Equivocal Structures in Some Gullah Complex Sentences

Equivocal Structures in Some Gullah Complex Sentences Author(s): Salikoko S. Mufwene Source: American Speech, Vol. 64, No. 4 (Winter, 1989), pp. 304-326 Published by: Duke University Press Stable URL: http://www.jstor.org/stable/455723 Accessed: 29/06/2010 01:32 Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/action/showPublisher?publisherCode=duke. Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

Duke University Press is collaborating with JSTOR to digitize, preserve and extend access to American Speech.

http://www.jstor.org

EQUIVOCAL STRUCTURES IN SOME GULLAH COMPLEX SENTENCES SALIKOKO S. MUFWENE Universityof Georgia T

HE SUBJECT MATTER OF THIS ESSAY IS SENTENCES such as the follow-

ing, which contain a subordinate clause starting with fa 'for' or se that':' 'say, la. [A w5 fo go] (common) 'I want to go' b. [A won Am fo peYmi mi mAnI] (JG) 'I want him to pay me my money' c. [A tel Am fo kAm teYkyu tu do dAkto] (elicited; original by LW without

[fa]) 'I told him to come and take you to the doctor' d. [A teYko naYffo kAt da mi:t] (common) 'I took a knife (in order) to cut

the meat' 2a. [A tEl Am se rabot teYk sik] (common) 'I told {him/her} that Robert fell

b.

sick' [A hie se rabotteYksik] (common)'I heard that Robert{fell/has fallen} sick'

Generally, sentences such as (la-d), to which most of this paper is devoted, have been assigned structural analyses by scholars which are similar to those of their English translations, with the embedded clauses analyzed as infinitival (e.g., Cunningham 1970, Nichols 1975, 1976, Jones-Jackson 1986; outside Gullah: Bailey 1966, Winford 1985).2 The morpheme {fo} has been equated with its English etymon for in clauseinitial position and has been identified as a complementizer.3 Sentences like (2a-b), discussed in part 2 (below), have received less attention. The few studies which discuss or mention the construction in the literature on Atlantic creoles (AC) either interpret [se] as a complementizer (Cunningham 1970, Bailey 1966) or at least recognize this as one of its functions (Alleyne 1980). One exception, Jones-Jackson (1986), treats it as a verb but does not mention its usage in sentences such as (2b). The other one, Bickerton (1981), claims simply that [se] is not a complementizer.4 This paper argues that, unlike their English translations, at least some of the sentences involving subordinate clauses starting with [fo] or [se] may be assigned more than one structural analysis. However, this structural equivocation is not tantamount to what has traditionally been characterized as "structural ambiguity" in transformational syntax. As will become apparent in parts 1 and 2, the meanings associated with the 304

EQUIVOCALSTRUCTURES IN GULLAH

305

alternative structures actually amount to more or less the same thing; the difference in the structural analyses of the sentences follows less from the semantics of [fa] or [se] than from the lexical category membership assigned to them. At present, there seems to be no evidence for favoring one analysis over other alternatives. This paper also addresses in its conclusions the significance of variation (both intra- and inter-individual) in grammar. Reflecting on the Saussurean conventionality of language, it raises the question of whether, in the first place, it is justified to assume that only one structural analysis must be assigned to nonambiguous sentences. 1. [fo] COMPLEMENTS 1. ARE CLAUSESINTRODUCED BY [fo] INFINITIVAL? A concomitant

to the

characterization of the clause-initial [fo] in sentences (la-d) as a complementizer is the assumption that the verb it precedes is infinitival. To my knowledge, no justification has ever been given in the literature on Gullah or other AC's for this analysis of the verb. Since the position on this issue is relevant to the main thesis of this paper, it should help to summarize below the main argument adduced in Mufwene and Dijkhoff (1986) for rejecting the finite/nonfinite distinction as part of the grammatical systems of AC basilects, hence against characterizing the second verbs of sentences (la-d) as infinitival.5 The infinitive is a morphological distinction which fits in a paradigm of inflections marking verbal forms differently depending on their mood, tense, person, and number.6 Yet, the form of the verb is invariant in the basilects of Gullah and other AC's. Mood, tense, and aspect distinctions are marked syntactically by preposing a free morpheme (e.g., [da] for DURATIVE or [bin] for ANTERIOR) to the verb. The other distinctions are indicated either through the subject pronoun for person and number, or through the subject noun for number (the person being invariably third in this case).7 Like isolating languages (e.g., Chinese), AC's lack agreement features (viz., number, person, and tense) on the verb, which in the vast majority of Indo-European languages determine whether a verb is finite or nonfinite. (See, e.g., Aoun 1985 and Chomsky 1986 on Government-Binding [GB] theory, where this has been of much concern in recent years.) Thus, while it is justified in Gullah and other AC's to speak of verbs or predicates because they can be delimited (at least semantically) in tense, mood, aspect, person, and number, it does not seem justified to assume a finite/nonfinite distinction, since these delimitations are realized periphrastically rather than inflectionally.

AMERICAN SPEECH 64.4 (1989)

306

In GB theory, languages without agreement features are also assumed not to have the PRO/prodistinction made in the identification of null subjects in languages such as Italian, Spanish, and, for that matter, English (although the literature has generally ignored it). According to this theory, PROstands either for a null subject with arbitrary reference, as in Johni is nice PROarb to talk to ti, or for the null subject of a nonfinite clause in the following: 3a. Johni wants PROi to leave. b. Johni stopped PRO,writing his letter. On the other hand, pro stands for the null subject of a finite verb as in (for example) the following sentences of colloquial English:8 4a. b. c. d.

pro having fun, isn't he? prowant some coffee? pro hope this is clear enough. pro hang in there.

In languages without agreement features, either pro or PRO alone (but not both) may be used for all null pronouns (see, e.g., Aoun 1985, Battistella 1985, Huang 1987, and Xu 1987). This position applies also to Gullah and other AC's. Thus, since there are, in addition to invariant verb forms, null-subject main clauses such as (5a-c), the argument that clauses where [fo] is followed by a null subject must be infinitival does not hold.9 5a. [o: IIsi brotsAmleYtlI?] (EL)-[pro do gud a waYl](JM) 'Oh, she brought some lately?-(It) has been a good while' b. [wen daenbrok do do da rum open en pro go on in] (JM) 'When Dan, broke the door, the room (was)opened, and he, went on in' c. [maenA planto1sot A tIl I pro plant poteYto I1pro plantbi:nz](JM) 'Man, I planted all sorts of things. (I) planted potatoes. (I) planted beans' Worse for this argument, there are also some sentences such as (6), where [fa] is followed by an overt subject pronoun, the form of which is just like in a main clause. 6. [A w5 fa hi peY mi mi mAnI] (JG) 'I want him to pay me my money'"0

