are no
complementizers. The examples illustrate:
(10)
a. xia#o li& xhi#dao$ ci$wei$ yo&u do#ng mia@n de xi@xi$ng
Lee know
hedgehog have hibernate DE characteristic.
Lee knows that hedgehogs hibernate.
b. ta#men que$xi$n sha#n li@n li& suo& you& de cai@fu$ do&u shu&yu@ sha#n she@n
They convince mtn forest in any DE treasure all belong mtn god.
They were convinced that all the treasure in the mountain forest belonged
to the mountain god.
Mandarin clausal complements appear as eventive, fact, and propositional complements
of verbs. For instance, the examples in (11) have event clausal complement arguments
the verbs chu#xi$n (occur) and fa#sheng (take place); (12) and (13) have fact and
propositional clausal complements, respectively.
(11) Event complements
a. jia#ti@ng ba$oli$ to#ngcha@ng shi$ zha$ngfu# o#uda& qi#zi& zui$ji$n she$nzhi$ chu#xia$n
[family violence usually be]
husbands beat wives;
recently even occur
qi#zi& o#uda& zha$ngfu#.
wife beat husband.
Husbands beat wives; recently it even occurs that wives beat husbands.
b. zuo@ri$ zho#ng wu& shi# e$r shi@ xu&, tu#ra@n fa#sheng
yesterday noon 12 o’clock around, suddenly take place
14
sha#n li@n huo& sha&o sha#n.
mountain forest fire burn mountain.
Yesterday, at about 12 o’clock, it suddenly took place that a forest fire burned on
the mountain.
(12) Fact complement
ni@n zuo$. wo& zha$n zhe mei@ gua#nxi#
You sit. I
stand ZHE not matter.
Please sit down. That I am standing doesn’t matter.
(13) Propositional complements
a. ge#nju$ xia$nyou& ke#ji$ gai&za$o huo& xi@ng shi$ ke&ne@ng de
on basis available science & technology transform Mars be possible DE
On the basis of available science and technology, it’s possible that Mars could be
transformed
b pe@iya&ng chua$ngza$o xi@ng re@nca@i shi$ ke&ne@ng de
Train
creativity talent be possible DE
That creativity could be trained with talent is possible.
The examples above show that we can’t rely on the distribution of clausal complements
to distinguish fact and propositional complements from event complements.
Substitution: Can we apply Peterson’s tests to Mandarin? Recall that Peterson
used that-clauses and indirect questions to distinguish event, fact, and propositional
complements. The closest equivalent to a that-clause is the clausal complement, as in the
examples above. Since such complement can appear with verbs of all relevant classes,
their distributional possibiities do not allow the substitution test.
15
Indirect question complements are problematic in Mandarin for a different reason:
they are not distinguishable by form from other types of phrases.14 Clauses with whwords may convey an indefinite rather than an indirect question, as in (14).
(14) ta# co@ngla@i bu$ xia#ngxi$n zi$ji& hui$ ai$sha$ng (she@i)
S/he never not believe self HUI fall in love (who)
S/he never believes s/he will fall in love with anybody
Thus clauses with the same surface form may be indirect questions or indefinite,
depending on the main verb, as (15) illustrates.
(15)a. john zhi#da$o wo& fu$qi#n chi#le she@nme
John know
I father eat-LE what.
John knows what my father ate
b. john xia#ngxi$n wo& fu$qi#n chi#le she@nme
John believe
I father eat -LE what.
John believes my father ate something.
In (15a) the complement is an indirect question; in (15b) it is a clause with an indefinite.
There are other difficulties with the notion of an indirect question complement as
a particular form in Mandarin. Topic structures are often preferred when a sentence has
an indirect question, but it’s not clear how such structures fit the Peterson pattern.
Sentences with indirect question complements are often odd becauses Mandarin strongly
prefers iconic word order. Finally, indirect complements appear only with a few verbs,
as Ren’s internet study shows. There are very few clear cases of indirect questions. Thus
we cannot use either of Peterson’s substitution tests in seeking distributional correlates
for abstract entities in Mandarin.
14
I thank Mary Erbaugh, Fei Ren, and Jianhua Hu for helpful discussion of this matter.
16
We need another notion of distribution to deal with this situation. The key fact:
verbs that are semantically proposition- and fact-embedding tend to appear with
complement clauses in actual use, where as verbs that are semantically event-embedding
do not. The adverbs that one finds in such sentences bear out this claim. Thus the
internet study shows that the distributional patterns of use distinguish proposition- and
fact-embedding predicates from event-embedding predicates.
I shall say that the actual distributional pattern of use in Mandarin of clausal
complements distinguishes between event, fact, propositional complements, although the
possible distributional pattern does not.
In her internet study, Ren looked at Mandarin texts of several genres, including
news reports, editorials, novels, descriptions such as instruction for installing software,
and informal texts such as conversation lines in BBS. Generally, news reports
predominated. Ren found that proposition-embedding verbs appear with complement
clauses most frequently (average percentage 33%); fact-embedding verbs appear with
complement clauses quite frequently (average 10%); event-embedding verbs appear
rarely with complement clauses (average 2% or less).15
15
A full discussion of Ren’s work will appear in Smith & Ren, forthcoming. For Ren’s
internet study, she looked up each word on ‘google’, and went through the first 50-300
items returned. The frequencies for clauses of different types are only approximate.
