Consideration of the problem of alternative agreement forms (§ 1) leads us to postulate a hierarchy of agreement positions (§ 2) which allows us to make predictions as to the possibility and relative frequency of semantic as compared to syntactic agreement. The hierarchy is justified by data from a variety of languages which permit alternative agreements (§ 3). Confirming evidence comes from a study of the instances where agreement in case and person is required in different languages (§ 4). The status of the hierarchy and the type of prediction it makes are discussed. It is suggested that the hierarchy determines the main divisions of a measure called SYNTACTIC DISTANCE and that other factors (word order, distance between controller and agreeing element, depth of stacking) determine relative degrees of distance within these main divisions (§ 5). The basic claim of the paper is that as syntactic distance increases so does the likelihood of semantic agreement.

At first glance Reflexivization in Italian seems to obey the same restrictions we find in English. Thus, if we consider the proposal that reflexive pronouns and their antecedents must be clausemates in English, we find in Italian that the same clausemate condition holds and that for any two coreferential pronouns within the same clause, the second of them is a reflexive pronoun. If, on the other hand, we consider the proposal that a reflexive pronoun and its antecedent may be indefinitely far apart provided that certain circumstances are not present, such as an intervening specified subject or the reflexive's being in a tensed S to which the antecedent does not belong (as in Chomsky (1973)), again the same conditions hold for Italian Reflexivization; as illustrated in (I)–(3).

In a number of works up to 1974 Jerrold Sadock advanced and defended the claim that there are a limited but distinct number of sentence structures in English which exhibit true illocutionary ambiguity. Structure (1) belongs to this limited number.

The expression is part of can be used in a number of different senses. One might cite as examples:A sheep is part of a flock. A priest is part of the clergy. Changing nappies is part of being a mother.A petal is part of a flower. But one does not ask What are the parts of the clergy? or What are the parts of being a mother? as one might ask What are the parts of a flower? or What are the parts of the human body? I shall refer to the latter as ‘structural parts’, to distinguish them from the rest; it is with the relation between structural parts and their wholes that the present paper will be concerned. This relation can also be expressed by sentences of the form A flower has petals, The human body has arms, etc. (The verb have can, of course, also be used in senses irrelevant to the present discussion.) The part-whole relation, thus conceived, is generally recognized as being one of the fundamental sense-relations structuring the vocabulary of a language.

Many contemporary linguists hold that an adequate description of a natural language must represent many of its vocabulary items as syntactically and/or semantically complex. A sentence containing the word kill, for instance, will on this view be assigned a ‘deep syntactic structure’ or ‘semantic representation’ in which kill is represented by a portion or portions of tree-structure, the lowest nodes of which are labelled with ‘semantic primitives’ such as CAUSE and DIE, or CAUSE, BECOME, NOT and ALIVE. In the case of words such as cats or walked, which are formed in accordance with productive rules of ‘inflexional’ rather than ‘derivational’ morphology, there is little dispute that their composite status will be reflected at most or all levels of linguistic representation. (That is why I refer, above, to ‘vocabulary items’: cat and cats may be called different ‘words’, but not different elements of the English vocbulary.) When morphologically simple words such as kill are treated as composite at a ‘deeper’ level, I, for one, find my credulity strained to breaking point. (The case of words formed in accordance with productive or non-productive rules of derivational morphology, such as killer or kingly, is an intermediate one and I shall briefly return to it below.)

Most linguists agree that there is a mismatch between the notions of grammaticality and acceptability. A widely held view is that there are several types of sentence which are technically well-formed, and consequently grammatical, but which are unacceptable to the ‘typical’ speaker. A more recent, and less widely held view is that the reverse situation is also possible: there exists a class of sentences which are ungrammatical, yet acceptable. In other words, it is claimed by some that the notions of grammaticality and acceptability cross-classify, since we get sentences which are:

In 1971–72 a lengthy discussion took place in Language concerning sentence stress (alias nuclear stress, alias intonation nucleus placement) and its relationship to syntactic structure (Bresnan, 1971, 1972; Lakoff, 1972; Berman & Szamosi, 1972; Bolinger, 1972). Bolinger had the last word in stating that placement of the nuclear accent ‘is not explainable by syntax or morphology’ but ‘is a matter of information, not of structure’ (Bolinger, 1972: 633). According to this view words with low informational content (i.e. highly predictable) in a particular context or situation are unlikely to receive a nuclear stress, whereas words with high information content (i.e. highly unpredictable) are likely to receive a nuclear stress. In addition, ‘the speaker will put the main accent as far to the right as he dares, when assertive pressure is high’ (Bolinger, 1972: 644).

1.Consider a string s = a1 a2 …an over a finite vocabulary Va, such that s is structured by an asymmetric binary relation R: i.e. there are ordered pairs of elements in the string, ai, and aj, where a, R aj.1 S is a simple structured string if (a) there is a unique element ak such that there is no a1 for which it is true that a1 R ak, and if (b) for all am am, ≠ akit is true that there is a unique a1 such that a1 R am. In graph-theoretic terms these conditions define a tree where a1, a2 …a are paired with nodes in the graph and each a, R a is represented by an arc, and where the tree has a unique root or centre associated with a. Figure i is an instance of such a tree, where k = 3. In Figure a, condition (a) is infringed, in that an arc converges on each node; i.e. for all a, there is an aj such that aj R ai. Figure 3, on the other hand, violates condition (b), in that two arcs converge on a2 i.e. there is not a unique ai for whch it is true that a1 R a2.

