Descriptive theories, explanatory theories, and Basic Linguistic Theory

than the sorts of grammars associated with generative grammar. It is not clear, in particular, that speakers make any generalizations across phrasal categories and hence it is not clear that there is any need for notions like ‘head’, ‘complement’, ‘adjunct’, or ‘specifier’. Again, the claim of this Descriptive theories and explanatory theories 221 paper is that the sort of grammars that use basic linguistic theory as their theoretical framework express a reasonable level of generalization. It is also worth pointing out the implications of this line of argument to the classic argument by Halle (1959) and Chomsky (1964: 88–89) against the phoneme, based on the parallels between the allophonic and morphophonemic rules in Russian that are required in an account assuming a classical level of phonemic representation. The argument was that these two rules could be collapsed into a single rule if such a level was abandoned. However, the alternative view is that the grammar of Russian does contain two separate rules, but their similarity is captured by an explanatory theory independent of the grammar. 8 4.5. An example involving economy A final example of a type of functional explanation that is independent of the grammar itself is provided by markedness phenomena. Consider the fact that the verbal paradigm in Lakhota has a zero for third person singular, and non-zero forms for the other combinations of person and number, as in (4). (4) wa-ka taka ‘I strike’ ya-ka taka ‘you (singular) strike’ ka taka ‘he strikes’ u -ka taka-pi ‘we strike’ ya-ka taka-pi ‘you (plural) strike’ ka taka-pi ‘they strike’ (Buechel 1939) This pattern in Lakhota reflects two markedness patterns: it is common for singular to be unmarked relative to plural and it is common for third person to be zero relative to first and second person. Note that the longest forms are those that are marked both for number (plural) and for person (first or second). One hypothesis for these (and other) markedness patterns is that they reflect discourse frequency: the most frequent values are usually unmarked. With respect to the markedness feature of zero expression, the use of zero for high frequency values is motivated by the fact that hearers will tend to interpret utterances with information unspecified according to what is most frequently used. Overall, a system which uses zero expression for most frequent values will be most “economical” (cf. Haiman 1983, 222 Matthew S. Dryer 1985; Croft 1990) and will conform to the principle of least effort (Zipf 1935). But again, the fact that using zeroes in third person singular will be most efficient may explain an aspect of the grammar of Lakhota but will not itself be part of the grammar. When speakers of Lakhota produce zero forms to express clauses with a third person singular subject and nonzero forms to express clauses with other subjects, they do so, not because it is more efficient to do so, but because that is what the grammar of Lakhota specifies is the form for particular person-number combinations. Again, the locus of the functional explanation is primarily at the level of language change: a language is most likely to lose an overt morpheme in a paradigm if it is associated with third person singular, since in its absence, hearers are most likely to interpret utterances in terms of the most frequent category, something they do all the time in interpreting utterances. If a language were to start to use zero expression for some other person-number combination, there would be a greater probability of utterances being misunderstood. Note, however, that as with other functional explanations, there may be a minority of languages which do not conform to the general pattern. The present tense in English employs an overt zero form in every form other than third person singular (I walk, you walk, he/she walks, they walk ). At the level of grammar, English is no different from Lakhota; in both cases, the grammar spells out where zero is used and where it is not used. We need an explanatory theory independent of the grammar to spell out why languages like Lakhota are much more common than languages like English, but this has no bearing on the descriptions of the two languages. And basic linguistic theory is adequate to provide the relevant sorts of descriptions. 4.6. Theories of form versus formal theories It is worth noting that arguments similar to those I have given here are given by Newmeyer (1998, 2002c) in arguing that functional explanations are outside of grammars and hence that functional explanations do not obviate the need for grammars. He points out for example (1998: 141–142) that certain properties of relative clauses in Swahili are plausibly explained by processing factors. However, the relevant processing explanation involves processing ease at an earlier stage of the language, when it was Descriptive theories and explanatory theories 223 SOV, and that the processing explanation no longer applies now that the language is SVO. From his arguments that functional explanations are outside of grammars, Newmeyer argues that even someone espousing functional explanations needs what he calls a “formal theory” of grammar (1998: 337). Newmeyer’s choice of the expression “formal theory” is unfortunate here because it is ambiguous. His arguments only show the need for a theory of linguistic