In the standard theory conception of phrase structure rules, they encode both dominance relationship and precedence relationships. Various linguists have suggested that these two ideas can be disentangled so that the PS-rules tell us only what the dominance relationships are, while a separate set of rules or principles tell us what the linear order should be. Within Lexical Functional Grammar, such ideas have been proposed by Falk (1983). They also figured prominently in GPSG (Gazdar, Pullum, Klein, and Sag (1985), continuing into HPSG (Pollard and Sag 1987). Within older styles of government-binding theory, such ideas were proposed by Farmer (1980, 1984) and Stowell (1981).
In sharp distinction to these approaches, much recent work within the Minimalist Program (Chomsky 1995) has adopted Kayne's (1994) Linear Correspondence Axiom, which can be summarized as follows
1) "...the human language faculty is in fact rigidly inflexible when it comes to the relation between hierarchical structure and linear order. Heads must always precede their associated complement position. Adjunctions must always be to the left, never to the right...This inflexibility extends to specifiers, too, which I argue to be an instance of adjunction. Hence, specifier positions must invariably appear to the left of their associated head, never to the right." (Kayne 1994:vii)
In Kayne's view, apparent counterexamples result from the application of movement rules.
In this paper, I will argue in favor of the position that PS-rules encode only dominance relationships. Linear precedence, I will suggest, is the result of violable optimality-theoretic constraints. My account will be framed within the optimality-theoretic implementation of Lexical Functional Grammar proposed by Bresnan (1998).
San Dionicio Ocotepec Zapotec (hereafter SDZ) is an Otomanguean language spoken in Oaxaca, Mexico. The basic word order of this language is VSO, with head initial NPs and PPs:
2) Ù-díny Juààny bèh'cw re' cùn yàg.
com-hit John dog that with stick
'John hit that dog with a stick'.
There are two preverbal positions for elements with special discourse functions. Topics and foci (contrastive or interrogative) appear in one position, negative foci occur in a second position. The overall phrase structure of SDZ is shown in figure 1.
Both subjects and objects may appear in [Spec, CP] when they bear a special discourse function (focus or topic):
3) Juààny ù-díny bèh'cw cùn yàg.John com-hit dog with stick
'John (topic/focus) hit the dog with a stick.'
4) Bèh'cw ù-díny Juààny cùn yàg.
dog com-hit John with stick
'John hit the dog (topic/focus) with a stick.'
Wh-phrases appear in the same [Spec, CP] position. Since only one phrase may appear in [Spec, CP], the following S is necessarily verb initial:(2)
5) a. ¿Cálóò gwíí Juáàny?
where com:go John
'Where did John go to? 5:3
b. *¿Cálóò Juáàny gwíí?
where John com:go
Wh-movement is obligatory, as shown by the following contrast:
6) ¿Túú ù-díny Juààny cùn yàg?
what com-hit John with stick
'What (anim.) did John with a stick?'(3)
*¿Ù-díny Juààny túú cùn yàg?
com-hit John what with stick
The Comp position is filled in yes-no questions by the particle lá, which cliticizes to the following word:
7) ¿Lá=Juáány gù Móòny zè'èh lò ìnyá'?
q=John or Ramón com:go to field
'Was it John or Ramón who went to the field?'
8) ¿Lá=bèh'ty Juáány bzììny?
q=com:kill John rat
'Did John kill the rat?'
Examples like (7) show that the Comp position precedes the [Spec, CP] position.
I assume that the Zapotec clause is dominated by S, which is a non-endocentric category in this language, though nothing in what follows hinges crucially on this.(4) Then the following linear precedence constraints will give us the right order:
9) X < Comp (heads precede complements)
X < Spec (heads precede specifiers)
SUBJ < OBJ < ADJ (less oblique arguments precede more oblique arguments)
Note that Spec and Comp are not differentiated from each other by phrase structure position.
The fact that wh-movement is obligatory suggests that SDZ shows the effects of a constraint like the following:
10) Align (IntF, L, CP, L)
Align the left edge of interrogative focus phrase with the left edge of CP.