Among the 50-300 items returned, some repeatedly occur, some contain more than one
sentence in which the given verb or phrase takes clausal complements, others may
contain the given word used as a noun, an adjective or an adverb. If a verb or phrase
allows clausal complements, but no indirect question complements were found, she
googled the verb with a ‘Wh word’ to see whether indirect question complements are
possible.
17
Ren, a native speaker of Mandarin, also constructed sentences of each type and
found that it was easiest to use proposition-embedding verbs with complement clauses,
more difficult to use fact-embedding verbs with complement clauses, and very difficult to
use event-embedding verbs complement clauses.
Further, in an informal study, my colleagues and I were able to distinguish
passages that express the argument mode from passages of narrative, report, description
and information with very good agreement.16 Argument mode passages in Mandarin
have relatively many complement clauses expressing facts and propositions, as well as
modal auxiliaries. The passages examplify Ren’s important findings about the frequency
of complement clases with verbs that take fact and propositional complements. I cite
here one passage - the beginning of an editorial from a Hong Kong newspaper:17
16) Editorial, December 6, 1999. Ming Pao Daily News:
Ga&ng fu#xu& ji@qu& xi#ya&tu@ sa#o lua$n jia$oxu$n
Kong government should learn Seattle riot lesson
Hong Kong government should learn lesson from Seattle
i) mei&guo@ xi#ya&tu@ ma$oyi$ zu& zhi# bu$zha&ng hui$yi$
America Seattle trade organization minister meeting
sui#ra@n qu& zho#@ng re@n sa$n, da$nshi$ ci& hui$yi$
although song end people depart, but this meeting
za$i xi#ya&tu@ yin&fa# de sa#olua$n que$ fa#re@n she#nxi&ng
at Seattle give rise to DE riot still give rise to people deep consider.
ii) yu@ lu$n pu&bia$n re$nwe@i, yi&nfa# sao# lua$n de
With discuss common recognize, give rise to riot DE
16
Colleagues Erbaugh and Ren, cf Smith & Erbaugh 2004 for discussion and examples.
17
More detail given in Smith & Erbaugh 2004.
18
yi@ge$ zho$ngya$o yua@nyi#n shi$ CL [ya# li$ tua@nti& ko&ngpa$
one
important reason is pressure group fear
[shi$ ma$o zu&zhi# tui# do$ng de ma$oyi$ zi$yo@uhua$
World Trade Organization promote DE trade free
CL
hui$ yi&nqi& ge$ng xua@nshu# de pi@nfu$ cha# ju$]].
could lead to even more disparate DE poor rich gap.
iii) wo&men re$nwe@i CL [ga#ng fu& yi$ng co@ng zho#ng ji@qu&
we believe
Kong government must from this learn
yo&u gua#n de jia$oxu$n zhe$ngshi$ wei$la@i fa&zha&n ji@
relevant DE lesson face squarely future technological development and
xia#nga&ng zhua&nxi@ng suo& hui$ yo&ufa# de di# ji$shu$
HK economic restructuring that which may arise DE low skill worker
la@ogo#ng shi# ye$ we$nti@ ji@za&o ca&iqu& you& xiao$ jie&jue@
unemployment issue promptly adopt effective resolve measure
cuo$shi# yi& CL [ba&ozhe$ng xia#ngga&ng she$hui$ de we&ndi$ng.]]
in order to guarantee Hong Kong society DE stable [stability].
Concerning the America-Seattle trade organization meeting, although the meeting is
over, the riots that ensued still give rise to serious thought. As commonly recognized,
one important reason the riots broke out is pressure group fear that WTO-advocated
free trade would produce a wider gap between rich and poor. We believe that the Hong
Kong government must learn from this the lesson (that it should) face squarely the issue
of low skill worker unemployment which may arise from future technological
development and economic restructuring of Hong Kong, (and) promptly adopt effective
measures in order to guarantee the stability of Hong Kong society.
This passage, and the evidence about individual sentences from Ren’s corpus study,
show that there is linguistic evidence for the argument mode in Mandarin, assuming that
actual distributional patterns count as evidence.
I conclude that there are linguistic correlates for all five discourse modes in
Mandarin, based on the main classes of situation entities. Eventualities and general
19
statives are distinguished by distributional possibilities: abstract entities are distinguished
by actual distributional patterns.
§3. Navajo
Navajo is an Athabaskan language, with the complex verb word that is
characteristic of this family.18 Verbs have agreement prefixes that agree with their
arguments.19 Nominal arguments are often optional; pronouns are overt only when
emphasized.
There are linguistic corrrelates for two of the main classes of situation entities.
Among eventualities, distributional correlates distinguish events and many statives; for
details see Smith 1991/7, 1996.
General statives include generics and generalizing statives. In generic sentences,
Navajo speakers tend to use an indication of plurality.20
(17)a.Ay¡n¶
t¬oh dei¬chozh.
buffalo grass plural-Impf-eat
Buffaloes eat grass.
18
I thank Ellavina Perkins for editorial work and careful proofreading of this section.
19
I shall assume here that nominals and clausal complements are arguments of the verb.
Navajo is sometimes analyzed as a ‘pronominal argument language’. In this approach
the subject-object verb prefixes are considered to be incorporated pronominal arguments,
while nominals and clausal complements are optional adjuncts (Willie & Jelinek 2000).
See Speas 1993/in press for a convincing critique.
