The examination of word order involves two distinct sets of questions. The first concerns the typological status of the language with respect to the observations of Greenberg 1966: What is the basic order of verb, subject and object? How are these constituents identified? Does their order correlate with that within sub-clausal phrases? The second concerns the derivation of permissible surface structures from some underlying representation: in the most straightforward derivation do clausal constituents remain in the same position or the same order?

Languages evidencing freedom of word order provide a challenge to the examination of either set of questions. Such languages include the non-configurational languages which generally evidence free word order, the use of syntactically discontinuous expressions, and extensive use of null anaphora (Hale 1983). Nahuatl, an Aztecan language of Mexico, meets certain of these criteria. Nahuatl appears to allow all possible orders of the basic meaningful elements, verb (v) subject (s) and object (o). The possible orders for a single clause is shown in (1).

1 a. no ikni'meh kichihchiwa'tih se kali' [North Puebla SMT]

my brothers 3S.3O.build.FUT a house

S V O

My brothers are going to build a house.

b. no ikni'meh se kali' kichihchiwa'tih

c. kichihchiwa'tih no ikni'meh se kali'

d. kichihchiwa'tih se kali' no ikni'meh

e. se kali' kichihchiwa'tih no ikni'meh

A house, my brothers are going to build.

f. se kali' no ikni'meh kichihchiwa'tih

A house, my brothers are going to build.

In addition, Nahuat makes liberal use of pro-drop, and in numerous clauses verbal arguments may be null, as in the clause in (2).

2 kichihchiwa'tih

3S.30.build.FUT

They are going to build it.

In true non-configurational languages verbal inflectional markers appear to function as verbal arguments, and associated nominals have properties generally associated with adjuncts (Jelinek 1984). For example the distribution of these. nominals do not show the subject/object asymmetries familiar from Binding Theory. In addition, there appear to be conditions on their functioning as syntactic operators such as a bare quantifier expression or fronted we-phrases. Nominals in the corpus examined here, however, behave consistently as arguments in configurational languages. and so I treat Nahuatl as configurational: nominals agreeing with verbal inflection are analysed in argument positions, and when such nominals are null I analyse the argument position as occupied by pro.

Word order variation (not due to null anaphora) in a configurational language is commonly attributed to the fronting of arguments to fixed topic and focus positions. In this paper I examine whether the availability of fronting itself provides a complete account of the Nahuatl word order data.

Nahuatl, also known as Aztec or Mexicano, includes a number of dialects, of varying degrees of mutually intelligibility. Here I examine data from fieldwork on a North Puebla dialect, spoken in San Miguel de Tenextapiloyan, plus data from the grammars collected in Langacker 1979, including Tuggy on the Tetelcingo dialect, Brockway on North Puebla, Beller and Beller on Huasteca, and Sischo on the Michoacan dialect.

The most common and least marked sentences consist of a predicate and one separately stated argument or simply an inflected predicate and its modifiers. Which argument is specified is generally determined by discourse and pragmatic factors as discussed in Givon 1985, Payne 1987, Du Bois 1987 among others. In clauses in which subjects and objects occur as separate phrases, indications of topic and focus are used to pin point basic word order following guidelines such as those suggested by researchers including Brody 1984 and England 1991. Thus researchers working on Nahuatl describe it as containing neutral clauses which are vso, svo, vos or some combination of these three.

3 Nahuat neutral word orders

vso

vos

svo

I do not evaluate the means by which topic and focus are differentiated with respect to the Nahuatl data here, but adopt the terms following usage in the grammars cited for data taken from each grammar.

The typological work on word order has long noted the impact of pragmatic considerations, and recently a division has been proposed between languages whose word order is primarily correlated with syntactic considerations and grammatical relations and those whose word order is primarily sensitive to the pragmatic status of arguments as given/new, definite/indefinite, or topic/comment.

I assume that there are some grammatical and some pragmatic factors at play in determining the word order of any language. The decisive role of operations which front focus and topic elements in determining the word order of Nahuatl suggests that, as with non-configurational languages, word order in this language is strongly determined by pragmatic factors. In this paper I explore some syntactic factors involved in surface word order flexibility in Nahuatl. Particularly I consider the derivation of clause negation and finite verbal information.

