¹ I am deeply indebted to Morris Halle and Edward S. Klima for valuable criticism of an earlier draft of this paper. My thanks are also due to Sheldon Sacks, James Sledd, and Robert P. Stockwell for many helpful suggestions.

² For these proposals and their corollary, see especially Morris Halle, ‘Phonology in generative grammar’, Word 18.64–8 (1962), and the revised version in Jerry A. Fodor and Jerrold J. Katz, eds., The structure of language: Readings in the philosophy of language 344–9 (Englewood Cliffs, N. J., 1964).

³ Word 18.66, ft. 12; Structure 346, ft. 13.

⁴ Word 18.66; Structure 346.

⁵ Lg. 40.2 (1964).

⁶ The notion of grammar is developed by Noam Chomsky, Syntactic structures ('s-Gravenhage, 1957).

Questions have frequently been raised concerning the feasibility of using this notion of grammar in historical analysis, in particular concerning the appeal to intuition. A linguist theorizing about a living language ideally has as a control his own native intuition and that of the speakers around him, or at worst the native intuition of speakers of a language foreign to him. Against such intuition he can test, among other things, degrees of grammaticality and types of ambiguity. With dead languages, howevre, the linguist can rely only on the limited data available to him, and at best on a secondary ‘native intuition’ which can arise only after several years of close association with the language. He can find very few, if any, syntactically minimal pairs from which to set up paradigms of grammatical versus ungrammatical sentences. Deviation and ambiguity are even more elusive. If we take in its strongest terms the requirement placed on linguistic theory that it should characterize and predict all and only the sentences of the language and also account for the native speaker's competence in producing and understanding utterances of the language, we might ultimately conclude that a grammar can be written only by a native speaker, not a foreigner, and that grammars of dead languages cannot be written at all. The degree of accuracy will naturally vary according to the degree of acquaintance with the language. But this does not mean that all investigation of language not native to the linguist must de facto be abandoned, any more than any theory of history, whether cultural or geological, must be rejected because we cannot recapture all and only the characteristics of previous eras. We may quite legitimately put forward a theory of a dead language, in terms of a grammar which fulfills the requirements of descriptive adequacy and explanatory power. This theory will be based on all observable data, and also on unobservable data when necessary, i.e. when the logical consequences of the model would not match the observable data without this hypothesis. As in analysis of a living language, that model will be the simplest which will characterize the sentences of the corpus, and so the infinite set of unobserved sentences which pattern with them. Within such a framework, deviance as well as grammaticality can tentatively be made explicit.

⁷ For fuller versions of these grammars, see Closs, Syllabus for English 110, History of English 11–6, 24–9, 34–7 (mim., University of California, Berkeley, 1964); Deep and surface structure in Old English (in preparation).

⁸ See especially Chomsky, ‘A transformational approach to syntax’ in Archibald A. Hill, ed., Third Texas conference on problems of linguistic analysis in English 131–2, 144–7 (Austin, 1962); Klima, ‘Negation in English’ in Fodor and Katz, eds., The structure of language: Readings in the philosophy of language 251–3 et passim; Robert B. Lees, A grammar of English nominalizations 19–20 et passim (Bloomington, 1960). For a discussion of the criteria by which the set of auxiliary verbs is set up, see James Sledd, A short introduction to English grammar 106–9 (Chicago, 1959).

⁹ Quotations for Old English are derived from Henry Sweet, ed., King Alfred's Orosius, EETS 79 (London, 1883), abbreviated Or.; and from Henry Sweet, ed., King Alfred's West-Saxon version of Gregory's Pastoral Care, EETS 45, 50 (London, 1871), abbreviated CP. References are to page and line numbers.

¹⁰ Recent detailed discussion of word-order problems include S.O. Andrew, Syntax and style in Old English (Cambridge, 1940); Paul Bacquet, La structure de la phrase verbale à l'époque Alfrédienne (Paris, 1962); C. R. Barrett, Studies in the word-order of Aelfric's Catholic Homilies and Lives of the Saints (Cambridge, 1953); Charles R. Carlton, Syntax of the Old English Charters 170–256 (unpub. doctoral diss., Michigan, 1958); David P. Harris, ‘The development of word-order patterns in twelfth-century English’ in Albert H. Marckwardt, ed., Studies in languages and linguistics in honor of Charles C. Fries 187–98 (University of Michigan, 1964); Bruce Mitchell, ‘Syntax and word-order in “The Peterborough Chronicle” 1122–1154‘, Neuphilologische Mitteilungen 65.113–44 (1964).

¹¹ (13), (14), (17) are examples of deviation from this rule.

