9. M.A.K. Halliday¹

9.1 Michael Alexander Kirkwood Halliday was a pupil of Firth's, and, with greater elaboration, has pursued similar precepts, above all that ‘linguistics' should `deal with meaning' `at all levels of analysis' and should study ‘texts' ‘in contexts of situation' (cf. 8.46f vs. 9.8, 22f, 38, 49, 107). Halliday finds ‘the question “what is language?”‘ unduly ‘diffuse’ and ‘disingenuous’, because ‘no one account of language will be appropriate for all purposes’ (IF xxix, EF 9) (13.22). ‘A theory being a means of action’, we must consider what ‘action’ we ‘want to take’ ‘involving’ ‘language’, so we know what is ‘relevant’ and ‘interesting’ for ‘the investigation or the task at hand’ -- ‘the nature and functions of language’, its ‘formal properties’, its ‘role’ ‘in the community and the individual’, its ‘relation’ to ‘culture’, and so on (EF 9, IF xixf) (cf. 9.111; 11.6; 13.58). We may inquire ‘what all languages have in common’ or how they ‘differ’, and how they ‘vary according to user’ and ‘function’ or ‘evolve through time’; or ‘how a child develops language, and how language may have evolved in human species’. Or, we may explore ‘the quality of texts’ such as ‘written and spoken’ or ‘literary and poetic’ (IF xxx). Or, we may seek ways to ‘help’ people ‘learning their mother tongue’ or a ‘foreign language’, or ‘training translators and interpreters’, or composing ‘reference works (dictionaries, grammars)’ or ‘computer software’ to ‘produce and understand’ ‘text’ and ‘speech’. Or, we may focus on ‘language and the brain’ to help in ‘the diagnosis and treatment of language pathologies’ (‘tumours’, ‘autism’, ‘Down's syndrome’), or in the ‘design of appliances’ for ‘the hard of hearing’.

9.2 Halliday compares two ‘depths of focus’ in ‘linguistics': ‘the more immediate’ ‘intrinsic’ aim to ‘explain the nature of language’, ‘implying an “autonomous” view’; versus ‘the further, extrinsic aim to explain features of the social structure’ through ‘language’, ‘implying an instrumental’ view (EF 69). He ‘stresses the instrumentality of linguistics’ (EF 96), but the two views are not really separable. ‘Autonomy’ can be only ‘conditional and temporary’ along the way to a ‘general account of language’ (EF 53). The ‘linguist’ who ‘insists on autonomy’ studies ‘grammar and phonology’ as ‘the “inner” strata of the linguistic system, the core of language’, but these are ‘contingent upon other systems’ depending on ‘extra-linguistic phenomena’ (EF 96; cf. EF 105) (cf. 13.40). ‘Grammatical phenomena’ are ‘related’ ‘to features of a culture’ in ‘extremely complex and abstract’ ways; ‘linguists’ who ‘avoid the language/culture issue’ blot out ‘an important area of research’ (IF xxxi).² Moreover, ‘criteria’ of ‘well-formedness’ are ‘not easy to find’ ‘within language’; ‘in “autonomous” linguistics’, ‘orthography’, ‘a codified form of idealization’, ‘usually’ ‘decides’ (EF 68) (cf. 9.43, 82, 9⁴¹). But in ‘a theory’ of ‘social structure’, ‘what is well-formed’ is ‘interpretable as a possible selection within a set of options based on some motivated hypothesis about language behaviour’ (EF 69).

9.3 Halliday draws a different ‘basic opposition in grammars of the second half of the 20th century’ than the one featured in ‘the public debates of the 1960s’ between ‘“structuralist”‘ and ‘“generative”‘ approaches (IF xxviii). On one side he places ‘paradigmatic’ or ‘“choice” grammars’ -- ‘the functional ones, with their roots in rhetoric and ethnography’ -- ‘interpreting language as a network of relations with structures’ that ‘realize³ these relationships’, ‘emphasizing variables among different languages’, and ‘taking semantics as foundation; hence grammar is natural and organized around the text or discourse’ (IF xxviii, xiii, xix). On the other side he places ‘syntagmatic’ or ‘“chain” grammars'-- ‘the formal ones with their roots in logic and philosophy’ -- ‘interpreting language as a list of structures’ connected by ‘regular relations’ (shown as ‘transformations’), ‘emphasizing universal features of language, and taking grammar (which they call “syntax”) as the foundation of language; hence the grammar is arbitrary’ and ‘organized around the sentence’. Despite some ‘cross currents’ and ‘borrowing’ of ‘insights’, the two sides have found it ‘difficult to maintain a dialogue’ or ‘exchange ideas’ (IF xviii).

9.4 Chomsky ‘called his own syntagmatic formal grammar “generative”‘ to ‘distinguish’ it from the ‘“structuralists”‘, on whom he was ‘building’ despite his ‘polemics’ , and to suggest that his was ‘written in a way which did not depend on unconscious assumptions of the reader’ but ‘operated as a formal system’ (IF xxviii) (cf. 7.91). In return, ‘the language has to be so idealized that it bears little relation to what people actually write, and still less to what they actually say’. The so-called ‘Chomskyan revolution’ was more ‘a shift of emphasis’ ‘from the anthropological to the philosophical standpoint’ (cf. 7.1ff). ‘The return’ to ‘discourse in the 1970s’, however, ‘restored the balance’ and reinstated ‘the ethnographic tradition’ (Malinowski, Whorf, Pike), of which ‘Chomsky seems to have been unaware’ (IF xxviiif).

9.5 In its ‘extreme form’, ‘the philosopher's approach to language’ encourages ‘linguistics’ to ‘idealize out all natural language as irrelevant and unsystematic, and to treat only constructed logical languages’ (EF 53) (cf. 13.50). ‘A lesser extreme’ ‘reduces’ ‘all sentences of natural language’ ‘to a “deep structure”‘ of ‘logical relations’ (7.62-66). Chomsky's ‘idealization’, by ‘reducing them to the same level as stutterings, false starts, clearings of the throat’, ‘irons out’ ‘behaviourally significant variations in language’, e.g., ‘features of assertion and doubt’ (cf. 7.12; 9.51f). ‘In a sociological context’, ‘the image of language having a “pure” form (“langue”) that becomes contaminated in the process of being translated into speech (“parole”) is of little value’ (EF 67) (13.47). So Halliday discards the ‘boundary between language and speech’, ‘“langue” and “parole”, or competence and performance’ (cf. 2.20; 5.7; 6.33, 46; 7.12; 8.30; 11.69; 12.12, 26, 47, 55, 67; 13.36).

9.6 Against ‘the Chomskyan notion of competence’, Halliday's notion of ‘meaning potential ‘is defined’ ‘in terms of culture’, not ‘mind': ‘what speakers can do’ and ‘can mean’, not what they ‘know’ (EF 52f, 25, 55, 57, 72, 110). We ‘force a distinction between meaning and function’ if we ‘characterize language subjectively as the ability or competence of the speaker, instead of objectively as a potential or set of alternatives’ (EF 25) (cf. 12.45). ‘“Can do”‘ is ‘related’ to ‘“does”‘ ‘as potential to actual’, whereas ‘the relation between ‘“knows”‘ and ‘“does”‘ ‘is complex and oblique’ (EF 52f) (cf. 9.12; 13.39). ‘A hypothesis about what the speaker can do in a social context’ ‘makes sense of what he does’, which might otherwise ‘appear merely as a random selection’ (EF 67). We must ‘pay attention to what is said’ and ‘relate it systematically to what might have been said but was not’ (cf. 9.21; 13.43).

9.7 ‘Language is the primary means of cultural transmission’ whereby ‘behavioural options’ are ‘typically realized’, ‘social groups are integrated, and the individual is identified and reinforced’ (EF 45, 107, 8, 48, 69f) (cf. 3.1; 4.10; 8.28). Therefore, Halliday concurs with Firth (who ‘introduced the term “sociological linguistics”‘ ‘in 1935’ [in ‘The Techniques of Semantics’, P1 7-33] that ‘language as social behaviour’ is ‘an acknowledged concern of modern linguistics’ (EF 48f) (cf. 8.50; Halliday & Hasan 1985:8). His own ‘orientation is to language as social rather than individual’ and is ‘aligned’ with ‘sociological rather than psychological’ research’ (IF xxx, EF 53f) (cf. 8.17, 28; 9⁴⁷; 13.15).⁴ Studying ‘social man’ ‘shifts the emphasis from the physical to the human environment’ (EF 48) (cf. 8.23). ‘The individual is seen as the focus of a complex of human relations which collectively define the content of his social’ and ‘linguistic’ ‘behaviour’ (EF 48, 52).

9.8 Whereas the ‘“context of culture”‘ ‘defines’ ‘the potential’, i.e., ‘the range of possibilities’, the ‘“context of situation”‘ determines ‘the actual’, i.e., the ‘choice’ that ‘takes place’ (EF 49) (cf. 6.11; 12.55f). Firth's ‘interest’ was ‘in the actual, the text and its relation to its surroundings’ -- not however, ‘in the accidental but the typical': ‘repetitive, significant, and systematizable patterns of social behaviour’ (EF 49, 26, 40, 43) (cf. 8.26f). Hence, the ‘actual’ is not ‘unique’ or a ‘chance product of random observation’; and ‘the analysis of language comes within the range of a social theory’ and leads toward ‘an account of semantic options deriving from the social structure’ (EF 49ff, 64; cf. EF 62) (13.14).

9.9 Research therefore demands a ‘theory of social meanings’ -- a ‘socio-semantics’, a ‘meeting ground of two ideologies, the social and the linguistic’ (EF 44, 56, 64). Moreover, ‘a behavioural semantics’ is needed to map out the ‘intermediate levels’ which ‘relate behavioural options’ ‘to the grammar’ (EF 55, 83). In this view, ‘the meaning potential’ of ‘language’ ‘realizes behaviour potential’ and is ‘in turn realized in the language system as lexicogrammatical⁵ potential’ (‘what the speaker “can say”’) (EF 51, 55, 69). ‘Meaning is a form of behaving’, and ‘“to mean” is a verb of the “doing” class’ (EF 55) (cf. 9.15).

9.10 ‘The connection between’ ‘the social functions of language and the linguistic system’ is ‘clearest in the case of the language of the very young child’ (EF 34, 31). Those ‘functions’ ‘determine both the options the child creates for himself and their realizations in structure’ (EF 44, 33, 29). ‘Language development’ is thus ‘the mastery of linguistic functions': learning ‘the meaning potential associated’ with ‘the uses of language’, i.e., ‘learning how to mean’ (EF 24, 7) (cf. Halliday 1975). ‘Insights into how language is learned’ ought to shed light on ‘the internal organization of language’ (IF 45) (cf. 7.24). We already have ‘important work’ in the ‘theory of social meanings’ based on ‘the socialization of the child’, e.g. that of Basil Bernstein (EF 44f, 63; cf. EF 8, 18f, 48, 52, 64, 68ff, 73).⁶

9.11 In ‘learning his mother tongue, a child’ ‘is, in effect, learning new modes and conditions of being’ (EF 7). He ‘first tends to use language in just one function at a time’; ‘structure’ and ‘internal form reflect’ a given ‘function’ ‘rather directly’, and ‘the utterance has just one structure’ (EF 7f, 34, 27, 97, 44) (cf. 5.32; 11¹⁶). The ‘two-level system with meanings coded directly into expressions (sounds and gestures)’ gets ‘replaced, in the second year of life, by a three-level system’ with a ‘grammar’, whereby ‘meanings are first coded into wordings and these then recoded into expressions’ (IF xviif) (cf. 9.36f). ‘This step’ ‘opens up’ ‘the potential for dialogue, the dynamic exchange of meanings with other people’, and ‘for combining different kinds of meaning in one utterance -- using language to think with and to act with at the same time’. ‘Later’ ‘in the evolution’ ‘of the system’, the ‘child’ ‘learns the principle of “grammatical metaphor”‘, ‘whereby meanings may be cross-coded, and phenomena represented by categories other than those evolved to represent them’ (cf. 9.97ff). Eventually, in ‘adult language’, ‘utterances are functionally complex': almost ‘every linguistic act’ ‘serves several functions’ that ‘interact’ in ‘subtle and complex’ ways (EF 34, 8) (cf. 9.14, 25).

9.12 Following Malinowski in aligning ‘ontogeny’ with ‘phylogeny’, Halliday speculates that ‘the developing language system of the child traverses, or at least provides an analogy for, the stages through which language itself has evolved’, and thus ‘opens up’ ‘a discussion about the nature and social origins of language’ (EF 34, IF xviii) (cf. 13.38). Having ‘no living specimens of its ancestral types’, we can gather ‘evidence’ ‘from studying the language and how it is learnt by a child’ (EF 23f) (cf. 5³⁵; 7.24ff, 87). ‘To judge from children's “protolanguage”‘, ‘language evolved in the human species’ from ‘an early stage’ ‘without any grammar’, the ‘meanings’ being ‘expressed through rather simple structures whose elements derive directly from the functions’ (IF xvii, EF 97) (cf. 9.39, 72).

