7/28/2019 sgrev
1/26
7/28/2019 sgrev
2/26
792 R. Hudson
phonology (especially intonation), lexis (which is roughly what others mean
by '[the theory of] the lexicon'), syntax, semantics, pragmatics (especially in
relation to cohesion and socially significant variation). He has also worked on
language acquisition and on the relations between language and social
structure and has been deeply involved for the last two decades in theapplication of linguistics to school education. He has published on all parts of
this broad spectrum, and interested readers will find a complete list ofreferences in the book by Butler reviewed below.
SG is for many synonymous with 'Hallidayan linguistics', which covers allthese many fields of interest. This is entirely to be expected (in view of the
way in which 'Chomskyan linguistics' is used as the name for one of the mainalternatives) and reflects the dominant influence which Halliday has held
since the start of the theory. This at least has the advantage, for the reviewer,
of making it relatively easy to sort out what counts as 'true' SG: that which
Halliday propounds. This is helpful because various other linguists have
I
''0Review article
Systemic grammar*
~ ,RICHARD HUDSON
.','
M. A. K. Halliday:An Introduction to Functional Grammar. London: Arnold,
1985. 387 + xxxv pp. 14.95.
C. S. Butler: Systemic Linguistics: Theory and Applications. London:
Batsford, 1985. 259+ix pp. 19.95.
Both the books under review are about systemic grammar (SG), a theoryunfamiliar to many readers, so the review starts with a brief historical account
of SG. This is followed by a review of the book by Halliday, in which I
concentrate on matters of coverage and presentation. The next sections deal
with a number of general issues which SG raises, and finally I turn to Butler's
book.
1. Historical background
'Systemic grammar' is a name which was coined in the mid-1960s for a theory
which Halliday was then developing. (The initiated pronounce it /sisti:mik/;
the name comes from the key concept, the 'system', which is just a set of
alternatives.) The precursor of systemic grammar was called 'scale andcategory' grammar (after the two main relations distinguished in Halliday1961). This was a rather different theory which was less well developed in
some important respects, so the difference of name is important. Morerecently, however, the name 'systemic functional grammar' is often used, but
as far as I can see the theories called 'systemic grammar' and 'systemicfunctional grammar' are more or less the same.
Halliday's ideas were based on those of J. R. Firth, under whom he studied
in London, and like Firth's they comprise a comprehensive theory of language
structure which applies to all levels. Halliday himself has made significantcontributions t6 work at all levels (except morphology):
Linguistics 24 (1986), 791-815 0024-3949/86/0024-:
7/28/2019 sgrev
3/26
developed Halliday's ideas in their own directions and have ended up with
very different theories which could be included under the label 'SG' only atthe cost of enormous confusion. (I myself am an example, and 1 have been
careful to minimize confusion by using names other than 'SG' for my ideas
once they became clearly incompatible with Halliday's.) The identification ofSG with Hallidayan linguistics has a disadvantage, however, which is that SG
then comes to include any area of linguistic work in which Halliday has
engaged, whether or not it has anything specifically to do with the theory oflanguage structure called SG. I have in mind here the work on cohesion
(Halliday and Hasan 1976) and much of the work on social structure andlanguage and on language in education. The question how much of this work
is 'really' part of SG is clearly not worth pursuing, but the fact remains thatfor many people there is no clear distinction between SG and the kind of
humane linguistic endeavor which Halliday advocates and practices.The school called SG is alive and well, socially speaking. Each year a
'systemic workshop' is organized and attended by a hundred or so linguists,
and in recent years these events have been retitled 'international systemic
workshops' in recognition of the fact that their participants come not onlyfrom Britain but also from North America and Australia (where Halliday is
now working) - indeed, several recent workshops have been held in NorthAmerica. Several linguistics departments are predominantly Hallidayan in
orientation, and 1 have the impression that the same is true of very many
language departments in teacher-training institutions. Not many schools oflinguistics can boast such organizational success (one thinks of GLOW as theChomskyan counterpart, but
7/28/2019 sgrev
4/26
~
there is no comparable standing organization for any other school that I know
of), so it demands explanation.One reason is certainly that a lot oflinguists are attracted by Halliday's
general philosophy, especially in contrast with the rather dry-as-dust and
daunting wares offered by the alternative theories. Halliday is interested inthe relations between language and people, and many people with similar
interests find his work stimulating. Another reason is that Halliday offers a
comprehensive analysis of English grammar which is relatively easy to applyto texts - something which no other school has to offer. This has obvious
attractions for applied linguists of various typesnot least for computationallinguists, who have taken SG very seriously since Winograd used it to make
SHRDLU work (Winograd 1972, 1983). Yet another reason for theorganizational success is the energy and efficiency of Fawcett, one of the
leading advocates of SG (Fawcett 1980) and the founder of the systemicworkshops.
And yet, for all this success, SG remains largely unknown except to the
initiated. Very few papers in refereed linguistics journals use SG, or even
refer to it as such (though the work on cohesion, for example, is quite oftenreferred to). Indeed, the bibliography in Butler's book contains only three
examples of such papers published since 1970 (Hudson 1973,1974; McCord1975). There has been no shortage of either books about SG or articles about
it in anthologies or in departmental working papers, but it is easy to
understand why most linguists confess to more or less complete ignoranceabout SG. Not surprisingly, SG has had little effect on other theories (exceptfor those which, like my own, are direct descendants), but the converse is
unfortunately also true: SG shows little sign of being influenced by work inother theories. Halliday complains (p. xxviii) that 'it is often difficult to
maintain a dialogue' with other theories, but a good deal more effort could
have been made by systemic linguists to establish this dialogue - by listeningand learning as well as by talking. If this had happened there might have"been more direct critical comment by outsiders, something which Butler (p.
77) regrets the lack of. Without such a change of attitude, SG looks set to sailon into the next century as an autonomous school of linguistics unaffected by
the developments in other schools.
2. Halliday'sIntroduction to Functional Grammar
This (henceforward: IFG) is an introductory book about English grammarwithin the framework of SG. It is introductory in that it assumes no
knowledge about linguistic analysis, so for example the first chapters
7/28/2019 sgrev
5/26
794 R. Hudson
introduce notions like 'constituency' and 'structure'; but as so often happens in
'introductory' books the pace accelerates rapidly and some of the later chaptersbecome very hard work. (He warns us [po xxxiv] that the writing becomes
dense in places; he is right.) It is mainly about English, with very fewreferences to other languages, and indeed rather few references to theoretical
issues which would justify the title of the book. It would perhaps have beenmore helpful to call it something like 'A functional introduction to English
grammar'. Facts about English grammar are not brought in to exemplifytheoretical matters; rather, the book sets out to give a more or less completeoverview of English grammar and is organized accordingly (with a chapter on
transitivity, one on the structure of the phrase, and so on).
