The present paper will assume, for the purposes of argument at least, the essential validity of the claims made on behalf of Case Grammar, in particular that ‘the case notions comprise a set of universal … concepts which identify certain types of judgements human beings are capable of making about the events that are going on around them, judgements about such matters as who did it, who it happened to, and what got changed’ (Fillmore, 1968: 24). We shall investigate in some detail the practicability of Fillmore's statement that ‘we should take the “case uses” as basic and regard the observable “case forms” as derivable from them by the rules of the grammar’ (Fillmore, 1971: 36), by attempting to establish, informally, some of the rules that would be needed to generate certain correct surface forms in Classical and Vulgar Latin and, in so doing, we shall examine, taking into account both the traditional distinction between ‘grammatical’ and ‘concrete’ cases (in particular, as propounded by Kurylowicz, 1964) and aspects of the localist hypothesis, most recently fully discussed by Anderson (1971), the extent to which the changes attested in Latin and early Romance are merely surface re-structurings, and to what extent they are more profound. Finally, certain implications for the theory itself, as at present formulated, are presented and briefly discussed.

This paper concerns the theory of SYSTEMIC GRAMMAR, developed by Halliday (1961, 1967a, 1967b 1968, 1970), Hudson (1967, 1970, 1971, 1973, forthcoming, ms. a, ms. b) and others. In Hudson (1971) a detailed generative version of the theory has been presented, and specifically I want to suggest modifications of this model which I believe improve it. Although a fair amount of the paper consists of criticism of, and comparison with, Hudson's model, the description of what I propose is self-contained and that portion can be read independently.

The law of least effort is of course constantly at play in language use. Redundant linguistic items are consistently reduced in size, replaced with a proform or simply left out. These phenomena are dealt with within the framework of a transformational-generative grammar through transformations involving reduction and deletion. In other grammars they may be handled differently; for example the ‘reduced’ sentences may be regarded as full sentences that may optionally be ‘completed’. But I am not concerned here to give a precise specification of the rules involved, but only to raise the more fundamental questions: what is the nature of the processes involved? And, under what conditions do they operate? In this discussion I shall use the term DELETION throughout although a more neutral term like OMISSION, or even NON-INCLUSION, might be preferred.

Charles Morris' division of signs into syntactics, semantics, and pragmatics has been handed down to linguists and philosophers of language, who have more or less accepted the validity of the categories. However, the generative semanticists have argued, persuasively in my opinion, that the line between syntactics and semantics is not sharp (cf., e.g. Lakoff, 1971). By examining a small but important class of adverbs, I will try to show that the line between semantics and pragmatics is also not sharp and that no non-arbitrary line can be drawn. What is needed for an adequate description of language is a theory that can predict and explain borderline cases without forcing them into pre-established categories arbitrarily.

This paper will be concerned with the problem of drawing a boundary, within a process or so-called ‘generative’ model of phonology, between what Matthews (1972) has called MORPHOLEXICAL phenomena and MORPHOPHONEMIC phenomena.