9.13 By this line of argument, ‘the lexicogrammar is a natural symbolic system’ (IF xviii) (cf. 9.3, 32). ‘Both the general kinds of grammatical pattern that have evolved in language and the specific manifestations of each’ ‘bear a natural relation to the meanings they have evolved to express’ (IF xvii). The early stage has a ‘relatively small range of meanings for which natural symbols can be devised’ (IF xviii). ‘In the later protolinguistic stage’, the ‘interface’ between ‘meaning’ and ‘sound’ ‘develops’ a ‘frontier of arbitrariness’ to make ‘communication’ less ‘restricted’; but the ‘interface’ between ‘meaning and wording’ ‘should not’ ‘become arbitrary’, since ‘such a system, by the time it got rich enough to be useful, would also have become impossible to learn’ (cf. 13.27). This account goes against ‘the psycholinguistic movement of 1960s’, ‘concerned primarily with the mechanism of language rather than with its meaning and function’, and focused on ‘the acquisition⁷ of sounds’ (‘articulation’, ‘phonology’) or of ‘linguistic forms (‘vocabulary’, ‘grammar’) (EF 24). Research measured ‘the size of the child's vocabulary’, ‘the relative frequency of different parts of speech’, plus ‘the control of sentence syntax in the written medium’. Later, work centred on ‘the acquisition of linguistic structures’ according to ‘the “nativist” view’ (cf. 7.22-28, 31f).

9.14 For ‘our conception of language’ to be ‘exhaustive, it must incorporate all the child's own “models”‘ (EF 17, 10). In ‘the instrumental model’, ‘language’ is ‘a means of getting things done’, and in ‘the regulatory model’, a means for ‘exercising control over others’ and ‘their behaviour’ (EF 11f, 31). In ‘the interactional model’, ‘language’ serves ‘the interaction between the self and others’ in ‘complex and rapidly changing’ ‘patterns’, and ‘defines and consolidates the group’ (EF 13). In ‘the personal model’, ‘the child’ becomes ‘aware of language as a form of individuality’ and of its ‘role’ in ‘the development of personality’ (EF 14). In ‘the heuristic model’, ‘language’ serves ‘to explore his environment’ and ‘investigate reality’, and in ‘the imaginative model’, ‘to create his own environment’ (EF 14f). ‘Finally’, in ‘the representational model’, ‘language is’ ‘a means of communicating about something, expressing propositions’, and ‘conveying a message’ with ‘specific reference’ to ‘processes, persons, objects, abstractions, qualities, states, and relations of the real world’ (EF 16) (cf. 13.24). ‘The ritual model’, with ‘language’ as ‘a means for showing how well one was brought up’, comes much later and ‘plays no part in the child's experience’ (EF 16f).

9.15 With all these facets, the child's total ‘“model” of language is highly complex’ (EF 11) ‘Most adult notions of language’, in contrast, even if ‘externalized and consciously formulated’, are ‘much too simple’, implying that ‘language’ is only for ‘transmission of content’ and that ‘the representational function’ is ‘dominant’ (EF 11, 16) (cf. 3.15; 8.47; 13.24). ‘We tend to underestimate the total extent and functional diversity of the part played by language in the life of the child’ (EF 11). Because, ‘for the child’, ‘language’ ‘has meaning in a very broad sense’ and ‘a range of functions which the adult does not normally think of as meaningful’, we have here a ‘vital’ domain for ‘redefining our notion of meaning’ to ‘include’ ‘all functions of language’ as ‘purposive, non-random, conceptualized activity’ (EF 18). ‘The young child’ ‘can be internalizing language while listening and talking’, and can ‘constantly ask questions’ to get ‘not merely facts’, but ‘generalizations about reality that language makes it possible to explore’ (EF 14f) (cf. 12.60). Also, ‘language in its imaginative function is not necessarily “about” anything’, not even ‘a make-believe copy of the world’; it may be for ‘pure sound’ and ‘linguistic play’ (EF 15f).

9.16 For effectual ‘language teaching’, ‘the teacher's own model of language’ should ‘encompass all that the child knows language to be’ and take account of the child's own linguistic experience’ in its ‘richest potential’ (EF 10, 19) (cf. 8.7). The model should also be ‘relevant’ to ‘later experiences’ and ‘to the linguistic demands of society’ -- where we are ‘surrounded’ not by ‘grammars and dictionaries or randomly chosen words and sentences’ but by ‘“text” or language in use’ in a ‘situation’ (EF 20) (cf. 11.86, 91; 13.31). ‘If the teacher's own “received” conception of language’ is ‘less rich’ and ‘diversified, it will be irrelevant to the educational task’, witness the ‘unhappy experience’ caused by ‘the view of language as primarily good manners’ (EF 10f, 19) (cf. 8.7). ‘In school’, ‘the child’ ‘is required to accept a stereotype of language contrary to insights’ from ‘his own experience’, as in ‘the traditional first “reading and writing” tasks’ (EF 11). ‘The old “see Spot run”‘ ‘reader’ ‘bore little’ ‘relation to any use of language’ (EF 12) (cf. 4.85).

9.17 Such issues are urgent because ‘educational failure is often’ ‘language failure’, due to ‘a fundamental mismatch between the child's linguistic capabilities and the demands’ being ‘made upon them’ (EF 18f) (Bernstein 1971-72) (cf. 4.85; 8.7). The problem lies not in ‘dialect or accent’, nor in ‘lack of words (vocabulary’ is ‘learnt very easily’ through ‘opportunity’ and ‘motivation’), nor in ‘an impoverishment of grammar’ or a ‘narrower range of syntactic options’ (EF 18).⁸ Instead, ‘the child’ ‘suffers some limitation’ in ‘linguistic models’ and some ‘restriction on the range of uses of language’; the ‘functions’ ‘have developed one-sidedly’, perhaps at the expense of ‘the personal function and the heuristic’, which do ‘not follow automatically from the acquisition of the grammar and vocabulary’ (EF 18f).

9.18 Halliday's concern for language development and pedagogy lends urgency to his broad social vision of language. He holds it to be ‘a universal of culture that all languages are called upon to fulfil a small set of distinct though related demands’ which, though ‘indefinitely many and varied’, are ‘derived ultimately from a small number of general headings’ (NT 3/207, EF 104f) (cf. 5.26, 84, 5²; 8.27). And it is ‘the nature of language’ to ‘have all these functions built into its total capacity’ such that ‘the social functioning of language’ is ‘reflected in’ ‘the internal organization of language as a system’ (cf. Malinowski 1923) (EF 23).⁹ Thus, ‘functional theories of language’ seek to ‘explain the nature’ and ‘organization’ of ‘the language system’ by asking which ‘functions it has evolved to serve’ ‘in the life of social man’ and how these are ‘achieved’ ‘through speaking and listening, reading and writing’ (EF 66, 42ff, 7).

9.19 Halliday now proposes a ‘functional grammar’ that can reveal how ‘the form of language’ is ‘determined by the functions’, and the ‘grammatical patterns’ by ‘configurations of functions’ (EF 7, IF x) (but cf. 3.16, 22, 24, 33; 4.47, 49; 7.63; 8.58; 9.28, 33; 12.25, 27, 50, 53-58). ‘Each element in a language is explained by reference to its function in the total linguistic system’; and most ‘linguistic items are multi-functional’ (IF xiii, 32, xxi) (9.25; cf. 13.43). Among the ‘many grammars that are functional in orientation’ is ‘systemic grammar’ (IF x).¹⁰ ‘Systemic theory’ ‘interprets’ ‘meaning as choice’ and ‘a language’ or ‘semiotic system’ ‘as networks of interlocking options’, in line with ‘Firth's category of the “system”‘ as ‘a functional paradigm'¹¹ and with his ‘polysystemic principle’ (IF xiv, xxvii, EF 55; cf. 8.30, 32). ‘A system is a set of features one’ ‘of which must be selected if the entry condition is satisfied’ (NT 1/37, EF 55). ‘Such a “selection expression” is then realized as a structure, the structural representation being fully derived from the systemic; each element of the structure is a point of entry into a further systemic network’ (NT 1/37). ‘Whatever is chosen in one system’ leads to ‘a set of choices in another’ as we move from ‘the most general features step by step’ to the ‘specific’ (IF xiv) .

9.20 But the Introduction does not ‘present’ ‘the systemic portion’ of ‘networks and realization statements’, which ‘is currently stored in a computer’ (IF x, xv) (cf. Mann & Matthiesen 1984). I would have liked to know more about the ‘semantic network’, for which extensive claims have been made. For language, it was to be ‘a statement of potential at that stratum’; ‘a representation’ of ‘paradigmatic relations’; ‘the input to the grammar’; and ‘a description of each meaning selection and an account of its relationship to all the others’ (EF 76, 79, 83). For human action, it was ‘the linguistic realization of patterns of behaviour’; ‘the bridge between behaviour patterns and linguistic forms’; and an ‘account of how social meanings are expressed in language’ (EF 79, 83, 65).

9.21 Instead, the Introduction presents ‘the structural portion’, ‘showing how the options are realized’ -- seeking ‘breadth before depth’ and not ‘making explicit all the steps’ (IF xv, x) (cf. 9.111). Though it ‘presents structures which are the output of networks’, the ‘grammar’ is not ‘“structural”‘ (nor ‘“structuralist”’ ‘in the American sense’), i.e. ‘syntagmatic’ (IF xvii) (cf. 8.31). ‘A systemic grammar is paradigmatic’; ‘describing something consists in relating it to everything else’ (9.7). The resulting ‘theory'¹² is ‘not parsimonious’, but ‘extravagant’, with ‘a wealth of apparatus’ (IF xix) (cf. 9.109f). The ‘grammar’ has ‘a round of choices and operations (a “system-structure cycle”) at each rank’; and ‘higher rank choices’ are ‘essentially choices in meaning without the grammar thereby losing contact with the ground’.

9.22 The Introduction propounds a ‘comprehensive view of grammar’ for ‘interpreting a text in its context of culture’ (IF xxxii, xvii) (cf. 8.91; 13.63). Every ‘interpretation of texts, of the system, and of elements of linguistic structure’ is based on ‘how the language is used’; ‘the uses’ ‘have shaped the system’ and must be studied, not just the ‘properties of the system as such’ (IF xxi, NT 3/207) (cf. 8.47; 13.36). But ‘whatever the final purpose’ or ‘direction’, the analysis must have ‘a grammar at the base’ (IF xvi) (cf. 13.54). ‘The study of discourse (“text linguistics”) cannot properly be separated from the study of the grammar’ (IF 345). Although ‘the text is a semantic unit, not a grammatical one’, ‘meanings are realized through wordings’ (i.e. through ‘sequences’ or syntagms’ of ‘lexical’ and ‘grammatical items’), and only a ‘grammar’ as ‘a theory of wordings’ allows one to ‘make explicit one's interpretation of the meaning of a text’ (IF xvii, xix, xxxivf, 19f; cf. 9.37f).

9.23 So ‘discourse analysis’ ‘provides a context within which grammar has a central place, and points the way to the kind of grammar required': ‘functional and semantic’, able not just to ‘characterize text in explicit formal terms’, but to ‘relate it to the non-linguistic universe of its situational and cultural environment’ (IF xvii) (cf. 9.109; 11.35). ‘The wheel has come full circle: when the mainstream of linguistics’ was in its ‘syntactic age’, Halliday ‘argued against grammar’ being ‘the beginning and ending of all things’; ‘now he insists on the importance of grammar’ lest ‘discourse analysis or text linguistics’ ‘be carried on without grammar’ (IF xvif). ‘A discourse analysis’ ‘not based on grammar is not an analysis at all, but simply a running commentary on a text’, appealing to ‘non-linguistic conventions’ or ‘trivial’ ‘linguistic’ ‘features’ ‘like the number of words per sentence’, whose ‘objectivity’ ‘is often illusory’.

9.24 Halliday himself proffers only ‘a thumbnail sketch’, ‘a minute fragment of an account of English grammar’ (IF 339, 286, 88, xiii). He grants that ‘many aspects of English’ should be ‘much more fundamentally reexamined’; ‘twentieth-century linguistics’ ‘has tended to wrap old descriptions’ inside ‘new theories’, whereas we really need ‘new descriptions’, e.g., ‘grammars for spoken language’ (IF xxxiv; cf. EF 57; 8.38, 67; 9.43; 13.8). Of course, no ‘account’ could be ‘“complete”‘ ‘because a language is inexhaustible’ (IF xiii) (13.22). We have ‘a finite body of text, written or spoken’, but ‘the language itself, the system’ ‘behind the text, is of indefinite extent’ (cf. 7¹⁶; 9.18; 13.43). Besides, ‘distinctions’ can be pursued only up to a certain ‘degree of fineness or “delicacy”‘ (IF xiii, 124; cf. IF xxvii, 286, EF 55, 58, 61, 75f, 94, IG 9).