In case the above gives the impression that IFG is a run-of-the-mill bookabout English grammar, I must stress that it is not. It is very unusual in most
respects one can imagine. The order of presentation is unusual - he starts withpoetry to illustrate the nature of linguistic structure, for example. The content
is unusual, precisely because this is the grammar that he has been developingover the last two decades, largely without influence from outside (so far as
one can tell). The book is stuffed with analyses which could not be matchedfrom other theories or which conflict with analyses offered in terms of other
theories. The coverage is unusual - he tends to continue his analyses at the
points where other linguists give up in despair, such as intonation,topicalization, adverbials, and the fuzzy area where sentence structure fadesinto discourse structure. And the presentation is unusual - for instance, his
examples are unusually well chosen not only for making the relevant point butalso for their inherent interest; a good many are fromAlice in Wonderlandbut
there is a fair sprinkling of extracts from taped conversations. IFG is achallenging book - it challenges those of us who are outsiders to see if we canproduce anything as impressive in its scope and internal consistency. It isimportant to bear this in mind in reading the following critical remarks,
because I know of only two other attempts to present a unified, theoreticallybased account of a comparably large section of English grammar, namely the
'UCLA' grammar of Stockwell et al. (1973) and Huddleston (1984), a work
which owes a lot to Halliday's influence. For one person to have produced IFG
is an achievement to be respected, whatever its shortcomings.The shortcomings are, unfortunately, numerous. Two are obvious at a first
glance. There is no index at all, which makes the book virtually impossible touse as a 'resource for the interpretation of texts' (p. xx). The lack of an index is
noted (p. xii) but we are told that the table of contents should do instead. Thisis certainly not the case for at least one reader
7/28/2019 sgrev
6/26
Systemic grammar 795
(me). For example, say I wanted to find out what the book has to say about
relative clauses, where do I look? Once you know the system, you know toturn up the sections labeled 'Experiential structure of the nominal group:
interpretation of ordering; the Qualifier' and 'Elaboration' (in the chapter on'The clause complex') - among others. But it would surely take several very
careful readings of the book to reach the point where this skill is acquired -and I doubt if many readers would miss the opportunity to build up their own
index on the way. It amazes me that the publishers agreed to publish the bookwithout an index, as it must halve the value of the book for its most importantpopulation of potential users, those who want a detailed grammatical analysis
which they can apply to texts. As it is, these customers will surely turn to one
of the more conventional grammars such as Quirk et al. (1972), although inmany ways Halliday's grammar is better- at least it is a serious attempt at afully integrated analysis.
The other obvious point about the book is the total lack of references to the
literature. (The one exception that I noticed was a short list of systemic workson p. 315.) All we are given is a bibliography of more or less relevant
systemic publications at the end of the book, without any guidance as to theareas they cover. We .are not even told that large portions of the book are
very similar to things published by Halliday himself elsewhere; for example,
chapters 3 to 5 are very similar in content to his seminal articles, 'Notes ontransitivity and theme', published in three parts as Halliday (1967a, 1967b);and chapter 9 summarizes the contents of Halliday and Hasan (1976).
Whether or not one takes the lack of scholarly references as a shortcomingis a matter of opinion, and even of taste. It is true that the book is an
introductory one, based on Halliday's lecture notes, and in this context itcould be argued that references are out of place. Nevertheless my opinion is
that it was a mistake to omit them. One reason is that other linguists are likelyto turn to IFG as a convenient source of information on the current state of
Halliday's thinking on English grammar (I certainly approached it myself inthis way). For such readers the lack of references is disastrous, as they lose all
time perspective. Is this all new, or has some of it been published before?
Those who are not familiar with Halliday's earlier publications have no way
of knowing that many of the ideas were first published up to 20 years ago andmight be forgiven for thinking that in some cases they are predated by work
which in fact was itself influenced by them. (I think for example of the workby Dik on semantic relations in Dik 1978.) The lack of references is also
unhelpful for beginners, who will get no sense of historical perspective andno encouragement to read further.
7/28/2019 sgrev
7/26
Since the material in this book has clearly been gestating for a long time,
one would expect the successive reworkings and revisions (which arementioned in the Foreword) to have ironed out all serious inconsistencies.
Curiously, this does not seem to be the case. For example,
- The Greeks started with morphology according to p. xiv, but with rhetoricaleffect according to p. xxiii.
- Every independent clause selects for mood, and some such clauses are minor,
for p. 44; but for p. 61 minor clauses do not select for mood. - In the sampletext on p. 65 he marks off two relative clauses but misses two more (he could
do about itand he could offer).- On page 229 he introduces the term 'metaphenomenon' as though it had not
previously been used ('We will refer to this as a "metaphenomenon".'); but itactually occurs frequently on previous pages (such as p. 227).
- There is a section headed 'Tonicity' (p. 273) which does not mention tonicity once,either by name or otherwise.
If a second edition of the book is produced I hope the publishers will consider
serious revisions, to cover slips like these, and the production of
an index at the very least.A deeper problem with the presentation is the level of argumentation
or rather, the lack of it. If one were asked to pick out the most importantdifferences between modem linguistics and traditional grammar, I think most
of us would agree that the list should include the fact that we now see that
analyses are not God-given but need to be justified, whereas traditionalgrammar tends to take the form of dogma passed on from generation togeneration. Any book, whether introductory or not, should show how the
analyses have been arrived at; indeed, this is arguably even more important inan introductory book than in a scholarly monograph, since one is trying to
bring the novice to some understanding of how linguists work. Otherwise we
are simply perpetuating the era of dogma in which traditional grammar
flourishes. But in IFG it is hard to find any passage which suggests that theanalyses presented are at all problematic, less still any which attempt to justify
these analyses in relation to the alternatives which are available.The ex-cathedra tone would perhaps be justified if the analyses were self-
evidently right, and it would be unreasonable to ask for step-by-stepjustification of every single analysis (I suspect this would lead to an infinite
regress and paralysis in the writer, not to mention boredom in the reader).However, this is hardly the case in IFG - in fact Halliday warns us (p. xxxiv)
that his descriptions break with the established tradition, so we might expect a
good deal of discussion of tJ:leir pros and cons. Let me again give some
examples of places where his analyses and claims seem to
7/28/2019 sgrev
8/26
be less than obviously right - in fact, where they seem to face fairly obvious
problems - and where some attempt at justification would have been in order.- He claims that speech functions are of four kinds, OFFER, COMMAND,
STATEMENT, and QUESTION, and that these are related on two
dimensions: giving versus demanding, and involving goods-andservicesversus information (p. 69). This classification means that OFFER and
QUESTION are maximally different (i.e. differ on both dimensions), so why
is it that both are realized, in contrast with the other types, as interrogativeclause structures?
- He tells us that common noun, adjective, numeral, and determiner are allkinds of NOUN (p. 164). This mayor may not be a good analysis, but at least
it conflicts both with the traditional analyses and with other current ones, sosome explanation is called for.
