A FIRST STEP INTO THE
M. A. K. Halliday
© Matthiessen & Halliday. Please do not copy or quote without authors' permission.
1. Into systemic-functional theory of grammar
1.1 General: [lexico]grammar & the study of grammar ('grammatics')
This is an introductory account of a particular theory of grammar, namely systemic-functional
theory. Grammar is one of the subsystems of a language; more specifically, it is the system of
wordings of a language. It is a phenomenon that can be studied, just like light, physical motion,
the human body, and decision-making processes in bureaucracies; and just as in the case of
these and other phenomena under study, we need theory in order to interpret it. So for
instance, the physical phenomenon of the atom has been interpreted theoretically in terms of
Democritus' theory, Rutherford's theory, Bohr's theory, and so on. We distinguish between the
phenomenon itself (the atom) and various theoretical models of it. What kind of thing the atom
is thought to be will of course vary considerably as we move from one theory to another.
Democritus' atom was very different from Bohr's atom, in that it was indivisible, not a
configuration of subatomic particles; that is, Democritus' theory allowed us to see much less of
the atom than Bohr's theory does. A well-known example of the way theory determines how
we interpret phenomena is light. Light can be interpreted either as particle or as wave; there are
two alternative theories. In this case, the alternatives turn out to be complementary, in the sense
that each reveals something about light that we need to account for. This situation is quite
typical in science: we need complementary theoretical perspectives to account for the rich
diversity of properties we uncover in the phenomena being studied.
Grammar as a phenomenon of study is thus interpreted according to different theories. So as
to maintain the distinction between grammar and theories of grammar, we shall call theory of
grammar grammatics. The distinction is analogous to that between language and linguistics, or
between society and sociology. The difficulty is that people often use the same term for both the
phenomenon and its study: e.g. we speak of the "grammar of English" (the phenomenon) but
also of "traditional grammar" (one theory of the phenomenon). We could clarify this situation if
we called the second "traditional grammatics". Our concern here is thus with systemic-
functional grammatics; and we shall illustrate how it can be used in the study of grammar with
examples from the grammars of Chinese, English, and Japanese.
Grammar (as a phenomenon) is part of language; it is the "system of wordings", as we put it
above. But how it is conceptualized will depend on our grammatics. In the history of thinking
about language in the West, there have been two somewhat different theoretical perspectives.
Both have their origins in Ancient Greece; there have been many variations, but we can still
trace these two strands of thinking today. In one, language is a set of rules — rules for
specifying structures; so grammar is a set of rules for specifying grammatical structures, such as
the construction of a transitive sentence with 'verb + object'. This perspective is that of logic and
philosophy, e.g. in the foregrounding of the sentence as the basic unit of language, organized
on a logical model into Subject + Predicate. Since the sentence is the basic unit, it is studied in
isolation. In the other view, language is a resource — a resource for making meanings; so
grammar is a resource for creating meaning by means of wording. This perspective is that of
rhetoric and ethnography, e.g. in the foregrounding of text (discourse) as the basic unit of
language, organized according to the rhetorical context. Since text is the basic unit, the sentence
is studied in its discourse environment.
The kind of grammatics that is usually presented in school is a diluted version of the
'grammar as rule' type of theory. It presents rules of grammar in terms of words in sentences,
with words serving functions such as Subject, Predicate, Object, and Adverbial. As a theory, it
falls far short of the demands that are now being made on grammatical theories. On the one
hand, it takes over too much from the European languages it was first applied to, starting with
Greek and Latin; hence it is of limited value in interpreting the grammars of non-European
languages such as Chinese, Japanese, Indonesian, Tagalog, Thai, Vietnamese or the languages
of other regions and continents. On the other hand, it builds in too little of the overall
grammatical system of language. It allows us to see only a small fragment of grammar and does
not provide us with a way of interpreting the overall organization of the grammar of a
language as a system of information. At this stage in history we need a richer theory of
grammar to meet the challenges of the age of information — e.g. in education (how to organize
and give access to knowledge) and in computation (how to achieve the automatic processing of
text). We are also in a position to learn more about grammar thanks to technical innovations:
the tape recorder allows us to store and examine spoken language, and the computer allows us
to manipulate vast amounts of text (spoken or written) for the purpose of grammatical study.
