Why Do Creoles Have Serial Verb Constructions?

Eric Schiller



Abstract: This paper is an updated version of Schiller (1993). The major changes involve adapting the terms and analysis presented in the earlier paper so that they conform to the ALGAE framework of autolexical grammar.  The data and conclusions remain unaltered.

Revision History:        Version 1.0 posted November, 1999

Technical notes:         This paper uses Lucida Sans Unicode to represent IPA. Systems lacking that font may not display the IPA properly. As superior technological solutions are found, the paper will be updated. For the present paper, such problems should not interfere with comprehension of the analysis.

Citation form: Schiller, Eric. 1999. Why do creoles have serial verb constructions. Posted on the Internet by Linguistics Unlimited (www.chessworks.com/ling). Since this paper is subject to modification, it is best to include the version number.


The origin of serial verb constructions in Creole languages has been the topic of considerable discussion, with the state of the art summarized in 1988 articles by Bickerton and Muysken in Newmeyer’s Cambridge Survey There are two basic positions reflected in these articles. The first is that serial verbs are found in these languages because the substrate languages contribute them to the inventory of syntactic devices available to the early stages of the Creole. The second, known as the Bioprogram hypothesis, takes the existence of serial constructions to be determined by characteristics of Universal Grammar, in the Chomskyan sense. Muysken, while not by any means convinced that the Substrate hypothesis is tenable, points out a number of flaws in the Bioprogram hypothesis, some of them answered by Bickerton’s rebuttal. My own position leans toward Bickerton’s, but differs in some important respects.

The Semantic Case Instantiation Principle and its predictions

In other work, and in particular my dissertation (Schiller, 1991), I have argued for the Semantic Case Instantiation Principle (1) and a Generalized Interface Principle (2) (Sadock and Schiller, 1993):


1)      Semantic Case Instantiation Principle (SCIP): Semantic Case relations are instantiated by the most concrete possible mechanism.


2)      Generalized Interface Principle (GIP): Paradigmatic and syntagmatic features at all levels should correspond as closely as possible.


By “concrete,” I refer to the accessibility of information in a language. The more concrete elements of the language will be reflected in the consistency of application by different speakers of the language. Thus although lexical semantics may seem concrete to an individual speaker, there will be less agreement among speakers about the meaning of a word than about the function of an inflectional ending.

Given the above, it will be predicted that semantic relations such as Instru­ment, Goal, Source and Location will be instantiated morphologically, if possible. If a language does not have the means for morphological instantiation or a specific Case relation, syntactic means involving the use of lexical items, generally verbs or prepositions[1] will be employed, usually via adpositional phrases. Failing that, a language may resort to Subordinating Serial Verb Constructions.[2]  Some languages, e.g. Kalam (Pawley 1980) do not even have that mechanism available. As Bickerton (in Bickerton & Muysken 1988:303) notes:

It is no accident, for instance, that the creoles that have few or no serial constructions are those that inherited the largest amount of superstrate morphology, while the creoles with the most serial constructions are precisely those that inherited least morphology from their superstrates.


I agree that this is no accident; it follows from the Semantic Case Instantiation Principle.

It is important to state that SCIP does not mean that a language will never have a mixture of devices. Sranan, for example, has both prepositional and serial instantiation of the Instrumental relation. This may be a result of the mixture of two dialects of Sranan, as Bickerton has suggested. In Haitian Creole, the bay serial construction has a prepositional counterpart with pu. These are not, how­ever, interchangeable constructions. In certain syntactic contexts the prepositional form is the only one acceptable, and native speakers confirm that pu is more of a pure marker of Benefactive, while bay retains some of the semantic content of give, the meaning of the homophonous verb and has perhaps some notion of direc­tionality encoded.[3]

The principles do not predict that serialization will never be found in lan­guages which make considerable use of prepositions. That is, the existence of a syntactic category of PP does not block serialization as a direct consequence of the grammar. From an Autolexical standpoint, when the lexicon contains lexical items which mark grammatical relations, then these lexical items are a more con­crete method of case marking than the indirect serial construction, and thus are to be preferred.

Verbs and Prepositions

The theoretical position articulated above receives support in the following examples. On rare occasions serialization will arise even in a prepositionally rich environment. Consider the case of Hawaiian Creole English. This fascinating lan­guage, which arose from the contact of English, Hawaiian, Portuguese, Korean, Japanese, Chinese, Samoan and several Philippine languages shows only a limited amount of verb phrase serialization, using English based prepositions for most case marking tasks. These include directional to, locative and temporal in, locative on, directional from and for, distance marker till. But there is (or perhaps, was) no manner preposition by or on (as in on foot in standard English). So we see instead a serial construction (4), from Bickerton (1977).