20
These and several other examples are from Fernald, Perkins & Smith 2003. They are
glossed as follows: the leftmost prefix indicates object and is noted 1,2,3 for person; the
2nd leftmost prefix indicates subject, noted the same way. Except for the Future mode,
Navajo does not have tense. There are 7 mode forms: 3 convey aspectual viewpoint
(perfective, imperfective, progressive); 1 conveys futurity, 1 optative; the other 2 are the
Customary and Iterative, discussed below. The modes usually have characteristic
conjugational prefixes and verb stem shapes;
20
b. N¡shdº¶ j¶igo a¬hosh.
mountain lion daytime 3-Impf-sleep
The mountain lion sleeps in the daytime.
For generalizing sentences that express a pattern of events, Navajo has two verb prefixes
that are in principle appropriate. They are the Customary or Habitual mode morpheme,
and the Iterative mode morpheme, available for event verb words.21 They differ formally
in that the Iterative has the prefix ná in Pos 2 of the verb word.
(18) a. Customary
Gohw¢¢h yidlªªh.
coffee 3-3-Cust-drinks
S/he usually drinks coffee.
b. Iterative
Gohw¢¢h n¢¶dlªªh.
coffee n¡-3-3-Iter-drinks
S/he repeatedly drinks coffee
Generalizing statives can also be expressed with the frequency adverb ¬eh (usually,
customarily) available to all verb words, event and stative. Other frequency adverbs
include: ¬¡h¡da (rarely), t’¡¡ ’ah££ (often), etc.
21
The status of the Customary and Iterative modes is unclear: they are unlike the other
modes because they do not have characteristic conjugational prefixes. The Customary
and Iterative modes appear with the ‘repetitive aspect’ stem shape. Nor are they like other
'aspects', or verb lexeme categories, which have distinct stem-sets and, often, special
prefixes (Young & Morgan 1987, Smith 1995). Following Kari 1990, I would analyze
them as ‘super-aspects’. Super-aspects have distinctive stem-sets and inherit the prefixes
of other aspects; the Customary and Iterative of Navajo fit this characterization. In fact
forms with customary meanings are super-aspects in Ahtna and Koyukon, two other
Athabaskan languages - cf Kari 1990, Axelrod 1993.
The modes are available only for event (non-stative) verb words. Verb words of both
these modes can also be used for simple repeated events (Young & Morgan 1987):
a Dishta’. (3-1-Cust-jerk)
I'm jerking it (e.g. my leg).
b Nisch’i¬. (n¡-1-Iter- blink)
I'm blinking.
21
(19)
a.Gohw¢¢h yishdl£™ ¬eh.
coffee 3-1-Impf-drink usually.
I usually drink coffee.
b.Kwe’¢ ¬¡h¡da nin¡h¡¬tªªh.
here seldom rep-3-Impf-rain
It hardly ever rains here.
The dedicated mode forms are not always used for generalizing statives. In fact,
dispositional and occupational generalizations are usually expressed with the
Imperfective mode, as the next examples illustrate. (19a-b) are dispositional and
occupational general statives, respectively; (19c) is both a generic and an occupational
generalization.
(20)
a.Mary chess yee naan¢.
Mary chess 3-to-her 3-Impf-play
Mary plays chess
b. J¡an b¢¢sh b™™h dah si’¡n¶ y¡ naalnish.
John Tribal Council 3pl-for 3-Impf-work
John works for the Tribal Council.
c. Din¢ s¡anii ak’idahi’ni¬¶ deit¬’º.
Navajo women blankets 3pl-3-Impf-weave
Navajo women weave blankets
These examples show that general statives can be expressed in several ways in Navajo.
Although the Customary or Iterative mode, or a frequency adverb, do not appear in all the
examples, these forms are always possible in such sentences and thus can function as
tests.
Abstract entities: Navajo complement clauses can be formed with the
complementizers – ígíí and –go; and with a zero (ø) complementizer (Elgin 1973,
Schauber 1979). These clauses are fully inflected and can stand on their own as
22
independent sentences. Clauses with –ígíí also appear as relative clauses; –go is a general
subordinator and also appears with subordinating adverbial clauses. The zero
complementizer usually appears with direct discourse object clauses. There are perhaps
subtle differences between the complementizer forms; I shall not discuss them here.
Complement and main clauses can appear in different orders, according to the
information structure of the sentence. Neither – ígíí, –go or ø complement clauses are
comparable to that-clauses because of their wider distribution and range of meanings.
Verbs that take complement clausal adjuncts are mostly fact and propositional
complement verbs, as in (21). In the glosses N indicates a stative verb stem; statives,
known as Neuters in the Navajo literature, do not allow different mode morphemes. The
following examples are from Schauber 1979, except for 21d.22
21) a. Hastiin Kin¬¡n¶d§§’ yºº’ an¡alwodgo baa ¡konisin
man Flagstaff-from away 3-Perf-run ø 3-of 1-N-aware
I’m aware that the man ran away from Flagstaff
b. Kee yºº’eelwod¶g¶¶ bi¬ b¢¢hºzin
K 3-Perf-runaway-IGII 3-with 3-about-N-know
S/he knows that Kee ran away
c. B¶l Kin¬¡n¶d¶ naalnishgo Mary bi¬ y¡’¡t’¢¢h
Bill Flagstaff-at 3-Impf-work-GO Mary 3-with 3-N-be good
Mary likes it that Bill works in Flagstaff
d. Din¢ bizaad b¶hoo¬’aahgo b¶ighah
Navajo language 3-2-Impf-learn- GO 3-N-possible/enough
You learning Navajo is possible
22
21d was supplied by a native speaker of Navajo at the summer 2006 meeting of the
Navajo Language Academy in Flagstaff, Arizona.