9.25 Halliday also leaves it open if his ‘introduction to the functional grammar’ is just for ‘English’, or if it's a ‘general’ one ‘using English as the language of illustration’ (IF xxxiv; cf. NT 3/209). He admits the ‘danger’ of ‘ethnocentrism': now that ‘more has been written about English than any other language’, ‘modern linguistics’ ‘tends to foist the English code on others’ as if they were ‘imperfect copies’, as once was done with ‘Latin’ (IF xxxiii, xxxi) (cf. 2.5; 3.50; 4.4; 5.24; 6.5; 8.5, 14; 7.79; 12.20f; 13.53). For his own part, ‘those features’ ‘explicitly claimed as universal are built into the theory’, notably his ‘hypothesis’ that three ‘“metafunctions”‘ ‘organize’ ‘the content systems’ ‘in all languages’ (IF xxxiv) (9.47f). ‘But the descriptive categories are treated as particular’; ‘it is far from clear just how similar a pair of features in different languages should be in order to justify calling them by the same name’.

9.26 To ‘attempt’ ‘a “grammar” of English’, we need to ‘treat the system as a whole’ (IF 372) (13.46). If ‘grammar is not specialized according to language use’ but applies to ‘all texts’, we could ‘cover all functional varieties of the language’ (cf. 9.40). ‘The components of the grammatical system’ thus ‘represent the functions of language in their most generalized form, as these underlie all the more specific contexts of language use’ (EF 67). ‘Grammar’ enables us to ‘mean more than one thing at once’, ‘combines’ ‘functionally distinct meaning selections’ ‘into integrated structures’, and thus ‘turns meanings into text’ (IF xxxv, EF 67, 42, 92f, 100).

9.27 Such a broad outlook raises substantial ‘problems for a grammatical theory’ when we ‘write about language’ and ‘turn’ it ‘back on itself’ (IF xxv, xxxiiif; cf. 8.33; 13.48). First, ‘the whole grammatical system hangs together, and it is difficult to break in without presupposing’ ‘what is still to come’; ‘the discussion of any one system’ or ‘component’ may ‘require frequent reference to others’ (IF xxxiiif, NT 2/215, 3/180). Second, to determine ‘what is systematic and what is irrelevant in language’, we need to ‘decide what are different’ entities and what are ‘instances of the same’; and this ‘question’ ‘is not determined by the system’, but ‘by the underlying social theory’ (EF 53, 49f) (cf. 4.26, 31; 5.15, 61, 65). Third, ‘categories of grammar’ are hard to ‘gloss in exactly equivalent wordings’, because ‘they have evolved to say something that cannot be said any other way’ and are on ‘a purely abstract level of coding with no direct input-output link with the outside world’ (IF xxvi, xxxv) (cf. 3.23). ‘The best one can do is display them at work, in paradigmatic contexts, so as to highlight the semantic distinctions they enshrine’ (IF xxvi) (cf. 4.55; 8.82). Fourth, ‘until linguistics begins to meddle’, ‘spontaneous speech’ has an ‘unconscious nature’, ‘performed without thinking’ (IF xxivf) (cf. 2¹⁶; 13.49). Our ‘generalizations’ are ‘statements about what actually happens subconsciously in natural speech’ (IF xxvi, 272). Also, the ‘unconscious’ ‘slices of meaning’ which ‘the categories of our language represent’ may not ‘correspond to our conscious structuring of the world’ (e.g. ‘the gender system in English’) (IF xxv) (cf. 3.23; 13.24). And ‘a category only existing in the unconscious semantic system’ is hard to ‘define succinctly or even discursively’ -- it may even be ‘threatening’ to bring it to ‘consciousness’ But in ‘a theory’ of ‘social structure’, ‘what is well-formed’ is ‘interpretable as a possible selection within a set of options based on some motivated hypothesis about language behaviour’ (EF 69).(IF xxvi). Like ‘“tone deaf”‘, people may be ‘“grammar deaf”‘ and ‘fail to recognize’ ‘subtle semantic distinctions’ or ‘even deny that they are possible’.

9.28 Accordingly, grammar entails ‘severe’ ‘problems’ in selecting ‘labels’, which ‘become reified’ when we ‘forget how we arrived at them’ (IF xxxiif). The ‘two significant ways to label a linguistic unit’ are to ‘assign it to a class’ (e.g. ‘adjective and noun’), or to ‘assign a function to it’ (e.g. ‘Modifier and Head’) (IF 27). ‘If all the members of a class always had’ ‘only one function, it would not matter which sets of labels we used’. But ‘class labels’ are ‘part of of the dictionary’ and ‘indicate potential’; ‘functional labels are an interpretation of the text’ and ‘indicate the part’ actually ‘played’ in a ‘particular structure’ as well as the ‘relation to the system of language as a whole’ (IF 29, 31f, i.r.; 13.39) (cf. 9.6, 8f; 13.39).¹³ Here, ‘description and analysis should not be distinct and unrelated operations’, but should ‘proceed side by side’, revealing each ‘structure’ as a ‘meaningful’ and ‘viable configuration of functions’ (IF 32, 37).

9.29 For his own labels, Halliday undertakes to deploy ‘familiar categories’ and ‘terms in general use’, ‘redefined, in part, to fit in with the total picture’ (IF xxxiv, 28) (cf. 8.40; 13.48). ‘Most of the labels’ he uses are ‘functional’, signalled by ‘beginning with a capital letter’.¹⁴ He does ‘refer to classes’ ‘in the discussion’, but notes that many ‘are defined’ on ‘mixed’ ‘criteria’; he offers ‘generalized glosses designed to suggest the core meaning of the category’ -- ‘basic semantic motifs’ rather than ‘definitions’ (IF 27, 31, 202). We must acknowledge the ‘high degree of indeterminacy’ pervading ‘language in its categories’, ‘relations’, ‘classes’, ‘types’, and ‘tokens’ (EF 108, IF 31) (cf. 9.19, 35, 9³², 38; 13.59). ‘There rarely are any sharp lines in language, since it is an evolved system, not a designed one’, witness the many ‘fuzzy lines’ and ‘borderline cases’ in the book (IF xix, 318, 171; EF 33; cf. IF 163f, 186, 209, 219, 267, 327; EF 32, 112,; NT2/223, 3/196)(cf. 5.47; 9.73; 11.22).

9.30 Halliday also reconsiders the ‘traditional linguistic’ ‘terms used for the levels or “strata” of a language -- the stages in the coding process from meaning to expression’ -- such as ‘phonology’, ‘semantics’, and ‘grammar’ (IF xiv) (cf. 4.71; 5.34f; 7.45; 8.51f; 11.16f, 35, 56; 12.82; 13.29). Though ‘phonological’ is named alongside ‘grammatical’ and ‘semantic’ as a ‘level’ of ‘options in natural language’ (cf. 7.56), Halliday's main ‘concern’ is ‘with grammar’; and as a ‘general principle’, ‘only those distinctions’ shown to be ‘meaningful’ in ‘the grammatical description’ ‘are represented in the phonological analysis’ (EF 55, IF 17f, IG 47) (9.39f, 53). As we'd expect, he views ‘phonological structures, such as syllable and foot’, as ‘configurations of functions’, but the ‘options’ are seldom ‘directly’ specifiable as ‘output’ for ‘options in the grammar’ (EF 94f). So aside from a gloss on the ‘phonometric structures’ of ‘spoken verse’ (IF 10-16; cf. Note 15), IF is concerned only with tone and key, which are crucial for signalling prominence in clauses and clause complexes (9.53). Hence, Halliday's model is among the few that did not treat phonology as the basic system and work from the smallest units (phonemes) on upward to grammar, stressing structure and constituency over function and meaning (cf. 13.27).

9.31 ‘Formal linguistics’ ‘replaces “grammar”‘ with ‘“syntax”‘, following ‘the philosophy of language, where syntax is opposed to semantics’ and ‘pragmatics’ is ‘a third term’; but in ‘functional linguistics’, ‘grammar consists of syntax and vocabulary’, plus ‘morphology’ ‘in languages which have word paradigms’ (IF xiv, EF 93) (cf. 9.4, 6f; 9⁶, 9¹⁴; 2.55; 13.28). Within the ‘direction’ in ‘Western linguistics’ since ‘ancient Greece’, the term ‘“syntax”‘ ‘suggests’ that ‘language is interpreted as a system of forms to which meanings are then attached’ (IF xiv) (cf. 13.54). ‘In functional grammar’, ‘the direction is reversed: a language is interpreted as system of meanings accompanied by forms through which the meanings are realized’. ‘The forms’ are a ‘means to an end’, not ‘an end in themselves’; we ask not ‘what do these forms mean?’ but ‘how are these meanings expressed?’ (IF xiv, 320).

9.32 Even so, Halliday says ‘grammar is the level of formal organization in language’ -- a ‘purely internal level’ and ‘the main defining characteristic of language’ (EF 98) (cf. 13.54). Yet ‘it is not arbitrary’; it is ‘natural’, having ‘evolved as “content form”‘ on ‘a functional basis’ (cf. 2.28ff; 9.3, 13, 35f, 51; 13.27). As the ‘complexity’ of both ‘linguistic function’ and ‘language’ increased, ‘the stratal form of organization’ ‘emerged’, ‘with a purely formal level of coding at its core’ to ‘integrate’ ‘complex meaning selections’ into ‘structures’ by ‘sorting specific uses of language into a small number of highly general functions’ (9.11ff, 37).

9.33 Halliday rejects ‘immediate constituent analysis’ by ‘maximal bracketing’, which ‘never allows more than two elements in a bracket’, in favour of ‘ranked constituent analysis’ by ‘minimal bracketing’, which takes as ‘constituents only those sequences that actually function as structural units in the item’ (NT 1/37, IF 22, 24, 26, 30) (cf. 4.59; 5.21, 50, 62; 7.37f, 63, 86; 13.26). In ‘trying to explain as much of grammar as possible in terms of constituent structure’, the ‘maximal’ way suggests an ‘order in which elements of a string are combined’, with some being ‘more closely bonded’; ‘it says nothing about the function’ of ‘any of the pieces’ (IF 30, 25, 22, 26f). ‘Function’ must be shown by ‘labelling parts’ and ‘nodes’ (in ‘the tree metaphor’) to ‘indicate’ the ‘configuration’ and ‘explain the value in relation to the whole’ (IF 27).

9.34 ‘A scale of rank for grammar’ can be ‘defined’ by ‘adopting’ ‘sentence, clause, group, word, and morpheme as a strict hierarchy of constituents’ (IF 25) (13.29).¹⁵ But ‘language’ ‘embodies a multiplicity of constituent hierarchies, coexisting in different parts of the system’ (IF 18) (cf. 5.36f, 39f).¹⁶ ‘Units of different rank tend to carry patterns of different kinds’; and ‘the functional specification of units of different functions is of fundamental significance in determining grammatical structure’. Almost every ‘constituent enters into more than one structural configuration’ ‘at a number of levels simultaneously’, and ‘has more than one function at a time’ (IF 32, 271, EF 44) (9.19, 59; cf. 13.57). ‘The choice of a word may express one type of meaning, its morphology another, and its position’ yet ‘another’ (EF 42) (cf. 9.37; 12.43, 72). The upshot is ‘infinite possibilities of matching them up in meaningful ways’ (IF 18) (cf. 9.18, 24).

9.35 Moreover, ‘since the relation of grammar to semantics’ is ‘natural, not arbitrary, and both are purely abstract systems of coding’, ‘there is no clear line between’ them; ‘functional grammar’ is ‘pushed in the direction of semantics’ (IF xix, xvii). ‘In principle, a grammatical system is as abstract (is as “semantic”) as possible given only that it can generate integrated structures’, i.e., ‘its output can be expressed in terms of functions mapped directly onto others’ to yield ‘a single structural “shape”‘ that is ‘multiply labelled’ (EF 95) (cf. 9.44). In this manner, ‘the combination of system and structure with rank leads’ to a ‘grammar’ whose ‘abstractness’ (‘“depth” in the Chomskyan sense’) we can ‘specify fairly accurately in theoretical terms’.