4)      dei don hæv no kaz, dei wawk fit go skul

         they do-not have no cars they walk feet go school

         They didn’t have any cars, they went to school on foot.


It is not surprising that the “missing” prepositions in Hawaiian Creole are those which are rarely used in standard English. The lexicon of Hawaiian Creole English does not contain a preposition which is used to express a manner of traveling on foot. Bickerton (1977:133) remarks that “It seems likely that if H[awaiian] C[reole] had not been subject to corrective pressure from its super­strate FROM THE START OF ITS EXISTENCE a factor that differentiates it from all other creoles that I am aware of it would have developed such struc­tures into regular parts of the grammar.”

It is reasonable to see this as a case of the existence of prepositions blocking serialization, except such cases where appropriate prepositions are not available in the lexicon. This has nothing to do with the existence of a syntactic category of PP, which is quite robust in HCE, but rather depends on the idiosyncratic lexical fact that HCE did not share with Standard English the Instrumental use of on which is generally used to mark a certain type of location. As a result, it resorted to a serial construction, which was formed with the verb walk.[4]

Another example involves Thai, a language where the category of PP is not so prolific. In fact there is almost no evidence at all of prepositional constructions, with the single exception to be discussed below. Unlike its typologically similar neighbor, Khmer, which makes great use of prepositional devices, Thai encodes its grammatical relations with serial verb constructions. Thai and Khmer are ei­ther unrelated or very distantly related languages which have had over 1500 years of close contact. Thai is not a creole, though like most SVO Southeast Asian lan­guages it shares a lot of syntactic and lexical characteristics[5] with creoles.

The following data will serve to illustrate serial vs. prepositional forms in Thai. In (5a) we have a Thai sentence which seems to parallel the Khmer prepositional form, given in (5b).



















‘He cut the meat with a knife.’




















‘He cut the meat with a knife.




















‘Prichaa cut the meat with a knife.’




















 ‘He cuts the meat with a knife.’



The Thai sentence in (Sa) is highly marked, as noted by Foley & Olsen (1985). At the 1989 Southeast Asian Summer Studies Institute I asked ten native speakers to provide translations for the English sentence He cut the meat with a knife. The prepositional form in the English sentence was used to encourage the production of (5a). Eight of the ten respondents failed to list (5a) at all, giving only the serial versions (5c,d) And the two who did list it also provided (5c) as an alternative, and informed me that the form (Sa) was used only to emphasize strongly that it was a knife, and not, say, a cleaver, that was used to cut the meat. One should also note that the lexical item used as a preposition in (5a) is not a native word and displays an unusual vowel. It is almost certainly a borrowing (Huffman 1986) from the Khmer word /daoy/ which is employed both as a verb meaning ‘use’ and as a preposition meaning ‘by’, though one must keep in mind that these are translations which may give an impression of greater semantic dif­ference than is perceived by native speakers.[6]

The point here is that the borrowing in a contact situation permits Thai to acquire the preposition, though it remains highly marked. If the category of prepo­sition were more robust, we might expect the sort of grammaticization we find in coverb phenomena elsewhere in Southeast Asia, and especially in Mandarin Chi­nese.[7] This grammaticization of coverbs follows the Semantic Case Instantiation Principle, and can be seen in various creoles as well. I have suggested that this process may be described as in the diagram in (6):

6)    Grammaticization of coverbs

Graphic representation of the analysis

The graph shows three independent developments—one involving the realm of syntactic category, starting from a simple verb, developing serial verb construc­tions, and winding up with a preposition via reanalysis. The meaning of the word develops various extensions via processes involving metaphor.[8] The only constant is the logico-semantic factor that the lexeme is a lexical predicate. The morpholo­gical properties of the lexeme may also change , but the change need not parallel the syntactic metamorphosis.[9]

Because this process takes place over time, one expects to find intermediate stages. And I think that much of the argument over the status of prepositions and coverbs in both Southeast Asian languages and Creoles is due to adopting a view whereby the changes take place in parallel. That is, the semantic richness of a lexeme is taken as a diagnostic for syntactic category. The more “verb-like” the semantics, the more likely the item is to be analyzed as a verb. If, however, the semantics are bleached, then the word is more likely to be assigned to the cate­gory of preposition.

