Tải bản đầy đủ - 0 (trang)
Chapter Eight. Constituent Order in Yagua: Conclusions and Implications

Chapter Eight. Constituent Order in Yagua: Conclusions and Implications

Tải bản đầy đủ - 0trang

2 4 0 8. Constituent Order in Yagua: Conclusions and Implications

The order of head noun and descriptive modifier could be argued

against (cf. Chapter 4), but given Hawkins' (1983:13) criteria for determining what is "basic", plus the evidence of naturally occurring

discourse data, I have argued that Head Noun + Descriptive Modifier is

the basic order. I have avoided using the term "adjective" simply

because what functions as a descriptive modifier is most often syntactically a nominal. The Genitive NP + Head Noun pattern is much more

frequent than the Set I clitic + Head Noun + Genitive NP pattern. NP

+ Postposition is similarly much more frequent than Set I clitic + Postposition + NP. There is no reason to assume that the latter orders have

become the norm when full noun phrases are used, either in terms of

frequency or communicative function. Thus, there is no reason to suppose

that the dependent noun phrases in adpositional and possessed noun

phrases are basically post-head, with the Set I clitic just being a sign of

inflectional agreement between the head and its post-head dependent.

Hawkins' (1983:65) Universal 2 states:


If a language has VSO word order, then if the adjective

follows the noun, the genitive follows the noun; i.e.

VSO --> (NAdj ~ > NGen).

This universal is given as an exceptionless one, ruling out the possibility that a language would have either of the following two occurrence

sets as its basic orders:


VSO/Prepositional/NAdj/GenN (Hawkins'Type 4)

VSO/Postpositional/NAdj/GenN (Hawkins' Type 8)

If we understand Yagua to be a VSO language, then it is in fact a

Type 8 language and Hawkins' Universal 2 cannot stand as exceptionless.

Given that Hawkins' proposals are based on a respectable sample of the

world's languages,1 it is worth investigating whether Yagua might not be

an SVO (or even SOV) language. Any order other than VSO would leave

the Universal as exceptionless, at least relative to currently attested


8.1.2. Set I clitic reference

In Chapters 3 and 4 we saw that subjects of Type 1 clauses, genitives in

possessed noun phrases, and objects of postpositions can all be expressed

by a noun phrase, a Set I clitic, or both simultaneously. The three different patterns are summarized in Table 8.1 for all phrasal categories. In the

A pattern a noun phrase precedes the predicate element or the head

8.1. Arguments in favor of SVO as basic 2 4 1

(either a possessed noun or a postposition). In the Β pattern a Set I clitic

precedes that element. In the C pattern a noun phrase follows what would

otherwise be the Β configuration. The term "verb" in Table 8.1 encompasses both semantically main verbs and auxiliaries.

Table 8.1.

Summary of Encoding Possibilities for Subjects of Type 1

clauses, Genitives, and Objects of Postpositions.


SubjNP + Verb

GenNP + Possd Ν

NP + Postp



Set I + Verb

Set I + Possd Ν

Set I + Postp

Set I + Verb + SubjNP

Set I + Possd Ν + GenNP

Set I + Postp + NP

What is the commonality uniting the patterns seen in Table 8.1? One

hypothesis is that the verb, possessed noun, and postposition are all heads

of phrasal categories and that the other element(s) encode the dependent

member of the phrasal category. However, in most frameworks the verb is

not taken to be the head of a verb plus subject constituent, in the same

sense that a head noun and postposition are heads of noun phrases and

postpositional phrases, respectively. The verb might, however, be

understood as the most head-like surface element in the clause in that it

is the constituent to which elements which have scope over the entire

clause may gravitate. 3

A second hypothesis is that the verb, possessed noun, and postposition

are in some sense predicates of their respective phrasal categories, and

that the other element(s) encode an argument of the construction. More

precisely, in each case the predicate is a one-place predicate. If we take a

transitive verb to be a two-place predicate, the addition of an object

argument results in a one-place predicate. In some sense, addition of an

object argument to make a one-place predicate must happen "prior" to

addition of the subject argument. This may be motivated on semanticosyntactic grounds, given that verbs have closer selectional restrictions and

interpretation requirements relative to their objects, as opposed to their

transitive subjects.4 It is important to note that not all one-place

predicates in Yagua take Set I clitics. In particular, arguments of S Q

clauses and of predicate nomináis are morphosyntactically treated in the

same way as objects of transitive verbs.5

If one were to posit SVO as the underlying syntactic order for major

clause constituents, then Set I clitic and noun phrase distribution across

all three phrasal categories could be accounted for simply and neatly by

rules such as the following (or their translations in whatever framework):

242 8. Constituent Order in Yagua: Conclusions and Implications









ARGUMENT : Set I clitic (NP)

Where "A : Β " is read as "A is instantiated by Β "

Rule (a) in (4) cannot be interpreted as a standard X-bar rule

(Jackendoff 1977). This is because when the predicate is a verb (or

auxiliary plus verb), then X cannot be understood as a verb phrase but

must be understood as a clausal category (C2, CI, or C; see Chapter 3).6

A further modification is needed for rule (4c) such that whenever the

"optional" NP occurs, it will be placed in post-predicate position.