- He analyzes constructions containing direct speech as cases of parataxis(two equal parts, without subordination); so Brutus said: 'Caesar was
ambitious'. has a structure similar to that of a pair of coordinated clauses (p.
228). If the two parts are equal, why is there an asymmetry between them,
such that the reported may be enclosed in the reported (e.g. 'Caesar,' said
Brutus, 'was ambitious'.), but not vice versa (* Brutus 'Caesar wasambitious', said.)? and why can't the reporting occur without the reported (asin *Brutus said.)?
Many of his analyses involve symmetrical tables whose very symmetry
could perhaps be taken as a justification of the analyses; but this symmetry issometimes questionable, as in the table on p. 113 whose symmetry dependson the assumption that the reversibility ofTom is the leader(cf. The leader is
Tom) is found also in The piano is Peter's, compared with Peter's is thepiano. But why, then, is it possible to say The
piano is Peter's piano,but not *Peter's piano is the piano (at least, not with
anything like the predicted meaning)?
He explains the position of the WH element at the start of a question bysaying that it is the 'theme' - that is, what the question is about - but omits tomention that precisely the same could (presumably) be said of the disjunctiveelement in an alternative question (such as Was the baby a boy or a girl?),which does not occur in initial position - so is the disjunctive element a theme in anodd position, or something which ought to be theme but isn't? Indeed, there is
no mention of alternative questions in the whole section on interrogatives,which is odd considering the amount of text that one assumes has been
analyzed according to his system.
It is true that Halliday recognizes that analyses must be justified, as witness
his claim (p. xx) that he could justify all the categories which he refers to byspelling out how they are manifested in terms of syntactic and
7/28/2019 sgrev
9/26
798 R. Hudson
lexical items. He refers to 'an important principle, namely that all the
categories employed must be clearly "there" in the grammar of the language.
They are not set up simply to label differences in meaning.' It is simplypressure of space that prevents him from including all the details of
realization, he says; the details are all 'stored in a computer' (p. xv). This isencouraging news, but surely it is not enough. We also need to know that for
each construction he has arrived at the best possible analysis, considering and
rejecting a reasonable range of alternatives. No doubt this has happened inHalliday's work too, but we have no evidence of it in IFL, nor is there anyattempt to anticipate predictable objections to the less standard analyses.
Another related problem arises, in connection with the justification ofanalyses. There are many places in the book where he makes a claim which I,
for one, was completely unable either to agree with or to disagree with
because the categories concerned were so vaguely defined that I could not
reliably identifiy instances of them. Take the important category 'theme', forexample. The following definitions are provided: 'The theme is what the
clause is going to be about' (p. 39). 'The Theme is what the speaker selects ashis point of departure, the means of development of the clause' (p. 53).
Presumably he finds these definitions adequate, because he is able to pick outthe themes in any clause and work out the general rules which govern their
use - for example, 'The Themeof any clause... extends up to (and includes)the topical Theme. The topical Theme is the first element in the clause thathas some function in the ideational structure (i.e. in transitivity...)' (p. 56).
Perhaps he is tuned into language in a way that the rest of us are incapable of,but those of us who can't easily pick out the parts of a clause which define
'what it is going to be about', or its 'point of departure' are simply unable to
decide whether any of his claims about themes are right or wrong. Whatcould one give as a counterexample? Indeed, one wonders what criteria he
himself is using to identify themes when one reads that the subordinatingconjunction thatis a theme in the clause it introduces (p. 51) - it is very hardto see it as defining what the clause is going to be about, and if it is 'the pointof departure' of the clause, it can only be in the sense of being the first
element. Somewhat confusingly, he himself seems to deny it the status of
theme when he says (p. 61) that 'if what comes first is "fixed", ... what comessecond may retain some thematic flavour' - the implication being,presumably, that what comes first (for example, that) lacks 'thematic flavour'?
One also wonders precisely how the computer holding all the details copeswith some of these problems.
Very similar problems arise with several other key categories, of which thefollowing are just a sample:
1
7/28/2019 sgrev
10/26
Systemic grammar 799
'subject': 'the element that is held responsible, in which is vested the success
of the clause in whatever is its particular speech function' (p. 36); 'somethingby reference to which the proposition can be affirmed or denied' (p. 76).
'given': 'information that is presented by the speaker as recoverable' (p. 277);'what you, the listener, already know about or have accessible to
you' (p.278) (compare 'definite': 'you know which one[s] 1 mean' - p. 292).'reference': something which 'can be taken as a reference point for something
which follows' (p. 288).The only way in which 1 could apply any of these categories in the analysis of
texts would be by looking at the (plentiful) examples which Halliday givesand working out my own rules of thumb. 1 doubt if at the end 1 could
honestly claim to have 'understood' either the language or the texts any better
for this exercise.
The use of ill-defined categories makes it very hard for Halliday to achieveone of his main goals, which is to explain in functional terms why English is
as it is. This goal is worthwhile - though 1 shall express some reservationsabout it below - but it can hardly be reached by means of explanations like the
following: 'The basic message of an imperative clause is 'I want you to dosomething' ... hence [sic] the unmarked Themeisyou... as inyou keep quiet'(p. 49). By what principle does the conclusion follow from the premise in this
case? The example is not isolated.These shortcomings are all the more regrettable because the book is in
many ways so impressive. One of Halliday's strengths is his ability to see
connections between apparently unrelated phenomena, and IFG is full of
cross-category generalizations. For example, he compares the ordering of
elements at the start of the clause with that at the start of the noun phrase (his'nominal group'):
So there is a progression in the nominal group from the kind of element that has the
greatest specifying potential to that which has the least; and this is a principle of
ordering that we have already recognised in the clause. In the clause, the Theme
comes first. We begin by establishing relevance: stating what it is that we are using to
introduce this clause into the discourse, ... In the nominal group, we begin with theDeictic: 'first I'll tell you which I mean' ... (p. 166).
The comparison may or may not be correct; but at least it is interesting andstimulating. Such unexpected comparisons proliferate, and similar remarks
apply to all of them. Halliday aims much higher than most of us and really
believes in the idea that a language is a system where everything
7/28/2019 sgrev
11/26
800 R. Hudson
hangs together. Everyone of his comparisons would provide the basis for a
series of testable hypotheses which could lead to some very interesting and
worthwhile research, and our understanding of English grammar wouldthereby have been deepened whether the research confirmed or refuted
Halliday's claims. Unfortunately, however, it seems unlikely that his ideaswill be widely exploited in this way simply because most linguists will be put
off by the shortcomings in their presentation.
