1 Introduction: Code-switching and its explanation
For many linguists, the goal of linguistics is to describe the mental representation of linguistic knowledge in the minds of individual speakers. While there is nothing wrong with that at all, I wish to draw attention to what I see as two problems with the way linguistic theorizing is generally conducted, and the specific shape they take in the code-switching literature. First, the concept of āknowledgeā is often understood in a needlessly limited way, excluding knowledge about language use and practice, about the social life of language, and about its functions. I will argue that a broader concept of knowledge is not just more ecologically valid, but also allows asking questions which now rarely rise to the surface, or only do so in separate fields that donāt communicate much with each other. Second, the concept of ālanguageā with which linguistics operates is relatively unclear. While linguistic knowledge resides in the individual speaker, we normally conceptualize language at higher levels of aggregation: what we generally describe is not really the knowledge of the individual speaker but the knowledge of clusters of speakers (e.g. the speakers of a particular dialect or language). Both problems are tackled by the currently ascendant usage-based approach in linguistics, and I will argue that adopting this approach in code-switching research can give the field a much-needed boost.
These two general problems are manifested in many different ways; here I will focus on three commonly made distinctions that are loosely related to them. Generally speaking, the issues all result from the tendency in modern linguistics to split rather than to lump. General linguistics as well as the study of code-switching (CS) has been characterized by a separation into sub-disciplines, while a lumping approach may at least be equally fruitful, or even, as I will argue, currently desirable. Each of the three distinctions I will focus on in this paper has helped advance knowledge, but at the same time has also held back further improvement. First, there is the separation of synchronic and diachronic issues. This has resulted in the strange dichotomy of ātheoretical linguisticsā versus āhistorical linguisticsā, which suggests that historical linguistics is not theoretical. In the code-switching literature, this separation shows up as the notorious and much debated distinction between code-switching and borrowing. I will argue in section 4 that a model that integrates synchronic and diachronic aspects of contact-induced language change can help resolve this debate. The second separation that has had unintended negative effects on contact linguistics is between lexicon and syntax, which has encouraged the separate study of code-switching and grammatical contact effects, such as interference or transfer. This modular separation has of course been a hallmark of linguistics, and characterizes most folk views of language, but much recent work in linguistics and psycholinguistics suggests that the basis for the distinction may not be that strong, as I will try to show in section 5. Finally, there is the divide between structural and sociolinguistic accounts of language, giving us rival linguistic approaches that focus on āformā and āfunctionā, respectively, and, in CS, two well established strands of research, one focusing on grammatical properties and constraints, the other on pragmatic functions. I will argue in section 6 that this presents a false choice: what we really need is an approach that doesnāt favor one over the other but instead sees form and meaning (including function) as the parts that any linguistic unit is made up of. Usage-based linguistics offers such an approach.
The next section introduces the usage-based approach, focusing on its application to language change. Section 3 reviews those aspects of code-switching that are relevant with respect to the value of the usage-based approach. Sections 4ā6 then investigate three topics in more detail. Section 4 reviews the debate about code-switching versus borrowing from a usage-based perspective. In section 5, code-switching is compared to other contact effects: it will try to make the case that contact linguistics should study the phenomena together, and construct a unified framework for analyzing them. Rounding off the main body of the paper, section 6 makes a case for the partial unification of pragmatic and grammatical studies of code-switching. The last two sections draw two types of conclusions. Section 7 extrapolates a research agenda from the foregoing sections, and section 8 is a general conclusion. The chapter is of a theoretical nature and will feature few examples; it is based, however, on many years of investigating language contact in the Turkish immigrant community in Holland and elsewhere in Western Europe.