Whenever X in rule (a) is a clausal category, the modification to rule (4c)

will have to ensure immediately post-verbal placement of the subject NP,

rather than placement following verb-plus-object. Some sort of simplicity

metric might suggest positing either (4b) or (4c) as the rule accounting

for basic, or syntactically underlying, order of the argument across all

categories. The alternative order might then be derived by a movement

transformation, yielding a more surface structure.

I believe there is major problem with stopping at rules such as those

outlined in the preceding paragraph (or their translations in whatever

framework). Despite their neatness, they ignore what speakers are actually

sensitive to when they (subconsciously) choose a variation such as (4b)

rather than (4c). The rules in (4a-c) might satisfactorily describe the

syntactic possibilities. But they say nothing about what is communicatively

basic. As I have argued in Chapters 4 and 7, the A and Β patterns in

Table 8.1 are communicatively basic for Genitive noun phrases and

postpositional phrases, but the Β and C patterns are communicatively

basic for subject - verb constructions. The A pattern is reflected in rule

(4b), the C pattern in rule (4c), and the Β pattern in rule (4c) without the

optional NP.

Perhaps what is largely at issue here is whether or not the more

typologically oriented tradition (as represented by Greenberg, Hawkins,

Mallinson and Blake, and Givón, to name just a few scholars), or the

syntactic possibilities tradition (as exemplified by much of X-bar syntax

and phrase structure theories) has an exclusive right to the term "basic

constituent [word] order". Clearly neither one does, unless we choose to

disagree with Humpty Dumpty who said: "When I use a word, it means

just what I choose it to mean - neither more nor less". That is, Humpty

Dumpty knew he had a (constitutional?) right to make a term mean

whatever he chose it to mean (unfortunately without regard to increased

8.1. Arguments in favor of SVO as basic 243

communication of his message). As long as we clearly understand what

various writers mean by their terms, perhaps we do not need to argue. But

in my own mind, it is not sufficient to stop with an understanding of

"basic constituent order" only in terms of syntactic possibilities, as

represented in (4).

Failing to explore what is communicatively most neutral or even most

frequent in naturally occurring discourse will leave us with an inadequate

understanding of the specific pathways by which syntactic change may

proceed, in that the grammaticization of discourse-pragmatic patterns

gives rise to a great deal of syntax. In order to account for historical

change, we must allow that languages can have points where they will not

be forced into simple and tidy generalizations. Change proceeds gradually

and does not necessarily affect parallel constructions to the same extent

simultaneously. In this particular instance, we need to allow that change

can proceed unevenly across the different rows in Table 8.1; we need not

assume that all of column A, or alternatively that all of column C, is

"basic". If we insist on simple and tidy generalizations at all points, we

might as well try to maintain that Natural Serialization Principle

(Lehmann 1973; cf. Chapter 2) was right to begin with. But clearly it was


In summary, if one were to posit the A pattern in Table 8.1 as

syntactically underlying for all categories represented, then just rules (4a)

and (b) would be necessary to account for all "basic" or underlying

syntactic orders, while rule (4c) accounts for non-basic orders. However, I

contend that it does not accurately reflect the cognitively and

communicatively basic orders across all categories.

8.13. Subject - object asymmetries

As discussed in Chapter 6, subject - object asymmetries have to do with

phenomena where either the subject or the object, but not both, evidences

certain privileges in terms of such things as order variation, "movement"

out of complement clauses, and control of person and number indices of

anaphoric elements. One possible way to account for subject - object

asymmetries is to posit a structural VP constituent containing the verb

and object noun phrase. This is particularly motivated if the subject object asymmetries in question could be argued to stem directly from the

fact that the subject is immediately dominated by the sentence (or clause),

whereas the object is immediately dominated by the verb phrase. Positing

SVO as the basic (underlying) constituent order would facilitate positing a

structural verb phrase in that the verb and object are then contiguous. In

2 4 4 8. Constituent Order in Yagua: Conclusions and Implications

Sections 3.8.2 and 6.3 I noted two possible subject - object asymmetries in

Yagua: potentially different strategies for questioning subject versus object

arguments of embedded clauses, and the fact that subjects but not objects

can determine the person/number/animacy index of the coreferential

clitics jfy- and -yù. With regard to the latter I have argued that the real

asymmetry is not between subject and object, but between Set I arguments

and object. Positing a structural verb phrase consisting of verb and object

does not help to resolve this problem. With regard to question formation

strategies, I do not have sufficient information to conclusively say that

subjects of embedded clauses are indeed treated differently from objects.