23
Complement clauses also occur with verbs that take eventive complements; such verbs do
not allow that-clauses in English. For instance, I elicited the following sentences from
Navajo speakers:
22) a. Yisk£™go Ted bidib¢ tadig¢¢sh¶g¶¶ haha¬zh¶¶sh
tomorrow Ted 3poss-sheep 3pl-3-Impf-shear+IGII 3-Fut-begin
Tomorrow the shearing of Ted’s sheep will begin
b. J¡an dzi¬di
yºº’’¶¶y¡ago ‘¡¬ah ‘aleeh yineest¬’ah
John mountain-at away 3-Perf-gooutofsight+GO meeting 3-3-P-delay
John’s getting lost in the mountain delayed the meeting
c. J¡an hataa¬go bi¬ n¡¡s ahoolzhiizh
John 3-Impf-sing+GO 3-with continuing 3-Perf-time pass
John’s singing went on and on.
These examples show that the distributional possibilities of clausal complements include
verbs taking event, factive, and propositional complements.
The -go complementizer appears in adverbial clauses, with a wide range of
meanings including because, when, if. An adverbial clause can indicate an evidential,
proposition-like frame comment to the main clause. The following example comes from
a Navajo text in which the author tells of her early life:23
(23) ‘‡¶d££’ naad££’ t’¢iy¡ agh¡ nahalingo daad££ `t’¢¢’
that time corn only more 3-N-seem+GO 3pl-Impf-eat past
It seemed like corn was eaten as the main staple at this time.
The adverbial clause precedes the main verb and is marked with the -go complementizer;
it functions as propositional, as the translation suggests.
This example is from a Navajo text by Nancy Woodman, nd: T’¡¡ Bita’¶gºº Njigh¡¡
Nt’¢ego Hahane’ (One person’s story about not having a place to g)o
23
24
Thus the distributional possibilities of complement clauses are not distinct for
abstract entities and eventives in Navajo.
What about the Peterson substitution tests? Recall that they rely on indirect
question complements and that-complements. Indirect question complements in Navajo
appear with and without an overt wh-word:
(24)a. J¡an shizh¢’¢ yiy¶¶y£’¶g¶¶ bi¬ b¢¢hºzin
1-father bread 3-3-P-eat+IGII John 3-with 3-N-be
John knows what my father ate.
b. M¢rii d¶kw¶¶shªª dib¢ nayiisnii’¶g¶¶ shi¬ b¢¢hºzin.
Mary howmany sheep 3-3-Perf-buy-GO 1-with 3-N-know
I know how many sheep Mary bought.
(24a) has no surface wh form; the direct object is coded on the verb with the prefix yi(from Schauber 1979). (24b) has the wh-word d¶kw¶¶shªª (how many).
The semantics of indirect questions precludes their appearance with propositional
complement verbs or event complement verbs, as shown above for Mandarin as well as
English. The same holds in Navajo: indirect questions are not possible with propositional
complement verbs or event complement verbs, as (25) illustrates for a propositional
comlement verb -dl£ (believe) and an event complement verb -y££ (eat).
(25) a. ?*J¡an shizh¢’¢ yiy¶¶y£’¶g¶¶ yoodl£
1-father bread 3-3-P-eat+IGII John 3-N-believe
John believes what my father ate.
b. ?*Ella ha¶ bi chid¶ baa nahashnii’¶g¶¶ wooshdl£
Ella who car 3-from pref-3-Perf-buy+IGII 3-1-I-believe
I believe who bought Ella’s car
c. Mary d¶kw¶¶shªª ¡ts££ yist’¢h¶g¶¶ J¡an yiy¶¶y££
Mary how many ribs 3-3-Perf-grill John 3-3-Perf-eat
I don’t know how many ribs John ate that mary grilled
25
(25c) is interpreted as a tacit indirect question, translated with know.24 The indirect
question sentences are odd in Navajo and English.
The Peterson substitution test comes out partly as predicted: complement clauses
and indirect questions are possible with verbs taking factive complements. With verbs
taking propositional complements, complement clauses are possible but indirect
questions are not. However, contrary to the Peterson predictions, verbs with event
complements allow the same forms of complement clauses.
Modals have propositional complements and thus can contribute to the Argument
mode. In Navajo modality can be expressed directly and indirectly by verbs of ability,
possibility, permission, and reporting; by the optative mode; and by sentence particles.
These forms are indeterminate in meaning. In some cases they expresss modality, in
others they do not; Willie 1996 provides detailed discussion and examples. Modal
meaning can also be implied by the future morpheme, as in the following:25
26) a T'¡¡ nih¶ ‘am¡dºº ‘azh¢h'¢ daniidl¶n¶g¶¶ niha'¡¬ch¶n¶
just we mothers and fathers 3pl-N-be+IGII 3+child
bee bich'•'
y¡deilti'go t'¢iy¡ t'¡¡ bi¬ nilª n¶dadoodlee¬
3-with 3-towards 3pl-I-speak-GO only just it-with respect 3pl-Fut-become
(Context: learning Navajo) We, who are mothers and fathers, (need to) talk to our
children about it, only (then) they will respect it.
24
Another Navajo speaker interprets (25c) as : Mary ate how many ribs John grilled
number unknown). Both interpretations are odd when translated directly in English,
because how many does not appear in English free relatives.
This example is from an article by Patricia Johnson, nd: A¬ch¶n¶ Baa ¡h¡y£ (Helping
our children).
25
26
The indirect nature of modal expressions, and the indeterminacy of others, make it
difficult to recognize them on the basis of form.