For those who might have thought of interpreting the distinction between the putative finite and infinitival verb uses syntactically (e.g., in expecting the pronominal subject of the infinitive to be in the objective

EQUIVOCALSTRUCTURES IN GULLAH

307

case, as in English), the subjective form in (6), unlike in sentence (lb), weakens their case. There is thus enough evidence for assuming that the Gullah clauses introduced by [fa] are not infinitival. 2. IS THE CLAUSE-INITIAL [fY] NECESSARILY A COMPLEMENTIZER? The clause-initial position of [fo] in the sentences discussed so far (with the exception of (lb)) as well as its preverbal position in especially sentence (1c) seem to have something to do with its characterization as a complementizer. It seems plausible to assume in at least the case of (la) and (lc) that a null subject pronoun occurs between the complementizer and the verb. This position may be disputed (as I do below) on the basis of sentence (Id), where [fo] could very well be characterized as a preposition (followed by no complementizer). It may, however, be claimed to find support both in sentence (6) and in the following alternative constructions to sentence (lb), which native speakers find equally acceptable: 7a. [A w5 fo, hi fo2 peY mi mi mAnI] (elicited) b. [A w5 0 hi peY mi mi mAni] (elicited)II

It could thus be argued that the clause-initial [fa] is optional, just like the post-subject [fo] seems to be. Since English (especially the nonstandard varieties to which the Afro-Americans must have been exposed in the seventeenth and eighteenth centuries) lexified Gullah, it is tempting to assume that the correlative pair [fol] [fKo]in (7a) is an analogue of the English for to, with the difference that [fo2] is used instead of to. This interpretation seems to be supported by the fact that in Appalachian and Ozark English (see, e.g., Wolfram and Christian 1976) and in Middle English (Lightfoot 1979) sentences such as I want for to go are common, with for to functioning as one correlative unit. Thus it could be argued that Gullah speakers have found significance only in [fo] and may repeat it, apparently redundantly, as in (7a). While the above considerations seem justified, sentence (Id), where the interpretation of [fo] as a complementizer is more obviously disputable, cannot be overlooked. [fo], the only marker of purpose here, must be characterized as a preposition. This analysis finds support in the existence of other sentences such as (8), where, incidentally, the verb [brlj] may not be translated by the infinitive in English: 8. [tejk yu fa brig do yAIJmaenfo si: mi] (AW)'Thank you for bringing the young man to see me' An argument could of course be made here for a case of homonymy, assuming two [fo]'s (more or less like for the two to's in the English

308

AMERICAN SPEECH 64.4 (1989)

examples John likesto fly and John is accustomedto flying), the first, in (lac), (6), and (7), being the complementizer and the second, in (Id) and before [brij] in (8), being the preposition. However, this position would be hard to defend in the absence of concomitant formal evidence. As a matter of fact, it is argued below that there is no solid evidence for ruling out the possible interpretation of clause-initial [fa] in sentences (la-c), (6) and (7a) as basically a preposition (albeit a bleached one which now plays only a syntactic function after future-oriented main verbs). It can be assumed that the clause-initial [fo] in the above sentences is somewhat comparable to the French pour both in its purposive function, as in especially (Id), and in its causal function, as in (8).12 Unlike English, where the purposive for takes an infinitival complement and the causal for a gerundial one, the French pour, as illustrated below, takes the infinitive in both cases. The only difference between the purposive and the causal complements lies in their tenses: present infinitive for the purposive complement and past infinitive (passe compose)for the causal complement. 9a. J'ai pris un couteau pour couper la viande. (cf. [Id]), 'I have taken a knife in-order (for-)to cut the meat' b. Mercipour avoiramen6lejeune homme (pour)me voir. (cf. [8])'Thank [you] for having brought the young man to see me' In Gullah, as illustrated by (Id) and (8), the form of the verb is invariant. Like French, however, Gullah allows the same tense markers to be used in clauses introduced by [fa] as in main clauses. In the following sentence, the same marker [bin] (like avoir in the French passe compose) is used:13 lOa. [dartswaeyu get fa bin teYkmi kYad](EL) 'That's what you get for {taking/havingtaken}my card' b. [A no yu bin teYkmi kYad](common) 'I know you {took/had taken} my card' The main differences between Gullah's clause-initial [fa] and its putative French counterpart may be stated as follows: (1) in French the preposition pour may be followed by either a finite or a nonfinite clause, whereas in Gullah the finite/nonfinite distinction does not apply; (2) in French a finite clause which is the object of pour must be introduced by the finite-clause complementizer que, as in ( lla), whereas in Gullah it may be argued that no complementizer is required in such a case. 1 a. J'ai pris un couteau pour qu'ilcoupe la viande 'I took a knife for him to cut the meat with'

EQUIVOCALSTRUCTURES IN GULLAH b.

[A teYk

309

a naYffa 0 hi kAt do mit] (elicited)

As speculative as the following conclusion based on meaning may be, the absence of the complementizer after the preposition [fo] in sentences (ld), (8), (10), and (1 b) may be responsible for the SAMEpreposition [fa] possibly being used (by extension) to introduce the complements of main verbs which are future- or purpose-oriented in their meanings. The prepositional meaning remains transparent in adverbial clauses, as in (Id), (8), (10), and (1 b), but rather bleached in nonadverbial clauses, as in (la-c), where, as some of the above data show, [fa] is used rather redundantly. (Other corroborative evidence for the latter claim follows in sections 1.4 and 2.1.) 3. ARE ALL [fO]'s PREPOSITIONAL? Before addressing

the question

of

equivocal structures, for which the above considerations have lain the groundwork, the question of the grammatical function of the second [fa] in (7a), which has been overlooked so far, must now be addressed. Contrary to traditional analyses which assumed the second [fa] (identified as fi2 for Jamaican Creole) to be an infinitival marker, Winford (1985) has justifiably proposed to analyze it as a modal verb, in contrast with the first ([fo]/fil), which is interpreted as a preposition-complementizer (PREP-COMP). This analysis is supported by the fact that, in English AC's, the preposition [fo] (or its regional variantsfi andfu) is also used in main clauses as a modal verb, as in (12a-b):14 12a. [ti fo stEYwi9 graxma](ET) 'T must stay with Grandma' b. [A fo tel yu] (AS) 'I must tell you' This modal use of [fa] may also be overtly tensed, as below: 13. [ti bin fo steYwil graema](elicited)'T had to stay with Grandma' What makes the identification of the two uses of [fa] in (especially) sentence (7a) relevant to this paper is particularly the relatedness of their basic meanings.

Both the PREP-COMP interpretation

of [fo], as in (la-c),

and its modal interpretation, as (12-13), are purpose- or future-oriented. The PREP-COMP use could have been extended from purely purposeoriented prepositional uses such as in sentence (Id). Based on the optionality of [fa] after the verbs [w5] 'want' and [traY]'try', it could be assumed that the usage has, in post-main-verb contexts, simply bleached from the clear prepositional function to the complementizer function. The bleaching could also account for its omission from some of the constructions which are analyzed in the next section as serial.

