Grammatical Theory in the United States from Bloomfield to Chomsky

From Psalms: Layer by Layer
Jump to: navigation, search
Matthews cover.jpg

Introduction

P. H. Matthews, Grammatical Theory in the United States from Bloomfield to Chomsky (Cambridge: Cambridge University Press, 1993).

In Grammatical Theory, Matthews, an established linguist whose own work spans decades,[1] takes the reader on a survey of the development of linguistics as a discipline from the 1930s well into the 1980s. Though the title highlights the two big players at the time, Bloomfield and Chomsky, their work is placed into historical context, as Matthews incisively traces their own predecessors' influences and how their work has been carried on by their followers.

Summary

In a sentence: Matthews shows how most of the development of the science of linguistics throughout the 20th century has consisted of wrestling with (read: rejecting, postponing, treating as a matter of deep structure) the place of meaning.

Matthews' historical survey, with minimal discussion of historical-comparative linguistics which dominated the 19th century, picks up the story from the emergence of linguistic structuralism in North America. No idea is formed in a vacuum, so Matthews carefully points out the primary motivation of the first generation of structuralists: namely, their work on North American languages. The so-called Chomskyan revolution, in seed form in Chomsky (1957),[2] but primarily from the 60s onwards, picked up on, adapted, and, only in some cases, rejected, the key characteristics (see below) of American Structuralism.

Outline

The first three chapters of Matthews (1993) trace the thinking of a certain area of study from Bloomfield to Chomsky. The fourth chapter primarily traces the generative paradigm from the initial stages of Chomsky's career to the 80s.[3]

  1. Introduction (general concerns)
  2. Bloomfield's Morphology and its successors (morphology)
  3. Distributional Syntax (syntax)
  4. Chomsky's Philosophy of Grammar

Key Concepts

Structuralist Linguistics

  • Descriptive
  • Synchronic
  • Language specific[4]
  • Autonomous
  • Only concerned with formal distribution[5]
  • Analysis limited to the sentence
  • Corpus-based
  • Both words and sentences are composed of binary pairs of oppositions.


Major shift: "Linguistics was no longer, in [generativists'] eyes, a ‘classificatory science’. It was a science of rules..." (1993, 89)

Early Generative Linguistics

  • Descriptive but also prescriptive
  • Synchronic
  • Language specific[6]
  • Autonomous, in the sense that grammaticality judgment and syntax had nothing to do with semantics[7]
  • Not only concerned with formal distribution[8]
  • Analysis limited to the sentence
  • Judgments of acceptability and grammaticality by intuitions of native speakers[9]
  • Competence the object of study, not performance[10]
  • Deep structure as where the semantic structure could be found, not on the surface structure.[11]
  • Simplicity at all costs[12]

Problems

  • How to define a word? Early Bloomfield was hesitant to draw a sharp line between morphology and syntax, because it was never entirely clear what constitutes a word. If, however, the morpheme was the minimal unit of lexicon (i.e., meaning), and the corresponding tagmeme only contained grammatical meaning - where to draw the line?
  • Allomorphs. If we would like to, strictly formally, distinguish between plurals horses [iz] and cats [s] due to their phonetic surroundings, what to do with the [ən] of oxen? Or man > men? (Even if 'zero' morphemes are called into action, these manifold alternations are not predictable by a general rule).
  • Distributionalism: Why not A box was murdered, when A man was murdered is acceptable?[13]


Bloomfieldian ideas picked up by cognitivists

  • Forms could and should be described with reference to meanings (episememe reminiscent of what came to be known as a construction CxN)[14]. Matthews later discusses Bloomfield's argument "that the meaning of a complex form depends on something other than the meanings of its component morphemes. ‘Every utterance’, as he puts it a page later, ‘contains some significant features that are not accounted for by the lexicon' (1935, 162; cited 1993, 143-144).
  • The focus on corpus as the data for language use

Avenues for neo-structuralists

  • The analysis must go beyond the sentence
  • The analysis must include meaning
  • The analysis must consist of actual language use
  • The analysis should be confirmed with typological understanding of world languages

Reviews

Definitions of the Two Major Schools of Thought

Structuralist Linguistics:

  • structural (adj.) A term used in linguistics referring to any approach to the analysis of language that pays explicit attention to the way in which linguistic features can be described in terms of structures and systems (structural or structuralist linguistics). In the general Saussurean sense, structuralist ideas enter into every school of linguistics. Structuralism does, however, have a more restricted definition, referring to the Bloomfieldian emphasis on the processes of segmenting and classifying the physical features of utterance (i.e. on what Noam Chomsky later called surface structures), with little reference to the abstract underlying structures (Chomsky’s deep structures) of language or their meaning. It is this emphasis which the Chomskyan approach to language strongly attacked; for generative linguistics, accordingly, the term is often pejorative. The contribution of this notion in linguistics is apparent in the more general concept of structuralism, especially as formulated in the work of the French anthropologist, Claude Lévi-Strauss (b. 1908), and others. Here, any human institution or behaviour (e.g. dancing, courtship, religion) is considered analysable in terms of an underlying network of relationships, and the structures demonstrated referrable to basic modes of thought. The crucial point is that the elements which constitute a network have no validity apart from the relations (of equivalence, contrast, etc.) which hold between them, and it is this network of relations which constitutes the structures of the system. Within linguistics, ‘structural’ will be found in several contexts in phonology, grammar and semantics. Structural(ist) grammar, as a general term, is now a largely dated conception of grammatical analysis, though the emphases which characterized it may still be seen in several areas of applied linguistic studies (e.g. in the structural drills of foreign-language teaching), and the term ‘structural’ is often given a special status as part of the exposition of a grammatical model, e.g. the notion of structural description in transformational grammar. Structural semantics is an influential contemporary position, which is still in its early stages of analysing the sense relations that interconnect lexemes and sentences. (David Crystal, A Dictionary of Linguistics and Phonetics, 2008).
  • Structural Linguistics: Any school or theory in which language is conceived as a self-contained, self-regulating system, whose elements are defined by their relationship to other elements. Structuralism originated above all in the posthumous work of Saussure, and by the mid-20th century was not only dominant in linguistics but was having an increasing influence on other disciplines, including anthropology and literary criticism. The term itself was in general use before the Second World War, with reference to a discipline in which the study of a language system was abstracted from the spoken and written use of languages and from their history: hence a growing divide between structuralism and the older branches of philology. But since 1960 usage has become in part confused. In particular, Chomsky and his followers identified structural linguistics with the local Post-Bloomfieldian school to which they were reacting in the USA; since they rejected Post-Bloomfieldian methods, they themselves were not ‘structuralists’. But in a wider conception, Chomsky's account of a generative grammar was clearly structuralist.


Generative Linguistics:

  • generative (adj.) (1) A term derived from mathematics, and introduced by Noam Chomsky in his book Syntactic Structures (1957) to refer to the capacity of a grammar to define (i.e. specify the membership of) the set of grammatical sentences in a language . Technically, a generative grammar is a set of formal rules which projects a finite set of sentences upon the potentially infinite set of sentences that constitute the language as a whole, and it does this in an explicit manner, assigning to each a set of structural descriptions. Related terms are generate and generation, referring to the process involved, and generativist, referring to the practitioner. Several possible models of generative grammar have been formally investigated, following Chomsky’s initial discussion of three types – finite-state, phrase-structure and transformational grammars. The term has also come to be applied to theories of several different kinds, apart from those developed by Chomsky, such as arc-pair grammar, lexical functional grammar and generalized phrase-structure grammar. There are two main branches of generative linguistics: generative phonology and generative syntax. (2) The generative semantics school of thought within generative linguistic theory was propounded by several American linguists (primarily George Lakoff (b. 1941), James McCawley (1938– 99), Paul Postal (b. 1936) and John Ross (b. 1938)) in the early 1970s; it views the semantic component of a grammar as being the generative base from which syntactic structure can be derived. One proceeds in an analysis by first providing a semantic representation of a sentence, and this single level is all that is needed to specify the conditions which produce well-formed surface structures. The subsequent syntactic rules are solely interpretive, and there is no intermediate level. This puts the approach plainly in contrast with the claims of Noam Chomsky and others (in the standard theory) who argued the need for a level of syntactic deep structure as well as a semantic level of analysis. ‘Generative’ in this phrase has, accordingly, a narrower sense than in ‘generative grammar’ as a whole, as it is specifically opposed to those models which operate with a different, interpretive view of semantics. (David Crystal, A Dictionary of Linguistics and Phonetics, 2008).
  • Generative grammar: A set of rules (2) which indicate precisely what can be and cannot be a sexntence in a language. Formulated by Chomsky in the 1950s as an abstract device interpreted as generating, or producing, a set of strings or sequences of units: a ‘sentence’ was formally a string so generated, and a ‘language’ defined as a set of sentences. (P. H. Matthews, The Concise Oxford Dictionary of Linguistics, 2nd Edition, 2007).
  • Generative Semantics: Theory of transformational grammar developed from the late 1960s to the mid-1970s. The original proposal was that a base component of a grammar should directly generate semantic representations of sentences, which would be converted to surface structures with no intervening level of deep structure. This was associated in particular with the view that lexical items were units only at the surface level. But in the 1970s it became clear that the proposed semantic representations could not be assigned by rules of grammar independent of the knowledge, beliefs, etc. of individual speakers. Therefore the break with what was later called the Standard Theory of transformational grammar was in reality far more radical. Leading proponents included G. P. Lakoff, J. D. McCawley, P. M. Postal, and J. R. Ross. Historically one precursor of what became known, more than a decade later, as Cognitive Linguistics. (P. H. Matthews, The Concise Oxford Dictionary of Linguistics, 2nd Edition, 2007).