In any case, subject - object asymmetries which may exist could

possibly be accounted for on the basis of closer semantic scope and

subcategorization relations holding between the verb and object, which do

not hold between the verb and (transitive) subject. There is no necessary

reason to argue that subject - object asymmetries have to be accounted

for on the basis of linear order within a structural verb phrase constituent.

8.2. Arguments against SVO as basic

There are several arguments against positing SVO as the basic order in

Yagua, at least within some of the more typological traditions outlined in

Chapter 2. First, the statistical evidence presented in Chapter 7 favors

basic postverbal order for both subject (both S and A) and object (O). As

discussed in Chapter 4 regarding order of genitives, we should not stop at

statistics but should investigate the principles motivating the observed

statistics. In the case of major clausal constituents, postverbal position is

unmarked whereas preverbal position is either pragmatically or

semantically marked. Part of what makes something marked is that it

occurs less frequently. If it were to become the statistical norm, by dint of

sheer frequency it would likely lose its marked status. But low frequency

is not the only thing which makes preverbal position of subject and object

pragmatically marked. As shown in Chapter 7, there are clear non-neutral

communicative intents which correlate with and motivate the preverbal


Second, if we consider the criterion of degree of presuppositionality,

SVO order is employed under conditions of greater presupposition than is

VSO order (and similarly for other orders with preverbal nuclear

constituents). VSO (or VS, VO, or V-Oblique) may be used simply to

introduce information where there is little or nothing presupposed at all.

8.2. Arguments against SVO as basic 2 4 5

But SVO (and all other orders where there are preverbal constituents) are

employed when it is assumed that there is some background

presupposition in the mind of the hearer which the speaker wishes to

modify in some way. This background presupposition provides a context

and raison d'être for focus of contrast, restatement, added detail

restatement, counter expectation, and other non-neutral communicative


Third, if we were to posit SVO as basic, it is still clear that objects,

postpositional and other oblique phrases, adverbs, and discontinuous

elements of noun phrases can also occur in preverbal position. If SVO is

basic, we might expect to be able to find cases of OSV, Oblique-SV,

Adverb-SV, Modifier-SV, etc. where the first element occurs in the

pragmatically marked (PM) position within the C 1 clause. However, these

do not occur. Whenever there are two preverbal constituents, one is

always in the non-nuclear delimiting position. This is shown partly by C 2

second position clitic placement and by resumptive use of Set I or II

clitics whenever the first element is coreferential with an argument of the

verb. If SVO is basic, we need to account in a motivated way for why the

non-occurring orders are missing. Why is there a limit of just one preverbal constituent within C*?

Related to the third objection is the fact that whenever a non-subject

constituent occurs in the preverbal position within C*, if a subject phrase

also occurs in the clause it must follow the verb. If we posit SVO as basic,

we then have to account for why the subject is "moved" or extraposed

whenever something else occurs in preverbal position. Keenan (1977)

claims that in verb medial (SVO) languages, there may be some form of

subject postposing either to the end of the clause, or just to postverbal

position when non-subjects are fronted. This is what one would have to

argue here.

However, there is a fourth objection. The fact remains that no

constituent need occur in preverbal position, and the subject is most

commonly postverbal even when there is no other preverbal constituent

(Chapter 7). What would motivate postposing in this case? If SVO is

basic and underlying, we are faced with the rather uncomfortable

distributional statement that the subject is extraposed to follow the verb

when it is pragmatically unmarked, but is retained in its preverbal position

and not moved whenever it is marked. Counter to this, whenever any nonsubject elements are pragmatically unmarked, they remain in their

underlying position, and are moved only when marked. It would be

simpler to have just one rule: when pragmatically marked, the constituent

in question (regardless of what it is) occurs in the preverbal PM position.

2 4 6 8. Constituent Order in Yagua: Conclusions and Implications

When pragmatically unmarked, the constituent in question remains in its

basic position.7

In Chapter 2 I noted Mallinson and Blake's stipulation that basic order

should be determined on the basis of transitive clauses where both

arguments are definite. Such a definiteness criterion does not distinguish

between SVO and VSO orders, since both configurations can be used to

express definite information or to introduce indefinite-referential

information. There is no necessary relationship between definiteness and

one order versus the other in Yagua. In actual fact, the number of clauses

with two overt noun phrases in natural text meeting either Mallinson and

Blake's or Givón's criteria is extremely small and does not allow us to

conclude anything with certainty.