Another positive feature of the book is the large number of individualpenetrating observations about particular areas of grammar. Some of them arequite general - for instance, with reference to the transitivity system of
English, he believes that English is organized around two conflicting patterns,the nominative-accusative and the ergative, and shows how the two patterns
interact with one another (chapter 5). An example of a valuable observation at
a less general level is the distinction that he makes between the 'attributive'
and 'identifying' uses of be (p. 114), exemplified respectively by Fred is
happy and Fred is the leader. These are often lumped together by other
linguists as 'the copula' although as Halliday shows there are fundamentaldifferences between them. Very particular points of fact abound too; for
example, he notes that He said so and He said that are both possible, andcould both refer back to the same preceding discourse, but present it
respectively in relation to its content and to its wording (p. 234).In summary, then, IFG is a disappointment. There are lots of good ideas
and insights which have been poorly developed and presented. lethe intended
user is the nonlinguist with a need to analyze texts grammatically, IFG maybe helpful, but I doubt it - it would be very hard to use as a reference manual,too many categories are vague, and too many analyses are debatable. If the
debatable analyses turn out to be wrong (in some sense), then I suppose itgoes without saying that it is worse than useless to apply them to texts. If on
the other hand the intended reader is the professional linguist or student oflinguistics, it is unlikely that IFG will make much impact.
3. The aims of systemic grammarians
I think it would not be unfair to say that most people who are attracted to SG
are people who enjoy doing things with language, rather than theorizingabout it at a distance. The theory makes this possible by providing a rather
flexible sorting device, the system network, which can be used for classifyingand cross-classifying. Armed with this machinery one can tackle portions of
language of any size or type and be guaranteed
7/28/2019 sgrev
12/26
to be able to systematize them to some degree. I see nothing wrong with this
enthusiasm for systematizing data in itself- and indeed, I confess to sharing itmyself. When I was a student of Halliday's I never failed to be impressed by
the classifications - usually cross-classifications - which he produced forotherwise unorganized facts, and I derived enormous satisfaction from doingsimilar things in my own work. It would, I think, have been very much harder
to achieve the same degree of satisfaction within any of the other frameworks
available at the time (the main alternative being a pre-1965 version oftransformational grammar).
If simple systematization is accepted as a goal, then it becomes quitereasonable to aim at an 'exhaustive' coverage of data - in fact, a 'complete'grammar of the language concerned. This is made even more reasonable byone of the general principles which has always informed work in SG, which is
that there is no need to aim for maximum 'delicacy' (that is, amount of detail).The systematization can stop once the main categories have been identified,
as there is no sense in which the subdivisions of these categories are
necessary for the validity of the classification. It immediately follows from
this that exhaustive textual analysis becomes possible, because it is possible toprovide a pigeonhole for every part of the text. And one of the main selling
points of SG hasalways been that it is easy to use for analyzing texts. (This is the main point,
for example, of Halliday 1964.) However, a number of questions arise about
the validity of this assumption.One is about the nature of the categories which should be used in thetextual analysis. SG makes a fundamental distinction between 'features' and
'functions' (in the sense of 'grammatical functions', that is, what some peoplecall grammatical relations such as 'subject'). What Halliday presents in IFG is
the functions (hence one part of the meaning of 'functional' in the title of the
book), but Butler argues repeatedly (for example, p. 197f.) that the value
ofSG for text analysis lies in the way that SG presents the 'meaningfulchoices' available to the speaker - which implies that the most relevantcategories are the features, because only these are related as meaningfulchoices. Functions are simply the surface
manifestation of these choices. It is odd that the practitioners themselvesdisagree on such a fundamental point.
Another question is about the issue which I raised at the end of theprevious section. Is there any point in applying a classification system unless
one is reasonably confident that it is in some sense 'right'? As we have seen,
Halliday himself stresses that the categories must in some sense be 'there' as
part of the grammar, so the question arises how one evaluates alternativesystematizations of the same data. It seems obvious that iftwo people analyze
the same text using different classification systems they
7/28/2019 sgrev
13/26
802 R. Hudson
may come to different conclusions about it. (And it goes without saying that
any analysis of text - other than a mere training exercise - must lead to someconclusion about the text, otherwise it is pointless.) A really broadminded
approach might seem to be to accept both conclusions as equally valid waysof viewing the data, but this is surely just woolly thinking when presented as
a general principle. For example, suppose we want to compare two texts Aand B, and one of the dimensions of comparison is their degree of
'nominality' (a topic on which Halliday has had much of interest to say - seeHalliday 1967a, 1967b), then it becomes crucial to decide correctly what
should count as a noun and what as a verb. Otherwise the ludicrous situation
could arise where one analyst claims that A is more nominal than B, and
another makes the reverse claim.The question then arises, how do we decide what counts as a 'true'
analysis? Or put less ambitiously, how do we decide which of two competinganalyses is the nearer to the truth? I don't think this is a question which has
been addressed seriously in SG theorizing. A number of different (butpossibly complementary) answers seem to have been given or assumed:
a. One based on elegance: An analysis is correct if it is elegant, and inparticular if it can be laid out as a matrix in which all the cells are filled. Such
tables are abundant in IFG and seem to be used at least implicitly as evidenceof validity. They can indeed be very impressive (and there is a particularly
impressive one on pp. 306-307 which brings together a vast number of
apparently disparate phenomena). However, as I pointed out in the previous
section, they are no better than the individual analytical decisions which liebehind the filled cells, so these decisions must be justified in their own right,
rather than by the fact that they conveniently fill an otherwise empty cell. Theconclusion then must be that elegance of this type does not vouch for validity.
b. One based on explicitness, orgenerativity: An SG grammar can be justas explicit as one formulated in terms of any other theory. Thus it is possible,
at least in principle, to write a SG grammar so that it is crystal clear whatstructures it generates and what it does not. Evidence for this comes partly
from the work of linguists (most obviously, perhaps, Hudson 1971), but more
especially from a variety of computer implementations of SG (Winograd1972; Davey 1978; Mann 1982). As noted earlier, Halliday says that the
grammar of IFG is the functional part of a larger grammar held in a computer,
a reference (I assume) to the system developed by Mann and his colleagues inCalifornia. However, it should be obvious that if you are clear it is just as
possible to be wrong as right; the only difference between clarity andvagueness is in the degree to which
7/28/2019 sgrev
14/26
Systemic grammar 803
it is possible to check for correctness. So explicitness, or generativity, in itself
is certainly a precondition for being right, but it is in no way a guarantee of
rightness. This elementary point is worth stressing because it appears to becalled into question by Halliday's remark in IFG (p. xx) that there is a
connection between being able to 'spell out all the steps from meaning towording' (that is, being explicit) and being sure that all the categories are
'clearly "there" in the grammar of the language' (that is, being right).
It is odd, incidentally, to find Halliday justifying his grammar by referringto the explicit rules known to the computer, in view of his remarks aboutChomsky's claim that a grammar can be generative (p. xxviii): v
By generative he meant explicit...His tremendous achievement was to show
that
this is in fact possible with a human language, as distinct from an artificial
'logical' language. But you have to pay a price: the language has to be soidealized that it bears little ,relation to what people actually write - and still
less to what they actually say.