9.36 Although we cannot ‘spell out all the steps from meaning to wording’, we should recognize the ‘principle’ ‘that all categories employed must be clearly “there” in the grammar of the language’, ‘not set up simply to label differences in meaning’ (IF xx) (cf. 9.62). Without some ‘lexicogrammatical reflex’, such ‘differences’ are not ‘systemically distinct in the grammar’. However firmly ‘based on meaning’, ‘a functional grammar’ is ‘an interpretation of linguistic forms': ‘every distinction’ -- ‘every set of options, or “system”‘ -- must ‘make some contribution to the form of the wording’ (IF xx, xvii) (cf. 9.22; 13.54). So ‘grammar’ is ‘a theory of wordings’, which ‘are purely abstract pieces of code’ to be ‘recoded in sound or writing’ before you can ‘see or hear them’ (IF xx, xvii, xix). This ‘recoding’, and not ‘the relation between the meaning and the wording’, is the domain of ‘arbitrariness’ (IF xviif).¹⁷

9.37 In the ‘grammar’, ‘meanings are accepted from different metafunctional inputs and spliced together to form integrated outputs or wordings’ (IF xxxivf) (cf. 9.32, 35). ‘The wording “realizes” or encodes the meaning’ and is ‘in turn “realized by” sound or writing’ (IF xx). We needn't ask ‘which determines which’ or what ‘each symbol as an isolate’ ‘means’; ‘the meaning is encoded in the wording as an integrated whole’. But in ‘a theory’ of ‘social structure’, ‘what is well-formed’ is ‘interpretable as a possible selection within a set of options based on some motivated hypothesis about language behaviour’ (EF 69).(IF xxvi). Like ‘“tone deaf”‘, people may be ‘“grammar deaf”‘ and ‘fail to recognize’ ‘subtle semantic distinctions’ or ‘even deny that they are possible’. ‘The choice’ of an ‘item’, its ‘place in the syntagm’, ‘its combination’ with another, and its ‘internal organization’ may each have a ‘meaning’; ‘the grammar’ ‘sorts out these possible variables and assigns them to their specific semantic functions’ (cf. 9.34).

9.38 Therefore, ‘a language’ ‘is a system for making meanings: a semantic system with other systems for encoding the meanings it produces’ (IF xvii) (cf. 8.46f). Halliday warns that ‘everyday terminology’ and ‘“meaning” in its lay use’ may ‘imply’ that many ‘areas of syntactic choice are not meaningful’; and he discards the term ‘content’ because it ‘calls to mind’ the ‘irrelevant’ ‘form/content opposition’ (NT 3/209).¹⁸ ‘“Semantics” does not simply refer to the meanings of words’ but to ‘the entire system of meanings’ ‘expressed by grammar’ and ‘vocabulary’ (IF xvii). And ‘semantics’ is ‘a stratum’ ‘intermediate between the social system’ (‘wholly outside language’) and ‘the grammatical system’ (‘wholly inside language’) (EF 96). . But though, we cannot give ‘a general account of English semantics’, but only for ‘a particular register’ or ‘body of text’ (IF 372, xx) (cf. 9.105; 13.31).

9.39 ‘Semantic systems’ ‘relate’ to ‘grammatical systems’ through the ‘pre-selection’ of ‘options’ (EF 98). Due to ‘indeterminacy between the strata’, we usually find not ‘one-to-one correspondences’ between ‘grammar’, ‘semantics’ and ‘phonology’, but rather ‘neutralization and diversification’ -- ‘many-to-many’ (cf. Lamb 1970) (EF 82, 93, 56f) (cf. 3.32; 5.48, 64; 13.54). ‘In some instances’, however, we can go from ‘semantics’ ‘directly to the “formal items”: to the actual words, phrases, and clauses of the language’, with ‘no need’ for ‘grammatical systems and structures’ (EF 83ff) (cf. 9.11f). True, this ‘happens only’ with ‘a closed set of options in a clearly circumscribed social context’, e.g., ‘a greeting system in middle-class British English’ or in a ‘closed transaction such as buying a train or bus ticket’ (EF 83f). In such cases, ‘the formal items’ ‘are rather like non-linguistic semiotic systems’, e.g., ‘traffic signs and care labels on clothing, where the meanings are directly encoded into patterns in the visual medium, with a minimum of stratal organization’. In genuine ‘language, such systems are marginal’, ‘a small fraction of the total phenomena’ among ‘much more open’ and ‘general settings’.

9.40 For similar reasons, it is not clear how far ‘grammatical and lexical properties of sentences used by the speaker in the speech situation’ ‘can be “predicted” from a semantics of behaviour based on social context’ (EF 90) (cf. 13.40). We will ‘find’ ‘a direct link’ between ‘features of the social structure’ and ‘forms of the language’ only ‘in odd cases’, e.g., ‘phonological variables’ of ‘dialect’ and ‘“accent”‘ (EF 65) (cf. Labov 1968). We can ‘specify’ ‘general’ or ‘principal grammatical features’ and ‘narrow down’ the ‘lexical set’ by ‘exploring areas of behaviour where the meanings are expressed through very general features’ ‘involved in nearly all uses of language’ (EF 91, 84) (cf. 9.26). But we might ‘not go very far in delicacy’, and might have to use ‘favourable instances’ and ‘restricted types of situation’, not ‘the whole of an individual's language behaviour’ (EF 92, 62). We would cover ‘only a small proportion’ ‘of the total’ ‘speech by educated adults in a complex society’ (EF 92).

9.41 Along another dimension of language versus context, ‘grammar is at once both’ ‘of the system’ and ‘of the text’ (IF xxii) (cf. 6.34; 13.39). Halliday follows Saussure's view of ‘the relationship between the system of language and instantiation in acts of speaking’, but not his ‘conclusion’ that ‘the texts can be dispensed with’ after ‘being used as evidence for the system’ (IF xxii) (2.19f; 13.36). ‘This mistake’ ‘haunted linguistics for much of the twentieth century’, ‘obsessed with system at the expense of text’, up to ‘the present swing’ ‘in the opposite direction’ (9.4).¹⁹ ‘An elegant theory of the system’ has ‘little use’ ‘if it cannot account for how the system engenders text’. So ‘discourse analysis must be founded on the study of the system’, which in turn ‘throws light on discourse’ and shows ‘the text’ ‘as process’.

9.42 ‘The ‘experience’ of a ‘process’ with ‘a continuous flow, without clear segments or boundaries’ -- ‘as text (mass noun) rather than as a text/texts (count noun)’ -- is best found in ‘speech’ (IF xxiii). ‘Speech’ is ‘important’ not just because ‘it comes first in the history of the race and the individual’, but because in it ‘the potential of the system is more richly developed’ and ‘fully revealed’, ‘its semantic frontiers are expanded, and its potential for meaning is enhanced’ (IF xxiiif, 201) (13.33). ‘Speech’ ‘responds continually’ to ‘subtle changes in the environment, both verbal and nonverbal’, and ‘exhibits a rich pattern of semantic’ and ‘grammatical variation’ not ‘explored in writing’ (IF xxiv). ‘Spoken language’ can ‘“choreograph” very long and intricate patterns of semantic movement while maintaining a continuous flow of discourse that is coherent without being constructional’ (IF 201f). These ‘systems that vary the form of the message’ get ‘neglected in grammars of English’ ‘because they are much less richly exploited in written language’. What ‘writing’ ‘achieves’ by ‘packing together lexical content’ in a ‘static and dense’ way, ‘speech’ ‘achieves’ through ‘grammatical frames’ in a ‘mobile and intricate’ way. That ‘spoken language is disorganized and featureless’ is a ‘folk belief’, sustained by ‘transcriptions’ in which ‘speech’ ‘looks silly’ ‘written down’, due to ‘the disorder and fragmentation’ in ‘the way it is transcribed’ without ‘intonation or rhythm or variation in tempo and loudness’ (IF xxiv).

9.43 Besides, ‘it being much harder to represent a process than’ ‘a product’, ‘the text’ is easily viewed as a ‘perceptible’ ‘object’, made most tangible ‘as a piece of writing’ (IF xxiif, 290) (cf. 9.2). And ‘traditionally, grammar’ has been ‘product grammar’ for ‘written language’ (IF xxiii). ‘In its earliest origins, classical Greek grammar’, ‘tied to rhetoric’, was for ‘speech’; ‘but Aristotle took grammar’ ‘into logic’, and focus shifted to ‘written discourse’ through ‘medieval and renaissance syntax’ up to the ‘received “traditional grammar”‘ of ‘today’ (8.17; 13.33). The latter is ‘unsuited to spoken language, which needs a more dynamic and less constructional form of representation’ (cf. 9.24). Yet ‘constructing’ a new ‘grammar for speech’ ‘from the beginning’ might ‘force an artificial polarization of speech versus writing’, deny ‘mixed categories’ (like ‘dramatic dialogue’), and make it ‘difficult to compare spoken and written texts’.

9.44 So Halliday calls for ‘a much more dynamic model of grammar’ he does not ‘offer here’, but hopes his treatment of ‘the clause complex’ may go ‘a little way in that direction’ (IF xxiii). He designates the ‘clause’ ‘the most significant grammatical unit’ and ‘the best example’ of ‘linguistic structure’ as ‘a means for the integrated expression of all the functionally distinct components of meaning in language’ (IF 101, EF 42) (cf. 5.55; 9.46ff, 50f). ‘The grammar of the clause’ ‘expresses’ ‘the semantic system of the language’, which in turn ‘sorts out’ the ‘“goings-on”‘ of ‘reality’ (IF 101) (13.24).

9.45 This process involves variations in ‘markedness’. ‘Within a systematic framework’ of ‘options’, the ‘typical form’ is ‘unmarked “with respect to” some other option’ (EF 58f, 79). Though ‘usually less frequent than an unmarked one’, ‘a marked option’ need not be ‘rare’; its ‘effectiveness’ comes from ‘contrasting with unmarked’, which is ‘less motivated than others in the same system and therefore selected unless there is specification to the contrary’ (NT 2/219, 213) (13.43). Halliday ‘invokes the “good reason” principle': ‘the “unmarked” one’ ‘is chosen unless there is good reason to choose otherwise’ (EF 57, IF 45, 287).

9.46 We can get some flavour of his approach from his treatment of the ‘Subject’, which so far lacks ‘a definitive account’ despite being a ‘basic’ ‘concept’ ‘in the Western tradition of grammatical analysis’ (IF 32f) (cf. 4.69; 5.55; 6.49; 7.63; 9.70; 11.34; 12.70, 79).²⁰ He sees ‘three broad definitions’ of ‘Subject': ‘the concern of message’, ‘the doer of the action’, and ‘that of which something is predicated’ (IF 33f, 102). Though these ‘definitions are obviously not synonymous’, they were usually treated as ‘aspects of one and the same general notion’ by sticking to ‘idealized clause patterns’ wherein they ‘coincide’ -- namely, in ‘the typical unmarked form’ of ‘the English declarative (statement-type) clause’ (IF 33f, 36, 77, NT 2/213) (cf. 4.68; 9.75, 109; 11.68; 13.50). But a full ‘account of natural living language’ requires that they be ‘interpreted’ as ‘three’ ‘distinct functions’, ‘subtly but significantly different in meaning’, which Halliday calls ‘Theme’, ‘Actor’, and ‘Subject’, respectively (IF 35f).

9.47 This trinity brings us to the centrepiece of Halliday's approach. He sees ‘the basis of the grammatical system’, and a ‘universal’ ‘feature’ of ‘language’, in a triad of ‘metafunctions’ -- ‘tendencies worked out differently in every language but clearly discernible in all’ (EF 66, IF xxxiv, 169, NT 2/243) (cf. 9.25). These contribute ‘three distinct principles of organization in the structure of grammatical units’ (‘as described by’ ‘the Prague school’, e.g. Danes 1964) ‘expressing three rather distinct and independent sets of underlying options’ (EF 66, IF 158). ‘Intersecting’ in ‘the clause’, these sets express ‘experiential meaning, speech function, and discourse organization’ (NT 2/243). Halliday's ‘grammar’ is extensively organized around this triad and the threes related to it (cf. IF xiii, xviif, 33, 35, 37f, 53, 78, 101, 128, 158ff, 321; NT 2/199 243, EF 66, 99, 105f) (and a photo of Henry Moore's sculpture of ‘Three Points’ appears on the cover of IF). Indeed, he expressly warns us when some aspect or structure is not seen these three ways (IF 158, 169, 176) (cf. 9.74, 78).

9.48 The three ‘metafunctions’ are ‘the textual’, ‘the ideational’, and ‘the interpersonal’ (EF 66, 99, 101, 105ff, IF xiii, xxxiv, 53, 158, NT 3/209).²¹ Through them, ‘three distinct structures are mapped onto one another to produce a single wording’, such that ‘the clause’ is ‘the simultaneous realization’ of three ‘meanings': a ‘message’ (‘meaning’ as ‘relevance to the context’), a ‘representation’ (‘meaning in the sense of “content”’), and an ‘exchange’ (‘meaning as a form of action’) (EF 42, IF 36f, 158, 53, xiii). The three aspects are not ‘discrete’ ‘components’ ‘expressed’ by ‘segments’ ‘we can point to’ (EF 42). They tend to be ‘embodied’ or ‘scattered’ throughout an ‘entire structure’, wherein they are ‘mapped onto one another’ in a `simultaneous’ ‘complex of structural roles’; a certain ‘alignment of roles’ may ‘represent a favourite clause type’ (IF 169f, 365; NT 2/215f, 243, 224).²² In terms of ‘the triad first proposed by Pike’ (5.31ff), Halliday envisions ‘textual meaning’ as ‘a wave-like pattern of periodicity’, with ‘peaks of prominence and boundary markers’; ‘experiential’ (i.e. ideational) ‘meaning’ as a ‘particle-like’ pattern of ‘building blocks’; and ‘interpersonal meaning’ as a ‘field-like’ ‘prosodic’ ‘pattern’ (IF 169).

