• No results found

D Generative studies of language

and of the vision of world is lost’.58

II.D. Generative studies of language

75. Generativism can designate an approach for relating language to the intuitive knowledge of speakers and to the mental capacities of humans at large. (Among various definitions of the term, this one seems dominant.) Historically, it purported to originate as a consummately modernist project more advanced than descriptivism, but ideologically it resonates partly with the ‘pre-modernist’ project of idealism in 17th century Europe,59 which addressed ideas more than language; and partly with the early modernist project of formal logic of the 20th century, one pillar of ‘unified science’ (I.81). Now, the ideology of mentalism, which had long before sponsored the dichotomies between ‘synchronic’ versus ‘diachronic’ and ‘langue’ versus ‘parole’ (II.40f), gained ascendancy.

76. Formal logic, as the term indicates, mainly concerns relations within and among arrays or strings of forms. This concern fit the centrality accorded to Syntax by generative linguistics60 when a ‘language’ was defined as a system (or ‘grammar’)61 of formal rules for generating an ‘infinite set of well-formed sentences’.62 This arcane definition should attract linguists who want to explain why the system of a language always remains open, and the set of instances is never complete; and who like to believe that the regularities in a language are due to ‘rules’, and that the sentence is the primary linguistic unit and the longest unit for linguistic study.63 In return, the definition implied some incisive evasions. ‘Generate’ does not at all mean ‘produce’ a sentence, but just means ‘assign it a structural description’ with no regard for its production; so despite the polemics,64 the ‘generative’ approach was after all a rarefied or diluted ‘descriptive’ approach. Besides, terms like ‘rule’ and ‘sentence’ carry different senses in formal logic than in studies of language. And an ‘infinite set’ in the strict sense of mathematics would eventually produce every combination, however wildly improbable, just as, in the familiar parable, a roomful of chimpanzees randomly pecking at with typewriters would, in infinite time, write the complete works of Shakespeare.65 By implication, the act of comprehending a sentence could require infinite processing time to search an infinite set of possible combinations.

77. Undeniably, generativism has sharply veered toward more theoretical con-ceptions of ‘language’ (II.67), until the term no longer designates a medium for real communication. Instead, the term designates an ‘internal’ and ‘universal’ system anchored in some mental or genetic human potential, of which any one language like English merely constitutes an ‘external’ instance.66 The relation might seem comparable to the relation between a language (or ‘langue’) and an individual discourse (or ‘parole’), but I doubt it. ‘Internal’ versus ‘external’ relates two fairly abstract systems, whereas ‘language’ versus ‘discourse’ relates an abstract (virtual) system with a concrete (actual) system (cf. II.113).

78. Meanwhile, the practices of generative linguists have been mainly argumentative, advancing large claims on the basis of small data sets,67 usually isolated, invented sentences, like ‘sincerity may frighten the boy’ adduced for ‘observing that the rules of grammar impose a partial order in terms of dominance’: [46] we can define the degrees of deviation[by] substituting a lexical item in the position of ‘frighten’.[…] Thus we should have the following order of deviance: (1) ‘sincerity may virtue the boy’; (ii) ‘sincerity

may elapse the boy’; (iii) ‘sincerity may admire the boy’. This seems to give a natural explication… (Noam Chomsky, Aspects)68

Expressions like ‘observe’ and ‘natural’ are incongruous for an approach declaring that ‘knowledge of the language, like most facts of interest and importance, is neither presented for direct observation nor extractable from data by inductive procedures of any known sort’.69 Anyway, the ‘rules’ are precisely not observable, and even the original sentence is hardly natural. None of the 295 occurrences of ‘sincerity’ in the British National Corpus appears at the start of a sentence; and precious few are Subjects of an Active Verb with a Direct Object, such as:

[47] Butler’s earnest sincerity made him a popular hero and leader, especially among the oil-field workers (Campaign for the Preservation of Rural Wales)BNC

Moreover ‘sincerity’ in the BNC data nearly always relates to active agents like the Trinidadian workers’ rights advocate Tubal Uriah Butler in[47]. The ‘deviant’ versions (i-iii) don’t support an ‘explication’ of ‘grammar’ because, whatever their status in any theory, they are equally gibberish in practice. And a linguistics that formulates a ‘hierarchy’ to rank gibberish is like a meteorology that ranks the ‘deviance’ of windless hurricanes, parched cloudbursts, and sizzling snowstorms.