So does IFG present part of a generative grammar, or not? If it does, how
does it manage t avoid the idealization which Halliday claims (wrongly, I
think) afflicts Chomskyan linguists? And if it does not, why mention the
existence of the rest of the grammar which provides the explicit links up tosemantics and down to surface structure?
c. One based on psychological reality: It is obvious, and accepted by
Halliday, that language is stored in people's minds, although the main interestof many systemic linguists (including Halliday) has always been in language
as a sociological phenomenon ('The orientation is to language as social ratherthan as individual phenomenon, and the origin and development of the theory
have aligned it with sociological rather than psychological modes of
explanation' p. xxx). It is not at all clear to me why it should make anydifference whether the linguist is more interested in 'society' or in the
individual; in either case one's object of study is the language of individuals,
and in either case those individuals acquire and use their language in relationto other members of their society. And in either case, the question arises as to
precisely what knowledge the individuals have which counts as theirlanguage. In this perspective, the ultimate criterion of success in linguistic
analysis must be whether the structure of the linguist's grammar matches thatof the relevant bits of the individual's knowledge. My own current work on
word grammar (Hudson 1984, 1985a, 1985b, 1986) is based on this criterion,but so is that of at least one systemic linguist, Fawcett (1980). It is true that
the structure of knowledge is hard to investigate in itself, but it is not totally
unknown
7/28/2019 sgrev
15/26
territory, and at least it is good to know what the grammarians are trying to do
even if we can't always be sure whether they are succeeding.There is one more question to be raised about the aims of systemic
grammarians, relating to the word 'function', Halliday says that the
'conceptual framework' on which IFG is based - by which he presumablymeans SG - is 'a functional one rather than a formal one' (p. xiii). The implied
contrast is between SG, which is functional (and is often called 'systemic
functional grammar', as noted earlier), and other theories in the Americanstructuralist tradition such as Chomsky's, which are 'formal'. His explanation
of the distinction points to three different senses of 'functional' (p. xiii), but ineach case I cannot see any important difference between his goals and those
of Chomskyans (though there is no doubt that his interests and emphases aredifferent):
a. SG 'is functional in the sense that it is designed to account for how thelanguage is USED'. But the same is true of a Chomskyan account of
language; they too want to explain language use, in terms of a theory of
performance of which a theory of competence is a part. Admittedly
Hallidayan and Chomskyan linguists fill in different parts of the picture:Halliday says a lot, for instance, about the effects of thematizationjtopi-
calization but nothing about unbounded dependencies (for example, What do
you think happened? versus * What do you think that happened?), and vice
versa for Chomsky; but does that amount to a theoretical difference, or just a
matter of personal research interests and priorities? Moreover, Halliday goeson, 'it is the uses of language that, over tens of thousands of generations, haveshaped the system. Language has evolved to satisfy human needs; and the
way it is organised is functional with respect to these needs - it is notarbitrary.' But it is at least arguable (and often argued) that humans have
needs not only as speakers and hearers, but also as language learners and
language storers; and that each of these pressures too has influence on the
structure of language (for example, pressures toward symmetry, simplicity,and so forth). Indeed, Halliday himself refers to the demand for symmetry as
'the most fundamental of all semogenic processes' (p. 251). So if the term'functional' is extended to include these other human activities, the difference
between Hallidayan and Chomskyan approaches dwindles to nothing.b. The second point is not too clear, but it seems to be that SG is based on
the assumption that 'the fundamental components of meaning are functionalcomponents' - that is, there are a small number of main parts into which thesemantic system of a language may be divided, and these parts are
distinguished by the 'very general purposes which underlie all uses of
language: (i) to understand the environment (ideational) and (ii) to act on theothers in it (interpersonal)'. This amounts to an empirical
7/28/2019 sgrev
16/26
Systemic grammar 805
claim about the nature of semantic structure, that it can be shown to fall
naturally into a small number of relatively independent components. Theclaim mayor may not be valid; but it is presumably one which could have
been made within any other theoretical framework. It is by no means obviousthat there is anything in SG which makes it specially suited to exploring or
formulating the claim.c. 'Each element in a language is explained by reference to its function in
the total linguistic system.' I take it that this is true of ANY account oflanguage structure, and it certainly does not distinguish SG from any other
theory.
I have labored the point about the meaning of 'functional' because it seems
to me that this is little more than a slogan for eliciting favorable reactionsfrom a particular kind of person - the kind of person who is struck by the
truism that language is a tool for communication, rather than an abstract andpointless formal system. SG seems to appeal to many people who find more
formal approaches harder to connect with their dayto-day experience, but thismay have just as much to do with the method of presentation (with real-life
examples and so on) and with the selection of areas of language to bediscussed as it does with the actual theories concerned. (Similar remarks
could be directed, incidentally, at the various other linguistic theories whosenames include the word 'functional'; a fair selection of these is included in
Moravcsik and Wirth 1980.)
4. The nature of grammatical structure
SG is a theory of language structure and (consequently) of sentence structure
as well. The most well developed areas of the theory are those to do with
rhythm and intonation; syntax; and certain aspects of semantic structure. Ishall concentrate in the following remarks on syntax.
As far as the view of sentence structure is concerned, SG is firmly withinthe American structuralist tradition started by Bloomfield. It has close
affinities with Pike's tagmemics (Pike 1982): a relatively flat version of
constituent structure, with each node carrying two labels (or sets of labels),one showing its class membership, the other its grammatical function. Onlysurface structure is recognized, and the part played in transformational
theories by underlying structures is played in SG by (a) the relatively richlabeling of constituents and (b) the relatively powerful mapping devices
which relate mothers to their daughters. I shall comment below on various
parts of this theory.
The units which are recognized in SG have structures of two types:'univariate' and 'multivariate'. Multivariate structures are those of single
7/28/2019 sgrev
17/26
806
words, phrases, or clauses. (I shall drop Halliday's distinction between
'phrase' and 'group', as nothing much seems to depend on it, and 'phrase' islikely to be more familiar to most readers.) Univariate ones are strings of
units in one of two relations to each other: 'parataxis' (for example,
coordination) or 'hypo taxis' - that is, subordination, or dependency. It isthislast type that I want to explore a little. Dependency is the relation which
Halliday recognizes, for example, between the following pairs of words (the
arrow points toward the dependent; this is not Halliday's notation):
very small (p. 173)
electric trains (p. 172)has eaten (p. 177)
is eaten (p. 178)
Now it is very hard to see any reason why the relation between
(say) has andfreckles, or that between is and happy, should not be
treated in the same way. As relations, they seem very similar tothose in the list above: in both cases, one word is clearly dependent
for its occurrence on the lexical properties of the other (namely,
have allows either a perfect participle or an object, and be allows
either a passive participle or an 'attribute'). However, the relations
between verbs and their objects or attributes are treated in SG interms of a completely different type of analysis: as cosisters of ashared mother, the clause. In this respect, of course, SG is a
completely orthodox variant of constituent-structure theory.