Better are distributional tests, but they are effective only at the extremes of the categorial continuum. From a semantic viewpoint, we expect to find a wide range of the degree of semantic bleaching, but, contra Riddle (1989) I do not see this as particularly surprising, or important.[10]

Creoles Revisited

If every language must instantiate certain semantic relations, then a creole is likewise going to have to do so. Initially, the creole will look to earlier, pidgin, stages of the language and select devices to mark grammatical relations from that source, to the extent that the available data are not too impoverished. As has often been noted, most pidgins are stripped of morphological markers and of function words as well. Note here that it is the pidgin that is in question, not the substrate or superstrate. If there is no appropriate lexical item or morphological marker in the pidgin, then some method of instantiation of grammatical relations must be found. In a process of creolization, two or more languages may be avail­able as sources for borrowing. Given that certain lexemes have a meaning which is compatible with a particular semantic relation (e.g. give, BEN; go, DIR), it may be more likely to be adopted as the marker of that relation. Whether it will be a preposition or a verb depends on the identification of the syntactic category of the word in the grammar of the Creole language.

So in the Hawaiian case, neither the pidgin nor the creole adopted the rare on preposition from English, but having chosen to use walk to express this relation, it did so in conformity with the perceived syntactic category of walk, to wit, a verb.

The Creole then has a stage in which serendipity plays a role in the acquisition and development of some markers, but in general SCIF seems to apply. A stable pidgin which contains a supply of prepositional devices is predicted to give rise to a creole which uses prepositions rather than serial verbs.

An important question remains. Why don’t all languages have serialization? This enters an area which is far too broad to discuss in this brief paper, and the interested reader is referred to Schiller (1990) for discussion of the typology of serial verb constructions and the connection with fundamental word order. Serial­ized verb phrases exist in languages which allow a particular structural configura­tion in the syntax, that is, concatenated verb phrases, to be associated with multi­ple predicates in the semantics. This is, of course, an automodular analysis. Generalized Phrase Structure Grammar (GPSG) (Sebba 1987) and Government and Binding (GB) (Baker 1989) require different analyses because of their imple­mentation of the rule-to-rule hypothesis and Projection Principle respectively, while Categorial Grammar (Welker 1990) involves lexical derivation rules. Not all languages allow concatenated VP’s, and this is especially true of SOV languages, as well as VSO and OSV languages.

This presents significant problems for the Universal Grammar based biopro­gram. Why not serial verb phrases (note: not serial verbs) in SOV languages? Baker (1989) offered some suggestions in a GB framework, but I have already noted empirical as well as conceptual problems with it in previous papers and in my dissertation.


I believe that the position that serial verb phrases are used in the absence of appropriate prepositions is basically correct, with the counterexamples proposed to date providing insufficient motivation to alter this view, as discussed in Schiller, 1991 From a Creole perspective, one must remember that a Creole usually grows in the presence of superstrate and substrate languages, and that these languages can have an effect even after pidginization, as Bickerton himself noted in the quote above. Perhaps this is a fuzzy situation, but it not clear to me why we should expect Creoles to be any cleaner than a fuzzy language like English.

Postscript (1999)

It is always a pleasure to revisit an old paper and see that little modification is needed, even after a considerable period of reflection. Some rather ad-hoc devices in the previous version have been eliminated because the Generalized Interface Principle can account for the same facts in a more straightforward fashion. I now think that aspect marking, which is not a serial verb phrase phenomenon but rather an auxiliary verb mechanism, can be explained in similar fashion, and this is the subject of ongoing work. The view here is a simple one, and while that may have the drawback of making the linguist seem a bit less brilliant and more of a common sense sort of fellow, but I believe (and that’s probably the best term) that language should be simple. To reprise the slogan of autolexical grammar: Once in a while you get shown the light in the strangest of places if you look at it right. (R. Hunter). Some of the most complex phenomena in language can be reduced to almost trivial observation, if an appropriate analysis is applied.


Baker, Mark. 1989. Object sharing and projection in serial verb constructions. Linguistic Inquiry 20 (4):515-54

Bickerton, Derek. 1977. Change and variation in Hawaiian English (Volume II). University of Hawaii.

Bickerton, Derek & Pieter Muysken. 1988. A dialog concerning the linguistic status of creole lan­guages. In Linguistics: The Cambridge survey. Vol. II, Linguistic theory: Extensions and implica­tions. Frederick J. Newmeyer, ed., 302-6. Cambridge: Cambridge University Press.

Dune, Mark. 1988. Verb serialization and Verbal-Prepositions” in Oceanic languages. Oceanic Lin­guistics 27(1):1-23.

