Tải bản đầy đủ - 0trang
Chapter Eight. Constituent Order in Yagua: Conclusions and Implications
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.
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 PREDICATEj _pj a c e
ARGUMENT : NP
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
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
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
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
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