The attention by ‘modern’ linguists to the problems of Old English diphthongal spellings and their phonetic realizations may perhaps be dated (without doing anyone an injury) from the study by Robert P. Stockwell and C. Barritt (1951). Since that time much ink has been spilled (or typewriter ribbons worn out) in attempts to specify in some precise and principled way the phonological structure of the entities spelled in Old English ea, eo, io, Ia and ie. I will spare the reader a review of these articles and commentaries since such a review would prove both tiresome and uninformative.

One of the more welcome developments in generative phonology has been the acceptance of the long-established and well-known concept that the morphophonemic structure of a language may be reorganized in an attempt to establish ‘paradigmatic coherence’, that is to say, a major rôle of analogy is to eliminate or minimize allomorphic variation within a paradigm. Of course it would be wrong to suppose that analogical change is restricted to the elimination of allomorphy; indeed, it is even possible to find examples of analogical change which increase allomorphy, as in the development of the plural imperative form onto:n of Greek einai ‘to be’ in place of the more regular esto:n.1 But, since analogical elimination of allomorphy plays such a crucial rôle in linguistic change, we may safely concentrate on that aspect of analogy in this paper without denying the many other types of change in which analogy plays a part.

In Comrie (1976: 108) we read: ‘Several languages have special inferential verb forms, to indicate that the speaker is reporting some event that he has not himself witnessed, but about whose occurrence he has learnt at second hand (though without, incidentally, necessarily casting doubt on the reliability of the information)’. Two pages later Comrie writes follows:

Serebrennikov (1960: 66) lists the following languages known to him where there is a close formal relation, down to identity, between the expression of perfect and inferential meaning: Turkic languages: the Uralic languages Nenets (Yurak-Samoyed), Finnish, Estonian, Man (Cheremis), Komi (Zyryan), Udmurt (Votyak), Mańśi (Vogul); Georgian; and the Indo-European languages Latvian, Bulgarian, and Albanian.

My main purpose here is to examine the basic colour vocabulary of modern standard French in the light of recent research on colour vocabularies and in particular to account for the fact that two highly salient terms, brun and marron, are used to designate a single colour category which I shall call BROWN. (The notion of saliency is defined in § 1.2 below.) Before the publication in 1969 of Berlin and Kay's Basic color terms with its universalist hypothesis, the fact that the English term brown has no one equivalent term in French would generally have been regarded as evidence for the structuralist view that languages are anisomorphic in their semantic structures (Lyons, 1968). Linguists before 1969 turned to colour vocabularies to demonstrate not only that languages are anisomorphic in their semantic structures but that the colour continuum is cut up in a completely arbitrary way in different languages (Hjelmslev, 1943; Gleason, 1955). Recent research, however, suggests that some modification of this view may be necessary.

Many linguists describe the underlying structure of sentences in terms of cases (e.g. Fillmore, 1968) or relations between the verb and nouns (Chafe, 1970), the, usually tacit, assumption being that this reflects the way we view the world. The underlying linguistic structure would accordingly be a formalization of cognitive structure. Elsewhere (Schlesinger, 1977: Ch. 5 ), I have argued that a distinction has to be made between cognitive structures and semantic deep structures. Cognitive structures correspond to the way we perceive the world.

In a recent article published in this journal (Faarlund, 1977), an attempt was made to account for the problem raised by the well-known fact that Scandinavian languages often use the neuter form of the predicative adjective in cases where the non-neuter form would seem to be called for by the ordinary rules of gender agreement.

Although there has been extensive research into the usage and interpretation of the modal auxiliary WILL (Leech, 1971: 52—54; Palmer, 1974: 104—115) sentences of the type (1) John will fight for justice. (2) Men will fight for justice. (3) John will be here. may still present problems of analysis.

Utilizing Palmer's system, we can interpret sentence (I) as implying the following: (i) simple futurity (ii) volition (iii) insistence (closely related to (ii) (Palmer, 1974: III)) (iv) characteristic activity.

Since Postal (1971: 27–31) first used the term Tough Movement, a great deal of material on this phenomenon has appeared in print. In all of this research, there has been a widespread (often tacit) assumption that the characteristic - in fact, exclusive - triggers for Tough Movement are adjectives (e.g. tough, hard, interesting) and nouns (e.g. bitch, bear, snap). For example, Lasnik and Fiengo (1974: 568) list 21 common Tough predicates. Although their list is not meant to be exhaustive, it is a characteristic one, especially in that adjectives outnumber noun triggers two to one and half the nouns are limited to vernacular usage (the most frequent use of noun triggers probably involves obscene usage).

Harris (1978) argues against the ‘descriptive’ interpretation of performative utterances on the ground that on a particular occasion of its use I apologize may not constitute a ‘genuine instance of apologizing’, although the use of that utterance on that occasion may be ‘correctly described’ by using the verb APOLOGIZE non-performatively, as in (these are Harris's examples): (1) S apologized to the lamp-post, and even the police sergeant laughed. (2) S apologized on T's behalf, although T himself flatly refused to.

In an earlier paper (Huddleston, 1977) I analysed the would of (I) If he had stayed in the army, he would have become a colonel as the unreal mood form of an epistemic modal verb WILL that figures in the underlying syntactic representation and falls within the semantic scope of a past element realized by HAVE. Palmer (1978) criticizes this analysis, claiming that WILL is ‘obviously’ semantically empty, a mere ‘dummy’ inserted to carry the mark of unreality – that would have become is ‘the Unreal and Past form of BECOME’.