Let us pursue the difference between these two treatments of
grammatical relations. In the constituency-based treatment, thesisters are not related directly to each other, but only to their shared
mother, whereas the dependency approach (in SG as elsewhere)does establish direct links between sisters. This means that each of
the terms of a dependency relation is free to be related, bydependency, to another item, without the need to recognize any
constituent structure. The following dependency structures aregiven by Halliday, for example:
rather more impressive(p. 171)
has been eating (p. 177)
Suppose we were to treat objects and attributes in the same way;
we would then have dependency structures like the following:
has been eating peanuts
has been happyvery
An analysis of this kind is clearly very much against the spirit not
only of SG but of all constituency-based theories. And yet it is hard
to see any
7/28/2019 sgrev
18/26
Systemic grammar 807
good reason why it should be rejected, once dependency analyses are
admitted as part of the descriptive apparatus. It is true that the relation of
happy to been is different from that of very to happy, but at the level of
delicacy where we are simply representing them as dependency relations,
there is no difference. (The differences cannot be ignored; I return to them
below.) The same applies to the difference between been happy and been
eating. So the question is how far we can or should develop these
possibilities. I have argued elsewhere that dependency should be taken as the
basis for analyzing all syntagmatic relations except coordination (Hudson
1984), because all such relations are asymmetrical and can be expressed as
relations between SINGLE WORDS. For example, in been very happy, it is
just the word happy that is related to been, and not the phrase very happy.
Similarly, we could extend the analysis above by adding in the subject:
He has been happy.very
What I am suggesting (without at this stage providing anything more thannegative evidence in support of the suggestion) is that everything that needsto be said about the relations among these words can be said about them
WITHOUT REFERRING TO ANY UNIT LARGER THAN THE WORD. Thus if somerelation involves the whole clause, it can be expressed as a relation to the
word on which all the other words in the clause depend, which is has. Forexample, we can add because at the beginning of the whole clause, as a link
to whatever clause precedes it; and in the main clause we can pick out the
verb again as the bearer of the relation :
He is lucky
because has been happy
he
Somewhat surprisingly we find ourselves back now in an orthodox positionvis a vis SG, because SG also treats relations between main and subordinate
clauses in terms of dependency. The following sentence is analyzed on p.193, for example:
It won't be surprising if people complain if they don't punish him if he'sguilty.
In the SG analysis, the clauses are related in a chain of dependencies. As we
have seen it is possible to refer just to the verbs and the subordinating
conjunctions instead of the clauses, so a dependency analysis ignoring the
irrelevant words would be like this:
won't if complain if don't if 's
7/28/2019 sgrev
19/26
808 R. Hudson
All these observations may strike the reader as simply perverse, because
one of the most widely accepted tenets of modern linguistics is the belief inthe fundamental importance of constituent structure. Replacing constituent
structure by dependency structure has many advantages (Hudson 1984), but Ishall simply make some observations which relate directly to the account of
SG presented in IFG. First, it is interesting to be reminded that SG alreadyallows dependency to play such an important role. Second, Halliday himself
repeatedly stresses that constituent structure is relatively unimportant; forexample, 'As one explores more deeply into language, constituency graduallyslips into the background...' (p. 18). Third, the constituent structures allowed
by SG are extremely unconstrained: constituents may be discontinuous (p.
81), they may overlap (p. 72), and several conflicting constituency analysescan be applied simultaneously to the same sentence (p. 121). Not surprisingly,
perhaps, very few of the constituents thus identified do any work in thegrammar - in particular, few of them carry any features, and few of themhave definable internal structures.
And fourth, no clear principle seems to govern the choice between
dependency and constituency as the basis for analyzing particular con-structions. Take the sentence which is analyzed on p. 194:
I don't mind if you leave as soon as you've finished,
A | B | C |as long as you're back when I need you.
D I E
The relations among the first three clauses are shown in terms of puredependency - in particular, no constituent is recognized for the pair B + C,though one is recognized for A + B + C. Clauses D and E, however, are takenas a constituent. It is not clear what principle, if any, underlies this
distinction. As far as dependency analysis is concerned, there is no need forit, so long as we show both Band D as dependents of A:
don't if leave as soon as finished
as long as 're when need
Indeed at times total confusion seems to reign in this part of SG, because
the dependency relation is given a completely different interpretation whenapplied to the internal structure of the noun phrase. Take the sequence very
small one; to judge by the structure given on p. 173, this should have verysmallas a constituent, although very depends on smalland smallon one in
precisely the way clauses A, B, and C above are related - where nointermediate constituent is recognized at all. When an unbracketed string of
dependents does occur within the noun phrase it is
7/28/2019 sgrev
20/26
Systemic grammar 809
interpreted as though there were bracketing 'from the top' (instead of 'from the
bottom', as in clause sequences). Thussplendid old electric trains is shown (p.
170) as having the following dependency structure:
splendid old electric trains
If we interpret this structure in the same way as we would a clause sequence,
it makes nonsense (with splendidmodifying just old, as though it were some
kind of adverb like very). Conversely, if we apply the interpretation demandedby this analysis to clause sequences, we again get nonsense. One conclusionwhich is reasonable is that perhaps it would be better if no constituent
structure were ever mixed up with dependency structure. If that had been the
case the problem could not have arisen.
The problem of sorting out the formal relations between dependency andconstituency looms large in SG, because both kinds of analysis are available
for showing that one clause is subordinate to another. The analyses offered inIFG are unimpressive in this respect. A subordinate clause which functions as
subject in the main clause is considered to be a part of its structure, but onethat is object is not (p. 196). One wonders (inter alia) how the rules for
passivization can be formulated under these circumstances. It would clearlybe more satisfactory for the relations to be treated in the same way unless
there is very clear evidence of differences, and one way of achieving thiswould be to abandon constituent structure entirely, leaving all the work todependency.
It would be misleading to imply that this step could be taken without far-
reaching ramifications for other parts of the theory. Other, equally drastic,changes would be needed - abandonment ofthe 'rank scale' (the hierarchy of unitsfrom 'word' to 'sentence'), and abandonment of realization rules (the ruleswhich map an item's features onto a suitable internal structure), for instance.
Sometimes major surgery is the only route to a fuller and richer life.