When there is no overt complementizer, I will assume that the Comp node is also absent from the tree, and that a wh-phrase in [Spec, CP] fully satisfies this constraint. This constraint suggests a simple tableau like the following:
11)
| Align
(IntF, L, CP, L) | ||
| a. | »¿Túú ù-díny Juààny cùn yàg?
(What hit John with stick) |
|
| b. | *¿Ù-díny Juààny túú cùn yàg?
(Hit John what with stick) |
* |
When we question the object of a preposition in this language, we get a surprising result:
12) ¿Xhíí cùn ù-díny Juààny bèh'cw?
what with com-hit John dog
'What did John hit the dog with?'
This pattern has been labelled 'pied-piping with inversion' (Smith Stark 1988), and it is found in all Zapotecan languages and in many other Mesoamerican languages as well (e.g. Tzotzil, Aissen 1996). (12) is the only grammatical order for this question in SDZ. It is ungrammatical to have either of the following:
13) a. *¿Cùn xhí ù-díny Juààny bèh'cw?
with what com-hit John dog
b. *¿Xhí ù-díny Juààny bèh'cw cùn?
what com-hit John dog with
The ungrammaticality of (13b) seems to be due to an undominated constraint in SDZ that forbids preposition stranding.(5) I will assume that this prohibition is one of a family of harmony constraints, which call for correspondence between the various representations of a sentence. In particular, the constraint needed here requires that an f-structure constituent (ADJUNCT) correspond to a c-structure constituent (in this case, PP) .
14) Harmony of Constituency -- ADJUNCT (HConADJ)
An f-str constituent with the function ADJUNCT must correspond to a c-structure constituent.
Since it is well-known from studies of extraction and island effects that different sorts of constituents show varying degrees of resistance to extraction, I will assume that (14) is one of a family of harmony constraints which may have different rankings in various languages. I will also follow Kaplan and Zaenen (1989) in assuming that the constraints on possible long-distance dependencies are best stated in terms of f-structure functions, rather than c-structure categories.
However, rather than capturing the effect of island effects through restrictions on uncertainty equations, as in Kaplan and Zaenen (1989), the approach advocated here treats the conditions on extraction through violable constraints.
The following ranking of constraints will select the correct candidate.
15)
| HCon-ADJ | Align (IntF,L, CP, L) | X < Comp | ||
| a. | ¿Ù-díny Juààny bèh'cw
cùn xhí?
(Hit John dog with what) |
*!*** | ||
| b. | ¿Cùn xhíí ù-díny Juààny
bèh'cw?
(With what hit John dog) |
*! | ||
| c. | ¿Xhíí ù-díny Juààny
bèh'cw cùn?
(What hit John dog with) |
*! | ||
| d. | »¿Xhíí cùn ù-díny
Juààny bèh'cw?
(What with hit John dog) |
* |
This result falls out easily from the optimality theoretic point of view, but is difficult to get in standard movement analyses.(6)
What is of special interest here is that the ordering of heads before complements is treated as a violable constraint.
4. Questioning specifiersIn SDZ, specifiers of NP normally follow the head:
16) x-pèh'cw Juààny 'John's dog'(7)
p-dog John
bèh'cw re' 'that dog'
dog that
We also see pied-piping with inversion for genitives and demonstratives. Compare the following statements and questions.
17) Juààny cù'á [NP x-pèh'cw Màríí].
John com:grab p-dog Mary
'John grabbed Mary's dog.' 4:214
*Juààny cù'á [NP Màríí x-pèh'cw].
John com:grab Mary p-dog
18) ¿[NPTúú x-pèh'cw] cù'á Juààny?
who p-dog com:grab John
'Whose dog did John grab?' 4:214
*¿[NPX-pèh'cw túú] cù'á Juààny?
p-dog who com:grab John
These sentences show that the possessor may not precede the possessed in a declarative. But in an interrogative, this is the only grammatical order. The following sentences make the same point for demonstratives:
19) Juààny cù'á bèh'cw re'
John com:grab dog that
'John grabbed that dog.'