9.49 ‘The textual’ metafunction enables ‘language’ to be ‘operationally relevant’ and ‘have texture in real contexts of situation’ (EF 42) (cf. 8.41). Here, ‘language becomes text, is related to itself and to its contexts of use’, including ‘the preceding and following text, and the context of situation’ (EF 44, IF 53). One division under the ‘textual’ heading is ‘information': ‘a process of interaction between what is already known or predictable and what is new or unpredictable’ (IF 274f).²³ ‘The information unit is a structure made up of two functions, the New and the Given’. The ‘“New”‘ is either ‘not previously mentioned’ or is ‘presented’ by ‘the speaker’ as ‘not being recoverable from the preceding discourse’ and thus as ‘textually and situationally non-derivable’; the ‘“Given”‘ is ‘what is not “New”‘ (NT 2/204f, 211; IF 277). ‘The idealized form’ of the ‘information unit’ has ‘a Given’ and ‘a New element’, but does not apply when ‘a discourse’ ‘starts’ nor when ‘the Given’ ‘refers to something already present in the verbal or non-verbal context’ (IF 275). So only the ‘New element’ is ‘obligatory’, while the ‘Given’ is ‘optional’ (IF 275, NT 2/204).

9.50 ‘The structure of the information unit contributes in large measure to the organization of discourse’, and ‘frequently defines’ ‘the domain of constituents’ more than does their ‘status in sentence structure’ (NT 2/210f). ‘The distribution of information specifies a distinct constituent structure’ on one ‘plane’, which is ‘then mapped onto the constituent structure’ of ‘clauses’ (NT 2/200, 242). This ‘distribution’ ‘determines’ the number of ‘points of information focus’ and ‘represents the speaker's blocking out of the message into quanta of information, or message blocks’, and deciding ‘where the main burden of the message lies’ (NT 2/202, 204). A ‘discourse’ with ‘much factually new material’ tends to have many ‘short information units, each with its focus’ (NT 2/205).

9.51 ‘Within each’ ‘unit’, ‘elements’ are selected as ‘points of prominence': ‘one primary point of information focus’, and possibly a ‘secondary’ one for ‘dependent’, ‘incomplete, contingent, or confirmatory’ ‘information’ (NT 2/203, 209). The ‘structure’ is ‘realized’ in a ‘“natural” (non-arbitrary)’ way, with ‘the New marked by prominence’ and ‘typically’ placed after ‘the Given’ (IF 275) (11.85). In ‘the unmarked or default condition’, ‘the information unit’ ‘corresponds to a clause’, but may be ‘more’ or ‘less’ in ‘marked’ cases (NT 2/201, 203, IF 59, 274, 287, 315, IG 19f).²⁴ Any ‘unit’ ‘less’ or ‘more’ than this is ‘marked’; ‘in continuous informal discourse’, ‘the average number of information units per clause lies between 1 and 2’ (NT 2/201).

9.52 ‘In any utterance in English, three distinct meaningful choices’ are made, which ‘usually are subsumed under’ ‘“intonation”': ‘“tonality”, “tonicity”, and “tone”‘ (IG 18, 30, 38). Such options suggest that a ‘notation’ is needed for ‘showing intonational and rhythmic structure’ which ‘has to be accounted for in a functional grammar’ (IF 286) (cf. 8.71; 9.42). ‘Discourse consists of a linear succession’ of ‘information units, realized by tonality, that is, as a sequence of tone groups’ (NT 2/211; IG 30; cf. IF 8, 59, 271, 273f). Within ‘each information unit’, the ‘choice’ of ‘focus’ is ‘realized by tonicity, the structuring of the tone group into a tonic’ (for ‘the general meaning’) ‘optionally preceded by a pretonic’ (for ‘more delicate distinctions’) but with ‘no separate post-tonic’ (NT 2/211, 205, 243, IF 283, IG 12f, 30).

9.53 ‘Tone’ concerns ‘phonological prominence’ allotted by ‘pitch movement’ and to a lesser degree by ‘duration’ and ‘intensity’ (NT 2/203, IG 14). In each ‘tone group’, ‘prominence’, and thus ‘information focus’, is given to ‘the element’ that ‘carries the main pitch movement: the main fall, or rise, or change of direction’ (IF 275). ‘The English tone system is based on an opposition’ between ‘falling’ and rising pitch’, and ‘the choice of tone’ yields the ‘semantic values of key’ (IF 281, IG 16f, 30).²⁵ This ‘opposition’ is so ‘fundamental’ that it ‘probably plays a part in the system of every language’ (IF 281). In English, ‘falling pitch’ means ‘certain’, while ‘rising pitch’ means ‘uncertain’ (IF 281f). Correspondingly, ‘falling-rising means “seems certain, but turns out not to be”‘; and ‘rising-falling means “seems uncertain, but turns out to be certain”‘. The ‘neutralized’ ‘level tone’ means ‘“not (yet) decided whether known or unknown”‘. ‘In normal conversational English’, ‘falling tone’ ‘is most frequent’, followed by ‘falling-rising’. ‘Rising tone’ is ‘more common in dialogue than in narrative’; ‘in formal speech and loud-reading’ ‘level tone’ ‘increases’; and ‘rising-falling tone’ is ‘characteristic of children's speech’.

9.54 As might be expected, ‘the unmarked realization of a statement’ ‘in a declarative clause’ ‘is falling tone’ (IF 281, 284, IG 25). The ‘other tones convey a statement’ with ‘additional’ ‘features’, e.g., ‘rising tone’ for ‘contradiction or protest’, ‘falling-rising’ for ‘reservation’, and ‘rising-falling’ for ‘strong assertion’ (IF 284, 281). A ‘falling tone’ is used for a ‘WH-question’ (in the sense of 9.58), and a ‘rising’ one for a ‘yes-no question’ (IF 281, 284). The ‘imperative’ has ‘two unmarked tones': ‘falling’ ‘for command’, and ‘level’ ‘for invitation’ (IF 284). ‘Minor clauses’ (in the sense of 9.74) ‘have varied tones depending on their function’, especially ‘calls (vocatives)’ (IF 285).

9.55 ‘Theme’ -- the other aspect of textual meaning alongside ‘information structure’ -- ‘is concerned’ with ‘the status of elements’ as ‘components of a message’ (NT 2/199; IF 36, 38f).²⁶ ‘The interplay of thematic and information structure carries the rhetorical gist’ (IF 280). Yet whereas ‘information’ and the ‘Given-New’ ‘dimension’ ‘determine the organization’ of ‘a text’ ‘into discourse units’, and whereas ‘information options’ are not ‘clause systems’, ‘theme’ affects the ‘organization’ and ‘sequence of elements of the clause in sentence structure’ (IF 287, NT 2/200, 223). ‘The choice of information focus’ ‘expresses the main point’ ‘of the discourse’; in ‘contrast’, ‘the choice of theme, clause by clause’, ‘carries forward the development of the text as a whole’ (IF 315). Whereas ‘information focus’ ‘favours the more “peripheral” elements, especially circumstances’, ‘thematic prominence’ favours ‘the more “central” among the clause elements (the participants’ in ‘the most active roles in transitivity’, cf. 9.60) (NT 2/214). Also, ‘information structure’ is ‘listener-oriented’ (‘“what I am asking you to attend to”’), whereas ‘thematic structure’ is ‘speaker-oriented’ (‘“what I am talking about”’) (IF 278, 316, 368). For all these reasons, ‘information’ and ‘theme’ are ‘independently variable': being ‘combinable in only one way’ would ‘curtail the potential of these two systems’ and remove an occasion for ‘meaningful choice’ (NT 2/205, 211f, IF 287) (cf. 13.50).²⁷

9.56 As shown by ‘the linguists of the Prague tradition’, who ‘explored’ ‘functional sentence perspective’ (cf. Vachek 1966; Firbas & Golkova 1976), ‘the sequence of elements in the clause tends to represent thematic ordering’ rather than ‘actor-action-goal’ (NT 2/205, IF 315; EF 107) (cf. 4.68; 7¹⁴; 9.46, 67). ‘Thematization’ ‘assigns to the clause a structure in terms of the functions “Theme” and “Rheme”‘ (NT 212). ‘If a clause is structured as two information units, the boundary’ ‘nearly always coincides with that between Theme and Rheme’ -- ‘a strong piece of evidence for construing the Theme’ this way (IF 40, 56). Since ‘Theme always precedes Rheme’, ‘the unmarked case’ ‘associates’ ‘the Theme with the Given’ and places ‘the focus of information’ (and ‘the New’) ‘within the Rheme’, though not always ‘extending over the whole of it’ (NT 2/205, 212, IF 60, 278). The ‘Theme’, ‘usually marked off as a tone group’, may ‘consist of just one element’ or of ‘two or more forming a single complex element’ (IF 40f).

9.57 ‘Typically’ ‘in a declarative clause’, ‘the Theme is conflated with the Subject’ -- a ‘mapping’ yielding ‘the unmarked Theme’ (IF 44f, 60) (cf. 9.46; 11.68). ‘In everyday conversation, the item most often’ used as such is ‘the first person pronoun “I”‘; ‘next’ ‘come the other’ ‘pronouns’, and ‘then’ ‘nominal groups’ (with ‘common’ or ‘proper noun as Head) and nominalizations’ (IF 45). In contrast, ‘a Theme’ ‘other than the Subject in a declarative clause’ is ‘marked’, ‘usually an adverbial group’ or a ‘prepositional phrase’ ‘functioning as adjunct in the clause’. ‘Thematic status’ makes ‘adjuncts’ of ‘time/place’, ‘cause, manner, etc.’ into ‘sentence adjuncts’; ‘their domain may extend over various levels of sentence structure’ (NT 2/220). Such cases show how ‘theme plays a part in the bracketing function of information structure’; if a ‘marked thematic element’ ‘occurs as a separate information unit’, its ‘domain extends over the whole of the next following information unit’, but if not ‘separate’, only over the ‘unit in which it occurs’ (NT 2/219f) (cf. 9.24). Also, ‘adjuncts’ ‘occurring obligatorily in initial position’ because they have ‘floated to the front of the clause’ during ‘the evolution of the language’, ‘do not take up the whole’ position of ‘theme’; the next element functioning as ‘Subject, Complement, or Adjunct’ is included too (IF 51, 53, i.r.).

9.58 ‘Interrogative clauses’ ‘embody the theme principle in their structural make-up': the ‘theme’ ‘element’ ‘comes first’ because of ‘the thematic significance attached to the first position in the English clause’ (IF 47f). In a ‘polar’ (‘yes/no’) question’ that ‘element’ is ‘the finite verb’, but in a ‘non-polar’ or ‘content’ (‘WH-’) ‘question’ (with ‘“who”, “what”, “when”, “how”, etc.’) it is ‘the element that requests’ ‘a missing piece of information’ (IF 47, 44, 85, NT 212f) (cf. 4.69). ‘The preference for the “inverted” interrogative structure in English’ confirms the ‘importance of thematic organization by sequence in the syntax of the English clause’; ‘interrogatives have a built-in unmarked theme’ (NT 2/214). ‘In the declarative the thematic pressure on the subject is much less strong, and marked themes are frequent in all registers’ to ‘foreground the speaker's point of departure’ (NT 2/215). The ‘imperative’, however, ‘commonly’ ‘with no subject or finite verb’, has ‘no explicit theme’; ‘the basic message is realized simply by the form of the clause’, which ‘consists of rheme only’ (IF 49) (cf. 9.52, 72, 96) (4.56).

9.59 ‘The ideational'²⁸ metafunction has a ‘vast and complex’ ‘meaning potential’ (EF 39). In Halliday's earlier scheme, it was two ‘separate’ metafunctions (‘components’), ‘the experiential’ and ‘the logical’ (NT 3/209, EF 106). Maybe he merged them because he likes threes (9.47), but his concern remains for the ‘basic logical relations’ in ‘natural language’, such as ‘in a univariate structure’ with a ‘recurrence of the same function’ and in a ‘multivariate structure’ with a ‘constellation’ of distinct functions’ (EF 66, IF 193, 172) (cf. 9.75, 80, 82f). Since these ‘relations’ form ‘part of the semantics of a language’, they do not ‘fit exactly into non-linguistic logical categories’ (understatement) -- ‘although since the latter derive from natural language in the first place there will obviously be a close resemblance’ (overstatement) (IF 202). However, Halliday frequently criticizes the reliance on ‘logic’ by linguists or grammarians, as when he warns against ‘problems’ ‘arising in linguistic analysis’ by ‘attempting to make the logical structure do duty for the other components’, or opines that ‘the logical element in the description of the clause appears to be, in English, entirely dispensable’ (NT 3/211f) (cf. 9.3, 5, 48; 8.5, 17; 13.17f).