I claim that free word order is derived by more than pragmatically driven movement of arguments. Notice that this pragmatically determined flexibility concerns only the position of arguments (nominal and adpositional categories) with respect to fixed predicate-information. The question arises whether only arguments exhibit word order Flexibility. If so, than word order flexibility. would seem to be always pragmatically determined. I argue here that this is not the case. The position of certain clausal information, particularly the predicate-information generally marked on the clausal head, is grammatically determined. I argue that languages can exhibit positioning Flexibility with respect to this information, and that this sort of Flexibility is a crucial characteristic of a configurational language such as Nahuatl.

This work is couched in terms of possible derivations of the Nahuat data within a version of Principles and Parameters syntax (Chomsky 1991, Chomsky and Lasnik 1993, Chomsky 1993). Within this theory, underlying order is no longer viewed as a defining characteristic of a grammar. A default underlying representation is assumed based on (i) the Internal Subject Hypothesis according to which subject and object arguments are generated within the vp and (ii) the work of Kayne 1993 concerning the branching of specifiers. The default representation is given in (4). The clausal predicate projects a head. If the predicate is verbal, the head is v. The object is sister to the predicate under a bar-level category, and the subject is in the specifier relation to v.

4

In Nahuat I assume an Ip projection above vp headed by inflectional features such as tense and agreement. The specifier of this position is generally conceived of as the position of the external argument; it is not a focus or topic position. A functional projection above Ip contains pragmatically marked elements such as fronted we-words which are reported as fronted cross- dialectally in Langacker 1977. Examples from the corpus are given in (5). Here the we-phrases ako'ni in (5a), tli in (5b), tlen in (5c), etc. occur in the initial position.

5 a. ako'ni kitak in kowat tech nochan [North Puebla SMT]

who saw the snake in my.house

Who saw the snake in my house?

b. tli ti-k-neki [Tetelcingo]

what 2s-3-want

What do you want?

c. tlen tonal ti-h-toh-ki [Huasteca]

what day 2s-30-plant-PAST

What day did you plant it?

d. kanki ti-yah-ki [Huasteca]

where 2s-go-PAST

Where did you go?

e. kalia no-koton ni-mic-maka-s [Michoacan]

which my-shirt Is-fLs-give-FUT

Which shirt of mine shall I give you?

Finally, I assume an agreement projection, AGROP, between Ip and vp. The specifier of AGRO iS the position in which the features of the object noun phrase are eventually checked. This position is justified by the presence of object agreement morphology on the Nahuatl verb, and is necessary to account for vos orders. Alternative accounts of vos - such as the possibilities that the specifier of vp optionally branches right or that there is a right branching strong topic position - are not considered since this paper is not primarily concerned with the position of arguments. The positions assumed are illustrated in (6). Non-dislocated subjects and objects do not occupy adjoined positions.

6

(6) is not a final analysis of Nahuat clause structure; it serves merely to anchor the discussion of possible word order derivations. Nonetheless, some support for the CP-IP structure is found in the examination of affirmative interrogative clauses. As shown in (7), arguments preceding a wH-word in interrogatives are interpreted as dislocated. Taha in (7a) precedes the interrogative tlika <why>, and receives a heavy topic interpretation. Yaha in (7b) likewise precedes a we-phrase and is interpreted with contrastive focus. The pragmatic markedness is expected by our analysis in which they are adjoined to cp.

7 a. taha tli-ka ti-k-maka-k [Tetelcingo]

you what-with 2s-3-hit-PERF

As for you, why did you hit him?

b. yaha tli-ka ti-k-maka-k taha [Tetelcingo]

3sm what-with 2s-3-hit-PERF you

Why did you hit HIM?

c. taha tli-ka o-ti-wala [Tetelcingo]

2s what-with PAST-2s-come

Why did YOU come?

d. no-nobia len ti-ya-lo ti-k-maka-lo [Michoacan]

my-fiancee what Ipgo-PL Ip-30-give-P

My fiancee, what are we going to give her?