Foley, WA. & M. Olson. 1985. Clausehood and verb serialization. In Grammar inside and outside the clause. J. Nichols & A. Woodbury, eds., 1-66. Cambridge: Cambridge University Press.

Huffman, Franklin. 1986. Khmer loanwords in Thai. In Papers from a conference on Thai studies in honor of William J. Gedney. R. J. Bickner, ed., 199-210. Ann Arbor: University of Michigan Press.

Muysken, Pieter. 1988. Are creoles a special type of language? In Linguistics: The Cambridge survey. Vol. II. Linguistic theory: Extensions and implications. Frederick J. Newmeyer, ed., 285-301. Cam­bridge: Cambridge University Press.

Pawley, Andrew. 1980. On meeting a language that defies Description. Linguistic Society of Papua New Guinea, Kivung Congress.

Riddle, Elizabeth M. 1989. Serial verbs and propositions in White Hmong. L.T.B.A. 12(2):1-13. Schiller, Eric. 1989. Syntactic polysemy and underspecification in the Lexicon. ELS 15:278-90. Schiller, Eric. 1990. The typology of serial verb constructions. Chicago Linguistic Society 26(1)

Schiller, Eric. 1991. An autolexical account of subordinating serial verb constructions. Ph.D. diss., Uni­versity of Chicago.

Schiller, Eric. 1992. Some autolexical solutions to problems in the description of Southeast Asian languages. In M. Ratliff et al. (eds) 1st Meeting of the Southeast Asian Linguistics Society. Tempe: Arizona State University Press.

Sebba, Mark. 1987. The syntax of senal verbs. Amsterdam/Philadelphia: John Benjamins.

Welker, Kate. 1990. Serial verbs in categorial Grammar. In When verbs collide: Papers from the 1990 Ohio State mini-conference on serial verbs. Brian D. Joseph & Arnold M. Zwicky, eds., 355-366. The Ohio State University Working Papers in Linguistics 39.

horizontal rule

[1]  Throughout the paper the term preposition should be taken to include postpositions.

[2]  For the distinction between subordinating and coordinating serial verb constructions see Sebba (1987).

[3]  There are other cases of prepositions in serializing languages, even when creolization has not taken place. One example is found in Thai, but the preposition involved is borrowed and highly marked. See Schiller (1990, 1991) for a discussion of this and other putative counterexamples. Byrne (personal communication) notes that judgments from Saramaccan are similar to those of Haitian, where the serial forms are felt to be more natural to Saramaccan.

[4]  The semantics of this equivalent of on foot are not so simple. The most common serial instantiation of instrumentality employs a word like take, but ‘He take feet go school’ is not, as far as I have been able to research, a common serial construction. Instrumental prepositions are not uncommon in this context, however, e.g., Khmer nɯŋ? cəəŋ ‘with foot’.

[5]  Thai, Lao, Khmer and many other Southeast Asian languages have undergone a great deal of lexical replacement, borrowing heavily from Sanskrit and its derivatives, especially Pali. Perhaps as much as 50% of Khmer consists of non-Mon-Khmer vocabulary.

[6]  See Schiller (1989, to appear) for a more detailed discussion of the question of polysemy and categories.

[7] Coverbs are lexical items which are derived historically from verbs but which function as prepositions synchronically, as in the Khmer example daoy mentioned above. Most of the examples discussed in the literature are derived from serial verbs (see Schiller 1991: Chapter 6).

[8]  In order for a lexical item to broaden its semantic scope, some mechanism must be involved. The model of metaphor and metonomy used by Lakoff (1987) is capable of explaining the semantic extensions in a principled manner.

[9] This model helps to explain the presence of verbal prepositions, which act like normal preposi­tions but retain some of the characteristics of verbs, e.g., taking inflectional agreement markers. They are primarily found in Oceanic languages. See Dune (1988).

[10]  Given the analysis presented above, one expects to find a wide range of possible relationships between the degree of semantic richness and syntactic category membership. If a syntactic preposi­tion is derived from a verb, there is no reason why it cannot retain a great deal of the semantic pro­perties of the verb. Conversely, if a verb participates in a serial structure, the semantics of the verb may weaken over time, to the point where it seems to be little more than a grammatical marker, while at the same time retaining its categorial status as a verb. There will, of course, be tendencies to conform to prototypical members of the category of verb or preposition, but deviations from the prototype are not problematic for a radically autonomous modular theory such as Autolexical Syntax.