Systemic-functional theory is one response to these demands. The theory was first developed
in work on the grammar of Chinese; and it has been used in educational and computational
contexts from an early stage. Unlike the theory of grammar that is still the received tradition in
school, systemic-functional grammatics takes the resource perspective rather than the rule
perspective; and it is designed to display the overall system of grammar rather than only
fragments. We hope to bring this out in the discussion which follows.
1.2 Grammar as resource; systems & their realization in structure
We use language to interact with one another — to construct and maintain our interpersonal
relations and the social order that lies behind them; and in doing so we interpret and represent
the world for one another and for ourselves. Language is a natural part of the process of living;
it is also used to 'store' the experience built up in the course of that process, both personal and
collective. It is (among other things) a tool for representing knowledge — or, to look at this in
terms of language itself, for constructing meaning.
Grammar is 'part of' this resource. But the relation of grammar to other 'parts' of the linguistic
system is not a part to whole relation; rather, it is a symbolic one. Grammar is a resource for
creating meaning in the form of wordings. Let us illustrate this point by reference to one
broad area of semantics and grammar — an area that we shall characterize as interpersonal:
this is one of three such general areas, the other two being ideational and textual.
In interacting with one another, we enter into a range of interpersonal relationships, choosing
among semantic strategies such as cajoling, persuading, enticing, requesting, ordering,
suggesting, asserting, insisting, doubting, and so on. The grammar provides us with the basic
resource for expressing these speech functions, in the form of a highly generalized set of clause
systems referred to as MOOD.
A system, in this technical sense, is a point of choice. In the grammars of Chinese, English,
and Japanese, the most general choice in mood is that between 'indicative' and 'imperative'
clauses. These two are the options or terms in the system. The following examples illustrate the
contrast between 'indicative' and 'imperative' in English:
systemic option (term)
the spy/ I/ you came/ comes/ will come in
from the cold;
who came &c in from the cold?;
did/ does/ will the spy/ I/ you come in from
[You] come in from the cold!
Any grammatical choice can be represented as a system with two or more alternative terms or
features , as shown graphically in Figure 1.
entry condition system name terms
Fig. 1: A system
This graphic representation shows (i) the system name (MOOD TYPE)1; (ii) the terms from
which one has to be chosen ('indicative'/ 'imperative'); (iii) the condition under which the
choice is available, the entry condition ('clause'). The full set of conventions for the systemic
representation is given in the Appendix.
How do we know that this system is part of the grammar of English? There are three parts to
the answer. (i) If we look at the wording of the examples given in the table above, we can see
that there are systematic differences between the 'indicative' ones and the 'imperative' ones. The
former have a Finite verb, whereas the latter do not; and the former have a Subject, whereas the
latter may or may not have one — it is typically absent. (ii) If we look at the system itself to
consider what choices are available for 'indicative' clauses, we find that they have a choice in
tense ('past/ present/ future'), expressed through the Finite verb; and also in person, expressed
through the Subject. In contrast, if we look at the system to consider the choices that are
available for 'imperative' clauses, we find that they have no choice in tense and the Subject can
(in principle) only be the addressee, 'you'. (iii) If we look at the distinction in meaning that the
system makes, we find that the choice has to do with the nature of what is being negotiated in
the dialogue: either information ( 'indicative', e.g. Did the spy come in from the cold? — Yes, he
did.), or goods-&-services ('imperative', e.g. Come in from the cold! — OK.). These three parts to
the answer illustrate three general angles of approach to any system in the grammar: (i) 'from
below', (ii) 'from around', and (iii) 'from above' — see Figure 2. (We return to this point below
in Section 3.3.) We now explore the system from different angles, beginning 'from below' —
from the point of view of how the systemic contrast is created in the wording.
1 The system name is not a formal part of the system; it is merely a convenient index.
(iii) 'from above'
(ii) 'from around'
(i) 'from below'
Fig. 2: Perspectives on a system
(i) Systemic contrasts are created by some aspect of the wording: the terms of the system are
differentiated by means of grammatical structure (e.g. the absence vs. presence of an element of
structure such as Subject), by means of grammatical or lexical items (e.g. the grammatical item
ka in Japanese indicating interrogative clauses), or, as a further step, by means of a phonological
feature (e.g. rising vs. falling intonation). We say that systemic terms, or features, are realized
(expressed, coded) by aspects of the wording. The choice in the MOOD system between
'indicative' and 'imperative' is realized structurally: only indicative clauses normally have a
Subject. We can indicate the presence of the grammatical function Subject in indicative clauses
as in Figure 3.