In addition, as expected with the Cp-IP strUCtUre, the subject is the only argument to appear between a WH-word or complementizer and the verb in examples in the corpus. Examples are provided in (8).

8 a. tlen ye'wa ok-kaEi ki-Eiwa-s [North Puebla]

what he yet-more 30-do-FUT

What else will he do?

d. kwando se ki-neki ki-Eia tapalkal... [Michoacan]

when one 30-want 30-make pottery

When one wants to make pottery...

There are numerous important aspects of derivations which I do not go into here particularly with respect to the position of arguments. Instead I focus on the question of whether the syntactic

movement of the predicate can be fixed in the derivation of the various word orders.

1 Affirmative Clauses

The word order facts.for affirmative clauses in (1) can be accounted for with a fixed predicate posi- tion. In (9) I outline the derivation for the six sentences with the verb in IO in each representation. AGROP is included only where necessary in the analyses roughly sketched here.

9 a. SVO: [Ip SUBJECT VERB [vp OBJECT]]

b. VSO: [Ip VERB [vp SUBJECT OBJECT]I

c. VOS: [Ip VERB [AGRoP OBJECT [vp SUBJECT ]]]

d. OVS: [Cp OBJECT [Ip VERB [vp SUBJECT ]]]

e. OSV: [cp OBJECT [rp SUBJECT VERB [vp ]]]

f. SOV: [Cp SUBJECT [Ip OBJECT VERB [vp ]]]

In deriving svo order both subject and verb move to the IP projection, (9a). In the verb-initial orders vso and vos only the verb moves to the IP projection, (9b,c). In the object-initial orders ovs and osv the object occupies the marked position, the specifier of cp. The analysis of the sov order in (9f) seems less plausible. With the verb fixed in IO, the subject occurs in the specifier of cp and the object must occur in the specifier of IO. If we wish to maintain the analysis of the finite in IO, we might seek to justify an analysis in which the subject is actually adjoined to cp, and the object in sentences with sov order are in the specifier of cp. However there is no evidence of strong pragmatic marking on the subject in examples of this sort in the corpus. (These are admittedly few, however.)

For this reason, an alternative view of the grammar, one in which the verb does not move to IO, (10), may be preferred.

10 SOV: [Ip SUBJECT [AGRoP OBJECT VERB [vp ]11

Notice, however, that the verb cannot remain in VP Or AGROP BS it is in (10) in the derivations of all other orders: the verb must move out of vp in any order in which it precedes the subject, e.g.vos. Thus this alternative is not consistent with a uniform analysis of finite verbs.

2 Negative Clauses

Negative clause data more clearly indicates the need for flexibility in the derivation of the predicate position. The marker of sentential negation in the dialects considered here are amo, a'mo, as: sometimes reduced to a. I use the term amo to refer to this set.

I assume that the sentential negator amo has a fixed position. Some support for this view is found in the fact that only two sorts of arguments occur prior to negation in the corpus: pragmatically marked phrases (11a,b,c) and subjects (11d,e). In the translations of the text examples in (11a,b,c) I provide in parentheses some of the context, the translations of surrounding clauses, so that the topic or focus interpretation is clear. For example in (11a), taken from the story of Coyote and Jackrabbit, Coyote literally says to Jackrabbit who is pretending to be a snake, "Rabbits can be eaten, but rattlesnake doesn't please me." and the pre-a'mo element in kaskabel 'rattlesnake' is interpreted with contrastive focus. This contrasts with the initial subject in miakten 'many' in (11d) Many did not go. which is not pragmatically marked.

11 a. in kaskabel a'mo nec-pak-ti-a ni-kwa-s [North Puebla]

ART rattler NEG 1SO-please-CAUS-PRES 1s-eat-FUT

(Rabbits can be eaten, but) I don't like to eat rattlesnake.

b. kapol.i a-weli ki-kwa tiek.tli [Tetelcingo]

berry NEG-can 3o-eat viper

(Jackrabbits eat berries, but)

vipers can't eat berries, (so you are a jackrabbit!)

c. wan newal amo ni-mo-kwepa ni-lehku-a in kwawi-mes [Mich.]

and Is NEG Is-REFL-return Is-climb-PRES the tree-FL

(She gave me just three whacks and said to me,)

("You go climbing trees again, and I shan't scold;)

(I'll deal with you this way lest you suddenly kill yourself!»)