310

AMERICAN SPEECH 64.4 (1989)

As for the relation of the preposition to the modal, this is supported by the fact that, in AC's (with the exception of Papiamentu), prepositions (hence prepositional phrases) are used predicatively without a copula.15 Since tense is not inflectional and seems to be associated with predicativeness rather than verbness, prepositions (like adjectives) are easily preceded by a tense marker in predicative function. The modal use of [fa] may thus not only be related etymologically to the same basic preposition [fo] through regular creole syntax, but also be analyzed synchronically as the same preposition [fo] in predicative use and having a verb phrase as its object. Byrne (1986a) proposes an alternative hypothesis which reduces both uses of [fo] to one verbal, rather than prepositional, case. The main reason for this analysis might be to accommodate the (overt) tense marker [bin] which may delimit [fa] in its modal-verbal function. However, as stated above, noninflectional tense may be associated with predicativeness rather than verbness. Both Byrne's analysis and the alternative presented here can account for the use of [fo] in nominal phrases such as [da haws fa Jin dAn bild] 'The house for Jeanne has been built'. For instance, according to Byrne's analysis, [fo Jin] in the above sentence may be interpreted as a contact relative clause. However, the following facts seem to favor the hypothesis that the prepositional function is the basic one from which the modal and PREP-COMP uses are both derived: (1) Not only is the English etymon a preposition, but also the function itself, which (subject to typological constraints) is crosslinguistically common, has been fairly well preserved in AC's as illustrated by the example just given. (2) Other prepositions are, like [fo], also used predicatively in AC's even though they do not have a modal function (and have not developed into complementizers), for example, [hi fram do siti] 'he [is] from the city'. (3) The diachronic development from the prepositional to the modal and complementizer functions (just within the context of the restructuring of English into a creole) need not be unidirectional, that is, from a preposition in English to a verb in creole and from there to a complementizer. It could, as hypothesized here, take diverging directions. That is, both PREP-COMP and the modal uses are derived directly from the preposition. The meaning of [fo] gets bleached where it is used redundantly after purpose- or future-oriented main verbs. Most of the equivocations discussed in the next section regard whether [fo] must be analyzed as a PREP-COMP or as a modal. It will be shown that, in a number of cases, either analysis is plausible. However, the equivocation is strictly syntactic and does not clearly result in ambiguity,

EQUIVOCALSTRUCTURES IN GULLAH

311

and the modal uses are extended from the same since the PREP-COMP basic meaning, as hypothesized above. 4. EQUIVOCAL ANALYSIS OF [fo] CLAUSES. Assuming, as in section 1.3, that there are a PREP-COMP and a modal [fa], that both are optional in clausal complements, and that their positions are constant, nothing precludes interpreting the [fa] in sentences (la-c) as a modal. After all, it is close to the verb and may be assumed to delimit this.16 Reformulated as a question against analyses which interpret [fa] solely as a PREP-COMP, since the subject is missing (on a less unanimous vote for sentence [lb] as shown below), how do we know for sure that it is the PREP-COMP [fa]I which occurs in these sentences, and not the modal? Under either the analysis of [fa] as a PREP-COMP or its interpretation as a modal, a null subject is allowed to occur; what test is there for determining which structural analysis is the more adequate here?

4.1. Let us start with sentence (Ib), since this allows us to bring more facts into consideration than does (la). An answer to the question [we yu w5] 'What do you want?' may be the following nonelliptical sentence: 14. [hi/*0 fo2 peYmi mi mAnI](elicited)'He/She must pay me my money' Here [fo] must definitely be assigned a modal interpretation, since it occurs after the subject; complementizers normally precede the subject, unless they are affixes or clitics. (Nothing suggests that [fa] may be a clitic in Gullah.) An alternative answer to the same question may be the elliptical clause below: 15. [faI hi

peY

mi mi mAnI] (elicited)'for him/her to pay me my money'

This is structurally more closely related to sentence (6) than is sentence (14), and it suggests that in its presubject position [ff] must be interpreted as a complementizer. A third alternative answer to the same question may also be the following elliptical clause: 16. [foahi fo2 peY mi mi mAni] (elicited)'for him to pay me my money' On the other hand, alternative (17) is excluded as a possible elliptical answer. Without the discourse frame provided by the matrix clause in sentence (7b), sentence (17) has a different, nonelliptical meaning, as indicated: 17. [hi peYmi mi mAnI] (elicited) 'He paid me my money'

Sentences (14-17) suggest that where Subject-Equi Deletion, or Subject-Control, must account for the interpretation of the null subject of

312

AMERICAN SPEECH 64.4 (1989)

clauses starting with [fa], as in sentence (la), either the modal or the PREP-COMP analysis of [fo] is conceivable.17 Only constructions whose interpretation is future-oriented are acceptable as answers to the question [we yu w5] 'What do you want?' As seen in section 1.3, both the modal and the COMPinterpretations of [fo] are future oriented; hence, there is no obvious justification for excluding either analysis. Both of the following analyses appear thus to be plausible; the choice of one or the other must be arbitrary:18 18a. [A^w5 [S'[COMPfa,] [s proi go]]] b. [A w5 [s'[coMP 0 ] [S proi f.2 go]]]

Tantalizing support for analysis (18a) might derive from its English infinitival translation. However, as I demonstrated in section 1.1, the finite/nonfinite distinction is not part of Gullah grammar. Even if Gullah had an infinitive, the status and function of this morphological category vary from language to language. For instance, as Mufwene and Dijkhoff (1986) show, French often uses a finite construction where English uses the infinitive, as illustrated by the following translation of sentence (lb): 19. Je veux qu'il me paie mon argent 'I want that he me pay-suBJUNCTIVE

my money = I want him to pay me my money'

The fact that, as in colloquial English, Gullah allows main clauses with null subjects when the referent can be inferred from the context (see section 1.1) is an argument in support of structural analysis (18b). Whether or not sentence (lb) is also structurally equivocal depends on the theoretical framework of one's analysis. In GB theory, where no raising of the embedded subject to a higher object position is assumed, as in (20a), [fo] must, according to the facts seen so far, be interpreted as a modal; the objective form [Am] must therefore be attributed to exceptional case marking by the higher verb. However, according to more traditional transformational frameworks, sentence (lb) is also structurally equivocal, since the surface-structure object of the higher verb may be derived from a position either before or after [fa], as in (20b-c), which could accordingly be interpreted either as a modal or as a PREP-COMP.Thus, the sentence

could be represented

either of the following ways:19 20a. [A won [s'[coMP0 ] [s Am fa2 PEYmi mi mAnI]]] b. [A won Am, [s' [coMP fi] [s ti peY mi mi mAnI]]] c. [A won Ami [s'[coMP 0 ] [s fO2 PeY mi mi mAnI]]]

structurally

in

EQUIVOCALSTRUCTURES IN GULLAH

313

4.2. The analysis of sentence (ic) is also equivocal; either one of the following structures seems plausible: 21a. b.

[A tel Ami [S'[CoMPf0i] [s proi kAm teYk yu tu do dAkta]]] [A tel Am, [s'[coMp 0 ] [s proif22 kAm teyk yu tu da dAkta]]]