9.60 The ‘experiential’ aspect covers ‘the interpretation and expression in language of different types of process of the external world, including material, mental, and abstract processes of every kind’, plus those ‘of our own consciousness’ (EF 39, IF 66). ‘A process’ can have ‘three components: the process itself, the ‘participants’, i.e. all things that ‘can become a Subject’, and the ‘circumstances’ (IF 101, 54, 114). ‘This tripartite interpretation’ of ‘how phenomena of the real world are represented as linguistic structures’ ‘lies behind the grammatical distinction of word classes into verbs, nouns, and the rest’ -- a ‘probably universal’ ‘pattern’ ‘among human languages’ (IF 102) (13.18). In ‘the preferred’ (‘unmarked’) ‘clause type’, ‘the initiating’ ‘element in the message’ is the ‘most closely associated with the process; and the culminating, information-carrying element’ is the ‘most remote’ (NT 3/214f).

9.61 ‘The grammar of the clause’ as ‘a structural unit’ for ‘expressing a particular range of ideational meanings’ is called ‘transitivity’ (EF 39). This domain is ‘the cornerstone of the semantic organization of experience’; it subsumes ‘all participant functions’ and ‘all experiential functions relevant to the syntax of the clause’ (EF 134; NT 3/182). ‘Endless variation is possible’ and ‘meaningful’; ‘the textual component provides’ ‘the means for distributing the experiential functions in every possible way over the functions Theme-Rheme and Given-New’ (NT 3/215) (13.43). The term thus denotes not the familiar ‘opposition’ between ‘transitive and intransitive verbs in English’, but ‘a set of clause types embodying a full range of possible transitivity distinctions’ (EF 39, NT 3/181f, 1/52, IF 103). ‘The potential distinction’ ‘between verbs which are inherently goal-directed or not is less useful as a generalization than the actual distinction between clauses’ which either have or lack ‘a feature of goal-directedness’ (NT 3/182) (cf. 9.67).

9.62 Halliday's way of classifying ‘processes’ instructively shows his balancing the plausible with the technical as well as the semantic with the grammatical (Table 9.1).

-- INSERT TABLE 9.1 ABOUT HERE --

Since ‘there are indefinitely many ways of drawing lines on purely semantic grounds’, we must inquire which ‘have systematic repercussions in the grammar’ (IF 108) (cf. 9.36; 13.54). We see a good illustration in the ‘criteria’ to show why ‘mental’ (i.e. ‘sensing’) and ‘material’ (i.e. ‘doing’) ‘processes’ ‘constitute distinct grammatical categories’ (IF 108, 102, 106). ‘Mental process’ is ‘distinct’ from ‘material process’ in: (a) having as ‘participants’ a ‘human Senser’ ‘endowed with consciousness’ and a ‘Phenomenon’ (as in ‘“I like the quiet”’), which ‘cannot be equated with Actor and Goal in a material process’ (as in ‘“the lion caught the tourist”’); (b) being ‘representable’ ‘as two-way’ or ‘bi-directional’ (as in ‘“Mary liked the gift”‘ versus ‘“the gift pleased Mary”’); (c) ‘that which is felt, thought or perceived’ being ‘a “Fact”‘ (‘a representation’ ‘ready packaged’, as in ‘“Jane saw that the stars had come out”’) as well as ‘a Thing’ (‘a phenomenon of our experience’, as in ‘“Jane saw the stars”’); and (d) having as ‘unmarked’ ‘tense’ ‘the simple present’ (as in ‘“I see the stars”’), whereas the ‘material process’ has ‘present in present’ (as in ‘“they are building a house”’) (IF 108-11, 136, 227f, 243).²⁹

9.63 ‘Mental processes’ have the ‘principal subtypes’ of ‘perception (“seeing, hearing”, etc.), affection (“liking, fearing”, etc.), and cognition (“thinking, knowing, understanding”, etc.)’ (IF 111). ‘Material processes’, in contrast, are divided into ‘dispositive’ (‘“doing to”’) and ‘creative’ (‘“bringing about”’), each of which may be either ‘concrete’ or ‘abstract’ (IF 103ff). Halliday recognizes as a third type ‘relational processes of being’ (IF 112). ‘Every language accommodates in its grammar a number of distinct ways of being’; ‘English’ has ‘intensive’ (i.e., ‘a relation of sameness’) (as in ‘“Tom is the leader”’), ‘circumstantial’ (as in ‘“the fair is on a Tuesday”’), and ‘possessive’ (as in ‘“Peter has a piano”’) (IF 112, 114). Each of these three ‘comes in two modes: attributive’ has the ‘functions’ ‘Attribute and Carrier’ (as in ‘“Sarah is wise”’), whereas ‘identifying’ has ‘Identified and Identifier’ (as in ‘“tomorrow is the tenth”’) (IF 113). Only ‘identifying clauses are reversible’ and have a ‘passive’ (as in ‘“Tom plays the leader”‘ and ‘“the leader is played by Tom”’); ‘attributives’ do not (as in ‘“the fair lasts all day”‘ but not ‘“all day is lasted by the fair”’), because an ‘Attribute is not a participant’ and so cannot ‘become a Subject’ (IF 114, 119f) (cf. 9.60).

9.64 Beyond these ‘three principal types of process found in the English clause’, Halliday sets up ‘three other subsidiary types’ (IF 128). ‘Behavioural processes’, both ‘physical and psychological’ (e.g. ‘“breathing, dreaming”’), are ‘intermediate between material and mental’, are usually ‘conscious’, and have the structure of ‘Behaver’ and ‘Process’ (as in ‘“the Mock Turtle sighed deeply”’) (IF 128f). ‘Verbal processes’ (e.g. ‘saying’) are ‘unlike mental processes’ in ‘not requiring a conscious participant’, and in having the structure of ‘Sayer’, ‘Receiver’, and ‘Verbiage’ (the ‘proposition’ or ‘proposal’) (as in ‘“he told me it was Tuesday”’) (IF 129f). ‘Existential processes’ (e.g. ‘existing’, ‘happening’) have the structure of ‘Existent’, ‘Process’, and optionally ‘Circumstance’ (as in ‘“there was an old woman tossed up in a basket”’) (IF 130f).

9.65 Even the six ‘process types’ cover only ‘participant functions’ ‘directly involved in the process’; ‘grammatically these are the elements that typically relate directly to the verb, without a preposition’ (IF 131). The ‘other participant functions’ for the ‘oblique or “indirect” participants’ that are more ‘optional’ than ‘inherent’ ‘in the process'³⁰ are ‘grouped’ under ‘Beneficiary’, including the ‘Recipient’ of ‘goods’ and the ‘Client’ of ‘services’ (e.g. ‘“John”‘ in ‘“I gave John a parcel”‘ or ‘“I painted John a picture”’); and ‘Range’, i.e., the ‘scope of the process’ (e.g. ‘“croquet”‘ in ‘“do you play croquet with the Queen today?”’) (IF 132ff).

9.66 Besides ‘participants’, Halliday has ‘circumstantial elements’, ‘the principal types’ of which, ‘in English’, are: ‘Extent and Location in time and space (including abstract space)’ (as in ‘“stay for two hours”‘, ‘“walk for seven miles”’); ‘Manner (means, quality and comparison)’ (as in ‘“beat with a stick”’); ‘Cause (reason, purpose, and behalf)’ (as in ‘“for want of a nail the shoe was lost”’); ‘Accompaniment’ (‘comitative’, i.e. ‘what with’, as in ‘“Fred came with Tom”‘, and ‘additive’, i.e. ‘what else’, as in ‘“Fred came as well as Tom”’); Matter (i.e. ‘what about’, as in ‘“I worry about her health”’); and Role (i.e. ‘what as’, as in ‘“I come here as a friend”’) (IF 137-42). This classification is drawn four ways: by meaning, by presupposed questions, by prepositions, and by illustrations (cf. Table on IF 148). Some interesting comparisons emerge. ‘Extent and Location’ show up the ‘close parallels between temporal and spatial expressions': having ‘standard units’, being ‘either definite or indefinite’, and being either ‘absolute or relative’ (IF 138). In return, ‘time is unidimensional’ and ‘moving’ whereas ‘space is ‘three-dimensional and static’; and only ‘time’ appears ‘in the tense system of the verb’ (IF 138f).

9.67 ‘From one point of view’, each ‘type of process’ ‘has a grammar of its own’ (IF 144f). Yet ‘from another point of view they are all alike’ and share ‘just one generalized representational structure’, based on one ‘ergative’ ‘variable’ of ‘causation': ‘is the process brought about from within or from outside?’ (IF 145, 147, NT 3/182). ‘The majority of verbs of high frequency in the language yield’ only ‘pairs’ of this kind (IF 145). Halliday attributes this ‘predominance’ in ‘modern English’ to ‘a far-reaching complex process of semantic change’ in the ‘language over the past five hundred years or more’ (IF 146). ‘The changes’ tend to ‘emphasize the textual function in the organization of English discourse’ over the ‘experiential function’, and within the latter function, ‘the cause-and-effect aspect’ over ‘the deed-and-extension’ or ‘actor-action-goal’ aspect (IF 146, EF 127; cf. IF 103; 9.56). The ‘waves of change’ indicate that ‘the transitivity system is particularly unstable in contemporary language’, due to ‘great pressure’ ‘for the language to adapt to a rapidly changing environment’ (IF 146) (cf. 12.59).

9.68 Halliday accordingly proposes ‘another interpretation’ of ‘the semantics of English’ vis-a-vis ‘the real world’, and of ‘the clause in its experiential function’ for ‘making generalizations about processes in the real world’ (IF 144-47). ‘Every Process’ has the ‘obligatory’ ‘participant’ or ‘element’ called ‘the Medium’, ‘through which the process is actualized’ (e.g. ‘“boat”‘ in ‘“the boat sailed”’) (IF 146).³¹ ‘The Process and the Medium together form the nucleus of an English clause’ that ‘determines the range of options’ for ‘the rest of the clause’ (IF 147). ‘The most general’ ‘option’, ‘turning up in all process types’, is the ‘ergative one': ‘the participant functioning as an external cause’, such that ‘the process’ is ‘represented as engendered from outside’ (e.g. ‘“Mary”‘ in ‘“Mary sailed the boat”’). We might need to ‘restructure our thinking’ to move from the ‘linear interpretation’ in terms of ‘transitive’, ‘emphasizing the distinction between participants’ and ‘circumstances’, to the ‘nuclear’ ‘interpretation’ in terms of ‘ergative’, allowing ‘a whole cluster of participant-like functions in the clause’ (IF 145, 149). These functions subsume further types of ‘causative agent’ -- ‘initiator’ (as in ‘“the police exploded the bomb”’), ‘inducer’ (as in ‘“the report convinced Mary”’), and ‘attributor’ (as in ‘“the sun ripened the bananas”’) -- which ‘in the transitive analysis’ would be ‘assigned different structural configurations’ (‘doing’ versus ‘making do’) (IF 152f).

9.69 ‘Probably all transitivity systems, in all languages, are some blend of these two semantic models of processes, the transitive and the ergative’ (IF 149). ‘Semantically, therefore, Agent, Beneficiary, and Range have some features of participants and some of circumstances’; ‘grammatically, also, they are mixed’ and may ‘enter’ ‘directly as nominal groups or indirectly in prepositional phrases’. The ‘choice’ to use a ‘preposition’ is thus not ‘random variation’, but ‘serves a textual function’; ‘a participant other than the Medium’ and having ‘prominence in the message’ -- i.e. ‘occurring either earlier’ (‘as marked theme’) ‘or later’ (‘as “late news”’) ‘than expected in the clause’ -- ‘tends to take a preposition’.³²

9.70 A related drift ‘away from a purely transitive type of symbol organization can be seen in the system of voice’, another major ‘resource of transitivity’ (IF 150, NT 3/203). Instead of labelling just ‘verbs’ as ‘active’ and ‘passive’, we might use ‘ergative terms’ and sort whole ‘clauses’ into ‘effective’ (with a ‘feature of agency’, as in ‘“the cat broke the glass”’) and ‘middle’ (without it, as in ‘“the glass broke”’). ‘The choice between active and passive’ is open only for an ‘effective’ ‘clause’, and ‘the reasons for choosing passive’ are: ‘to get the Medium as subject’ and thus as ‘unmarked theme’, or ‘to make the Agent either “late news” by putting it last’ in the slot for ‘unmarked’ ‘information focus’, or else ‘implicit by leaving it out’ (IF 151, 118; NT 3/205, 2/215, 217).³³ ‘In spoken English the great majority of passive clauses are, in fact, Agent-less’ (IF 151f) (cf. 7.53).