...And I for sure will never climb trees again!

d. in miak.ten a'mo o-ya-'ke [North Puebla]

ART much.PL NEG PERF.go.PL

Many did not go.

e. i-pelo-hmes amo ki-kwah-kwa-lo-aya [Michoacan]

his-dog-FL NEG SO-RDP-eat-PL-PAST

His pets didn't bite her.

In addition, the fixed position of the sentential negator is evidenced in data such as (12). Here we see that amo occurs before, not after the finite verb. If amo is post-verbal, as in (12d,e), either sentence final or before some other constituent, the sentence is ill-formed.

12 a. amo tikchihchiwatih se kali [North Puebla SMT]

NEG 2pSU-3OBJ-build-FUT a house

We're not going to build a house.

b. kanaya as kwali nopa oh-tli [Huasteca]

in.places NEG good that road-ABS

In places the road was bad.

c. pa kamo niC-ahsi-ski-aya no-nsn-cin [Michoacan]

so.that NEG IsO-reach-FUT-PAST my-mother-HON

(When I saw her grab a stick, I ran to an open space)

so that my mother couldn't reach me.

d. *tikchihchiwatih amo sekali

2pSU-30BJ-build-FUT NEG a house

We're not going to build a house.

e. * tikchihchiwatih sekali amo

2pSU-30BJ-build-FUT a house NEG

(We're not going to build a house.)

Of course, when amo does not mark sentential negation, this fixedness is not observed. Amo forms other negated constituents and these may precede the verb, as in (13a,b,c), or follow it, as in (13d,e). These data, together with the example in (14), also establish that amo does not have the distribution of a clitic.

13 a. [amo tech nochan] In Mariah kitak in kowat [North Puebla SMT]

NEG in my house Marish saw the snake

Not in my house did Maria see the snake.

b. [aS kema) ki-neki tekiti-s [Huasteca]

NEG when 30-want work-FUT

(He never wants to work.)

c. Roberto [aS Wan] ki-wika-k mo-tomin (Huasteca]

Roberto NEG Wan 30-take-PAST your-money

(It was Roberto, not John, that took your money.)

d. o-ki-ci'-Ciw wan [a.yak.mo] yohke [North Puebla]

PERF-30-RDP-do CONJ NEG.yet.NEG same

(He fixed it, but it's not the same anymore.)

e. aSan o-ni-k-tek kaEi [a'momiak] [NorthPuebla]

today PERF-ls-30-cut more NEG much

(I didn't pick as much today.)

14 (wan) tla k-oni noso a'mo... (North Puebla]

(and) it 30-drink or NEG...

and if he drinks or not...

Amo occurs after WH-constituents as shown in (15). In the ungrammatical (15a) amo precedes the WH-phrase.

15 a. *amo akoni kinmaka 7at a in tapialmeh [North Puebla SMT]

NEG who 3SU-30BJ-give water to the animals

Who isn't giving water to the animals?

b. akoniamo kinmaka 7at a in tapialmeh

who NEG 3SU-30BJ-give water to the animals

Who isn't giving water to the animals?

Finally amo co-occurs with complementizers. In the sentence in (16) amo co-occurs with mah 'that', tagen 'if' and xe 'if'.

16 a. nechilih mah amo nik'ita aken [North Puebla SMT]

3sS.lsO.tell.PAST that NEG 3sS.30.see someone

He told me not to see anyone.

b. tagan nookniw amo kwalki nipili, niye'tos kwala'ntok

if my sister NEG 3sSu-bring her child 1sSu-be-FUT upset

If my sister doesn't bring her child, I will be upset.

c. xe amo tika'ki, ximopahpaka monakaspahpa'ka!

if NEG 2sSU.hear IMP.wash your.ears

If you can't hear, wash your ears!

These facts may be captured if amo, the element marking sentential negation, originates as the head of NEGP above vp and lower than cp.