In sum, the distributional possibilities do not distinguish propositional and fact
clauses from each other, nor from eventive clauses in Navajo. The substitution test with
indirect questions does separate propositional complements from others.
To investigate Navajo written discourse, I studied a relatively small group of
texts. They vary in genre: 6 are narratives, 3 are written version of addresses about the
teaching and learning of Navajo, 3 are articles concerning current topics.26 The culture
does not have a strong written tradition and I was not able to find many texts. There are a
few collections of older texts and some speeches and articles from the 1990s. Navajo
speakers and scholars are working to find older and current texts.27
Only a few predicates with complement clauses actually appear in these texts, the
same few recurring frequently. The most frequent complements are: -n¶i (say, tell), jin¶
(they say), -kees, n¶zin (think), -éego (to be), y¡’¡t’¢¢h (good). Although in principle
there are many possibilities, they are not used - at least in these texts.
I did not find any passages with a predominance of propositional and factive
clausal complements. Such passages signal the Argument mode. The reason for this is
almost certainly cultural: in the Navajo culture direct expressions of opinion are not
highly valued. This is true for many other native American cultures, as the conference
discussion brought out.
26
The texts were translated, morpheme by morpheme, by Lorene Legah, a native speaker
of Navajo.
27
For instance, the Navajo Language Academy has collected some texts; and there are
texts in the possession of individual scholars.
27
To illustrate, I reproduce here a Navajo text. It involves a number of factive,
propositional, and modal notions, but has with relatively few complement clauses. The
glosses: person indicated by 1, 2, 3, 4 (the fourth person is a polite form, similar to ‘one’
in English); pl - plural; -go and –¶g¶¶, complementizers; PRT indicates a particle, usually
not directly translatable into English;
(27) Veterans Organization ataa yit’éego Baa Nts¶n¶kees
Veterans Organization how 3-is-GO 3-about 2-Impf-think
How do you think the Veterans Organization is doing?
By Julius Harvey
a) Kwe’¢ Ts¢ Ntsaa Deez’¡h¶gi silao¬tsoo¶ dah ahooih¶g¶¶ ¢¶ ¡din.
Here Rock point-at
Veterans associate (org) PRT none
Here at Rock Point there is no veterans organization.
b) Jº Din¢ sil¡o¬tsoo¶ nil¶n¶g¶¶ t’¡¡ b¶ doo hazhº’º bi¬
so Navajo veterans 3-N-be+IGII 3-own neg good 3-with
because Navajo beterans are not well-informed about their organization
c) ch’¢t’aah da dºº doo kºt’¢ daan¶I da.
Impf-makeknown and neg it-is 3+-Impf-say neg
and they are not speaking up for themselves.
d) Kodºº Rock Point ¢¶ naakigo nihisil¡o¬tsoo¶ nilª•go nabid¢¢¬kid
from-here Rock Point-from 2-of
our-veterans 3-N-be+GO 3pl-1-Perf-ask
I’ve asked two of our veterans from here at Rock Point
e) HW Begay dºº L Begay ¢¶ t’¡¡ a¬ah ch’¢¢h sil¡o¬tsoo¶ dah ahoojih¶g¶¶
HW Begay and L Begay PRT both trying veterans associate+IGII
HW Begay and L Begay, who jointly tried to organize the veterans association
f) yin¡hodit’¡¡h, sil¡o¬tsoo¶ danil¶n¶g¶¶ doo b¡ yaa n¶daat’ª• da.
3-start-plan, veterans 3+be+IGII neg 3-for 3-about 3pl-Impf-bother neg.
but other veterans are not giving them any support.
g) HW Begay ¢¶ kwe’¢ ¡¬ah n¡’¡dleehj¶ binaanish naat’i’.
HW Begay
here gather meet-side 3-work extend.
As for HW Begay , he works at the Chapter House here.
h) T’ah hod¶¶na’ªd££’ ¢¶ ¡¬ah n¡’¡dleehdºº bee n¶’diij¢ego
28
still earlier+past
gather meet-from it-with 3-Perf-stand-GO
Some time ago, a vote was passed (for a veterans organization);
i) HWBegay n¶l¢¶ Ts¢gh¡hoodz¡n¡di sil¡o¬tsoo¶ ¡¬ah n¶daadleehgo ¡kºº
HW Begaythere Window Rock-at veterans gather 3+Perf-meet+GO there
HWB went back and forth to attend the veterans organization meetings in Window Rock.
k) a¬n¡n¡d¡ahgo ¡¡d§§’ baa n¶dahast’••d¶g¶¶ kodi yaa n¡h¡lne’ doolee¬ y§¢.
rep-Perf-go+GO there-from it-about 3+Perf-discuss-IGII here-at again-Perf-tell fut past
and he would have shared any issuees that were discussed there but
l) sil¡o¬tsoo¶ danil¶n¶g¶¶ doo ¡¬ah n¡’¡dleehj•’ nidakai da;
veterans
3-N-be+IGII neg gather meet-to
3+Impf-go neg
other veterans never came to the chapter meetings (to be informed)
m) bee nin¡h¡t’¡ahgo ch’¢¢h dabi’dºkeed y§∞ ¡¬t’aa bich’•’ an¡hººt’i’go
3-with plan-GO trying 3+-Impf-ask past after-all 3-toward problem-GO
when they were asked to attend yet when they found themselves in need
n) ¡¬ah n¡’¡dleehj•’ de¶lyeed ¬eh, sh¶k¡ ‘at’oolwo¬ dan¶igo.