5. Categories
As noted earlier, two different kinds of category are distinguished in SG:features and functions. I consider this to be one of the most positive
characteristics of the theory, and certainly one which has contributed a lot tomy own thinking. A feature is a paradigmatic category, which relates the item
in question to other items in the language which are similar in the relevantrespect. A function is a syntagmatic category which, as its name implies,
shows the function which the item carries out in the sentence in which itoccurs. For example, an item might have the features [singular,
7/28/2019 sgrev
21/26
810R. Hudson
nominal] and the function SUBJECT. Every item (except the main clause itself) has at least one feature and at least one
function, even if the two are in a one:one relation (as is often the case).In some respects the SG treatment of categories is superior to what we find in other current theories. Take features.
Syntactic features are one of the weakest areas of many theories, because of the lack of constraints on them. In the1970s it was common to invoke syntactic features as a deus ex machina for fixing up a transformational grammar - for
example, for dealing with exceptions - but they hardly seem to play any part in the current version of transformationalgrammar, government and binding theory (Chomsky 1981), except as the latter-day incarnation of the major word-class
categories (N, V, etc.). And yet the same apparatus of features still seems to be available in principle. The theory inwhich features play the most prominent part is generalized phrase structure grammar (Gazdar et al. 1985). Several
different types of feature are distinguished, some of which are allowed to have rich internal structures, but there is no
discussion of restrictions on the combinability of features.
In SG, on the other hand, features have been one of the main focuses of interest since the early days, and the 'system',the cornerstone of the theory, was developed specifically to describe the ways in which features can combine with one
another. Systems can be subjected to various formal constraints which make it impossible to generate certain types offeature combination. If two features occur as part of the description of an item, it is always because a system network
has permitted them to cooccur; features cannot be slipped in as ad hoc rescue devices. Nor can they have any kind ofinternal structure - their names are always arbitrary and unanalyzable. Linguists of other persuasions who are concernedabout the unconstrained ways in which features are used might well find interesting ideas in the SG literature.
Functions are comparable with the 'grammatical relations' of other theories, and SG is one of the theories which
indicates these explicitly (in contrast with both transformational grammar and GPSG). One of the positive contributionsthat SG can make to the pool of ideas from which linguistic theories can draw is the idea that a single item can have
more than one function - to take an obvious example, who in Who came? can be taken simultaneously as subject and asinterrogative theme (that is, 'complementizer', in other people's terminology). It is generally assumed in other theories
which deal in grammatical relations that they should not combine with one another (and a comparable assumptionapplying to semantic relations is explicit in government and binding theory in the form of the 'theta criterion' - Chomsky
1981: 36). Once again SG seems in advance of other theories, since multiple functions clearly are needed.
7/28/2019 sgrev
22/26
811
Take for example a sentence like He drinks coffee black. There is at least a prima facie case for saying that coffee hasa dual function in both the syntax and the semantics: as object ofdrinks and as subject ofblack. We return to this point
below.
I have two suggestions for improving the SG treatment of functions. One is to allow them to be relatedhierarchically (as indeed was the case in the very earliest version of the theory, 'scale and category' theory - see
p. 19 of Butler). For instance, the function INDIRECT OBJECT would be shown as a particular case of OBJECT
which in turn would be a kind of COMPLEMENT. These hierarchical relations could easily be shown by means ofsystems, just like the hierarchical relations among features; at present this is not permitted by the theory but I see no
reason in principle why it should not be. A number of advantages would follow from a hierarchical organization offunctions, notably that the functions higher in the hierarchy could always be predicted from those lower down; so as
soon as some element is defined as INDIRECT OBJECT it would automatically follow that it was also OBJECT andCOMPLEMENT. At present I think it would be necessary to specify all three functions independently.
Given this proposal, what would we put in the functional hierarchy above categories like COMPLEMENT? In thelight of my previous suggestion about dependency, the answer is obvious: all functions are a particular case ofDEPENDENT. Moreover, we could even extend the hierarchy up higher still, by introducing 'relational category' toinclude both HEAD and DEPENDENT, and maybe even 'concept' above this. (These suggestions have in fact beenimplemented in my own recent work, but I shall have to leave them as programmatic ideas in this review.) The resultinghierarchy looks something like the following:
My second suggestion is that functions should indeed be treated as relational categories, in line with this proposal. It
could be objected that this is already how they are treated, by virtue of being distinguished from features, but this is not
so. It is true that in SG grammars (such as the one
concept
relationalcategory
feature
headdependent
subject complement
object predicative
direct indirect
7/28/2019 sgrev
23/26
812 R. Hudson
in Hudson 1971) functions are introduced in a different way from features
(functions are introduced by rules which realize features, and features by ruleswhich realize functions), but as far as the sentence structures thus generated
are concerned their status is just the same. Each could be represented as a one-place predicate (for example, 'is a nominal', 'is a subject'). In some other
theories, such as relational grammar and dependency theory (including myown word grammar), relational categories are treated differently from
features, as two-place predicates. Thus instead of saying simply 'X is asubject', we should say 'X is the subject of Y'. This approach is clearly morein line with the relational nature of relational categories, and it brings various
advantages with it. In particular it allows any rule which refers to a relational
category to have access to BOTH of the relata. This means, for example, thatwe can distinguish between the notion 'modifier of a common noun' (such as
nice (person]) and 'modifier of an indefinite pronoun' (such as [someone]nice) without having to introduce any additional relational category for the
purpose. Had I treated functions in this way in my 1971 book, I should haveneeded far fewer than the outrageous 90 functions that I had to postulate
there!We now return to my earlier observation about the need for multiple
functions, in connection with the example He drinks coffee black. Let us
assume that functions are truly relational categories, whose specificationinvolves reference to a second element, and further that they are a type ofdependent whose relata are single words. In order to distinguish the different
kinds of dependent from one another we can write an appropriate symbol onthe dependency arrow ('S' for 'subject', 'A' for 'attribute', and so on). The
analysis of the sentence is as follows:
He S drinks O coffee S
black
A
The analysis with old-style one-place functions would be more complicated.First, it would be necessary to distinguish two different kinds of attribute,
according to whether they had the subject or the object of the verb as theirown subject; otherwise there would be no way of disambiguating sentences
likeHe left her angry (it was in lectures by Halliday that I first became awareof this problem). And then rules would be needed for interpreting this
distinction in terms of semantic links between the attribute and the relevantother element.
These changes in the SG treatment of functions have brought us to a
sophisticated dependency-based theory in which (for instance) multiple
dependencies are permitted. I have pointed out just a few of the manyadvantages of adopting such a theory, and I am not aware of any
7/28/2019 sgrev
24/26
Systemic grammar8 I 3
compensating losses. It has always been a characteristic of SG to favor an
approach where the hierarchy is relatively flat rather than the typically many-layered structures of phrase-structure grammar (p. 24 of IFG; Hudson 1967);
the move toward dependency can be seen as the logical conclusion of thistendency, where the constituent hierarchy is totally flat. I leave it to others to
present counterarguments.