There is a more limited arrangement of arguments with respect to negative verbs than affirmative ones. These orders are summarized in (17). Through out the corpus we find examples of amo immediately preceding the verb which is followed by its arguments, (17a,b), and examples in which amo precedes the preverbal subject, (17c). In addition we find examples, such as those already examined in (11), in which the subject or a pragmatically marked argument precedes the negative element .

a. NEG VSO

b. NEG VOS

c. NEG SVO

d. S NEG V(O) where S may be pragmatically marked

e. O NEG V(S) where O must be pragmatically marked

It is also the case that in all examples in which amo directly precedes a fronted object, the object, not the clause, is interpreted as the negated constituent. This is evident in the examples in (18) in which amo is analysed as a modifier of se kali7 'a house'.

18 a. amo se kali7 kichihchiwa>tih Xwan wan nik ikni'meh [N.P.SMTJ

NEG a house 3pS.30.make.FUT Juan and his brothers

It isnY a house that Jzlan and his brothers are going to build.

b. amo se kali7 Xwan wan ni ikni'meh kichihchiwa'tih

NEG a house Juan and his brothers 3pS.30.make.FUT

It is not one house Juan and his brothers are going to blLild.

For the purposes of examining predicate positions the crucial examples are given in (19). The negative clauses in (19) are all analysed with the verb obligatorily in IO. NEG-SVO examples such as (19b) require that the NEGP headed by amo dominates Ip. However, this forces us to analyse the subject in examples like (19c) in a topic position, i.e. the specifier of cp. This does not accord with the neutral interpretation of the sentence. The preferred position for the subject in miakten is the specifier of IO, but positioning it here requires that NEGP OCCUL. below Ip. Two positions for NEcP are thus required, one above Ip and one below it. It should be noted that the S-NEG-VO order in (19c) is found in numerous examples in four of the five dialects in the corpus. It is not a dialect specific construction.

19 a. amo kitak inxwan inkowat [North Puebla SMT]

NEG 3S.30.see.PAST the Juan the snake

NEG [Ip V [VP S 0]1

JZLalZ didn>t see the snalce.

b. amo in mariah kinmaka 'at a in tapialmeh

NEG the Maria 3S.30.give water to the animals

NEG [Ip S V [VP 0]1

MaTia isn>t giving water to the animals.

c. in miakten amo oya'keh

the much.PL NEG PERF.go.PL

S NEG [rp V]

Alternative: [IPS NEG-V]

Many did not go.

If v-to-IO movement is characterized as optional, the data in (19) can be analysed using a single instance of NEGP: NEGP OCCUTS aS the sister to IO. This is shown in (20). With the lower base position, amo must adjoin to the finite verb and move with it when v-to-IO movement occurs.

20 a. amo kitak inxwan inkowat (North Puebla SMT]

NEG 3S.3O.see.PAST the Juan the snake

NEG [AgrP V [VP S O]]

[IP NEG-V [vp S O]]

Juan didn't see the snake.

b. amo in mariah kinmaka 'at a in i tepialmeh

NEG the Maria 3S.30.give water to the animals

NEG [vp S V O]

Maria isn't giving water to the animals.

c. in miakten amo oya'keh

the much.PL NEG PERF.go.PL

[rp S [NegP NEG [vp V]]]

Many did not go.

The conclusion we reach is not entirely new: non-configurationality involves optionality, but in this case optionality describes the syntactic movement of the predicate head and not (only) the placement af arguments in accord with their pragmatic function. The pattern in verb movements required to capture all the facts of both affirmative and negative Nahuat clauses are stated in (21):

21 a. When no argument moves, the verb raises to IO

b. If the object argument moves, the verb raises (and checks it) in AGROP

c. Otherwise the verb does not move

3 Conclusions

The pragmatic and discourse-based aspects of word order generally address the distribution of arguments with respect to their predicate. The positioning of the predicate - which crucially defines other clausal positions - is determined by syntactic and morpho-syntactic matters. This is the case even if the syntactic notion of basic word order is not a defining characteristic of the grammar of a language. The syntactic conditions on word order are not limited to defining structural input to derivation.

