A Serial Verb Construction in Tzotzil*
For Jorge, who has always liked a puzzle and a straightforward
|'Then the man was left there wearing the skirt.' (OCK:49)|
This sentence occurs in a story about a man who ends up inside his wife's skirt on his wedding night (Chamulan Tzotzil skirts are wide cylinders of cloth, tightly bound during the day with sashes, but loosened at night). The wife slips out of her skirt in the morning, leaving the man inside.
From a linguistic point of view, (1) is intriguing because it seems to involve two intertwined predicate domains, one consisting of ikom 'remained' (intransitive) with its (sole) argument li vinike 'the man', the other of slapoj 'wore' (transitive) with its external argument li vinike 'the man' and its internal argument tzekil 'the skirt'. Under this construal, the transitive verb slapoj is separated from its internal argument tzekil by the verb ikom, as diagrammed in (2).
Nowhere else in Tzotzil is an internal argument separated from the verb which licenses it by a verb to which it bears no relation. The question then is why such a discontinuity would arise here.
While the analysis of this construction is by no means obvious, what we can do here is show that (2) is incorrect. The dependencies shown in (2) are based on the translation of (1), which is taken directly from Laughlin's collection of texts (1977). But that translation turns out to be misleading, at least from a grammatical point of view. There are good reasonsboth syntactic and semanticto conclude that the sole argument of ikom 'remained' is not li vinike 'the man', but tzekil 'skirt'. Thus a closer translation might be the skirt was left, worn by the man. This is shown in (3):
While tzekil 'skirt' is still separated from slapoj by the predicate ikom, its presence in the structure is also licensed by ikom, which is contiguous to it.
The analyses in (2) and (3) lead to distinct predictions along several dimensions. The following makes reference to the labels in (4) (the labels NPinternal and NPexternal are meant to suggest what I think is the right analysis, but they can be regarded simply as tags).
The first prediction concerns agreement on V2, the position filled by ikom 'remained' in (1). V2 is the finite verb in this constructionit carries aspect marking for the clause as a whole and it might reasonably be expected to agree with its argument. The representation in (2) predicts V2 should agree with NPexternal, while (3) predicts agreement with NPinternal. The example in (1) does not tell us much, however, since both NPs are 3rd person, and 3rd person absolutive agreement in Tzotzil is Ø. Indeed, V2 carries no overt agreement, in accord with both (2) and (3). However, (2) predicts that making NPexternal 1st or 2nd person will induce appropriate agreement on V2, while (3) predicts that V2 will remain without overt agreement. In fact, changing the person of NPexternal induces 1st or 2nd person (ergative) agreement on V1, but V2 remains uninflected for agreement:
|'You were left wearing the skirt||(better: the skirt was left, with you wearing it).'|
|'I left, carrying my bag||(better: I carried my bag away).'|
This supports (3) over (2). Per (3), the sole argument of V2 is NPinternal, which is 3rd person in both (5) and (6). If this line of reasoning is right, we should be able to force overt agreement on V2 by making NPinternal 1st or 2nd person. This is correct:
|'They pursued me||(more literally: they pushed me, I went).'|
Grammatical agreement then strongly supports the analysis in (3), in which it is NPinternal which functions as the sole argument of V2, not NPexternal.
The analyses in (2) and (3) also make distinct claims about the construal of examples like (1). According to (2), V2 applies to NPexternal; according to (3), it applies to NPinternal. While this difference should have detectable semantic consequences, it is hard to discern them in (1): if the man is left wearing the skirt, then the skirt is left with the man wearing it. This problem is not limited to (1). V2 is most commonly filled by a verb of motion (or absence of motion) and movement by NPexternal usually entails the same movement by NPinternal. In (6), for example, if I go, carrying my bag, then my bag also goes.
However, some verbs of motion carry presuppositions which can help clarify which of the arguments in (4) V2 applies to. If the analysis in (3) is correct, then those presuppositions should apply to NPinternal, not to NPexternal. This can be tested by instantiating V2 by a verb like return which asserts that its argument A reached a destination D and presupposes that A was at D at some earlier (discontinuous) point in time. The linguistic intuitions of Tzotzil speakers are instructive here. Translating (8) as they returned wearing their hats, in line with the analysis in (2), does not account for the fact that (8) is appropriate only if the men set out in the first place with their hats, i.e. the hats must return.
|'They wore their hats back again.'|
Speaker CKH suggested that an appropriate scenario for (8) would be one in which they men went out to sell their hats, but failing to do so, returned with them. Likewise, translating (9) as the man returned carrying his bag fails to account for the fact that the man must have set out with his bag:
|'The man carried his bag back.'|
Speaker MP suggested that (9) would be appropriate if the man brought a bag of fruit to his girlfriend (fruit is a traditional offering in Zinacantec courtship). The girlfriend, however, does not accept the fruit, and the bag must be carried back.