Fig. 3: System with associated realization statement
The arrow («) represents the relation of realization: the feature 'indicative' is realized by the
presence of the function Subject, stated as +Subject. The different types of realization statement
are summarized in the Appendix. The presence of Subject is one step in the specification of the
function structure of an indicative clause, i.e. of the organization of the clause as a
configuration of functions.
(ii) When we come to explore 'from around', we find that, through their entry conditions, a
number of systems come together as an inter-related set, called a system network. We can
illustrate again from the grammar of MOOD. The choice between 'indicative' and 'imperative' is
the most general one in this area of the grammar; but each alternative leads to further choices.
For instance, indicative clauses are either 'declarative' (they rode horses) or 'interrogative' (did
they rode horses; who rode horses); declarative clauses are either 'untagged' (e.g., they rode horses) or
'tagged' (e.g., they rode horses, didn't they), and interrogative clauses are either of the wh- type
(e.g., who rode horses?) or the yes/no type (e.g. did they ride horses?). See Figure 4.
Fig. 4: Network of MOOD systems (realization statements in boxes)
In the diagram in Figure 4, the grammatical resources are represented as a network of inter-
connected systems, each of which is a choice point. The systems in the network are ordered
from left to right, starting with the most general option and moving towards more specific
ones: if 'clause', then 'indicative' or 'imperative'; if 'indicative', then 'interrogative' or
'declarative'; if 'declarative', then 'tagged' or 'untagged'; if 'interrogative', then 'yes/no' or 'wh-'.
This is the scale of delicacy (degree of detail, specificity, granularity).
In the example in Figure 4, each entry condition is a simple feature, 'clause'; but entry
conditions can also be complexes of features, involving conjunction and/or disjunction. Such
features likewise are always terms in other systems. Let us illustrate disjunction in an entry
condition. Consider again the MOOD grammar of Figure 4. It has one system, MOOD TAG,
whose entry condition is 'declarative'. However, this system is actually not restricted to
declarative clauses; it is also open to imperative ones (e.g., [you] saddle the horses, will you; let's
saddle the horses, shall we). Consequently, we need to be able to state "if either 'declarative' or
'imperative', then 'tagged' / 'untagged'". That is, we need a disjunctive entry condition: see
Fig. 5: Disjunctive entry condition
The same systemic feature or complex of features may occur as the entry condition to more
than one system in the system network. In this case, the systems are simultaneous. For
example, the primary MOOD system (MOOD TYPE) is simultaneous with the system
POLARITY — the choice between 'positive' and 'negative' clauses: see Figure 6.
Fig. 6: Simultaneous systems
Two simultaneous strands in a system network define a two-dimensional paradigm. It is often
useful to present examples in the form of a matrix table, with one system represented by the
columns and another by the rows. Thus MOOD TYPE and POLARITY intersect as follows:
The spy came in from the Come in from the cold!
The spy didn't come in
Don't come in from the
from the cold.
Such matrices can be used in probing the accuracy of a complex system network: if it is not
possible to find examples for one or more of the cells of a matrix, this means that the system
network predicts a combination of systemic terms that does not exist.
Let us summarize what we have shown about the concepts of system and structure, and the
relation between them. These concepts theorize the axes of organization in language, the
paradigmatic and syntagmatic. The systemic, paradigmatic, axis is primary in the particular
sense that it defines the overall organization of the grammar of a language; and the structural,
syntagmatic, axis is secondary in the particular sense that it is specified locally in the
environment of the various terms of the systemic axis. Figure 7 shows the intersection of the
two axes in the grammar of MOOD, with the systemic axis providing the overall organization.
This bifurcation into the paradigmatic axis and the syntagmatic axis makes it possible for the
system to relate both to what is above and to what is below — in other words, both to what the
system realizes and to what it is realized by.
(iii) Looking at the system from above, we are asking what it means: in other words, what
semantic features are being realized by this particular set of options in the grammar. As already
noted, in the case of MOOD the meaning has to do with the negotiation of speech-functional
roles in dialogue: with basic categories such as statement and question (exchange of
information), command and offer (exchange of goods-&-services), and the complex network of
variable and more delicate categories of verbal interaction. We shall not pursue the semantic
analysis here; but we may note that the resources and methods for representing semantic
categories are formally identical with those used in the lexicogrammar.