¹² ‘Independent clauses’ here include ‘demonstrative clauses’ introduced by demonstrative adverbs þa ‘then’, þonne ‘then’, þær ‘there’ in which the finite verb or one helping verb usually precedes the subject (cf. Andrew, Syntax and style in Old English 3). Both independent clauses with demonstrative adverbs and those without share the main features of verb order under discussion.

¹³ The rules are particularly interesting in that they are basically similar to those suggested by Emmon Bach for German, ‘The order of elements in a transformational grammar of German’, Lg. 38.263–9 (1962).

V_ix in Rule 4.4 stands for the class of all V_i that are not V_imove. It includes verbs homonymous with the members of V_imove.

¹⁴ Quotations for Middle English are taken from Hans Kurath, Sherman Kuhn, John Reidy, eds., Middle English dictionary (Ann Arbor, 1954-); James Gairdner, ed., The Paston letters 1422–1509 (London, 1904), abbreviated PL., with references to volume, page, and line numbers; and Geoffrey Chaucer, The text of the Canterbury Tales, ed. John M. Manly and Edith Rickert (Chicago, 1940).

¹⁵ Clear loss of identity as MV is indicated by the occasional interchange in different MSS of gin - Inf and do - Inf; cf. Cursor Mundi, Göt. 2009 (c. 1400): A neu liuelad gan he bigin ‘He began a new kind of life’, with MS variants con, cun (reduced forms of gan) and also dud. A summary and bibliography of studies on do and gin is provided in Tauno F. Mustanoja, A Middle English syntax I: Parts of speech 600–15 (Helsinki, 1960).

¹⁶ On some of the problems in accounting for specifically poetic deviance, cf. Samuel R. Levin, ‘Poetry and grammaticalness’, in Horace Lunt, ed., Proceedings of the ninth international congress of linguists 308–15 ('s-Gravenhage, 1964).

¹⁷ Data for Early Modern English are derived from the Oxford English dictionary; and Ronald B. McKerrow, ed., The works of Thomas Nashe (Oxford, 1958), abbreviated N., with references to volume, line, and page numbers.

¹⁸ The latter is a modern construction which did not come into general use until the nineteenth century. The first clear instance of a passive of this type cited by Fernand Mossé, Histoire de la forme périphrastique étre + participe présent II: Moyen-anglais et anglais moderne par. 263 (Paris, 1938), is from a letter by Robert Southey: A fellow whose uppermost upper grinder is being torn out by a mutton-fisted barber. For detailed discussion of the history of the passive progressive, see Mossé, ibid., pars. 231–81.

¹⁹ Thomas Deloney, Works, ed. Francis O. Mann (Oxford, 1912).

²⁰ A synchronic grammar cannot account for these changes, except so far as it treats different dialects, or different reflexes of different changes. When Klima says the order in which he describes the rules for pronouns in different dialects reflects the historic order of change, he is actually referring to the order of mutations, not innovations. Each set of rules for each dialect requires different ordering of basically the same rules. Each set has its own unique relationship to the rest in the structure of the language, and cannot be collapsed under the same grammar except as a discrete subset of the grammar. It has been suggested that grammars should provide rules accounting for synchronic relatedness between grammatical systems, such that different systems may be regarded as modifications or extensions of a given basic system. This is essentially what Klima's grammar does for pronouns. In addition, it has been suggested that grammars should provide rules accounting for diachronic relatedness between grammatical systems, also such that the different systems may be regarded as modifications or extensions of a given basic system. Such grammars would reveal with great clarity the similarities and differences between stages of the language, and would provide in simpler, i.e. more compact, form the same information that separate grammars of different stages of the language provide. They cannot, however, specify actual change or provide historical perspective. A grammar of the actual changes would be a kind of algebra accounting in the simplest way possible for all relevant changes, in their chronological order.

²¹ See F. Th. Visser, An historical syntax of the English language 93–138 (Leiden, 1963).

²² Visser, ibid. 131, suggests that this ambiguity was one of the factors leading to the transitivization of intransitives.

²³ Further subdivisions may or may not be made according to the particular model of grammar adopted. Grammars like Lees's Grammar of English nominalizations allow for certain groupings in the phrase structure according to sets of subcategorizations; Charles Fillmore's study ‘The position of embedding transformations in a grammar’, Word 19.208–31 (1963), specifies groupings for two-string vs. one-string transformations. In the latest models, however, such as Chomsky's blocking grammar and Klima's nonblocking grammar (cf. Klima, ‘Current developments in generative grammar’, forthcoming in Kybernetika I, Prague), the phrase-structure component is minimal and cannot be subject to groupings. Context restrictions and subcategorizations are largely specified in a lexicon in which the only significant groupings are the overall categories N, V, Adj, etc.; only in the filter transformations do we find areas in which the concept ‘point of break’ is significant for syntax.