To summarize, Hawkins* Universal 2 (Section 8.1.1) should not be

taken as a reason to prefer SVO over VSO as the basic order. The

Universal should be based on data and not vice versa. Subject - object

asymmetries (if such exist in Yagua; Section 8.1.3) could be accounted for

on the basis of subcategorization and semantic scope relations, rather

than positing a structural VP constituent. The primary motivation internal

to the grammar of Yagua for positing SVO as basic concerns simplicity of

description relative to distribution of noun phrases and Set I clitics that

code arguments of (certain) one-place predicates, as in Rules (4a-c). But

balanced against this is complication of description when it comes to

conditions of use. According to criteria such as those suggested by

Hawkins (1983:13), Givón, Mallinson and Blake, and those discussed in

Chapter 7, I conclude that VSO is the most basic order whenever full

noun phrases are used.

8.3. Summary of typological traits

Table 8.2 summarizes the verb initial features found in Yagua, according

to the verb initial norm (VIN; Chapter 2). Some of these features, such as

agglutinative and polysynthetic morphological structure, are not

exclusively verb initial characteristics.

8.3. Summary of typological traits 2 4 7

Table 8.2. Summary of Verb Initial Features in Yagua




















Basic constituent order is VSO (order of direct object

and oblique may vary).

Fronting of NPs (or other elements) to the left of the verb

is a possibility under pragmatically marked conditions.

There is a tendency to move given information to end of the

clause (relative to the order of direct object and oblique).

The language is agglutinative and polysynthetic.

There is essentially no nominal case marking for subject and

object (but Set II clitics have case/pronominal features).

Relative clauses are post-head.

Descriptive modifiers are post-head.

Relativization may be by deletion or by retention of a

Set I or Set II clitic in the position relativized.

Manner adverbs generally follow the verb.

Auxiliaries precede the verb.

The dominant negative particle néé precedes the verb.

(Some) modal formatives are affixal to the verb.

Embedded verbs generally follow the embedding verb.

Clausal objects follow the main verb.

There is no overt copula.

Placement of the yes/no question particle is specified with

reference to the beginning of clause (second position within


In information questions, the questioned NP is fronted

("movement" of questioned NPs from embedded clauses is

also a possibility at least for subjects).

Some adverbial and complement clause types follow their

main clause (though conditionals and other -tiy clauses

precede their main clause).

Complementizers precede their clause.

Table 8.3 summarizes features found in Yagua which are not

characteristic of verb initial languages, according to VIN. Some of these

features are not exclusively characteristic of any one constituent order

type, however. For example, suffixing is much more common crosslinguistically than prefixing.

248 8. Constituent Order in Yagua: Conclusions and Implications

Table 83. Summary of Non-Verb Initial Features in Yagua














The language is almost exclusively suffixing.

There are postpositions and no prepositions.

Demonstratives and numerals are pre-head.

Genitive expressions are pre-head.

There is some agreement between the head noun and other

constituents of the noun phrase (numerals and


Adverbs precede descriptive modifiers.

Relative pronouns occur.

There is a rich variety of means for nominalizing verbs,

particularly using classifiers.

There is no productive specifically passive construction.

The verb agrees with just one argument (though two are

potentially referenced in the clause).

In the comparative construction the comparative precedes the

standard (though comparison is most commonly done by


The coordinate particle jaryeey follows the coordinated


-Daryájy 'because' and -tyifnu 'while' are subordinating

suffixes (rather than prefixes).

8.4. Implications for head-dependent ordering principles and

Hawkins' Universals

If we look at Yagua in terms of head-dependent ordering, then it is not a

well behaved language. At least features 1, 6, 7, 10, 13, 14, 18, and 19 in

Table 8.2 could be described as evidencing head-dependent order. But at

least features 2, 3, 4, and 13 in Table 8.3 could be described as evidencing

dependent-head order. Certain other features could be said to follow from

one or the other of these ordering patterns in accord with Lehmann's

"primary concomitant" principle (Lehmann 1973). This principle states

that modifiers of a basic syntactic element stand on the opposite side of

that element from its primary concomitant. For example, the object is

purportedly the primary concomitant of the verb. Since the object follows

the verb, other verbal modifiers should precede the verb. This would

motivate pre-verbal positioning of the (primary) negative particle.

8.4. Implications for head-dependent ordering principles and Hawkins' Universals 2 4 9

However, there are discrepancies in the order of noun phrase elements.