1.3 Example: MOOD
The grammar of a language is a very rich and complex system; the grammatics must bring out
that richness and complexity, and not obscure it. This means recognizing the different vectors
along which the complexity is ordered, and exploring one step at a time.
Here we have introduced only one 'corner' of the grammar, and only in the most general
terms: the primary systems of MOOD, as these are found in English. Because grammar is
viewed as a resource rather than as a set of rules, it is interpreted in systemic-functional theory
as a system network; this represents the grammatical potential available to the language user.
The system network allows us to map out the overall organization of the grammar of a
language, with delicacy as the main principle for ordering the various systems relative to one
another. Naturally such networks soon get very large; in the systemic grammars of English
stored in computers, there are somewhere around 1000 systems. We have illustrated such a
map of the grammar of English with fragments from the MOOD grammar, as in Figures 4 and
The partial English MOOD grammar we have presented is a systemic-functional description
of one particular language, cast in the theoretical terms of systemic organization with
associated structural realizations. That is, while the type of organization embodied in the
system network is part of the theory, and is a general feature common to all languages, the
particular systemic features and structural realizations are part of our descriptive interpretation
of English. They are not part of the general systemic-functional theory of grammar (see further
Section 3.3 below).
Fig. 7: The systemic (paradigmatic) and structural (syntagmatic) axes intersecting
As a descriptive generalization about interpersonal grammar, we can assume that all
languages have a system of MOOD: i.e. grammatical resources for the interaction between
speaker and addressee, expressing speech functional selections in dialogue. Further, the
semantic categories of giving information (statement), demanding information (question), and
demanding services or goods (commands) are very likely enacted in the grammars of all
languages. However, the organization of the MOOD system, and the realizations of the various
options, differ from one language to another. For instance, the degree to which there is a
distinct grammatical category corresponding to commands is variable: there may or may not be
a distinct form of the imperative, and even where there is, there are usually many other
possible realizations. Similarly, while all languages probably have a basic opposition between
statements and yes-no questions (polarity questions), which it is often (though not universally)
possible to express by means of the distinction between falling and rising intonation, questions
demanding a specific element of information (other than the value of the polarity) may be
grouped systemically either with statements or with yes-no questions. It is easy to see why:
they are like statements in that their polarity is certain, but at the same time they are like yes-no
questions in that they demand information. Different languages organize their MOOD
grammars around different generalizations in this way. Furthermore, languages differ
considerably with respect to more delicate options, such as those concerned with how
interactants position one another in dialogue (e.g. by indicating expected responses) and with
how they assess the information being exchanged (e.g. by indicating degree of probability or
source of evidence).
At the least delicate end of the grammar, Chinese, English, and Japanese have similar MOOD
systems. All three distinguish 'indicative' vs. 'imperative' clauses, and within the former,
'declarative' vs. 'interrogative', with one interrogative subtype for querying elements and
another for querying polarity. Examples are tabulated below:
Tailang shang xue qu Taro is going to
Taroo wa gakkoo e
elemental Tailang dao nali qu?
Where is Taro
Taroo wa doko e ikimasu
Tailang shang xue qu Is Taro going to
Taroo wa gakkoo e
Shang xue qu!
Go to school!
Gakkoo e ike!
But while the three MOOD systems are congruent up to the point in delicacy shown in the table
above, they also differ from one another in more delicate terms. For instance, in (Mandarin)
Chinese, 'polar' interrogatives are further differentiated according to the speaker's expectation
regarding the polarity of the proposition: they are biased (positive or negative) or unbiased; e.g.
'Do you want it?', positive bias Ni yao ma?, negative bias Ni buyao ma?, unbiased Ni yao buyao?.
English has only the biased forms: positive (semantically neutral) did you see him?; negative
(semantically, positive bias) didn't you see him? English has no unbiased form, other than the
highly marked (peremptory) did you see him or not?