The equivocation is supported by a number of facts, the first of which is sentence (22), where [fo] is omitted without affecting the meaning conveyed by (Ic): 22. [A tel Am kAm t?Ykyu tu da dAkta] (LW) 'I told him to (come and) take

you to the doctor' This spontaneous sentence, whose pattern is quite common and for which alternative (1c) was elicited, can be analyzed as a serial construction. This analysis is suggested by the absence of a complementizer between [tel Am] and [kAm] (disregarding the [kAm teYk]sequence) as well as by the sharing of an NP (at least referentially) by the sequenced verbs. Sentence (22) is relevant if the following other facts are taken into account. As illustrated by the examples in (5) (above), sentences may have null subjects regardless of whether they are dependent or main clauses. The absence of an overt subject for the second of two consecutive verbs is in part what makes verb serialization possible.20 In addition, as shown earlier in sentences (12-13), [fo] may have a modal function in any kind of clause. Thus, since there is no way of telling a priori that sentence (lc) does not involve serialization of predicates, the [fa] before its second verb could very well be either the PREP-COMP or the modal verb. After all, if serialization is interpreted syntactically as sequencing of verbs or predicates sharing an NP and connected with each other by no complementizer or conjunction, nothing precludes predicates with a modal function from occurring in such sequences. Possible elliptical answers to the question [we A tel Am] 'What did I tell him?' might be either of the constructions in (23), which also indicates that both analyses in (21) are plausible:21 23a. [fa, kAmteYkyu tu da dAkto](elicited)'to (come and) take you to the doctor' b. [hi fa2 kAmteYkyu tu do dAkto](elicited)'He had to (come and) take you to the doctor' Although the interpretation of [fo] in sentence (Ic) as a PREP-COMP might be attributed to language-external considerations, such as its infinitival translation in English, there is no strong reason for doing so.

AMERICAN SPEECH 64.4 (1989)

314

Especially with [Am] functioning as the dative object, [fa] can plausibly be interpreted as a PREP-COMPpreceding a null subject of the verb [kAm]. In the absence of a decisive argument for ruling out either the serial or the subordinate-clause analysis of the clause introduced by [fa], we must simply acknowledge that the structural analysis of sentence (1c) is equivocal. The question to address in this case is whether the different analyses correspond to different semantic interpretations of the sentence, hence whether it would be justified to speak of ambiguity here. The indirect evidence considered indicates that the semantic difference is slight, follows primarily from the lexical category membership assigned to [fo], and does not seem to affect the overall interpretation of the sentence. (Recall that both the modal and PREP-COMPuses of [fa] are future oriented and putatively derived from the purpose interpretation of the preposition [fa].) For instance, native speakers take the following two constructions to mean more or less the same thing, in spite of the differences in both the position of [fa] and of its interpretation (in terms of our linguistic analysis) as either a preposition or a modal (ADVL stands for adverbial clause): [A sen dis pleYt [ADVL[PREP fa] [s Jin sen mi SAm fud]]] (elicited)'I sent this plate for Jeanne to send me some food' b. [[s Asen dis pleYt][s Jin (bin) fo2 sen mi sam fud]] (elicited)'I sent this plate.Jeanne had to send me some food'

24a.

Even though there is no reason for considering [fo] a complementizer in (24a), the analyses (24a) and (24b) are quite parallel to those assigned to sentence (Ic), with the support of elliptical answers (23). It is not clear what the punctuation should be between the two clauses of (24b) taken as a paratactic utterance. But the modal interpretation of [fa] here indirectly supports the modal analysis of [fo] in (Ic). As for (24a), [fo] is not identified as a complementizer because it does not introduce a complement of the higher verb. The complement of the higher verb is [dis pleYt], and the role of [fo] here appears to be that of establishing the relation of the event of sending the plate to that of Jeanne sending some food, which is purposive-adverbial. Unlike complementizers, whose meaning derives at least partly from that of the matrix verb (for instance future, intent, or nonfactual orientation of verbs which take a clausal complement with a bleached [fo]), [fo] has an independent lexical meaning in (24a); there is nothing else in the sentence which identifies the purpose/intent. Thus [fo] must be a preposition (allowed to take a clausal complement without a complementizer) in (24a).22 The following other considerations justify interpreting [fo] as a prepo-

EQUIVOCALSTRUCTURES IN GULLAH

315

sition in (24a). The questions [we yu sen da? pleYtfo] 'What did you send that plate for?' and [we yu du da9 fo] 'Why did you do that?' may be appropriately asked in relation to the [fa]-clause of (24a), as of sentences (8) and (10), but not of (for example) sentence (6) or (7a). It is also possible to form a sentence such as (25a) on the basis of sentence (24a) whereas (25b) is impossible on the basis of (6): 25a. [fojini sen yu da fud yu fa2 sen Ami sAm mAnI tu] 'In order for Jeanne

to send you the food you must send some money too' b. *[f3 hi peYmi mi mAniA wo] 'For him to pay me my money I want' On the other hand, one can also see from the above discussion why the complementizer function of [fa] (as of its English etymon for) must have derived from that of the preposition (see, e.g., Emonds 1976, Lightfoot 1979, or Chomsky 1986 for English). What really seems to make the difference between identifying [fa] as a preposition and identifying it as a complementizer is whether or not the clause which it precedes is an adverbial adjunct or the complement of a higher verb. Since [fo] is normally used as a complementizer to subordinate a clause to a main verb which is semantically future oriented, the future orientation of the preposition may (as suggested in section 1.3) have something to do with its selection; the redundance following from the future-orientation of both the verb and the preposition must have led to the semantic bleaching of the latter and its generalized usage as a complementizer in other nonfactual cases. (This speculation may have to be demonstrated more convincingly in future work.) 4.3. As for sentence (Id), it may, like sentence (24a), be analyzed plausibly as follows: 26. [AiteYka nAYf[ADVL [PREP fa] [s proi kAt da mit]]] (elicited)

However, nothing precludes analyzing it, by analogy to (Ic), (2 b), and (22), as a serial (27a) or paratactic (27b) construction, with a modal [fa] after the first verb, as below:23 [VP [vp teYk a naYf] [VPf2 kAt da mit]]]24 b. [s Ai [vp teYka naYf]][s proi [vp fa2 kAt da mit]].