6. Butler's Systemic Linguistics: Theory and Applications
Butler's book (SLTA) is very different from IFG. It sets out to present a
factual and critical account of the development of the theory of SG, sodescriptive matters come in only incidentally. The whole of Halliday's
analysis of English is contained in a single chapter - a very useful summaryfor anyone planning to embark on IFG. SL T A aims to cover all work in SG,
and not just that of Halliday, so it includes a chapter on the work of Fawcettand myself and sections on such topics as computer implementations ofSG
and the discourse work of Sinclair and Coulthard.
Butler is himself one of the main advocates of SG, but the book is
extremely well balanced and nonpartisan. It relates work in SG to movements
in other theories and includes a good number of references to work by non-SGlinguists. Moreover it contains a large number of critical comments about SG,all of which are well taken. Some of the criticisms contained in the present
review duplicate criticisms which Butler makes, and SLTA contains a goodnumber of other negative remarks about SG. As mentioned earlier Butler
complains of the lack of critical comment from outsiders (p. 77). Maybe thisdoes not matter as long as insiders like Butler are as honest and objective
about the theory in their criticisms. (Another penetratingly honest review ofSG by an insider is Berry 1980.)
I have few complaints of substance about SL T A. In my opinion the last
chapter could well have been omitted - it is an attempt to evaluate SG by the criteriawhich Moravcsik proposed at the comparative conference on currentapproaches to syntax (Moravcsik 1980), and I found it rather unhelpful. It
largely duplicated comments made earlier in the book and was hard to readthrough as continuous prose because it followed Moravcsik's structure too
closely.In the list of descriptions of English based on SG (p. 191) he could have
included Huddleston et al. (1968) and Huddleston (1971), which is based onthis; and of course he could also have mentioned the work of Quirk and his
colleagues (such as Quirk et al. 1972), which was heavily
influenced by Halliday.
In his account of my own work I found only one point where he got my
7/28/2019 sgrev
25/26
814 R. Hudson
present position wrong (though I can't blame him for that, as I have writtenthings which lend support to his interpretation). On p. 128 he implies that I
see 'lexis' as a separate kind of linguistic patterning from syntax, in contrastwith another view prevalent in sa which is that lexis is just 'most delicate'
syntax. In fact it is the second view which is nearer to my current position -that is, I believe that individual lexical items are at the bottom of the same
hierarchy which contains the general syntactic categories like 'noun' and
'auxiliary verb' (Hudson 1984: 24).These complaints are clearly very minor indeed. In short, I thoroughlyrecommend Butler's book. I recommend it to outsiders who want a clear and
fair presentation of sa theory. And I recommend it (particularly) to insiders,who will find in it an excellent account of the theoretical issues and problems
which confront them. The problems are clear and can hardly be ~gnored; it isup to the insiders to solve them.
Received26March 1986 University College London
Note
* I should like to thank Chris Butler for comments on an earlier version of this review.Correspondence address: Linguistics, University College London, Gower Street, London
WCIE 6BT, England.
References
Berry, M. (1980). 'They're all out of step except our Johnny': a discussion of motivation (or
the lack of it) in systemic linguistics. Unpublished mimeo. Chomsky, N. (1981). Lectures on
Government and Binding. Dordrecht: Foris. Davey, A. (1978).Discourse Production: A
Computer Model of Some Aspects of a Speaker.
Edinburgh: Edinburgh University Press. Dik, S. C. (1978).Functional Grammar. Amsterdam:
North Holland. Fawcett, R. P. (1980). Cognitive Linguistics and Social Interaction: Towards an
Integrated
Model of a Systemic Functional Grammar and the Other Components of a Communicating
Mind.Heidelberg: Groos.Gazdar, G., Klein, E., Pullum, G., and Sag, I. (1985). Generalized Phrase Structure
Grammar. Oxford: Blackwell. Halliday, M. A. K. (1961). Categories of the theory of
grammar. Word17, 241-292.
---(1964). Syntax and the consumer. In C. I. J. M. Stuart (ed.),Monograph Series on
Languages and Linguistics 17, 11-24. Washington, DC: Georgetown University Press. ---
(1967a). Notes on transitivity and theme in English, part 1.Journal of Linguistics 3,37-82. ---
(1967b). Notes on transitivity and theme in English, part 2.Journal of Linguistics 3,
199-244.
---(1968). Notes on transitivity and theme in English, part 3.Journal of Linguistics 4,
179-216.
7/28/2019 sgrev
26/26
Systemic grammar 815
-, and Hasan, R. (1976). Cohesion inEnglish. London: Longman.
Huddleston, R. D. (1971). The Sentence in Written English: A Syntactic Study Based on an
Analysis of Scientific Texts. London: Cambridge University Press.
-(1984). An Introduction to the Grammar of English. London: Cambridge University Press. -,
Hudson, R. A., Winter, O. E., and Henrici, A. (1968). Sentence and clause in scientific
English. Unpublished mimeo.
Hudson, R. A. (1967). Constituency in a systemic description of the English clause. Lingua
18,225-250.
-(1971). English Complex Sentences: An Introduction to Systemic Grammar. Amsterdam:
North Holland.-(1973). An 'item-and-paradigm' approach to Beja syntax and morphology. Foundations of
Language 9, 504-548.
-(1974). Systemic generative grammar.Linguistics 139, 5-42.
-(1984). Word Grammar. Oxford: Blackwell.
-(1985a). The limits of subcategorisation.Linguistic Analysis.
-(1985b). A psychologically and socially plausible theory of language structure. In D .
Schiffrin (ed.),Meaning, Form and Use in Context: Linguistic Applications, 150-159.
Washington, DC: Georgetown University Press.
-(1986). Frame semantics, frame linguistics, frame. ... Quaderni di Semantica 7, 95-111. McCord,
M. C. (1975). On the form of a systemic grammar. Journal of Linguistics 11,
195-212.
Mann, W. C. (1982).An Introduction to the Nigel Text Generation Grammar Computer
Program. Information Sciences Institute, Technical Report. University of Southern
California.
Moravcsik, E. A. (1980). Introduction: on syntactic approaches. In
E. A. Moravcsik andJ. R. Wirth (eds.), Syntax and Semantics 13: Current Approaches to Syntax, 1-18. New
York: Academic Press.
-, and Wirth, J. R. (eds.) (1980). Syntax and Semantics 13: Current Approaches to Syntax.
New York: Academic Press.
Pike, K. L. (1982).Linguistic Concepts: An Introduction to Tagmemics. Lincoln and London:
University of Nebraska Press.
Quirk, R., Greenbaum, S., Leech, G., and Svartvik, J. (1972). A Grammar of Contemporary
English. London: Longman.
Stockwell, R. P., Schachter, P., and Partee, B. H. (1973). The Major Syntactic Structures of
English.New York: Holt, Rinehart and Winston. Winograd, T. (1972). Understanding Natural
Language.New York: Academic Press. -(1983). Language as a Cognitive Process I: Syntax.
Reading, Mass.: Addison-Wesley.