*Juààny cù'á re' bèh'cw.
John com:grab that dog
20) ¿Túú bèh'cw cù'á Juààny?
which dog com:grab John
'Which dog did John grab?'
*¿Bèh'cw túú cù'á Juà:ny?
dog which com:grab John
Note that the interrogative túú is the equivalent of both 'who, whose, what (animate)' and 'which (animate)' in SDZ. Within a NP, the 'whose' reading is differentiated by the presence of the /x-/ possessive prefix on the noun.The same is true for xhíí, which is the equivalent of both 'what (inanimate)' 'whose (inanimate)' and 'which (inanimate)'.
As in English, it is ungrammatical to attempt to extract either a determiner or a possessive from the NP without pied-piping the NP:
21) *¿Túú cù'á Juààny bèh'cw?
which com:grab John dog
(Which did John grab dog?)
22) *¿Túú cù'á Juààny x-pèh'cw?
who com:grab John p-dog
(Whose did John grab dog?)
These facts are easily handled if we posit the following constraints:
23) X < Specifier
X precedes the specifier within XP
24) Harmony of Constituency -- OBJ
An f-structure constituent with the function OBJ must correspond to a c-structure constituent.
Then the following tableau (for the possessive case) shows how the correct candidate is selected.
25)
| HConOBJ | Align
(IntF,L, CP, L) |
X < Spec | ||
| a. | ¿Túú cù'á Juààny x-pèh'cw?
(Whose John grabbed dog?) |
*! | ||
| b. | »¿Túú x-pèh'cw cù'á
Juààny?
(Whose dog grabbed John?) |
* | ||
| c. | ¿X-pèh'cw túú cù'á
Juààny?
(Dog whose grabbed John?) |
*! |
We see nearly identical effects in sentences with negative focussed elements. Focus negation is always accompanied by fronting of the focussed object to preverbal position.
The negative agrees with the fronted XP in animacy--rútèh'cà is used with an animate focus; xtéh'cà is used with an inanimate focus. The negative focussed NP is preceded by an interrogative/indefinite pronoun. Túú is for animates; xhíí is for inanimates.
25) Juáány rú-tèh'cà túú bzììny be-'ty-bí.
John anim-not any mouse com-kill-3
'John didn't kill any mice.' 4:175
26) X-téh'cà xhíí gèhht ù-dáòw-à.
inan-not any tortilla com-eat-1s
'I didn't eat any tortillas.' 4:241
This fronting is obligatory, like the obligatory fronting of wh-phrases:
27) *Juáány rú-tèh'cà be-'ty-bí túú bzììny.
John anim-not com-kill-3 any mouse
(John didn't kill any mice.)
28) X-téh'cà ù-dáòw-à xhíí gèhht.
inan-not com-eat-1s any tortilla
(I didn't eat any tortillas.)
The negative and negative focus positions come before the verb; the subject may either be in situ after the verb or may appear in the topic/focus position.
29) X-téh'cà xhíí gèhht ù-dáò Juààny.
inan-not any tortilla com-eat John
'John didn't eat any tortillas.' 5:33
30) Juààny x-téh'cà xhíí gèhht ù-dáò.
John inan-not any tortilla com-eat
'John didn't eat any tortillas.' 5:33
If the negative focussed element is the object of a preposition, we get pied-piping with inversion, just as in the questions:
31) Rú-tèh'cà túú lò ù-déhhdy Màríí cààrrt.
anim-not anyone to com-give Mary letter
'Mary didn't give the letter to anyone.' 5:35
*Rú-tèh'cà lò túú ù-déhhdy Màríí cààrrt.
anim-not to anyone com-give Mary letter
32) Rú-tèh'cà túú x-pèh'cw ù-díny Màríí.
anim-not anyone p-dog com-hit Mary
'Mary didn't hit anyone's dog.'