9.71 ‘The interpersonal’ meta-function concerns ‘forms of interaction’ and ‘embodies all use of language to express social and personal relations’, ‘personalities, and personal feelings’, as well as ‘the speaker's intrusion into speech situation and speech act’ (EF 41, 66, 106, NT 3/210). This ‘function’ ‘extends beyond’ the ‘rhetorical’ by ‘expressing both the inner and the outer surfaces of the individual’, and is thus ‘personal in the broadest sense’ (EF 107). ‘The speaker’ (‘a cover term for both speaker and writer’) ‘expresses his comments, attitudes, evaluations’, ‘adopts’ a ‘speech role’, and ‘assigns the listener a complementary role’, ‘the most fundamental’ being ‘giving’ and ‘demanding’ (IF 68, EF 106, NT 3/210). But ‘we can recognize an unlimited number’ of ‘specific’ ‘socio-personal’ ‘uses of language': ‘ask and answer’, ‘approve and disapprove’, ‘greet, chat up, take leave’, ‘express belief, opinion, doubt’, and ‘feelings’, ‘include in’ or ‘exclude from the social group’, and so on (EF 41) (cf. 8.42). ‘The act of speaking’ might well ‘be called an “interact”‘ (IF 68).

9.72 The ‘interpersonal function of the clause is that of exchanging roles in rhetorical interaction’ (IF 53). ‘Goods-&-services’ are also ‘exchanged’ via ‘offers and commands’, wherein ‘language functions simply as means’ toward ‘non-linguistic ends’; these uses have ‘priority in the ontogenetic development of language’ and ‘serve as a point of entry to a great variety of different rhetorical functions’ (IF 68, 70f) (cf. 9.14). ‘Information'³⁴ is ‘exchanged’ via ‘statements and questions’, wherein ‘language is the end’ and ‘the means’, and ‘the clause takes on the form of a proposition’ ‘that can be affirmed or denied, qualified’, ‘regretted, and so on’ (cf 3.36, 44f; 8.55; 9²⁴; 11.39-50). ‘Propositions’ are ‘useful to look at’ because they ‘have a clearly defined grammar’ with more ‘special resources’ (IF 70).

9.73 ‘Mood represents the organization of participants in speech situations’ and ‘speaker roles’, such as ‘informing’, ‘confirming’, ‘contradicting’, etc. (NT 2/199). In ‘the clause as domain’, if ‘theme is the grammar of discourse’ and ‘transitivity is the grammar of experience’, then ‘mood is the grammar of speech function’. The ‘choice’ of an ‘element’ ‘as theme’ may ‘depend on the choice of mood’; and ‘some options are on the borderline of theme and mood’ (IF 44, NT 2/243n). ‘Any thematic element’ not ‘derived from the mood of the clause’ must be ‘a “marked Theme”‘ (NT 2/223). Yet ‘unlike the Theme’, which ‘carries forward the development of the text as a whole’, ‘the Mood element has little significance beyond the immediate sequence of clauses’ (IF 98).

9.74 In the grammar, ‘the Mood’ is the ‘constituent’ formed by ‘Subject and Finite’ ‘closely linked together’, ‘the remainder of the clause’ being ‘the Residue’ (IF 73f). ‘Every ‘major clause’, ‘whether independent or not’, ‘selects for mood’; ‘those which do not’ are ‘minor clauses’ (e.g. in ‘calls, greetings, and exclamations’) (IF 44, 61, 63) (cf. 9.54). The ‘independent major clause’, in which ‘the constituent specified by the mood systems’ ‘is obligatory’ and which ‘exhibits the options of theme in its full interpretation’, is either ‘indicative or imperative in mood’ (IF 44, NT 2/213, 221). As a ‘general principle’, ‘the indicative’ is ‘used to exchange information’, either by ‘statement’ in ‘the declarative’ (with ‘Subject before Finite’), or by ‘question’ in ‘the interrogative’ (with ‘Finite before Subject’, unless a ‘WH-element is the Subject’) (IF 74). Another ‘subcategory’ of ‘declarative clause’ is ‘the exclamative’ with a ‘WH-element as theme’ (e.g. in ‘“what tremendously easy questions you ask!”’) (IF 47).

9.75 ‘Below the clause’ is ‘the grammar of the group’, ‘interpreted as a word complex’ with ‘Head’ and ‘modifying element’ (IF 158f, 192) (compare Firth's ‘piece’, 8.55). ‘In the Western grammatical tradition, it was not recognized as a distinct structural unit; instead, simple sentences’ ‘(clauses in our terms) were analysed directly into words’ (IF 158f). ‘Such an analysis’ requires ‘confining our attention’ to the ‘idealized isolated sentences that grammarians have usually dealt with’ (e.g. ‘“John threw the ball”’) and ‘ignoring several important aspects of the meanings’; ‘and in the analysis of real-life discourse it leads to impossible complexity’ -- like ‘describing a house’ as ‘bricks’ without ‘intermediate structural units’ such as ‘walls and rooms’. So ‘the group’ should be ‘recognized’ ‘as a distinct rank in grammar’ with its own ‘multivariate constituent structure’, even if it ‘no doubt evolved by expansion outwards from the word’ just as the ‘sentence’ did from the ‘clause’ (IF 192, 159) (cf. 9.82; 13.54). This factor divides the group from the ‘phrase’, which has ‘roughly the same status on the rank scale’ but ‘is a contraction of a clause’.

9.76 In ‘the group’, ‘the three’ metafunctions are ‘represented’ not as ‘separate whole structures, but rather as partial contributions to a single structural line’ (IF 158). This ‘difference between clause and group’, though ‘only one of degree’, allows us to ‘analyse the group in one operation, not three’ (IF 158, 169, 176; cf. 9.46ff, 80). However, Halliday does ‘split the ideational’ back into ‘experiential’ and ‘logical'³⁵, the latter showing ‘the group’ as ‘a word complex: ‘a combination of words built up on the basis’ of ‘generalized logical-semantic relations’ ‘encoded in natural language’ (IF 158f, 170) (cf. 9.57).

9.77 The ‘main classes of group’ are ‘nominal’, ‘verbal’, and ‘adverbial’ (IF 159). ‘Interpersonal meanings’ in ‘the nominal group’ are ‘embodied in (a) the person system’, (b) ‘the attitudinal’ ‘Epithets’ (like ‘“splendid!”’), (c) ‘the connotative meanings of lexical items’, and (d) ‘prosodic features such as swear-words and voice quality’ (IF 169f). ‘The experiential structure of the nominal group’ includes ‘the functional elements Deictic, Numerative, Epithet, Classifier, and Thing’ (IF 160, 164). ‘The Deictic’ ‘indicates whether or not some subset’ of ‘a class of things’ is ‘intended’ (e.g. ‘“all”’, ‘“some”’), and, used ‘demonstratively’, can stipulate ‘proximity to the speaker’ (e.g. ‘“this”’) or ‘possession’ (e.g. ‘“your”’). The ‘Numerative’ ‘indicates some numerical feature of the subset’ and can be ‘quantifying’ (e.g. ‘“two”’) or ‘ordering’ (e.g. ‘“second”’), matching the familiar classes of ‘cardinal’ and ‘ordinal numerals’, plus ‘inexact number’ (e.g. ‘“many”’). ‘The Epithet indicates some quality of the subset’, with ‘no hard and fast line’ between ‘objective property of the thing’ (‘experiential in function’) (e.g. ‘“old”’) versus ‘the speaker's subjective attitude’ (‘interpersonal’ in function) (e.g. ‘“silly”’) (IF 163) (cf. 13.24).³⁶ The ‘Classifier indicates a particular subclass of things’ in terms of ‘material’, ‘scale’, ‘origin’, and so on (e.g. ‘“wooden”’) (IF 164). Finally, the ‘Thing’ is ‘a phenomenon of our experience’ and ‘the semantic core of the group’, usually ‘realized’ as a ‘noun’ (e.g. ‘“nose”’) (IF 108, 167, 164). We could thus have the sequence ‘“your two silly old wooden noses”‘.

9.78 The ‘ordering’ of ‘the nominal group’ is thereby ‘interpreted’ in terms of an ‘experiential pattern’ (IF 165). ‘The progression’ goes from ‘greatest specifying potential to’ ‘the least’ (IF 166). ‘The Deictic’ ‘starts by relating to the speaker in the context of the speech event’, and then come ‘elements with successively less identifying potential’ and more concern for ‘permanent attributes’. Hence, we ‘begin with the immediate context’ and ‘go on to quantitative features’ (‘order and number’), then ‘qualititative features’, and ‘finally’ ‘class membership’ (IF 165f). ‘We should beware, however, of assuming that the taxonomic order of modification always corresponds to something in the extra-linguistic universe’ (IF 171) (13.24).

9.79 ‘The verbal group’ has the ‘structure’ of ‘Finite’ plus ‘Event (with an optional Auxilary’ if ‘the Finite’ is not ‘fused with the Event’, as in ‘one word verbal groups such as “ate”’) (IF 175) (cf. 9.74).³⁷ ‘Finiteness’ is ‘expressed by means of a verbal operator which is either temporal or modal’ (IF 75; cf. 8.59, 64). This ‘finite element’ gives ‘the proposition’ ‘a point of reference in the here and now’, ‘relates’ it ‘to its context in the speech event’, and ‘refers’ either ‘to the time of speaking’ (via ‘primary tense': ‘past, present, or future’) or ‘to the judgment of the speaker’ (via ‘modality': ‘probabilities’, ‘obligations’, ‘desires’) (IF 75f, 86).³⁸

9.80 Beyond this ‘experiential’ aspect, ‘the verbal group’ (like the nominal) needs no ‘separate analysis’ for the other two metafunctions (IF 176) (cf. 9.76). ‘Textual meaning is embodied in the ordering of the elements’; ‘interpersonal meaning resides in the deictic feature associated with finiteness (primary tense or modality)’ and in ‘attitudinal colouring’ of ‘the lexical verb’. However, ‘the logical structure’ carries ‘most of the semantic load’, and in a way having ‘no parallel in the nominal group’, where the main issue is ‘the recursive aspect of the modifying relation’ for ‘generating long strings’ ‘in univariate’ and ‘multivariate structures’ (IF 175, 172) (cf. 9.59). In ‘the verbal group’, in contrast, the ‘logical’ issue is ‘the recursive tense system’ (IF 176f). ‘The primary tense’ is ‘relative to speech event’ (‘past, present, future’), and ‘the secondary tenses’ are ‘relative to time selected in previous tense’. Though a ‘recursive’ ‘system’ has ‘no longest possible tense’, ‘in practice, the total set’ of ‘finite tenses’ is ‘limited’ to ‘thirty-six’ by ‘“stop rules”': ‘future’ and ‘present occur only once’ and ‘the same tense does not occur twice consecutively’ (IF 179).³⁹ Also, ‘the system varies for different speakers’ and ‘is tending to expand all the time, although it has probably just about reached its limits’ (IF 184).

9.81 Albeit only ‘the elements of verbal group are purely grammatical’ in that ‘the options they represent are closed’ rather than ‘open-ended’, Halliday sees a major ‘parallelism’ between ‘the verbal group as the expansion of a verb’ and ‘the nominal group as the expansion of a noun’ (IF 175, 178). Both ‘Finite’ and ‘Deictic’ ‘relate’ ‘to the “speaker-now”‘ (IF 176, 160). ‘The Event’ ‘is the verbal equivalent of the Thing’ in that ‘both represent the core of the lexical meaning’, although since ‘Things are more highly organized than Events’, the ‘nominal’ has ‘additional lexical elements’ (IF 176, 184n).⁴⁰ In sum, ‘both verbal and nominal groups begin with the element that “fixes” the group in relation to the speech exchange; and both end with the element that specifies the representational content’ (IF 176). This makes sense: ‘initial position is thematic’, the ‘natural theme’ being the ‘relation to the here-and-now’, whereas ‘final position is informative’, the natural place for ‘the newsworthy’ (cf. 9.51f; 7¹⁴). ‘So the structure of groups recapitulates, in the fixed ordering of their elements, the meaning’ ‘incorporated as choice in the message structure of the clause’ (IF 176, 166).