27a. [s A

Analysis (26) is supported by sentences such as 28. [s A teYka nAYf [ADv fa, rabat kAt da mit]] (common) 'I took a knife {so that Robert [can]/in order for Robert to} cut the meat'

316

AMERICAN SPEECH 64.4 (1989)

On the other hand, analyses (27) are supported by the following serial sentence, although, as pointed out by Bickerton (1984), it has an implicative meaning, which is different from that of sentence (Id): 29.

[A teYk

a nAYf kAt da mit] (common)'I took a knife [and] cut the meat'.

Evidence for the above equivocations comes also from the following sentence, which is shown in Mufwene (1986) to have two slightly different, but related, interpretations. The modal interpretation of the second [fo] is corroborated by the fact that it can be delimited by an overt tense marker: 30. [de: haedno fud fa, d3 c6i0n(bin) fo2 it] 'They had no food {for the children/whichthe children had} to eat.25 A counterpart of the paratactic interpretation (27b) is analysis (31), where the second VP has an independent, nominal subject: 31. [s A [vpteyk a nAYf]] [s rabat[vpf32 kAt do mit]] A reason for rejecting the serial analysis of (27a) could be a semantic constraint which might not allow modal verbs to follow other verbs in a serial construction. This is possible on account of the fact that the literature on serialization does not include such series. Otherwise, there seems to be no compelling evidence for preferring one structural analysis over other alternatives in regard to subjectless clauses whose first overt constituent is [fa], as illustrated by (la-d). There are no significant semantic differences to be correlated with the alternative analyses, either. The differences in analysis seem to follow essentially from our conception of lexical categories and their structural implications. 2. [sE] COMPLEMENTS 1. STRUCTURAL

EQUIVOCATION

OF CLAUSES STARTING

WITH [SE].

In the

following examples, one of which is a copy of sentence (2b), [se] functions undoubtedly as a complementizer: 32a. [A hio (se) rab3tteYksik] (common)'I heard (that Robert fell sick' b. [hi no (se) rabotteYksik] (common)'He {knows/knew}that Robertfell sick' However, while the complementizer analysis may be the only plausible one for the above constructions, such may not be the case for other sentences. [se] can also be analyzed serially in sentence (33a), and the serial analysis seems in fact to be the most adequate one for (33b-c),

EQUIVOCALSTRUCTURES IN GULLAH

317

unless more than one grammatical meaning is assigned to the putative complementizer:26 33a.

rabotteYksik] (common)'I told him {that Robertfell sick/ "Robertfell sick"}'

[A tel Am se

b. [A aeks Am se we yi mama do] (elicited) 'I asked him, "Where's your

mother?"'

c.

[A aeksAm SEYyi mamada hom] (elicited)'I asked him, "Isyour mother

home?"' The presence of the quotative, or direct, reported speech in (33b-c) is significant here. This style, with very little reorientation of deictics, is quite common in Gullah reported speech. This means that [se] may be interpreted here as a verb which is followed by a null complementizer and whose proproxsubject 'a la Byrne is coreferential with that of the preceding verb. There is some phonological evidence for the verbal interpretation of [se]. It is normally fully stressed and a pause may occur between it and the preceding verb. While the pause suggests parataxis rather than serialization, one can see where the distinction between the two construction types might be blurred due to the relatively free distribution of null subjects. The serial analysis is suggested mostly for the many cases where the pause does not occur (which is more natural) but [se] is fully stressed. The quotative style adds complexity to the subject matter of equivocal structures. Since postverbal [se] and [fo] qua PREP-COMP may both be omitted, not only may seemingly simple sentences such as (34a) be interpreted quotatively, as in (34b), but also they may be related to fuller sentences with either [se] or [fa] as complementizers and thus be analyzed as involving indirect reported speech subordinated with a null complementizer, as in (34c):27 34a. [A tel Am stap] 'I told him/her to stop' b. [s A tel Ami [S-IMPER pro stap]] c. [s A tel Ami [s'[C oM 0] [s proi stap]]

This hypothesis is supported by the synonymy of sentences such as the following, the equivocal analysis of (35b) notwithstanding: 35a. [tel yObrAdoget yo suz fa yu] (PR)'Tellyour brotherto get your shoes for you.'28 b. [tel yo brAd^fa get yo guz fa yu] (elicited). Sentence (35a) itself is an interesting serial construction, which must now be analyzed analogically to (34c) only because of the deictic reorientation encoded in the pronouns [yo] and [yu]. It differs from sen-

318

AMERICAN SPEECH 64.4 (1989)

tence (33b-c) in that the subject of the second verb is here the object of the first verb. Coming back to sentences (33) and the embedded clause-initial use of [se], the serial analysis is, like that of (35a), itself supported by the independent existence of other serial constructions in the language, for example, sentence (29). However, a close examination of sentences (32) and (33) reveals that neither the complementizer nor the serial analysis of [se] is general or powerful enough to account for all its uses as the first overt constituent of the following clause. The complementizer analysis is the only adequate one for sentences (32), the serial analysis is the only adequate one for sentences (33b-c), and sentence (33a) may be analyzed either way. The characterization of the grammatical status of [se] in such complex sentences must thus be disjunctive, including both the serial verb and the complementizer functions. 2. THE BASICFUNCTION OF [se]. A question of interest following

from

the above discussion is, which of the two uses of [se] is the more basic one? It may be helpful to start with a cross-linguistic observation. A complementizer is not always required for subordinating a clause to a higher verb as its complement. There are languages such as Chinese which provide no evidence of having a complementizer (see, e.g., Li and Thompson 1981, Li 1984). Re-examining the data of Lakoff (1968) and contrary to her analysis, it also seems that Classical Latin (unlike Vulgar Latin) had a very restricted usage of anything similar to a complementizer. Most subordinate clauses are introduced by a null complementizer, which is sometimes preceded by the purposive conjunction ut(i) 'in order' or ne 'in order not', which require the subjunctive (where English uses to plus infinitive or that plus the subjunctive). Otherwise eitherfor the subordinate verb is in the infinitive, as illustrated by the following sentences from Lakoff:29 36a. Vol+o Marc+um i+re 'I want Marcus to go' (want+ lsg Marcus+ Acc go + Infin) b. Dic+ o me esse consul+ em 'I say (that)I am a consul'(say+ lsg I + Acc be + Infin consul+ Acc) c. Imper+ o Marc+o ut e + at 'I order Marcusto go' (order+ Isg Marcus+ Dat in-order go + Subjunctive) d. Revort+ i ut(i) me purgar+ em 'I returned{in order to/so that I may} clear myself' return+ Isg in-order I + Acc clear+ Subjunctive+ Isg Closer to home, English omits the complementizer optionally in some cases, as in (37a), or obligatorily, as in (37b):