*Rú-tèh'cà x-pèh'cw túú ù-díny Màríí.
anim-not p-dog anyone com-hit Mary
These facts suggest that there is a constraint which favors candidates in which the left edge of a negative focussed constituent aligns with the right edge of the negative element, along the following lines:
33) Align(NegF, L, Neg, R)
Align the left edge of a negative focussed item with the right edge of the negation.(8)
Substituting this constraint for Align(IntF, L, CP, L) in the tableau above will yield the right word order for these sentences.
6. ComplicationsWhat happens when we attempt to question the specifier of the object of a prepostion in SDZ? The following examples show that there are two grammatical results:
35) Ù-díny Juààny bèh'cw cùn yàg ré'.
com-hit John dog with stick that
'John hit the dog with that stick.'
36) ¿Cùn xhíí yàg ù-díny Juààny bèh'cw?
with which stick com-hit John dog
'With which stick did John hit the dog?'
37) ¿Xhíí cùn yàg ù-díny Juààny bèh'cw?
which with stick com-hit John dog
(Which with stick did John hit the dog?)
Three other logically possible candidates are ungrammatical:
38) *¿Xhíí yàg cùn ù-díny Juààny bèh'cw?
which stick with com-hit John dog
39) *¿Yàg xhíí cùn ù-díny Juààny bèh'cw?
stick which with com-hit John dog
40) *¿Xhíí yàg ù-díny Juààny bèh'cw cùn?
which stick com-hit John dog with
Exactly the same facts are found with the possessive:
41) ¿Cùn túú x-cyàg ù-díny Juààny bèh'cw?
with whose p-stick com-hit John dog
'With whose stick did John hit the dog?'
42) ¿Túú cùn x-cyàg ù-díny Juààny bèh'cw?
whose with p-stick com-hit John dog
(Whose with stick did John hit the dog?)
43) *¿Túú x-cyàg cùn ù-díny Juààny bèh'cw?
whose p-stick with com-hit John dog
44) *¿X-cyàg túú cùn ù-díny Juààny bèh'cw?
p-stick whose with com-hit John dog
45) *¿Túú x-cyàg ù-díny Juààny bèh'cw cun?
whose p-stick com-hit John dog with
We find the same two possibilities for negative focus:
46) Rú-tèh'cà túú lò chèh' ù-dèhhdy Màrìì cààrrt.
anim-not anyone to husband com-give Mary letter
'Mary didn't give the letter to anyone's husband.'
47) Rú-tèh'cà lò túú chèh' ù-dèhhdy Màrìì cààrrt.
anim-not to anyone husband com-give Mary letter
'Mary didn't give the letter to anyone's husband.'
The two grammatical possibilities for such questions, with suggested labels, are shown below:
48) a. Prep Wh N ......... (P-initial question)
b. Wh Prep N ......... (Wh-initial question)
The P-initial questions are easier to deal with, so I will address them first.
The key to explaining why P-initial questions occur is understanding the constraints that rule out alternative candidates. Consider again the following pair (repeated from above):
49) ¿Cùn túú x-cyàg ù-díny Juààny bèh'cw?
with whose p-stick com-hit John dog
'With whose stick did John hit the dog?'
50) *¿Túú x-cyàg cùn ù-díny Juààny bèh'cw?
whose p-stick with com-hit John dog
The candidate sentence in (50) requires a structure like the following for PP:
Note that in this structure, two ordering constraints-- X < Comp and X < Spec--are violated. We have seen that individually both of these constraints are outranked by Align (IntF,L, CP, L).
However, it seems that a candidate which simultaneously violates both ordering preferences is completely ruled out. To account for this, I will follow the now standard assumption in phonology that the conjunction of constraints may be separately ranked. The ungrammaticality of the following example also shows us that of the two ordering preferences, X < Comp outranks X< Spec:
51) *¿X-cyàg túú cùn ù-díny Juààny bèh'cw?
p-stick whose with com-hit John dog
We can account for success of the P-initial candidate relative to the other candidates with the following tableau:
| HCon-ADJ | X <Spec & X < Comp | Align
(IntF,L, CP, L) |
X < Comp | X < Spec | ||
| a. | ¿X-cyàg túú
cùn ù-díny
Juààny bèh'cw?