9.82 ‘One step above the clause’ is ‘the clause complex’, which has ‘the typical sequence’ of ‘Head (dominant) clause plus Modifying (dependent) clause’ or, for ‘thematic’ ‘motives’, ‘the reverse order’ (IF 57, 192). The ‘clause complex corresponds closely to a sentence of written English’ and has in fact ‘led to the evolution of the sentence in the writing system’ while ‘the sentence’ ‘evolved’ ‘over the centuries’ by ‘expansion outward from the clause’ (IF 288, 192f) (cf. 9.75; 13.54). Yet ‘the sentence’ does not qualify ‘as a multivariate constituent structure with its own range of functional configurations’ (IF 192). It does not have ‘elements that are distinct in function, realized by distinct classes, and more or less fixed in sequence’. Instead, ‘the tendency is’ ‘for any clause to have the potential for functioning with any value in a multi-clausal complex’. Halliday accordingly makes ‘the “clause complex”‘ ‘the only grammatical unit above the clause’ and ‘assumes’ it ‘enables us to account in full for the functional organization of the sentence’ (IF 193). ‘The sentence’ will be not ‘a distinct grammatical category’ but ‘an orthographic unit’ ‘between full stops’ -- ‘a constituent of writing, while a clause complex’ is one of ‘grammar’.⁴¹

9.83 ‘The relations between clauses’ are again referred to ‘the logical component’ (IF 193; cf. 9.59). In one ‘system’, ‘expansion’ makes ‘the secondary clause’ (the later one) relate to ‘the primary clause’ (the earlier one) by ‘elaborating’ (as in ‘“John didn't wait; he ran away”’), ‘extending’ (as in ‘“John ran away and Fred stayed behind”’), or ‘enhancing’ it (as in ‘“John was scared, so he ran away”’), whereas ‘projection’ makes the ‘primary clause’ ‘instate’ ‘the secondary clause’ as a ‘locution’ (‘a construction of wording’, as in ‘“John said, ‘I'm running away'”’) or ‘an idea’ (‘a construction of meaning’, as in ‘“John thought he would run away”’) (IF 195ff). ‘Expansion and projection form the basis of the English clause complex’, and ‘generally’ ‘recur throughout the semantic system’ and ‘the lexicogrammar’ (IF 378). In another ‘system’, ‘parataxis’ links ‘two elements of equal status’ in ‘sequence’, ‘one initiating and the other continuing’ but both ‘free’ (i.e., able to ‘stand as a functioning whole’), whereas ‘hypotaxis’ links two of ‘unequal status’ in ‘dependence’, the ‘dominant’ one being ‘free’ but the ‘dependent’ one not free (IF 193, 195) (cf. 4.68; 5.53). ‘Parataxis and hypotaxis’ are ‘the two basic forms of logical relations in natural language’ and can ‘define’ ‘univariate structures’ in ‘complexes of any rank’ -- ‘word, group, phrase, and clause alike’ (IF 198, 193) (cf. 9.59). In ‘the tone system’, for example, they appear as ‘tone concord’ (‘two or more instances of same tone’) and ‘tone sequence’ (different tones), respectively (IF 285).

9.84 A familiar type of hypotaxis is ‘relative clauses’, which Halliday again divides in two groups. ‘Defining’ ones ‘define subsets’ (as in ‘“the only plan which might have succeeded”’), whereas ‘non-defining’ ones ‘(also called “non-restrictive”, “descriptive”)’ ‘add a further characterization of something’ ‘taken to be already fully specific’ (as in ‘“inflation, which was necessary for the system, became also lethal”’) (IF 204, 167, 379, NT 2/209). This division is ‘clearly signalled in both speech and writing’ (IF 205). In ‘speech’, the ‘defining’ relation is marked by ‘tone concord’, whereas ‘the non-defining relative forms a separate tone group’. In writing, only the ‘non-defining’ ‘is marked off by punctuation, usually commas but sometimes’ ‘a dash’. Halliday sees ‘an analogy’ between ‘identifying process’ and ‘defining relative’ on the one hand, and between ‘attributive process’ and ‘non-defining relative’ on the other (IF 379; cf. 9.63).

9.85 ‘In the clause complex’, ‘dependent clauses may be finite or non-finite’ (IF 199, 204ff). The ‘finite’ kind is well known, though Halliday's conception of ‘finiteness’ is unusually elaborate (9.58, 74, 79f). What he calls the ‘non-finite clause’, however (e.g. ‘“selling office equipment”‘, IF 206), has often not been counted a clause at all, but a participial modifier (cf. 4.69). He admits it may ‘occur’ without a ‘marker’ or ‘indication of its logical-semantic function’ or its ‘category’ (IF 217f). ‘The best solution here is to find the nearest finite form’ and classify that. But ‘in most instances the Subject is left implicit, to be presupposed from the primary clause’; it can be ‘difficult to identify’ because ‘the non-finite’ ‘makes it unnecessary to decide’ (IF 207). Or, we could suspend the problem by not postulating a clause at all, and assuming the presupposed material to be semantic (agent), not grammatical (subject), but that would violate the principle that all catgories be ‘there in the grammar’ (9.36).

9.86 Beyond the clause complex lies the domain of ‘cohesion’, based on ‘additional relations within the text’ which ‘hold across’ ‘gaps of any distance’ and ‘link items of any size’ ‘from single words to lengthy passages’ (IF 288f) (Hasan 1967; Halliday & Hasan 1976) (cf. 10¹⁵). ‘Typically any clause complex in connected discourse will have from one up to half a dozen cohesive ties with what has gone before it’, plus ‘some purely internal ones’ (IF 290). These ties are termed ‘non-structural forms of organization’ on the grounds that ‘the clause complex’ sets ‘the upper limits of grammatical structure’, and that not ‘words and structures’ but ‘ongoing semantic relationships’ ‘make text’ (IF xxi, 288, 318, 291; cf. IF 380f; NT 2/206) (cf. 4.67; 9.3, 22, 91f, 95). Yet since the whole grammar is to some degree text-based and semantic, this division is troublesome, as we shall see (9.95).

9.87 ‘English’ has ‘four ways by which cohesion is created': ‘conjunction’, ‘reference, ellipsis’, and ‘lexical organization’ (IF 288, 313). ‘Conjunction’ covers a ‘cohesive bond’, ‘expressed’ by a ‘conjunction’ (like ‘“but”’) or a ‘conjunctive adjunct’ (like ‘“however”’) between ‘two’ ‘typically contiguous elements’, ranging from ‘clauses’ up to ‘paragraphs or their equivalent in spoken language’ (IF 303f, 289) (cf. 11.48). The ‘semantic relations’ ‘are basically of the same kind’ as ‘between clauses’ in a ‘clause complex': ‘elaboration, extension, enhancement’ (IF 289, 303f) (cf. 9.81). But Halliday now breaks them down further into ‘categories’ of ‘apposition’, ‘clarification, addition’, ‘variation’, ‘temporal’, and ‘causal-conditional’; these are further subdivided, yielding around 30 final categories, some with erudite names like ‘adversative addition’ (‘“however”’), ‘verifactive clarification’ (‘“actually”’), and ‘punctiliar temporal conjunction’ (‘“at this moment”’) (IF 303-08). The ‘categories’ are mainly commonsensical (but not all, e.g., ‘distractive’ and ‘dismissive’ are listed as ‘clarifying’) and ‘may be found useful in the interpretation of texts’ (IF 303f, 308f).

9.88 ‘Conjunction’ is ‘a way’ of ‘achieving texture’ by ‘setting up logical-semantic relations’ ‘between messages’ ‘in the absence of structural relationships’ (IF 308, 301, 289, 317). Such relations are obviously ‘cohesive’ when ‘expressed’ by ‘words’ (e.g. ‘“consequently”‘ expressing ‘cause’); ‘implicit conjunction’ applies where ‘the semantic relationship is clearly felt to be present but unexpressed’ (IF 301f, 308). We could ‘recognize’ these cases ‘by the possibility of inserting a conjunction without changing the logical-semantic relation’; or we could ‘treat them as semantically unrelated’ because ‘if the speaker had wanted to relate’ them ‘he could have done so’ (IF 217). ‘Including’ these ‘relations’ ‘in the analysis leads to a great deal of indeterminacy’ about ‘whether’ they are ‘present’ and ‘which’ ‘kind’ ‘hold between pairs of adjacent sentences, or between each sentence and anything that precedes it’ (IF 308f). So we should be ‘cautious in assigning implicit conjunction’ by ‘noticing’ where it is ‘recognized’, yet ‘characterizing the text without it: to see how much we still feel is being left unaccounted for’. ‘The presence or absence of explicit conjunction’ is a ‘principled variable in English discourse’ and should not be ‘obscured’. We should look also to ‘other forms of cohesion’ to assist ‘our intuition’ about the ‘pattern of conjunctive relationships’.

9.89 ‘Reference’, usually termed ‘co-reference’ in text linguistics,⁴² is ‘a relation between things or facts’, ‘usually’ ‘single elements that have a function within the clause -- processes, participants, circumstances’ (IF 289). Probably, ‘reference first evolved’ as ‘exophoric reference': ‘linking “outwards” to some person or object in the environment’ (IF 290) (cf. 13.24). ‘We may postulate an imaginary stage in the evolution of language when the basic referential category of person was deictic’, ‘referring to the situation here and now’ (IF 291) (cf. 9.77f, 80f). ‘First and second person’ pronouns ‘retain this deictic sense’ (4.70), as do the ‘demonstratives “this/that”‘; but ‘third person’ pronouns are ‘more often than not anaphoric’, i.e. ‘pointing’ ‘to the preceding text’, or, more rarely, ‘cataphoric’, pointing to ‘the following text’ (IF 291ff). Whenever ‘the listener has to look elsewhere’, the ‘effect’ is ‘cohesive’, ‘linking the two passages into a coherent unity’. ‘If the pronoun and referent are in the same clause complex’, we have ‘already one text by virtue of the structural relationship between the clauses’. If not, ‘cohesion’ is ‘the sole linking feature and hence critical to the creation of text’. So ‘the cohesive relationship’ ‘carries a greater load’ beyond the ‘clause complex’.⁴³

9.90 ‘Ellipsis’ occurs when ‘a clause, or a part of a clause’ or ‘of a verbal or nominal group’ is ‘presupposed at a subsequent place’ via ‘positive omission’ -- ‘saying nothing where something is required to make up the sense’ (IF 288). Again, what is missing depends upon one's grammatical expectations, and here Halliday takes a rather extreme view. In ‘clausal ellipsis’, as is ‘typical in a dialogue sequence’, ‘everything is omitted’ ‘in a response turn’ ‘except the information-bearing element’, so that ‘the listener’ must ‘supply the missing words’ ‘from what has gone before’ (IF 300). ‘It is always possible to “reconstitute” the ellipsed item’ and make it ‘fully explicit’ (more overstatement); ‘the exact wording’ is ‘taken over’, aside from ‘reversal of speaker-listener deixis’ and ‘change of mood where appropriate’. In a ‘question-answer sequence’, mere ‘“yes”‘ and ‘“no”‘ are taken to be ‘elliptical’ for ‘the whole clause’, as are ‘“Why?”‘ and ‘“Who?”‘ in responses to statements (IF 297).

9.91 Ellipsis doesn't fully match the other means of cohesion. It is a relation not between two actual passages in a text, but between an actual passage and a virtual or theoretical complete version. Also, though it ‘contributes’ to ‘the semantic structure’, it ‘sets up a relationship that is not semantic but lexicogrammatical’ -- ‘in wording’, not ‘meaning’ (IF 296, 300). For instance, since ‘every independent clause in English requires a Subject’, ‘the listener will understand the Subject’ removed by ‘ellipsis’ -- e.g., ‘“I”‘ ‘in a giving clause (offer or statement)’ (e.g. ‘“carry your bag?”’) , or ‘“you”‘ ‘in a demanding clause (question or command)’ (e.g. ‘“play us a tune”’) (IF 90f). In this view, despite ‘most accounts of English grammar, the imperative’ is not ‘a special case’ but ‘an instance of this general principle by which a Subject is understood’ (cf. 4.56).

9.92 ‘A substitute serves as a place-holding device, showing where something has been omitted and what its grammatical function would be’ (IF 297; cf. NT 2/239ff). ‘“Do”‘ is a ‘verbal substitute’, and ‘“one”‘ ‘a nominal substitute’, both ‘derived by extension from an item in the full, non-elliptical group’ (IF 300f). Although differing by ‘environments’, ‘ellipsis and substitution are variants of the same type of cohesive relation’, in that ‘the missing words’ listeners presumably ‘retrieve’ ‘must be grammatically appropriate’ for being ‘inserted in place’ (IF 297, 301f). ‘This is not the case with reference’, where ‘the relationship’ is ‘semantic’ and has ‘no grammatical constraint: the class of the reference item need not match that of what it presupposes, and one cannot normally insert the presupposed element’ (IF 302). For that ‘reason’, ‘reference’ ‘can reach back a long way, whereas ellipsis - substitution is largely’ ‘confined to closely contiguous passages’, e.g. ‘“adjacency pairs” in dialogue’ (IF 302, 289, 317). In return, ‘reference’ usually ‘refers to the same thing’ while ‘ellipsis - substitution’ need not (IF 302).⁴⁴

9.93 ‘Lexical cohesion’ ‘selects items’ ‘related in some way to previous ones’ and creates whole ‘referential chains’ whose ‘interaction’ ‘gives the text its coherence’ and ‘dynamic flow’ (IF 316, 310, 289) (cf. 11.30, 45). Subtypes include ‘repetition’, where a repeated item is the same word or some ‘inflectional’ or ‘derivational variant’ (e.g. ‘“dine - dinner”’); and ‘synonymy’ (e.g. ‘“sound - noise”’), along with its ‘variants': ‘hyponymy’ of ‘general’ and ‘specific’ (e.g. ‘“vegetation - grass”’), ‘meronymy’ of ‘whole’ and ‘part’ (e.g. ‘“bottle - stopper”’), and ‘antonymy’ of ‘opposites’ (e.g. ‘“fell asleep - woke”’) (IF 310ff). Also, Firth's ‘collocation’ is included as a ‘“co-occurrence tendency”‘ having a ‘sema