EQUIVOCALSTRUCTURES IN GULLAH

319

37a. John said (that) he was sorry. b. John said, "I am sorry." One of the outcomes of the restructuring which has resulted in Gullah (and many of the pidgins and creoles commonly discussed in the literature) may be the loss (partial or complete) of the complementizer. Its function is then assumed by a different syntactic strategy, for example, verb serialization. Cross-linguistically, one is more likely to come across a language without a complementizer than across one without a verb meaning 'say'. It may thus be safe to assume that in Gullah the verbal usage of [se] is more basic than its complementizer usage, which may have developed by analogy, perhaps after speakers realized its equivocal status in sentences such as (33a). It is significant that it may be omitted in sentences (32). However, this diachronic position does not resolve the question of the synchronic analysis of the clause-initial [se]. As far as sentence (33a) is concerned, it will undoubtedly be necessary to consider other facts of the language before any conclusive position is taken. For instance, note that the distribution of [se] as a complementizer is rather restricted compared to that of [fa] or the null complementizer, which both occur also in relative clauses (see Mufwene 1986). This situation militates against considering the complementizer function of [se] as the more basic; it may constitute evidence for treating such use as a reanalysis of its apparently serial verb usage after verba dicendi. From this reanalysis it would have been extended to verbs of perception and the like. In other words, if [se] were basically a complementizer, it would very likely be attested in some nonpurposive relative clauses and in cleft constructions, more or less like the rather general, nonpurposive complementizers that in English, que in French, and the like in other languages. Based on Mufwene (1986 and 1987b), the general, nonpurposive complementizer in Gullah seems to be null, which, as indicated above, actually alternates with [se] qua complementizer and occurs also in relative clauses as well as in cleft constructions.30 The serial verb usage of [se] must thus be the more basic or unmarked function from which the complementizer usage has been extended, although the possibility remains that different speakers may use it differently or that the community has always used it alternately as a complementizer or as a serial verb. As more and more different aspects of Gullah 3. PARTIALCONCLUSION.

(and other AC's) are being investigated, we should be able to take a better-justified position soon on the above issue, based on a more com-

320

AMERICAN SPEECH 64.4 (1989)

prehensive analysis of the relevant facts. The primary concern of this paper has been to demonstrate that the structural analyses of some complex sentences in Gullah are equivocal. This grammatical state of affairs, which should not be ignored, has been shown to be the case in more than one type of sentence. The awareness of this should help us better understand not only the ways creole grammar develops complex sentential structures but also the significance of variation in at least the creole speech community. 3. CONCLUSIONS As this paper comes now to its conclusion, it is opportune to consider briefly a related question which also deserves attention here: is it a requirement of the Saussurean conventionality of language that different speakers must assign identical or similar syntactic structures to sentences which are assigned the same or similar (linguistic) meanings? Is the concern of this paper not an artifact of an analytical framework that may assume more of the sharing of rules and structures than is warranted by facts? It seems that, like computer systems, which can yield the same or similar final outputs using different algorithms, it should be plausible for native speakers of a language to have a perhaps very limited number of alternative syntactic analyses which assign more or less the same meanings to the same sentences. Such an assumption seems consistent with the findings of current research on language variation. It should in fact also account for cases where the different structural analyses assigned by different speakers to the same sentences yield different meanings. A case in point here may be of the English sentence Tom promisedSally to return beforenightfall, where, according to Comrie (1986), the subject of the embedded clause may be understood to be Sally for a minority of English speakers. Although the suggestion of equivocal structural analysis (distinguished from the thesis of ambiguous structure) seems to be an easy way out of a dilemma, it should help to keep the option open for the type of sentences discussed in this paper. In a language where the membership of words in a particular lexical category is determined on syntactic rather than inflectional criteria, the danger of deciding by fiat whether a particular item is a verb, a preposition, or a complementizer can hardly be overstated, especially where there appears to be some internal inconsistency, as in the case of [se] treated either as a verb only or as a comple-

EQUIVOCALSTRUCTURES IN GULLAH

321

mentizer only. Further research will undoubtedly tell us more about this aspect of the grammar of Gullah and other Atlantic creoles. Note, however, that the kind of structural variation assumed here actually makes language behavior quite consonant with many other aspects of social behavior where variation is assumed to be natural.

NOTES

The research on which this article is based was supported by the National Science Foundation,grant #BNS-8519315. Earlierversions of this paper were presentedat the 1987 Symposiumon InflectionalMorphologyand Syntax,Universityof North Carolina,Chapel Hill, and at the Spring 1987 meeting of the SoutheasternConferenceon Linguistics(at GeorgetownUniversity).I am grateful to both NancyCondonand to my anonymousAS referees for their comments and questions, which have helped me clarify my statementsand positions in a number of places. All the remainingimperfectionsare my responsibility. 1. The Gullahdata in this paper are identified in the followingways:"common"for a sentence patternwhichis commonlyheard and which the author has reproduced without looking for a specific citation;"XY"for the initialsof the native speaker from whose spontaneous speech the constructionis cited; and "elicited"for a sentenceobtainedthrough elicitationfrom some nativespeakers. All the sentencesdiscussedhere have been double-checkedwith nativespeakers. 2. As far as studies of creoles lexified by Indo-Europeanlanguages (henceforth IE creoles)are concerned,Bickerton(1981, 1984)and, for Atlanticcreoles, Mufweneand Dijkhoff(1986) are the exception to the rule. 3. It is assumedhere, as in some versionsof transformationalgrammar(e.g., McCawley1988), that, in sentences such as the following, the deep-structure to;for is putativelydeleted complementizeris the same correlativepairfor in the surface structure of subordinateclauses where the subject is null (viz., PRO in

(ii)):

i. I would likefor Maryto stay home ii. Ii would like PROjto stay home 4. DiscussingGuyaneseCreole (105-6), Bickertonclaimsthat se may not be omitted and that the following pair of sentences mean different things: i. mi hia se i a kom 'I heard (that)he was coming' ii. mi hia i a kom 'I heard him coming' If his claimis true of GuyaneseCreole,it does not applyto Gullah.The following sentence without [se] is simply ambiguous: iii. [A^ hie rabatdo kAm]'I heard Robert (was)coming' The only reasonwhy the followingcounterpartof sentence (ii) is not ambiguous is that in Gullah the pronominalform [(h)i] is restrictedto the subjectiveand possessivefunctionsand its allomorph [Am]to the objectivefunction:

322

AMERICAN SPEECH 64.4 (1989) iv. [A hie hi do kAm] 'I heard that he was coming'