79. Among the largest claims, advanced with no data, was that ‘Syntax’ or ‘Grammar’ — the two terms evasively used interchangeably — is ‘autonomous’ and ‘independent of meaning’.70 This claim handily promised to suspend, at a stroke, the tough problems of accounting for meaning, such as those I pointed out in descriptive Morphology and Semantics (II.54ff, 70ff). A transformational grammar71 would only present a set of ‘rules’ for ‘transforming’ any one sentence structure into another, prospectively without changing its meaning. The paraded example was the ‘passive transformation’, which ‘transforms’ a riveting sentence like ‘the man hit the ball’ into ‘the ball was hit by the man’ (but cf. II.126).

80. In such a ‘grammar’ ‘describing the structure’ of any given sentence equals using ‘rules’ that fit it into the total set of ‘well-formed sentence structures’ in a ‘language’. In contrast to the distinct correspondences between theoretical and practical units in Phonology and Morphology (II.45, 49), the ‘sentence’ evidently doubles interchangeably in both roles (II.68), although the artificiality of invented data camouflages this practice. In theory, a modest set of rules should apply to the total set via ‘transformations’ from a minimal set of maximally simple structures sometimes called ‘kernels’72 — perhaps vaguely inspired by the minimal Phonemes and Morphemes of descriptive linguistics? In practice, generative linguists would work out the set of rules and confirm their applicability.

81. But this practical work must have seemed unattractive, and, as far as I can discover, was never achieved for any language. Moreover, few generative linguists seem to be seriously pursuing it nowadays. Instead, many are pursuing luxuriant elaborations on the theoretical side. The practical but untenable claims were quietly shelved that Syntax is independent of meaning, and that all sentences are ‘transformed’ from a base of ‘kernels’. A more encompassing scheme was pro-posed under fresh labels, having a base with a ‘deep structure’, and any actual sentence with a ‘surface structure’. The ‘deep structure’ gets ‘generated’ and then ‘submitted to the semantic component for semantic interpretation’, whilst the ‘sur-face structure enters the phonological component’ for ‘phonetic interpretation’.73 Morphology simply didn’t appear. Later on, a ‘pragmatic interpretation’ was also postulated,74 but was not integrated into official generative theory, doubtless because pragmatics deals with language use (II.72).

82. And precisely language use had been set aside by another large claim, namely that ‘linguistic theory’ in general and a ‘generative grammar’ in particular should describe ‘competence, the speaker-hearer’s

knowledge of his language’, in a programmatic dichotomy with ‘performance, the actual use of language in concrete situations’; ‘only under the idealization’ of the ‘speaker-hearer’ ‘is performance a direct reflection of competence’, but not ‘in actual fact’.75 Moreover, this ‘com-petence was attributed to an ‘ideal speaker-hearer in a completely homogeneous speech-community, who knows its language perfectly’.75 ‘The study of competence abstracts away from the whole question’ of ‘why speakers say what they say, how language is used in various social groups, how it is used in communication’.76 These claims expediently dismissed all the problems of language variation due to such factors as social class, education, and culture; and culminated in the trenchant claim that the ‘observed use of language’ ‘surely cannot constitute the subject-matter of linguistics’ as a ‘serious discipline’.75 We might well recall the early pronouncement that ‘speech is a ‘heterogeneous mass’ of ‘accessory and accidental facts’ (II.40). Overtones of hopeless utopia resound again, when we are told that ‘from the point of view of the theory’, ‘much of the actual speech observed consists of fragments and deviant expressions’.77