gather meet-to 3pl-Impf-come usually, 3-Impf-get help 3pl-Impf-say-GO.
they readily ran to the chapter house saying they need assistance.
o) T’ºº ¬a’ sil¡o¬tsoo¶ dah ahoojih¶g¶¶ hºl≠– le’ hwiinidzin nidi ts’¶d¡
of some veterans
up bunch-IGII
exist wish 4pl-N-think but just
We wish that there were a veterans’ organization but it takes
p) d¶kw¶i da sil¡o¬tsoo¶ yilt’¢ego ahi¬ nidaalnishgo t’¢iy¡ ¬ah doon¶¶¬.
several veterans number+GO together 3+Impf-work+GO just some 3-Fut-do.
a good number of veterans working together to make this possible.
q) D¶¶ sil¡o¬tsoo¶ dah ahoojih¶g¶¶ hºl≠–go ¢¶ ¬eeh hwii’n¶¬¶ t’¡¡ sahdii
this veterans
up bunch-IGII 4-N-exist+GO dirt put-away separately
If there were a veterans’ organization it would have been possible to have a piece
r) b¡ k¢yah hadazdoozoh, hosil¡otsoo¶ ho¬ danilª•go.
it-for land up-3+Impf+draw, their veteras it-with 3+-N?- respect+GO
of land set aside for a cemetery with the veterans’ support system.
Here at Rock Point there is no veterans organization, because Navajo beterans are
not well-informed about their organization and they are not speaking up for
themselves. I’ve asked two of our veterans from here at Rock Point. HW Begay
and L Begay, who jointly tried to organize the veterans associatio .but other
veterans are not giving them any support. As for HW Begay , he works at the
Chapter House here. Some time ago, a vote was passed (for a veterans
29
organization); HWB egay went back and forth to attend the veterans organization
meetings in Window Rock. He would have shared any issues that were discussed
there but other veterans never came to the chapter meetings (to be informed)
when they were asked to attend yet when they found themselves in need they
readily ran to the chapter house saying they need assistance. We wish that there
were a veterans’ organization but it takes a good number of veterans working
together to make this possible. If there were a veterans’ organization it would
have been possible to have a piece of land set aside for a cemetery with the
veterans’ support system (if the veterans were working with them).
Note that line (k) conveys a future in past - a modal notion - without an overt modal
form. Similarly, the if-then clauses in lines (q, r) are expressed without overt modals.
I conclude that the Navajo language has resources for clausal complements of a
factive and propositional nature, although they are not grammatically distinguishable
from event complements. However, these resources are not exploited in texts, at least
those that I was able to examine. Thus I failed to find text passages in the argument
mode, although such passages are possible in the language.
§4. French
French is an Indo-European language with a long written tradition. It is
reasonably close to English in structure. Not surprisingly, it has distributional correlates
for situation entities, and passages of the argument mode are easily found. This section
discusses briefly former briefly, and demonstrates the latter, a passage of the Argument
mode.
There are distributional and semantic tests that distinguish clauses of events and
specific states. They are given in Smith 1991/7; I shall not repeat them here.
30
General Statives: generic sentences have a definite article, singular or plural
(28a,b). Generalizing sentences tend to have frequency adverbs. Generalizing, or
habitual sentences are in present tense (28c) or the imparfait, a past tense.
(28) general statives
a. Les oiseaux migrateurs ne sont pas rares.
Migratory bird are not rare.
b. L’avare est malheureux
The miser is unhappy. (can have specific or class reference)
c. Sabine met la table (tous les jours).
Sabine sets the table every day.
As Kfrika et al 1995 note, it’s always possible to add a frequency adverb or to paraphrase
with a general nounphrase, but general statives do not always have particular linguistic
forms that identify them.
Correlates for Abstract Entities: French has verb complement clauses with the
complementizer que, which correspond quite closely to that-clauses. There are no other
complement clause forms. The language has complex nominal clauses with and without
head nouns. The nominals have a wide distribution, appearing with predicates of many
types, like their counterparts in English, Mandarin, and Navajo. (29) illustrates two
complex nominals (a,b) and a clausal complement (c).
(29) Complements in French
a. Le fait que Marie a/ait refusé l'offre a été/était significatif.
The fact that Mary refused the offer was significant.
b. Le refus de l'offre par Marie a été/était significatif.
The refusal of the offer by Marie was significant.
31
c. Que Marie refuse l'offre était incohérént/inconsistant.
That Mary refused the offer was inconsistent.
The subjunctive is required with certain predicates, as in (29a; (29c) is a factive
complement.
Generally, complement clauses appear as complements of fact and propositional
verbs, illustrated in (30). There are also a few event predicates that (causer, retarder)
allow them, as in (31):28
(30)
a Que Marie refuse l'offre présentait des contradictions.
That Mary refused the offer was contradictory.
b. Que Marie refuse l'offre a été/était significatif.
That Mary refused the offer was significant.
c. Que Marie refuse l'offre a été/était cohérent.
That Mary refused the offer was consistent.
d. Que Marie refuse l'offre a été/fut révélateur.
That Mary refused the offer was revealing.
(31)
a. J’ai oublié que Marie a refusé l’offre
I forgot that Mary refused the offer.
b. Que le bateau a coule était de longue durée
It took a long time for the boat to sink.
c. Que Chirac a gagné l'élection ait retardé les pourparlers.
That Chirac won the election delayed the discussions.
d. *Que Chirac a gagné l'élection s'est passé hier.
That Chirac won the election happened yesterday
28
I thank Knud Lambrecht, Sabrina Parent, and her colleagues for native speaker
judgments on these and other sentences in French.