form, or what is called here a descriptive theory (or theoretical framework); to that extent, his arguments are congruent with the arguments of this paper regarding the need for descriptive theories. However, the expression “formal theory” is generally interpreted in the field, not as a theory of form, but as a theory of a particular theoretical orientation, where “formal” contrasts with “functional”, or with “substantive”, or with “informal”. Newmeyer concludes, apparently playing unconsciously on the ambiguity of “formal theory”, that functionalists need some version of generative grammar as a theory of forms, to supplement their functional explanations. However, all that his arguments demonstrate is that functionalists need a theory of linguistic form, and he ignores the fact that functionalists already have a theory of linguistic form, namely basic linguistic theory. Basic linguistic theory is an informal theory of form, and none of Newmeyer’s arguments argue for a formal theory of form. 5. Stipulative analyses In much formal theoretical work, much of the effort expended in arguing for or against particular analyses is motivated by attempts to provide what are viewed as explanatory analyses. This often involves abstract principles or abstract analyses, the details of which are not motivated for each language but are posited on general theoretical grounds. The sorts of analyses posited in basic linguistic theory are viewed by many generative linguists as inadequate because they are seen as overly stipulative or ad hoc, requiring rules or aspects of rules that are motivated only for the phenomenon in question. Under a functionalist view, the fact that an analysis may be stipulative or ad hoc does not present a problem: the analysis can be stipulative and ad hoc without implying that there are not explanatory principles or motivations that underlie them. In fact, under a functionalist perspective, a highly stipulative analysis may in fact be a 224 Matthew S. Dryer virtue, since the evidence often suggests that speakers learn complex sets of grammatical rules without any awareness, conscious or unconscious, of the underlying motivations that led to those rules being the way they are before the speaker ever started learning the language. Another way to make the same point is to say that grammars are an emergent phenomenon and are not reducible to explanatory principles, whether they be parameter settings, rankings of universal constraints, or functional principles. They reflect underlying explanatory principles of various sorts which resolve themselves, as competing motivations, at the level of language change. But the grammar itself is a complex system with considerable idiosyncrasies of various sorts. The goal of a descriptive theory is to provide a set of tools and concepts for providing adequate descriptions of each language in all its complexity and idiosyncrasy. The reason why Chomskyan generative approaches have found basic linguistic theory inadequate is not in general that it fails to provide adequate descriptions, but that it fails to provide adequate explanations. But if the description of a language is independent of explanation, then this is not a problem for basic linguistic theory. 6. Description as explanation In describing the contrast between description and explanation here, I am following common parlance among functionalists in restricting the term explanation to explanations for why languages are the way they are. But there is a sense in which what I am calling descriptions are themselves explanatory at a different level. Namely, if the grammar is a representation of what is inside speakers’ heads and hence what underlies linguistic behaviour, then the grammar itself can be viewed as part of the explanation for linguistic behaviour, and the grammar serves as an explanation for particular facts of the language. The reason that speakers of English do not say things like *My house is a house blue rather than things like My house is a blue house is because the grammar of English states that attributive adjectives precede the noun, and there is a sense in which the grammar of English explains this fact about language use. A similar point is made by Greenberg (1968: 180): In descriptive linguistics, even grammatical rules of the conventional sort are explanatory of particular phenomena ... . If, for example, a student who Descriptive theories and explanatory theories 225 is just learning Turkish is told that the plural of di (tooth) is di ler while that of ku (bird) is ku lar, he may ask why the first word forms its plural by adding -ler while the second does so by adding -lar. He may then be told that any word whose final vowel is -i takes -ler, while one in which the final vowel is -u takes -lar. This may be considered an explanation, insofar as further interrogation has to do with classes of words that have -i and -u as their final vowel rather than with the individual forms di and ku . What I am describing in terms of description and explanation can thus be equally well described in terms of a distinction between explanations for linguistic behaviour or specific facts about a language and explanations for why languages are the way they are. My claim that we need both descriptive theories and explanatory theories can thus be equally well described in terms of a need for two different types of explanatory theories. Part of the reason why