1 This work was part of research supported by a grant from the Department of Scientific and Industrial Research. An earlier version was read by M. A. K. Halliday, R. M. W. Dixon, J. O. Ellis, W. Nelson Francis, R. A. Hudson, R. Quirk, and J. M. Sinclair: I am most grateful to them for their numerous helpful suggestions; they are not of course responsible for any errors that remain.

2 I cannot here give a full bibliography; the works on scale-and-category theory that are most directly relevant to this paper are M. A. K. Halliday, ‘Categories of the theory of grammar’, Word 17.241-92 (1961); ‘Class in relation to the axes of chain and choice in language’, Linguistics 2.5-15 (1963); ‘Intonation in English grammar’, TPS 1963.143-69; ‘Syntax and the consumer’, Georgetown monograph series on language and linguistics 17.11-24 (1964); M. A. K. Halliday, Angus McIntosh, Peter Stevens, The linguistic sciences and language teaching (London 1964); Halliday, ‘Some notes on deep grammar’, paper read to the Linguistics Association of Great Britain, Spring Meeting 1965.

3 A much fuller account of the concept of delicacy is given in Halliday, ‘Syntax and the consumer’.

4 For rankshift, where this hierarchical ‘consists of’ relationship is broken, see §1.1 below.

5 Present in present in the sequent tense system is exemplified in (He asked why his mother) was reading (that book again).

6 Broadly speaking, rank is concerned with the syntagmatic delimitation of constituents, delicacy with the labeling of the constituents. Thus a move from group to word is a move in rank, for a group consists of (or may consist of) more than one word in a syntagmatic sequence, whereas the moves from noun to substantive to countable substantive and from modifier to ordinator to cardinal ordinator are moves in delicacy: noun is more comprehensive than substantive paradigmatically but not syntagmatically. Contrast Postal, who claims that the distinction between rank and delicacy is arbitrary and unmotivated: P. Postal, Constituent structure 103 (The Hague, 1964).

7 Diamond brackets signify inclusion: ‘S’ is included in ‘P’. This order is expounded by the sequence: first word of verbal group, then subject, then remainder of verbal group, as in have you spent (any money). When the verbal group consists of a single word that is not always and necessarily final in verbal group structure, the structure 'P <S>' may be expounded by the sequence predicator then subject. Thus have you (any money) has the structure 'P <S>', whereas said John exemplifies the structure 'P S'.

8 These examples of structural exponence are oversimplified. For the theory does not require that the exponence relation be biunique. Thus who came has the property interrogative although its structure is ‘S P’, whereas (hardly) had I finished is affirmative although its structure is '... P <S>' It is of course necessary that such 'exceptions' be explicitly accounted for, and this will mean complicating the exponential statements. The latter could obviously be simplified by assigning to who came the property affirmative and to (hardly) had I finished the property interrogative, but this would merely put the complexity elsewhere in the total description, i.e. in the grammatical semantics.

9 Phonological description and notation as in Halliday, ‘The tones of English’, Archivum linguisticum 15.1-28 (1963).

10 Cf. R. E. Longacre, Grammar discovery procedures 16 (The Hague, 1964): 'The most obvious way in which constructions on different levels of structure (not levels of representation) are related within a language is by virtue of grammatical hierarchy. Immediate constituent analysis yielded ad hoc hierarchies specialized overmuch in terms of particular sentences. Immediate constituent analysis failed, therefore, to uncover hierarchically arranged patterns of maximum relevance and comparability.' Tagmemic 'level' corresponds fairly closely to scale-and-category ‘rank’.

11 Cf. Halliday, ‘Categories’ 253; this requirement of total accountability at all ranks is not maintained in tagmemics: see Longacre, Grammar discovery procedures 17 fn. 14. For further discussion see §2.1 below.

12 Where ‘m’ stands for modifier, ‘h’ for head. Note that the nominal group (23) has three modifiers, not just one: the two old is not a single constituent. The move from ‘m’ to ‘d’ is a move in delicacy, not rank, for ‘d’ is not a constituent of ‘m’. See fn. 6 above, and contrast Postal, Constituent structure 102.

13 This example is taken from R. B. Lees, review of Harris, String analysis of sentence structure, IJAL 30.418 (1964).

14 There may of course be nesting of paratactic structures within either or both of the two constituents, as in both John and his wife, and the doctor. (33) is thus not ungrammatical if spoken with an intonation marking some such nesting; indeed in some varieties of English it may be not ungrammatical as a three-constituent structure.

15 Much of what has often been treated as sentence structure I am accounting for here at the rank of clause complex. If all sentences consist of a single (paratactic or hypotactic) clause complex, then there would be no reason to distinguish the two ranks; though if this were so the sentence would be, in my terms, a supplementary unit.

16 Nondefining relative clauses are probably best regarded as belonging to the dependent class even though they may contain an interrogative tag (interrogative being normally a property applicable only to independent clauses). Such a complex as They're going to Alford, which is near Skegness isn't it? seems clearly hypotactic. Note that in this analysis the relative clause is not rankshifted: Alford, which is near Skegness isn't it? is not a constituent at any place in the tree; defining relatives however are rankshifted to the element ‘q’ in nominal group structure, as in (36).

17 The theory requires that each unit should have at least one compound structure; cf. ‘Categories’ 256. Whether or not a basic unit has also one or more simple structures is a descriptive variable.

18 Where both the men and the women are old. If old applies only to men, then (48) is a group complex.

19 Or rather, to any item in the text. An item can be defined syntagmatically as that stretch which is a constituent at some layer in the tree. So that in He puts it on the table, s, put, table, the, the table, on, on the table, it, puts, he, he puts it on the table are all items, but it on, on the, etc. are not. For a discussion of the paradigmatic definition of item, see ‘Syntax and the consumer’.