32
e. ?*Que Marie a refusé l’offre était suivi par un silence.
That Mary refused the offer was followed by silence.
I emphasize the grammaticality of the event complements in (31a-c) because they are
somewhat surprising. The ungrammaticality of (31d-e) is what we expect; in the
overwhelming majority of cases, que-complements are factive or propositional.
The substitution tests apply to French. The pattern of fact, proposition, and event
complements holds: fact complements can be que-clauses or indirect questinos;
propositional clauses allow que-clauses but not indirect questions; event complements
allow neither. These are Peterson’s examples (1997: 96):
(32) a. Il était important que le bateau ait coulé rapidement. (Fact)
It was important that the boat sank rapidly.
b. Il paraît que le bateau a coulé rapidement. (Proposition)
It appeared that the boat sank rapidly.
c. Comment le bateau a coulé était mysterieux. (Fact)
How the boat sank was mysterious.
d. *Il paraît ce qui a coulé rapidement. (Proposition)
It appears what sank rapidly
Thus French has fairly reliable linguistic correlates for abstract entities.
Passages in the argument mode are not difficult to find; they appear in books,
magazines, newspapers. I present one short example here. Like the examples from the
other languages, it is a newspaper editorial.
(33) Le Monde Feb 4, 2006
33
Déstabilisation
Les deux raisons du succès du Hamas sont connus. Du retrait israelién de Gaza,
The two reasons for the success of Hamas are known. From the Israeli retreat from Gaza,
après 5 ans de guerre, la plupart des Palestiniens, exsangues, ont conclu
after 5 years of war, most Palestinians, exhausted, concluded
que les Israéliens ne se retirent jamais d’un territoire conquis que devant la force.
that the Israelis would never leave conquered territory except by force.
Ils ont credité les islamistes de cette cohérence. Et ils ont voulu “punir” leur Autorité
They credited the islamists with this logic. And they wanted “to punish” the Authority
pour son incurie, son clientélisme et sa corruption. …Israéliens, Américains et
for its carelessness, its pandering, and its corruption. Israelis, Americans and
Européens tous disaient souhaiter le rédemarrage de la “feuille de route”
Europeans all said that they hoped the re-starting of the “road map”
devant mener à la instauration d’un futur Etat palestinien. Tous espéraient
would lead to the institution of a future state of Palestine. All overtly expected
overtement la défaite des islamistes. Jérusalem et Washington ne voulaient
the defeat of the islamists.
Jerusalem and Washington did not want to
pas les voir au governement.
see them at the helm of government.
The two reasons for the success of Hamas are known. From the Israeli retreat
from Gaza,after 5 years of war, most Palestinians, exhausted, concluded that
that the Israelis would never leave conquered territory except by force. They
credited the islamists with this logic. And they wanted “to punish” the Authority
for its carelessness, its pandering, and its corruption. Israelis, Americans and
Europeans all said that they hoped the re-starting of the “road map” would lead
to the institution of a future state of Palestine. All overtly expected the defeat of
the islamists. Jerusalem and Washington did not want to see them at the helm of
government.
This passage is typical of the argument mode, with several factive and propositional
complement clauses.
34
§5. Conclusion
I have looked at linguistic correlates for three classes of situation entities in
Mandarin Chinese, Navajo, and French. Eventualities and general statives can be
distinguished on classic distributional grounds in these languages.
For the expression of fact and propositional complements, the languages differ
significantly in the linguistic forms available and the grammatical correlates of those
forms. English and French have distributional correlates for the conceptual situation
categories that are quite reliable. The possible distributions of the relevant forms pattern
according to these categories.29
Mandarin Chinese does not have such distributional possibilities; nor are the
forms that realize these categories distinct. However, there are actual distributional
patterns which distinguish fact and propositional complements from event complements.
On the basis of these patterns we can recognize passages of the Argument discourse
mode. I suggest that we recognize actual distributional patterns as a type of linguistic
evidence. Such patterns complement the classic type of distributional evidence, which
relies on possible forms rather than actual occurrence.
In Navajo the situation is different: the forms vary but not according to the
situation entity categories. And there are neither distributional possibilities nor actual
distributional patterns that are distinct for fact and propositional complements.
The Argument mode appears to be robust in English, French and Mandarin;
29
This is apparently true of the other languages that Peterson investigated. They all have
a relatively strong inflectional component, which may be relevant, as Mary Erbaugh has
pointed out to me.
35
but not in Navajo. This may be due to the fact that there is not a long tradition of written
texts in Navajo.
Appendix
Examples of English text passages that realize the five discourse modes. These examples
are discussed at length in Smith 2003.
Narrative
1 A few days later I called on Dr P and his wife at home, with thescore of the
Dichterliebe in my briefcase and a variety of odd objects for the testing of perception. 2
Mrs. P showed me into a lofty apartment, which recalled fin-de-siècle Berlin. 3 A
magnificent old Bösendorfer stood in state in the centre of the room, and all around it
were music stands, instruments, scores. 4 Dr. P came in, a little bowed, and advanced
with outstretched hand to the grandfather clock, but, hearing my voice, corrected himself,
and shook hands with me. 5 We exchanged greetings and chatted a little of current
concerts and performances. 7 Diffidently, I asked him if he would sing.