2 Usage-based approach
The three problems identified above featured in the general criticism of linguistic theory that led to the creation and development of the usage-based approach (Langacker 1987), which underlies a lot of work in Cognitive Linguistics, e.g. Cognitive Grammar (Langacker 2008), Construction Grammar (Goldberg 2006) and Exemplar Grammar (Bybee 2010). Under this approach, mental representation (the domain of general linguistics) is based directly on usage (the domain of sociolinguistics) and on general human cognitive properties (the domain of psycholinguistics). Usage and cognition may further be interrelated in the sense that language use will at least partially be directed by the properties of human cognition. This view of how representation is built up inside every individualās mind implies that mental representation varies from person to person, and changes all the time (Croft 2010). The division of linguistics into the mentioned sub-disciplines has precluded an integrated explanation of this characteristic variability and changeability, and perhaps even general recognition of the fact that they are crucial design features of language to begin with, features that any linguistic theory needs to explain.
Such usage-based perspective forces the linguist to consider the relationship between the synchronic and diachronic planes of description. This relationship is relatively unimportant for generative approaches to linguistic theory, which assumes children are born with a linguistic competence they make use of all their life in generating synchronic utterances. Diachronic change is assumed to occur across idiolects, often at great time distances, but not within the competence of a native speaker. Under a usage-based approach, however, this competence is subject to constant change, and while the diachronic changes that the competence of each person undergoes may be unspectacular, they do need an explanation, and it is sought in synchronic usage. Basically, every synchronic act is assumed to have diachronic implications.
A fairly detailed articulation of the usage-based view of change is Croft (2000). Faced with the task of conveying a message, speakers have the choice between saying something new (āaltered replicationā) or something old (ānormal replicationā ). A full utterance is virtually always new, since we rarely store whole utterances with their specific meaning. This, of course, is the feature often referred to as linguistic creativity, another design feature of language, and one of the cornerstones of generative grammar. However, utterances contain smaller sequences, such as words and word combinations, and many of these are stored as such in the speakerās mental representation. Using them, therefore, is the result of an act of normal replication. What makes their selection possible is entrenchment: they are stored in memory and are retrieved when needed. To be sure, at these lower levels, too, new combinations are formed all the time, so that any utterance will combine instances of altered and normal replication. At the lowest level, though, there will be few cases of saying something new: we do not as a habit use new words all the time. One notable exception, and a very relevant one for us, is the use of loanwords from another language.
Whenever a new unit is used, it can be committed to memory by speaker and hearer, and repeated usage will entrench it ever further. If it is used often enough, the selection of such units will no longer represent altered replication: it has become normal replication. Its entrenchment level in any individual speaker may fluctuate, though, and it may well be in competition with one or more other established ways of saying the same thing. Selecting any of them will be a case of normal replication, but one is more ānormalā than the other one: initially the older form will probably be better entrenched than the incoming one. Whenever we observe such alternation between units, we are looking at ongoing language change. If the process ever goes to completion, the older unit will have fallen into disuse, and the once new form has become the default option: its synchronic selection will then be a canonical case of normal replication.
The frequency with which a unit is used therefore plays an important role in language change. Full utterances tend to be unique and will often be used with an exact frequency of just one occurrence. This low frequency, but perhaps even more so its complex meaning (itself perhaps the ultimate cause for low frequency) and its long and complicated form, keeps it from getting entrenched very well, if at all. At the levels of individual words and combinations of two or a few morphemes, however, there are wildly fluctuating frequencies, and these are hypothesized to be responsible for fluctuating degrees of entrenchment as well. As a result, language change is conceptualized as mostly involving the constant waxing and waning of degrees of entrenchment for particular units. The limiting case may be such complete disuse of an old unit that it has effectively disappeared from the language: in such cases entrenchment for individual speakers has gone down to zero. Cases of completed change involve the complete discontinuation of the use of the older variant.
The above scenario is conceptually easiest to illustrate at the word level. Contact situations often figure competition between an inherited native word and an incoming loanword, and at some point the loanword may have completely replaced the native word. However, usage-based approaches tend to adopt the same mechanisms for all levels of language. If words and structures change in the same way, there is no need to make a strict division between lexicon and syntax, and indeed, Cognitive Linguistic theories tend to blur the distinction. Mental representation is seen as a matter of stored knowledge, and units that get stored may be of any level of complexity, and be specific or schematic or a combination of both. It is important to emphasize that the unit that gets stored is never simply a form: all linguistic units are combinations of a form and a meaning. Schemat...