The basic MOOD system we have discussed is concerned with (i) the nature of the commodity
being exchanged (information vs. goods-&-services) and (ii) the orientation of the exchange
(giving vs. demanding). But there are other aspects of the exchange that may be
grammaticalized in this part of the grammar, in particular aspects of the tenor of the
relationship between the interactants engaging in the exchange, i.e. between speaker and
addressee. In Japanese, this area is perhaps more highly codified in the grammar than in either
Chinese or English. For instance, alongside the "plain" imperative (as in Hanase! 'Talk!), there
are also polite options for situations where the speaker is superior to the addressee (as in
Hanashi-nasai!) or inferior to the addressee (as in Hanashite-kudasai! ). The elaboration of the
grammar of Japanese in the areas of politeness and honorification is well-known. It is an
important characteristic of the grammatical system — one that makes very good sense in terms
of the interpersonal metafunction. At the same time, we have to recognize that the grammars of
both Chinese and English also have created considerable potentials for enacting a wide range of
subtly different tenor relationships. These potentials are perhaps not immediately obvious
because they rely to a large extent on a cryptic feature of the system, viz. grammatical
metaphor. Thus alongside the congruent Come in from the cold!, there are also various
metaphorical variants where the command is realized not as an imperative clause but as if it
was a statement or a question. For example: I'd like you to come in from the cold; I want you to come
in from the cold; you should/ must /will come in from the cold; Would / Could you come in from the cold.
Such expansions of the system are of course characteristic of Japanese as well.
What generalizations can be made about the realization of systemic options in mood? MOOD
options are typically realized in various ways, including intonation (direction of pitch
movement), mood particles, relative sequence of elements (usually involving a finite verb), and
special verbal categories. It seems that interpersonal systems in general tend to be realized by
some prosodic mode of expression; and the realizations of MOOD that we find across
languages can often be shown to be prosodic (e.g., interpersonal mood particles that serve as
juncture prosodies). These particulars are not, of course, part of the general theory of grammar
— they are empirical descriptive generalizations covering a number of different languages.
And here Chinese, English, and Japanese illustrate nicely a general principle of crosslinguistic
similarity. While their basic mood systems are congruent with one another, their systemic
contrasts are created in different ways, deploying somewhat different subsets of the
realizational resources. The basic patterns are tabulated below (leaving out realization by
intonation, which is used by all three languages):
+ Mood (Subject, Finite)
Subject ^ Finite
+ Negotiation = ka ^ #
Finite ^ Subject
+ Wh; # ^ Wh ^ Finite
+ Negotiation = ma
As the realizational table indicates, English differs from Chinese and Japanese in its mood
structure. It has a Mood element, which consists of Subject + Finite. This Mood element plays a
central role in the realization of mood options, in terms of both its presence and its internal
organization. In the unmarked cases tabulated above, it is present in 'indicative' clauses (e.g.
Mood: You will / Will you + come in from the cold), but not in 'imperative' ones (e.g. Come in from
the cold). Further, 'declarative' clauses are distinguished from 'polar' ones by the relative
ordering of Subject + Finite — Subject ^ Finite (will you ) and Finite ^ Subject (you will),
respectively. The significance of the Mood element in English is also shown e.g. in tags, where
the Mood element is picked up at the end of the clause as the Moodtag, consisting of Tagfinite ^
Tagsubject (e.g. You will come in from the cold, won't you?) Neither Chinese nor Japanese has a
distinct Mood element. It follows that they do not rely on the sequence of Subject + Finite in
realizing mood options. In fact, neither language has a separate function Finite in the mood
structure of the clause. Chinese has no system of verbal finiteness at all, and Japanese does not
separate out finiteness from the rest of the verbal group in its clausal structure as English does.
Instead, both languages deploy mood particles at the end of the clause serving the function we
have called Negotiation, since it determines the clause's negotiatory value in dialogic
interaction. The difference is that Japanese Negotiation = ka is a property of 'interrogative'
clauses in general, whereas Chinese Negotiation = ma is a property of 'polar' interrogatives in
particular. (Chinese also has another type of 'polar' interrogative, where the Predicator is
repeated with a negator as in shi bu shi. We referred to it above when we discussed differences
in more delicate mood systems.) In fact, these mood particles are part of more extensive sets of
interpersonal particles in both languages, including ne, ba in Chinese and ne, yo in Japanese;
the closest equivalent of the English option of tagging a clause is a particle of this kind. The
generalization is that the grammars of Chinese and Japanese provide the resource for indicating
how the speaker intends the addressee to take his/ her move in the dialogue as s/he is about to
'hand over' to the addressee. (Such interpersonal particles are common around the languages of
the world; for example, we find them clause-initially in Arabic (hael, ?a) and in French (est-ce
que), and we find them in various (South-)East Asian languages, e.g. in Korean, Thai (clause
final máy, rii; na etc.), Vietnamese (clause final phong, a, u, chu, di etc.). Such particles may also