I have presented some evidence in the course of this discussion that free negative morphemes can also be grammatically fixed, and that examination of the distribution of clausal arguments, adverbials and modifiers with respect to the unit of negation is also an insightful means of examining word order. I extrapolate the results here to claim that Word Order cannot be established solely on the basis of affirmative clauses; rather, negative clauses must be taken into account.

Finally, I have argued here that non-configurationality is observed not only in the distribution of noun phrases or arguments, but in the positional flexibility of the crucial units defining argument positions - the position of the finite verb and negation in this discussion. I have presented some evidence from Nahuat that these grammatically fixed units do not occupy the same position in every clause.

The obvious question to ask is whether the grammatical motivation for the movement of predicate information is affected by clausal pragmatics. Is there a uniform explanation for (21)? I suggest that the pragmatics only indirectly affects predicate derivation. For example, the Nahuat verb raises to AGRO when an object argument moves outside the VP. This verb checks the grammatical Case features of the object in the appropriate Spec-head configuration when the object is in the specifier of AGRO. In the derivation of other clauses, the checking of the grammatical feature of the object is delayed: object and verb occupy AGROP only covertly at LF. However it is conceivable that delay in cases of object fronting would lead to ungrammaticality if we assume that in order to interpret the object in a pragmatically marked position, the object may not reconstruct into AGROP at LF. In this way the object function indirectly enters into the calculation of syntactic verb movement.

Selected References

Aissen, Judith. 1992. 'Topic and Focus in Mayan', Language, 68:1, 43-80.

Brody, Jill. 1984. 'Some problems with the concept of basic word order', Linguistics, 22, 711-7363.

Chomsky, Noam. 1991. 'Some Notes on the Economy of Derivntion and nepresentation', in R. Freidin, ed., Principles and Parameters in Comparative brammar, M~T Press, Cambridge.

______________: 1993. 'A Minimalist Program for Linguistic Theory', in Kenneth Hale and Samuel Jay Keyser, eds., The View from Building 20, MIT Press, Cambridge.

Chomsky, Noam and Howard Lasnik: 1993. 'Principles and Parameters Theory', in J. Jacobs, A. van Stechow, W. Sternefeld and T. Vennemann, eds., Syntax: An International Handbook of Contemporary Research, Welter de Gruyter, Berlin.

Diesing, Molly. 1990. 'Verb Movement and the Subject Position in Yiddish', Natural Language and Linguistic Theory, 8, 41-79.

Du Bois, John. 1987. 'The Discourse Basis of Ergativity', Language, 63:4, 805-855.

England, Nora. 1991. 'Changes in Basic Word Order in Mayan Languages', IJAL, 57:4, 446-486.

Givon, Talmy. 1985. 'The Pragmatics of Word Order: Predictability, Importance and attention', in Edith Moravcsik, ed., Language Typology and Language Universals, Amsterdam: Benjamins.

Greenberg, Joseph. l966. 'Some Universals of grammar with Particular Reference to the Order of Meaningful Elements', in Joseph Greenberg, ed., Universals of Language, Cambridge: MIT Press.

Hale, Ken. l983. 'Warlpiri and the Grammar of Non-Configurational Languages', Natural Language and Linguistic Theory, 1, 5-47.

Jelinek, Eloise. 1984. 'Empty Categories, Case, and Configurationality', Natural Language and Linguistic Theory, 2, 39-76.

Kayne, Richard. l993. The antisymmetry of syntax. Msc, Graduate Center, CUNY. Law, Howard. l966. Obligatory Construction of Isthmus Nahuat Grammar. The Hague: Mouton.

Langacker, Ronald. 1977. Studies in Uto-Aztecan Grammar 1: An Overview of Uto-Aztecan Grammar, Dallas: Summer Institute of Linguistics.

___________: 1979. Studies in Uto-Aztecan Grammar 2: Modern Aztec Grammatical Sketches, Dallas: Summer Institute of Linguistics.

Payne, Doris. 1987. 'Information Structuring in Papago Narrative Discourse', Language, 63:4, 783-804.

Zanuttini, Raffaella. 1991. Syntactic Properties of Sentential Negation. A Comparative Study of Romance Languages. University of Pennsylvania dissertation.

Robin Schafer

Department of Linguistics 0108

UCSD

La Jolla, CA 98093