functionalists tend to avoid applying the term explanation to grammatical descriptions is that they associate such talk with the sorts of analyses that formal linguists propose in which an analysis is described as explanatory, not only in the sense in which a grammar explains particular linguistic facts, but in the further sense in which the analysis is intended to be part of an explanatory theory of why languages are the way they are. Because Chomskyan generative linguists typically attempt both types of explanation in their analyses, they do not always distinguish the two types of explanation. However, once one distinguishes these two sorts of explanation, then there should be no problem with functionalists recognizing descriptions as explanations in the lower-level sense of explanations for particular linguistic facts. 7. Is basic linguistic theory a good descriptive theory? In arguing here that basic linguistic theory is a theory, it does not follow that it must be a good theory. One might accept the notion of a descriptive theory, but claim that basic linguistic theory is inadequate as a descriptive theory. Unfortunately, the question of the adequacy of basic linguistic theory as a descriptive theory is rarely addressed. It is implicit in much Chomskyan generative work that basic linguistic theory is inadequate, but as discussed above, the implicit argument is that it is inadequate as an explanatory theory, which I have argued is irrelevant to the question of its adequacy as a descriptive theory. Another argument implicit in early 226 Matthew S. Dryer generative grammar and in formal approaches like Head-Driven Phrase Structure Grammar (Pollard and Sag 1994) is that basic linguistic theory is inadequate in that it is too imprecise and too vague. There is little question that basic linguistic theory is an informal theory, and it is clear that some linguists demand a higher level of precision than one often finds in basic linguistic theory. It is worth emphasizing, however, that precision should not be confused with the use of formalism: a description in English can be quite precise if the meaning of the terms is clear, and it is easy to find much work that assumes basic linguistic theory that is fairly precise. Conversely, the use of formalism does not guarantee precision if the formalism is not precisely defined. 9 A more important question for the purposes of this paper is whether basic linguistic theory is adequate as a descriptive theory for the purposes of functionalists. It is certainly possible that some functionalists view descriptions in basic linguistic theory as adequate as descriptions of the language, but inadequate as representations of what speakers of a language know, a distinction I have ignored here. Some functionalists have developed functionally-oriented theoretical frameworks (e.g. Functional Grammar (Dik 1978, 1989, 1997) and Role and Reference Grammar (Van Valin 1993, Van Valin and LaPolla 1997)), but they do not directly address the question of how their theoretical frameworks might be better than basic linguistic theory: in arguing for their theories, they contrast themselves with formal generative approaches rather than with basic linguistic theory. It is not clear what arguments might be given by practitioners of such theories against basic linguistic theory. Furthermore, as with other theoretical approaches, basic linguistic theory is an overall theoretical framework encompassing different points of view, and criticisms of specific practices within basic linguistic theory can often be construed as theory-internal disagreements as easily as criticisms of basic linguistic theory itself. My view (Dryer 1997) that grammatical relations (and other grammatical notions) are ultimately language-particular and that crosslinguistic notions are no more than convenient fictions is clearly a minority view within basic linguistic theory. But I do not view this position as a criticism of basic linguistic theory, but rather as an issue within basic linguistic theory. There are various other respects in which I believe much descriptive work paints a distorted view of language. For example, grammatical descriptions tend to concentrate on regularities and to play down lexical idiosyncracies and lexicalized grammatical constructions. I believe that word classes in particular languages are often Descriptive theories and explanatory theories 227 not as well-motivated as descriptions sometimes suggest, and that word class systems are often highly complex. But again, I view these not as criticisms of basic linguistic theory, but simply as criticisms within basic linguistic theory of applications of the theory. 10 The improvements in basic linguistic theory over the past twenty-five years have not been prompted by specific attempts to improve it, since most linguists have failed to recognize its status as a theoretical framework. Developments have been the side effect of work in typology, and there is every reason to believe that further developments will continue in coming decades, both because of work in typology and quite possibly from new ideas from some other quarter. However, further improvements might develop if more functional, typological, or descriptive linguists recognized the status of basic linguistic theory as a theory, and addressed the question: how could we make the descriptive grammars we are writing even better than they are now?