Wherever the pronoun does not vary, the construction is simply ambiguous. 5. As noted in Mufwene (1987a), the basilect is just a useful construct consisting, in the case of English creoles, of a disjunctive set of all the most unEnglish features attested only partially in various speakers. That [tu] is attested in the speech of some speakers (regardless of whether or not they are nearbasilectal) does not necessarily invalidate the assumption (made by Bickerton 1973 first for Guyanese Creole) that English creole basilects do not have an infinitive. After all, there may have never been any speakers of this idealized lect, which has been postulated essentially as a working tool for dealing with grammatical variability. As argued in Mufwene (1987c), grammatical variability must have been typical of pidgins and creoles since their inception. The fact that future-oriented clauses begin with [fa], with [tu], or with a null marker need not be attributed to decreolization. The prefix post in DeCamp's (1971) post-creole speechcontinuumwas unwarranted. 6. English, with which many readers of this article may be most familiar, is by no means as typical a case as Latin or German are with regard to the form of the infinitive. Many of its verbal inflections have been lost. Given an incomplete paradigm in which its inflection is now -0, its users rely now on syntactic context, ) to and subordination to another verb, including the complementizer (for to determine which 0-inflected verb form is infinitival. An argument against assuming that to is the infinitival marker is based on the following examples, where the verb form helps determine the status of to itself: i. John likes to fly + 0. ii. John is accustomed to fly + -ing. Invoking to in order to determine whether the verb is infinitival is circular, especially since verbs are sometimes used in the infinitive without to. 7. The subject of verbs in the imperative is assumed to be a null pronoun, discussed below in the main text. 8. See Drachman (1975) for a more comprehensive discussion of such sentences. 9. My choice of pro here and in the rest of this paper for the null subject in Gullah is arbitrary, perhaps only siding with the majority. The adoption of either PROor pro raises for the analytical framework a question of metalinguistic inconsistency, viz., why use the symbol where it may not be used in a language with a finite/nonfinite distinction? This is, however, peripheral to the substance of this essay. 10. Sentences (6) and (lb), produced by the same speaker in the same discourse, are, as tested with other speakers, equally well formed. Similar variation has been perceived in the spontaneous speech of other native speakers. 11. Clauses are often subordinated without a complementizer to [w5] 'want' or [traY]'try' as main verbs. Native speakers find sentence (7b) quite well-formed. The following pattern with a null subject is even more common: [A w5 0 go] 'I want to go'.

EQUIVOCAL STRUCTURES IN GULLAH

323

12. It is significant that in English the preposition for is also ambiguous between a purposive and a causal meaning, as evidenced by the following sentences: i. She did it for approval. PURPOSE iia. She did it for love. CAUSE or PURPOSE iib. Thank you for {helping/your help}. CAUSE Due to the absence of verbal inflections in Gullah, the morphosyntactic reflexes of the CAUSE/PURPOSE distinction which English conveys by inflecting the verb in either the gerund or the infinitive have been obliterated. The distinction is thus purely semantic, just as with the nominal complements in English. 13. In English, past tense is expressed differently in finite and nonfinite clauses; compare I brokethe rule; I apologizefor having brokenthe rule; and I assume the rule to have been broken. 14. See also Mufwene (1983, n. 4) for Jamaican Creole, Byrne (1984) for Saramaccan. 15. Papiamentu may be one of the rare AC's where the counterpart of [fa], pa, does not have a modal-verb usage. This preposition is not attested in postsubject position in subordinate clauses. 16. Winford (1985) indeed proposes this modal analysis of the Jamaican Creolefi for a similar construction with the Subject-Equi/Control verb [traY]'try'; I find no reason for rejecting it. Even Byrne (1986a), who disputes some of Winford's assumptions, agrees with this analysis. 17. The facts discussed in this paper are consistent both with the traditional analysis of some null subjects in terms of an Equi Deletion rule and with GB theory's analysis in terms of PRO and control. I consider both models equivalent in this respect (differing essentially in the mechanisms they invoke to account for coreference), and I take no position for either approach here. 18. Here and in the remainder of the article, initial and final brackets indicate phonetic notation; internal, labeled brackets indicate syntactic analysis. 19. ti in (20b-c) indicates the position from which [Am] has been raised. No position need be taken here between the Raising and non-Raising analyses. 20. This assumption is consistent with Byrne's (1985) hypothesis that the subject of each subsequent verb is a proprox pronoun, which is coreferential with an NP argument of the preceding verb. 21. Even though the English translation in (23b) would not be an appropriate answer to the English translation of the question, native speakers of Gullah with whom these constructions have been checked find both (23a) and (23b) acceptable. According to them, there is virtually no semantic difference between the two sentences. I assume that whatever difference I may have failed to elicit regards modality. 22. Gullah must thus apply here the same principle followed by English in not using a complementizer to introduce a clause which is the object of a preposition. There are languages such as French which require a complementizer in such cases, for example Apresque je sois parti,Jean est arrivg 'After 0 I left, John came'. 23. Some may prefer to sustain a distinction between serialization and consecutivization. According to such a view, a serial construction represents one

324

AMERICAN SPEECH 64.4 (1989)

single event whereas a consecutivized construction represents separate events. There are a number of reasons for rejecting the distinction. First, it has no syntactic correlate. Secondly, it is based on no independent principle of what is a primitive event, hence of when two sequenced events count as a single complex event or more than one. Thirdly, this semantic distinction seems irrelevant for a phenomenon which, even in being syntactic, deals strictly with how verbs or predicates follow one another in their linear, but not structural, arrangements. The syntactic functions of subsequent verbs to preceding ones vary a lot. (See, e.g., Tai 1985 for Chinese, Byrne 1987 and Sebba 1987 for creoles.) In a different vein, due both to the fast speech tempo of Gullah and to the fact that main clauses may be subjectless, it is not always easy to tell a paratactic from a serial construction, hence the equivocation between the two kinds here. Future research might shed light on this. 24. This representation of serial verb constructions as dominated by one major VP node is rather arbitrary, just as were Bickerton's (1984) and Byrne's (1985) decisions to represent the serial as a sentential complement derived from a higher VP node which also dominates the first verb (and its object, if there is one). The subject matter calls for further investigation, particularly since the relationship of each following verb to the preceding one is variable. Studies in languages such as Chinese (see, e.g., Tai 1985) indicate that some serial constructions involve a complement-type relationship, agreeing neither with the Bickerton-Byrne analysis nor with that proposed in (27a); some others involve just the kind of representation given here (see also Lightfoot 1979); and still others involve an adverbial type of relation, corresponding in my view to the Bickerton-Byrne analysis and to more or less that of what Hornstein and Lightfoot (1987) term "small clauses," e.g., John swam naked. 25. In order for the modal analysis to be plausible it must be assumed, as in note 19, that as a preposition [fao] takes a sentential complement introduced by no complementizer. 26. The following analysis inspired by Byrne (1985) may be assigned to sentence (33a): [A, [vP [vP tel Am] [s proproxi[vP se [, rabot teYk sik]]]]]. I see, however,

nothing so far which precludes an analysis parallel to (27a), as in the following: [A [vP [vP tel Am] [vPse [s rabat teYk sik]]]]. This question of which structural analysis represents a serial construction most adequately may be resolved elsewhere. What matters here is whether [se] must be interpreted as a complementizer or as a serial verb. 27. Unlike in (34c), coreference in (34b) is established only pragmatically, not syntactically; pro in the embedded imperative of (34b) stands for second person singular null subject. 28. The possibility of interpreting the embedded clause in this and other exercitive sentences as a direct quotation is weakened by the deixis reorientation in the form [yo], which the second speaker (the addressee of the main clause), whose words are presumed quoted, would not use in reference to himself. This and the facts of serialization may reflect the general development from parataxis (with null subjects) toward a complex sentential structure with embedded sentences. But this development is at this point only a conjecture which still must be demonstrated in some future work. 29. It could be argued, as in Lakoff (1968), that the infinitive marker is the