Since the object noun phrase is said to be the primary concomitant of the

verb and follows the verb, modifiers of a noun should follow the noun.

This might account for post-head positioning of descriptive modifiers and

relative clauses, but it does not account for pre-head positioning of

demonstratives, numerals, and genitives. Even if we look at Yagua in

terms of Hawkins' framework, which does not predict the limited number

of co-occurrence types that Lehmann (1973) does, we find that Yagua

does not behave. In particular, by his own criteria as to what is basic,

Yagua stands as a counter example to Hawkins' Universal 2.

Consequently, we conclude that the universal is wrongly stated as an

exceptionless one. It may hold true with overwhelmingly more than

chance frequency, but absolute agreement with the Universal is not


The data base on which Hawkins' proposed universals are founded has

two problems. It is not a random sample, and there are language types

not represented in the sample (cf. Doris Payne 1985b). The second feature

is particularly important in an adequate data base for drawing the type of

exceptionless universals Hawkins proposes. The universals must in fact

reflect all actually occurring co-occurrence types. Insofar as even one

language's co-occurrence set is not represented in the sample, one might

draw the erroneous conclusion that certain universals are exceptionless.

This is apparently what has happened in the case of Universal 2.

No data base can ever be perfect because we can never be sure that

languages which have died out might not have been exceptional to some

proposed universal. Perhaps one counterexample does not constitute much

of an exception and we can still say the universal is "nearly

exceptionless". But the degree to which it stands as "nearly

exceptionless" awaits further research. Additional data from the Amazon

area should contribute significantly to such study. There is at least one

pocket of verb-initial/postpositional languages in the western Amazon

area. This includes Yagua, Taushiro (genetic relationship unknown;

Alicea 1975), and the Arawakan languages Baure (Keenan 1978),

Machiguenga (Betty Snell, personal communication), Nomatsiguenga

(Wise 1971), Caquinte (Kenneth Swift, personal communication), some

Asheninca dialects (David Payne, personal communication), and

Amuesha (Martha Duff Tripp and Mary Ruth Wise, personal

communication). It is worth pointing out that Hawkins has already noted

that his Universal 3 must be taken as statistical in its basic form. This

universal is: PREP~> (NAdj~>NGen). 8 The fact that both Universals 2

and 3 are best taken as statistical (in their simplest formulation) suggests

that the degree to which all of the proposed universals stand as

2 5 0 8. Constituent Order in Yagua: Conclusions and Implications










8.5. Yagua as a head marking language

Although head - dependent serialization principles do not make much

sense out of the Yagua data, and although Yagua should not exist

according to Hawkins' Universal 2, there is another framework which may

account for at least some of the co-occurring features seen in Tables 8.2

and 8.3. This is Johanna Nichols' notion of head marking versus

dependent marking languages (Nichols 1986). Head marking and

dependent marking have to do with the presence and location of overt

morphological marking of syntactic relations: are such relations marked

on the head or on the dependent element in a syntactic phrase? Briefly

put, a head marking language marks dependency relations on the head

element in a given construction. A dependent marking language marks

such relations on the dependent element. Languages may evidence a

mixture of head and dependent marking. At certain points of the grammar

they may be neutral with regard to head versus dependent marking, or

they may mark both the head and the dependent of certain constructions.

As discussed in Section 4.2.3, Nichols defines the head of a

construction as "the word which governs, or is subcategorized for, or

otherwise determines the possibility of occurrence, of the other. It

determines the category of its phrase. " This definition yields

indeterminate conclusions when it comes to differentiating head and

modifying nouns in Yagua noun phrases, and I have amplified it with the

discourse based notion of "pragmatic head". At the clause level, Nichols

considers the verb and/or auxiliary verb to be the head, perhaps because

it is the verb which determines the possibility of occurrence of subject and

object (and other) relations. That is, in naturally occurring discourse,

occasionally noun phrases can be simply juxtaposed in a paratactic way to

other constituents with ellipsis of understood predicates. But when such

phenomena occur, the grammatical relations of the overt elements are

potentially unclear if not non-existent. In this sense it is the presence of a

verb or predicate which guarantees or forces the assignment of

grammatical relations to accompanying noun phrases. Nichols does not

suggest that the verb plus object preferably form a syntactic constituent

separate from the subject. This accords well with the facts of Yagua, both

in terms of its VSO order, and the difficulties in trying to motivate the

Tài liệu bạn tìm kiếm đã sẵn sàng tải về

Chapter Eight. Constituent Order in Yagua: Conclusions and Implications

Tải bản đầy đủ ngay(0 tr)