Finally, while verbs of motion are the most common as instantiations of V2, speakers have accepted other verbs. The meaning of (10) is unambiguous: it is the child who is crying, not the man.
|'The man carried his son crying.'|
Here, V2 clearly applies to NPinternal, not to NPexternal. Furthermore, speaker reactions to examples like (11) are very sharp:
|('The man carried his bag, dancing/working.')|
CKH and MP had exactly the same metalinguistic reaction to the examples in (11): namely, that these examples are unacceptable because bags do not dance, neither do they work. These speakers clearly take NPinternal, not NPexternal, to be the argument of V2.
Happily then, the semantic evidence converges with the evidence from agreement: the argument of V2 is NPinternal. Thus while tzekil in (1) is separated from slapoj, the element which separates them itself licenses tzekil. The analysis in (3) raises its own puzzles, but it does not entail the discontinuity of (2).
Having clarified (some of) the basic facts, we now face the question of how to analyze them. In certain important respects, (1) and (5)(10) look like transitive clauses, with the characteristic VOS order of Tzotzil. What is special about them is that each contains two predicates which jointly determine the argument structure of the clause. In (1), tzekil 'skirt' is licensed twice, being simultaneously the internal argument of both V1 and V2; NPexternal, on the other hand, is licensed only by V1. Likewise for the other examples.
This construction appears to fall squarely in the category of what have (somewhat loosely) been called 'serial verb constructions'. Generative work on such constructions has focused on the questions of structure, of semantic role assignment, and on what feature or parameter distinguishes languages with serial verb constructions from those without (see, for example, Baker (1989) and the papers in Lefebvre (1991)).
With respect to the question of parameters, Larson (1991) makes the interesting suggestion that serial verb constructions are a type of secondary predication and that 'serial verb languages' are just those which allow verbs to function as secondary predicates (English is not such a language). This suggestion provides a good starting point for the analysis of the Tzotzil construction. If there is a single main verb in this structure, it is V2, the verb which carries aspect marking for the whole clause. (V1 is inflected in the perfect, which seems to function here as a non-finite form.) Significantly, the position in which V1 occurs is the position in Tzotzil reserved for secondary predicates of various types, including a range of verbal predicates. The construction under discussion then appears to be nothing other than a clause with a transitive secondary predicate. If this is right, then what we have is a main verb (V2) which is intransitive and which licenses the internal argument, and a transitive secondary predicate (V1) which licenses the external argument. This seems odd, but is consistent, at least in spirit, with a growing body of work that sees the determination of clausal argument structure by several distinct heads as more the rule than the exception (Borer 1994; Kratzer 1996; Jelinek 1998; Ritter and Rosen 1998).
Baker, M. (1989). Object sharing and projection in serial verb constructions. Natural Language and Linguistic Theory 20(1): 513553.
Borer, H. (1994). The projection of arguments. In E. Benedicto and J. Runner, (eds.), Functional Projections, University of Massachusetts Occasional Papers 17, Amherst, Massachusetts: GLSA, University of Massachusetts.
Jelinek, E. (1998). Voice and transitivity as functional projections in Yaqui. In M. Butt and W. Geuder, (eds.), The projection of arguments: lexical and compositional factors, Stanford, CA: CSLI, pp. 177206.
Kratzer, A. (1996). Severing the external argument from its verb. In J. Rooryk and L. Zaring, (eds.), Phrase structure and the lexicon, Dordrecht: Kluwer, pp. 109137.
Larson, R. (1991). Some issues in verb serialization. In C. Lefebvre, (ed.) Serial verbs: Grammatical, comparative and cognitive approaches, Amsterdam: John Benjamins, pp. 185210.
Laughlin, R. (1977). Of cabbages and kings. Washington, D.C.: Smithsonian Institution Press.
Lefebvre, C., Ed. (1991). Serial verbs: Grammatical, comparative and cognitive approaches. SSLS #8. Amsterdam, John Benjamins.
Ritter, E. and S. T. Rosen (1998). Delimiting events in Syntax. In M. Butt and W. Geuder, (eds.), The projection of arguments, Stanford, CA: CSLI, pp. 135164.
* This squib is based on work undertaken principally in 1983 and 1990 with two speakers of Tzotzil (Mayan), Chep Kontzares Hernantis and Manvel Peres, both of Zinacantán, Chiapas Mexico). I am grateful to them for their help. [Back]
1 Abbreviations used in glosses: OCK = Laughlin 1977, E = ergative agreement marker, A = absolutive agreement marker, 1/2/3 = 1st/2nd/3rd person, PF = perfect aspect, CP = completive aspect, ENC = enclitic, PL = plural. [Back]