EQUIVOCAL STRUCTURES IN GULLAH

325

abstract complementizer, which has been suffixed to the verb by a transformation. While this abstract analysis has had a certain appeal (note the treatment of the English gerundive -ing in transformational grammar), it could also be counterargued that in the case of Latin the infinitive is governed by the presence of the higher verb (which calls for this particular construction), just as the subjunctive is governed by the presence of ut(i)lne (or the higher verb). Thus, the infinitival marker need not be a complementizer. 30. This is an interesting reversal of the situation in its lexifier, English, where the general nonpurposive complementizer is that, which also occurs in relative clauses.

REFERENCES

Alleyne, Mervyn C. 1980. ComparativeAfro-American:An Historical-Comparative Study of English-BasedAfro-AmericanDialects of the New World. Ann Arbor: Karoma. Aoun, Joseph. 1985. A Grammarof Anaphora. Cambridge, MA: MIT P. Bailey, Beryl Loftman. 1966. Jamaican CreoleSyntax:A TransformationalApproach. Cambridge: Cambridge UP. Battistella, Edwin. 1985. "On the Distribution of pro in Chinese." Natural Language and Linguistic Theory3: 317-40. Bickerton, Derek. 1973. "On the Nature of a Creole Continuum." Language 49: 640-69. . 1981. Roots of Language. Ann Arbor: Karoma. . 1984. "The Language Bioprogram Hypothesis." TheBehavioraland Brain Sciences 7: 173-221. Byrne, Francis X. 1984. "Fi and fu: Origins and Functions in Some Caribbean English-Based Creoles." Lingua 62: 97-120. 1985. "Proproxin Saramaccan." Linguistic Inquiry 16: 313-20. 1986a. Note. Journal of Pidgin and CreoleLanguages 1: 147-52. . 1987. GrammaticalRelations in a Radical Creole.Amsterdam: Benjamins. Chomsky, Noam. 1986. Knowledgeof Language: Its Nature, Origin, and Use. New York; Westport, CT: Praeger. Comrie, Bernard. 1986. "Markedness, Grammar, People, and the World." Markedness.Ed. Fred R. Eckman, Edith A. Moravcsik, and Jessica R. Wirth. New York: Plenum, 85-106. Cunningham, Irma Aloyce Ewing. 1970. "A Syntactic Analysis of Sea Island Creole ('Gullah')." Diss. U of Michigan. DeCamp, David. 1971. "Toward a Generative Analysis of a Post-Creole Speech Continuum." Pidginizationand Creolizationof Languages. Ed. Dell Hymes. Cambridge: Cambridge UP, 349-70. Drachman, Gaberell. 1975. "The Syntax of Casual Speech." SalzburgerBeitrage zur Linguistik,vol. 1. Salzburg: Verlag Gunter Narr, 273-95. Emonds, Joseph E. 1976. A TransformationalApproachto English Syntax: Root, Structure-Preserving,and Local Transformations.New York: Academic.

326

AMERICAN SPEECH 64.4 (1989)

Hornstein, Norbert, and David Lightfoot. 1987. "Predication and PRO."Language 63: 23-52. Huang, C.-T. James. 1987. "Remarks on Empty Categories in Chinese." Linguistic Inquiry 18: 321-37. Jones-Jackson, Patricia. 1986. "On the Status of Gullah on the Sea Islands." In Montgomery and Bailey, 63-72. Lakoff, Robin. 1968. AbstractSyntaxand Latin Complementation.Cambridge, MA: MIT P. Li, Baizheng. 1984. "A Contrastive Analysis of Complementizer in English and Chinese." MA thesis, Tianjin Normal U. Li, Charles N., and Sandra A. Thompson. 1981. Mandarin Chinese:A Functional ReferenceGrammar.Berkeley: U of California P. Lightfoot, David W. 1979. Principlesof DiachronicSyntax.Cambridge: Cambridge UP. McCawley, James D. 1988. The SyntacticPhenomena of English. Chicago: U of Chicago P. Montgomery, Michael B., and Guy Bailey, eds. 1986. Language Varietyin theSouth: Perspectivesin Black and White. University: U of Alabama P. Mufwene, Salikoko S. 1983. "Observations on Time Reference in Jamaican and Guyanese Creoles." English World-Wide4: 199-223. . 1986. "Restrictive Relativization in Gullah." Journal of Pidgin and Creole Languages 1: 1-31. . 1987a. Review of Michael Montgomery and Guy Bailey (1986). Journal of Pidgin and CreoleLanguages 2: 93-110. . 1987b. "An Issue on Predicate-Clefting: Evidence from Atlantic Creoles and African languages." VariaCreolica.Ed. Philippe Maurer and Thomas Stolz. Bochum: Brockmeyer, 71-89. . 1987c. "Pidginization and Creolization: An Evolutionary Biology Analogue." Lecture given at Northwestern U, Evanston, IL. Mufwene, Salikoko S., and Marta B. Dijkhoff. 1986. "Notes on the So-Called 'Infinitive' in Creoles." Sixth Biennial Conference of the Society for Caribbean Linguistics, St. Augustine, Trinidad. Publ. Lingua 77 (1989): 297-330. Nichols, Patricia C. 1975. "Complementizers in Creoles." WorkingPapers on Language Universals 19: 131-35. . 1976. "Linguistic Change in Gullah: Sex, Age, and Mobility." Diss. Stanford U. Sebba, Mark. 1987. The Syntax of Serial Verbs.Amsterdam: Benjamins. Tai, James H.-Y. 1985. "Temporal Sequence and Chinese Word Order." Iconicity in Syntax. Ed. John Haiman. Amsterdam: Benjamins, 49-72. Winford, Donald. 1985. "The Syntax of fi Complements in Caribbean English Creole." Language 61: 588-624. Wolfram, Walt, and Donna Christian. 1976. AppalachianSpeech.Arlington, VA: Center for Applied Linguistics. Xu, Liejiong. 1987. "Free Empty Category." LinguisticInquiry 17: 75-93.