Mentioned Works

Journals

References

  1. ↑ See, for example, his Amazon publications list!
  2. ↑ Chomsky, N. 1957. Syntactic Structures. The Hague: Mouton.
  3. ↑ Note that, published in 1993, Matthews meticulously represents Chomsky's work after forty years of development, but notes that he was barely reaching retirement age. For example, a landmark publication representing a major shift in generative thinking, (Chomsky, N. 1995. The Minimalist Program. Cambridge, MA: MIT Press) was not to be published for another two years.
  4. ↑ “One central structuralist idea is that every language has a structure of its own, in which individual elements have a role distinct from that which superficially similar elements have in other languages” (Matthews 1993, 6).
  5. ↑ That is, there was a complete disregard for meaning, as this was considered a parallel area of investigation, that of metalinguistics. The fear was that including meaning into the discussion would be pseudo-scientific, since this rather reflects social and cultural interaction and communication, rather than verifiable and falsifiable descriptive data of, for example, phonemics. So went the argument.
  6. ↑ Chomsky's work was largely limited to English until the 80s.
  7. ↑ Chomsky's famous Colourless green ideas sleep furiously apparently shows that well-formed syntax can be semantically nonsense, thus the autonomy between the two areas. As per Chomsky: “language as ‘an instrument or tool’, whose structure should be described ‘with no explicit reference’ to its use (1957, 103). This formal description ‘may be expected to provide insight’ into its use — ‘i.e., into the process of understanding sentences’ (ibid.). But ‘systematic semantic considerations are apparently not helpful in determining it in the first place’ (108), cited in Matthews (1993, 140).
  8. ↑ RULES were introduced, so that so-called grammars could be evaluated and compared. Which best represented good use of the language? In contrast, this description could only be reached in the compilation of data in the Structuralist scheme, and no evaluation was permitted. Now instead of simply describing language structure, rules were pursued which could generate, like a mathematical formula, grammatical sentences. The primary problem to be tackled was: how can a limited set of phonemes in a language be combined to produce an infinite number of different sentences? Furthermore, how can speakers of a language judge an entirely novel sentence which they have never heard before as grammatical or not? As expressed by Hockett, ‘just as the speaker can produce any number of new utterances from essentially the same system of underlying habits’, our description ‘must be capable of producing any number of new utterances, each capable of passing the test of casual acceptance by a native speaker’. Likewise it must match the speaker’s ability to understand ‘an utterance he has never heard before’ (Hockett, 1952b: 98).
  9. ↑ In light of the supposed separation between syntax and semantics, grammatical does not necessarily imply meaningful.
  10. ↑ A speaker's competence refers to their tacit knowledge of their language and thus capability to perform grammaticality judgment; performance refers to their actual use of the language, which was not deemed sufficiently scientific to make up the content of analysis. There is a similar distinction made in de Saussure's early European Structuralism, between langue and parole, which does not seem to have been widely adopted in American Structuralism.
  11. ↑ Transformations became the rage from the late 50s well into the 70s, at which point it was largely abandoned. Instead, movement and tracing were new methods of accounting for different surface structures with the same semantic meaning.
  12. ↑ Explanatory elegance meant, between positing two different semantic structures for two different surface structures, transformations were used to explain phenomena such as passive/active alternations of the same state of affairs (such that John read the book and The book was read by John have the same truth-value semantics, only differing in one surface structure undergoing the passive transformation, while the other does not).
  13. ↑ Harris had to put this down to "culturally determined limitation" (Harris, Z. S. 1951. Methods in Structural Linguistics. Chicago, IL: University of Chicago Press, 253). Chomsky's work through the 70s tackled this issue under the rubric of selectional restrictions. His work, along with Jackendoff (1977. X-bar Syntax: a Study of Phrase Structure (Cambridge, Mass.: MIT Press) developed the X-bar system, whereby NP can behave like nouns, etc.
  14. ↑ "Constructions, to repeat, are tagmemes and tagmemes have meanings, called episememes, just as morphemes have meanings, called sememes" (1993, 125).