83. Such claims might further imply that young children, who are exposed only to ‘performance’ and ‘observed use’, should encounter severe problems in learning their ‘language’, which they don’t. So the ‘grammar’ was also claimed to represent a ‘genetic’ capacity of human beings, whereby a ‘language’ is ‘acquired’ through an ‘innate language acquisition device’ programmed to select the ‘grammar’ of the native language from a ‘universal grammar’.78 As evidence, the straightforward fact was adduced that young children acquire their native language without being expressly taught, and can soon produce and understand many sentences they have never heard or seen. But this fact does not prove that children acquire their language by unconsciously filtering down from a ‘universal grammar’ of ‘deep structures’. Far more plausibly, they acquire it from experiencing a great deal of text and discourse in meaningful situations that display the creative openness of the language. The ‘acquisition device’ papered over the patent unlearnability of the types of ‘grammar’ (or ‘syntax’) that generative linguistics has consistently pro-pounded; one idealisation was devised to salvage another.

84. Whether the concept of ‘ linguistic universals’79 can be linked to substantive evidence might depend on which aspects of language are addressed. Doing so would be easiest for ‘phonological universals’ reflecting the construction of the human vocal apparatus, than for ‘semantic universals’ invoked as a presumptive basis for semantic units and features (cf. II.71). An ecologist approach would postulate a dialectic between universality and diversity in linguistic, cognitive, and social evolution. Universality would rest more on theory, and diversity more on practice; cognitive and social parallels can favour linguistic parallels, but easily may not, as any sensitive learner of a foreign language can confirm. 85. All in all, generative studies have emphatically reversed the realism and behaviourism instated by

descriptive studies, and promoted a radical idealism and mentalism. By now theory (language) has run much further away from practice (discourse) than in the early dichotomy of ‘langue’ and ‘parole’, yet the key problem is the same: the practices of the linguists themselves are left in a vacuum, enjoined to ‘study language’ but not ‘actual speech’ — to produce a theory without observing the practices, and to create theory-driven, top-down input without recourse to data-driven, bottom-up input (cf. I.84) — like a science providing explanations of its own abstractions (II.40). We cannot look for data from an ‘ideal speaker-hearer’ or a ‘completely homogeneous speech-community’, since neither can ever exist. In effect, a ‘theory of language’ in the generative sense is immune to refutation (cf. II.33).80 ‘Language’ is defined a priori to be just what the theory states it is, no matter what masses of the evidence indicate it is. 86. These waves of idealisation are utterly unlike the idealisations of prescriptivism and purism. A generative grammar aspires to describe ‘competence’, whose ‘rules are ‘unconsciously’ and ‘automatically’ applied; a sample is ‘ungrammatical’ if its structure cannot be described. A prescriptive

grammar aspires to intervene in performance, and with ‘rules’ that need to be applied with conscious effort; sam-ples are ‘ungrammatical’ if some language guardian objects to them.

87. Nonetheless, dismissing the ‘observed use of language’ as the ‘subject-matter of linguistics’ (II.82) makes the generativists resemble the prescriptivists in offering their own variety of the language as the sole model (II.18), which they access by ‘intuition’ and ‘introspection’. Indeed, the offer is far more radical when the linguists represent both the ‘ideal speaker-hearer’ who ‘knows the language perfectly’, and the entire ‘speech-community’ if the latter is indeed ‘completely homogeneous’ (cf. II.82, 85, 116, 144).

88. Matters were aggravated by the generativist reservation that speakers are not ‘aware of the rules’, nor even able to ‘become aware’ of them, nor are their ‘statements about their intuitive knowledge’ ‘necessarily accurate’; ‘any interesting generative grammar will be dealing, for the most part, with mental processes’ ‘far beyond the level of actual or even potential consciousness’.81 This reservation should logically weaken the reports given by generative linguists making ‘state-ments’ or presenting ‘rules’. Otherwise, they would need superhuman powers of introspection, apparently conferred by academic degree in linguistics (cf. II.145 ).

89. Taken together, these theoretical and practical moves characterise generativism as an updated pre- modern project closer to idealism than to formal logic. Over the years, theories have traversed a bewildering gallery of technical transmutations with portentous names, leading to a ‘minimalist program’82 that defiantly advertises doing as little possible. Obviously, such an approach offers no substance or support for the ecologist agenda advocated in this book, such as social or educational applications; and we are bluntly told that ‘your professional training as a linguist’ ‘just doesn’t help you to be useful to other people’.83

Related documents