(Oliver Sacks, 1970. The Man who Mistook His Wife for a Hat. New York: Harper & Row)
Report
1 A week that began in violence ended violently here, with bloody clashes in the
West Bank and Gaza and intensified fighting in Southern Lebanon . 2 Despite the
violence, back-channel talks continued in Sweden. 3 Israeli, Palestinian and American
officials have characterized them as a serious and constructive effort. 4 Israel is offering
as much as 90 percent of the West Bank to Palestinians.
(Barak fights on many fronts. The New York Times May 20, 2000)
Description
...1 On the big land below the house a man was ploughing and shouting admonitions to
the oxen who dragged the ploughshares squeaking through the heavy red soil. 2 On the
track to the station the loaded wagon with its team of sixteen oxen creaked and groaned
while the leader cracked his whip that reached to the horns of the leader oxen and yelled
on a note only they understood. 3 The birds twittered and sang. 4 The wind sang not only in
the wires, but through the grasses, and the wires vibrated and twanged.
(Doris Lessing 1994. Under My Skin. New York: Harper, p 107.)
Information
1 When people try to get a message from one individual to another in the party
game telephone, they usually garble the words beyond recognition. 2 It might seem
surprising, then, that mere molecules inside our cells constantly enact their own version
of telephone without distorting the relayed information in the least. 3 Actually, no one
could survive without such precise signalling in cells.
36
(John Scott & Tony Pawson, 2000, Cell Communication. Scientific American.)
Argument
1 The national outpouring after the Littleton shootings has forced us to confront
something that we have suspected for a long time: the American high school is obsolete
and should be abolished. 2 In the last month, high school students present and past have
come forward with stories about cliques and the artificial intensity of a world defined by
insiders and outsiders, in which the insiders hold sway because of superficial definitions
of good looks and attractiveness, popularity and sports prowess.
(Leon Botstein, Let Teenagers Try Adulthood. The New York Time, June 1999.
References
Asher, Nicholas, 1993. Reference to Abstract Objects in Discourse. Dordrecht: Kluwer.
Axelrod, Melissa, 1993. The Semantics of Time. Lincoln: University of Nebraska Press.
Baker, Carl L., 1989. English Syntax. Cambridge, Mass.: MIT Press.
Dowty, David, 1979. Word Meaning and Montague Grammar. Dordrecht: Reidel.
Elgin, Suzanne, 1973. Some Topics in Navajo Syntax. Ph.D dissertation, University of
California, San Diego.
Fernald, Theodore, Ellavina Perkins, & Carlota S. Smith, 2003. Generalizing sentences
in Navajo. Athabaskan Language Conference, Humboldt State University.
Kari, James. 1990. Ahtna Athabaskan Dictionary. Fairbanks: Alaska Native Language
Center. University of Alaska, Fairbanks.
Kiparsky, Paul, & Carol Kiparsky, 1970. Fact. In M. Bierwisch & K.E. Heidolph (eds),
Progress in Linguistics. The Hague: Mouton.
Krifka, Manfred, Francis J. Pelletier, Gregory Carlson, Alice ter Meulen, Gennaro
Chierchia, & Godehard Link, 1995. Genericity: An introduction. In G. Carlson & F.J.
Pelletier (eds), The Generic Book. Chicago, Illinois: University of Chicago Press.
Levinson, Stephen, 1979/1992. Activity Types in Language. Linguistics 17:356-99.
Peterson, Philip, 1997. Fact, Proposition, Event. Dordrecht, Holland: Kluwer.
Ren, Fei, 2005. Report, Mandarin Chinese verbs and clausal complements. Unpublished
paper, University of Texas.
37
Schauber, Ellen, 1979. The Syntax and Semantics of Questions in Navajo. New York:
Garland.
Smith, Carlota S., 1983. A theory of aspectual choice. Language 59.3, 479-501.
Smith, Carlota S., 1991/7. The Parameter of Aspect. Dordrecht: Kluwer.
Smith, Carlota S., 1996. Aspectual categories in Navajo. International Journal of
American Linguistics, 62:227–263.
Smith, Carlota S., 2003. Modes of discourse. Cambridge: Cambridge University Press.
Smith, Carlota S., 2004. The domain of tense. In J. Guéron and J. Lacarme (eds), The
Syntax of Time. Cambridge, Mass: MIT Press.
Smith, Carlota S., & Mary S. Erbaugh, 2005. Temporal interpretation in Mandarin.
Linguistics 43 713-756.
Smith, Carlota S., Ellavina Perkins, & Ted Fernald, in press (2007). Time in Navajo:
Direct and indirect interpretation. International Journal of American Linguistics.
Smith, Carlota S. & Fei Ren, forthcoming. Complements in Mandarin Chinese.
Speas, Margaret, 1993/in press. Economy, agreement and the representation of null
arguments. In P. Ackema (ed), Agreement and Argument Structure. Oxford: Oxford
University Press.
Vendler, Zeno, 1967. Linguistics in Philosophy. Ithaca, New York: Cornell University
Press.
Vendler, Zeno, 1972. Res Cogitans: An essay in Rational Psychology. Ithaca, New York:
Cornell University Press.
Willie, Mary Ann, 1996. On the expression of modality in Navajo. In E. Jelinek, S.
Midgette, K Rice, & L. Saxon (eds), Athabaskan Language Studies. Albuquerque, New
Mexico: University of New Mexico Press.
Willie, Mary Ann & Eloise Jelinek, 2000. Navajo as a discourse configurational
language. In T. Fernald & P Platero (eds), The Athabaskan Languages. Oxford: Oxford
University Press.
Young, Robert & Willie Morgan, 1987. The Navajo Language. Albuquerque, New
Mexico: University of New Mexico Press.
38