(Stick whose with hit John the dog) |
* | *! | |||
| b. | »¿Cùn túú x-cyàg ù-díny
Juààny bèh'cw?
(With whose stick hit John the dog) |
* | * | |||
| c. | ¿Túú x-cyàg
cùn ù-díny
Juààny bèh'cw?
(Whose stick with hit John the dog) |
*! | * | * | ||
| d. | ¿Túú x-cyàg ù-díny Juààny
bèh'cw cùn?
(Whose stick hit John the dog with) |
*! | * |
The more puzzling pattern is the one seen in examples like the following:
52) ¿Túú cùn x-cyàg ù-díny Juààny bèh'cw?
whose with p-stick com-hit John dog
(Whose with stick did John hit the dog?)
I will tentatively assign the following c-structure to this example:
This structure puts the initial wh-word in [Spec, PP]. Such a structure is possible, but should be marked, because túú 'who' and xcyàg 'dog (possessed)' form an f-structure constituent, but not a c-structure constituent.
In accordance with the other harmony constraints, we can posit the following:
53) Harmony of Constituency -- OBJTh
An f-structure constituent with the function OBJTh must correspond to a c-structure constituent.
With this constraint we can now consider the full range of candidates:
| HCon-ADJ | X < Spec
X < Comp |
HCon-OBJTh | Align
(IntF,L, CP, L) |
X < Comp | X < Spec | ||
| a. | ¿X-cyàg túú
cùn ù-díny
Juààny
bèh'cw?
(Stick whose with hit John the dog) |
* | *! | ||||
| b. | »¿Cùn túú
x-cyàg ù-díny Juààny
bèh'cw?
(With whose stick hit John the dog) |
* | * | ||||
| c. | ¿Túú x-cyàg
cùn ù-díny
Juààny
bèh'cw?
(Whose stick with hit John the dog) |
*! | * | * | |||
| d. | »¿Túú cùn
x-cyàg ù-díny Juààny
bèh'cw?
(Whose with stick hit John the dog?) |
* | * | ||||
| e. | ¿Túú x-cyàg
ù-díny
Juààny
bèh'cw cùn?
(Whose stick hit John the dog with) |
*! | * |
The constraints HCon-OBJ and Align (IntF, L, CP, L) are unranked with respect to each other and so two candidates end up tying for best. (This should be indicated with a dotted line between the two columns, but is not in this HTML version of the paper.) Each violates one of these two constraints.
8. ConclusionIn this paper, I hope to have demonstrated that an approach which accounts for word-order variation through violable constraints is able to provide a simple explanation of some otherwise difficult facts in the syntax of San Dionicio Ocotepec Zapotec.
References
Aissen, Judith. 1996. Pied-piping, abstract agreement, and functional projections in Tzotzil. Natural language and linguistic theory 14:447-491.
Black, Cheryl. 1994. Quiegolani Zapotec syntax. PhD thesis. University of California, Santa Cruz.
Bresnan, Joan. 1998. Optimal syntax. Ms. Stanford University. (Available at http://www-lfg.stanford.edu/lfg/bresnan/pt3.ps.)
Bresnan, Joan. forthcoming. Lexical functional syntax.
Broadwell, George Aaron. 1999. Focus alignment and optimal order in Zapotec. Proceedings of the 35th Chicago Linguistics Society. (Also available at http://www.albany.edu/anthro/fac/broadwell.htm)
Chomsky, Noam. 1995. The minimalist program. Cambridge, MA: MIT Press.
Falk, Yehuda. 1983. Constituency, word order, and phrase structure rules. Linguistic Analysis 11:331-360.
Farmer, Ann. 1980. On the interaction of morphology and syntax. Ph.D. thesis, MIT.
Farmer, Ann, 1984. Modularity in syntax: A study of Japanese and English. Cambridge, MA: MIT Press.
Gazdar, Gerald; Ewan Klein; Geoffrey Pullum; and Ivan Sag. 1985. Generalized phrase structure grammar. Cambridge, Ma: Harvard.
Grimshaw, Jane. 1991. Extended projection. Ms. Dept. of Linguistics and Center for Cognitive Science, Rutgers University.
Grimshaw, Jane. 1997. Projection, heads, and optimality. Linguistic Inquiry 28:73-422.
Jackendoff, Ray. 1990. Semantic structures. Cambridge, MA: MIT Press.
Kaplan, Ronald and Annie Zaenen. 1989. Long-distance dependencies, constituent structure, and functional uncertainty. pp. 17-42 in Alternative conceptions of phrase structure, ed. by Mark Baltin and Anthony Kroch. Chicago: University of Chicago Press.
Kayne, Richard. 1994. The antisymmetry of syntax Cambridge, MA: MIT Press.
King, Tracy Holloway. 1995. Configuring topic and focus in Russian. Stanford: CSLI.
Pollard, Carl and Ivan Sag. 1987. Information-based syntax and semantics. Stanford:CSLI.
Sadock, Jerrold. 1991. Autolexical syntax. Chicago: University of Chicago Press.
Smith Stark, Thomas. 1988. 'Pied-piping' con inversion en preguntas parciales. Ms. Centro de estudios lingüísticos y literarios, Colegio de México y Seminario de lenguas indígenas.
Stowell, Timothy. 1981. Origins of phrase structure. PhD thesis, MIT.
van Riemsdijk, Henk. 1978. A case study in syntactic markedness: The binding nature of prepositional phrases. Dordrecht: Foris.
Van Valin Jr., Robert D. and Randy J. LaPolla. Syntax: structure, meaning, and function. Cambridge: Cambridge University Press.
1. An earlier version of this paper was given at the Chicago Linguistic Society. Thanks are due to Peter Austin, Lee Bickmore, Joan Bresnan, Yehuda Falk, Javier Gutiérrez Rexach, Ed Keer, Jerrold Sadock, and Robert Van Valin for helpful discussions of this paper. Special thanks to Luisa Martínez, who supplied all the data.
The orthography for SDZ is adapted from the practical orthographies for other Zapotec languages spoken in the Valley
of Oaxaca. In the SDZ orthography,
Glosses use the following abbreviations: anim = animative, com = completive aspect, inan = inanimate, p = possessed, q
= question.
A few features of this paper have been altered to accomodate the display constraints of current HTML. The winning candidate in a tableau is shown with », rather than the usual arrow or finger. I have also omitted the now standard shading of cells in a tableau. OBJTh is the nearest approximation available in HTML for OBJ with a theta subscript.
2. The S can, however, be preceded by negation and negative focussed phrases, as discussed below.
3. SDZ uses the wh-words xhíí 'what, which' for inanimates and túú 'who, what, which' for animates (both people and
animals). I've glossed the examples with the appropriate English wh-word.
4. As suggested by King (1995), Bresnan (1998), Universal Grammar allows clauses with both endocentric (IP) and
lexocentric (S) organizations. Some VSO languages are best analyzed with the LFG analogue of head-movement to
INFL; others show a flatter syntax.
5. Although these examples use a preposition borrowed from Spanish (cun 'with' < Spanish con), the facts described
work identically in sentences with native prepositions.
6. See Broadwell (1999) for a more extended discussion of the difficulty of deriving these facts under a movement-based account.
7. In SDZ alienable possession, the possessed N has a /x-/ prefix, and the initial consonant of the noun stem is
devoiced. In a few cases there are irregular changes to the initial consonant, e.g. yàg 'stick', x-cyàg Juààny 'John's
stick'.
8. This constraint is not completely parallel to the constraint on interrogative focus, since that constraint aligned the
interrogative focus with the left edge of CP. There do not appear to be any clauses in SDZ with both an overt complementizer and a wh-phrase. If we posit an
empty Comp, we could restate the constraint on interrogative focus as Align(IntF, L, Comp, L), but empty nodes are
not in keeping with Economy of Expression (Bresnan, forthcoming).