Professional Documents
Culture Documents
Combining Clauses Into Clause Complexes: A Multi Faceted View
Combining Clauses Into Clause Complexes: A Multi Faceted View
net/publication/235928932
CITATIONS READS
38 1,812
1 author:
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Christian M.I.M. Matthiessen on 20 January 2019.
1 Introduction
the clause complex after the point where he stopped — … so the government should plant
more trees and stop people cutting down our trees. The relational structure of the clause
complex is set out in Figure 1-1. The conventions are those of Halliday (1965; 1985/ 1994:
Ch. 7) and are given in Table 2-1, Section 2.1. But the details are not important at this point.
What is important is the impressive intricacy of the structure that emerges as the speaker
expands the clause complex. This is dynamic grammar. It is the grammar of everyday, casual,
spontaneous conversation: while the clause complex that emerges in the conversation above
is a combination of 10 clauses, there is nothing unusual about this degree of intricacy — I
will give examples below of far more intricate clause complexes.
x
1 β If there weren’t trees on the earth, um earth,
α α we would all be dead,
x
β ‘cos there wouldn’t be oxygen;
+
2 1 1 1 trees make oxygen,
x
2 so we can breathe,
x x
2 β so if we had heaps of trees around us,
α 1 it produces heaps of oxygen,
x
2 so we can breathe;
x
2 α so trees, big trees, are really good
x
β because heaps of oxygen comes out of them.
Figure 1-1: Analysis of clause complex from casual conversation
gives us conceptual tools crucial in dealing with clause combining — e.g. the concept of
clines of categoriality instead of dichotomies (Hopper & Thompson, 1984) and the
parametric or systemic framework developed in the modelling of transitivity (Hopper &
Thompson, 1980). Typological issues will be explored in particular in Section 3.3.
Another strand in Sandy’s work on clause combining has been the explanation of
features of clause combining — in particular, the placement of dependent clauses — in terms
of the “flow of information” in discourse: Thompson (1984), Ford & Thompson (1986).
These studies show among other things that dependent clauses are placed before the
dominant clauses they are dependent on for thematic reasons: they serve to specify a local
context within which what follows is interpreted. Thus the complex in Figure 1-1 sets up a
local, highly motivated pattern of information distribution. For example, hypotactically
dependent clauses of condition, concerned with the existence of trees (if they’re weren’t trees
on earth; if we had heaps of trees around us), are presented as thematic in the argument,
scaffolding it; but hypotactically dependent clauses of reason, concerned with the presence of
oxygen (‘cos there wouldn’t be oxygen; because heaps of oxygen comes out of them), are
presented as rhematic. Textual considerations will be mentioned in Section 2.2.1.
Yet another strand in Sandy’s work has been the investigation of clause combining in
relation to the “flow of events” in discourse. Thompson (1987) shows a clear, but intricate,
relationship between the mainline of the flow of events and “subordination”: events that fall
on the mainline can be — but tend not to be — construed in hypotactically dependent and
embedded clauses. (Here “subordination” does not differentiate between hypotaxis and
embedding. If these two are differentiated in the light of Matthiessen & Thompson (1988),
the correlations identified by Sandy can be revised to sharpen the picture.)
A fourth strand in Sandy’s work on clause combining is the interpretation of clause
combining as the grammaticalization of the rhetorical-relational organization of text — of
what I will call rhetorical complexing: Matthiessen & Thompson (1988). See further Section
3.2 below.
Relevant to Sandy’s work on clause combining is also one of her metalinguistic
contributions: she is one of the few linguists who has made links between the concerns of
West-Coast functionalism (cf. Cumming & Ono, 1997) and systemic functional linguistics —
not only in Matthiessen & Thompson (1988), but also in e.g. Thompson (1984) and
Thompson (1990). Just as her work on clause combining, this is a long-term project and the
two are related since clause combining has also been a concern in the systemic functional
work by Halliday and other systemic functional linguists (e.g. Halliday, 1965, 1985/94: Ch.
7; Huddleston et al, 1968; Martin, 1988; Matthiessen, 1995a: Ch. 3; Nesbitt & Plum, 1988;
Bateman, 1989).
ideational [logical & experiential], interpersonal, and textual modes of meaning) and
instantiation (the cline from text instances via text types / registers to the general systemic
potential of language). Each dimension defines different vantage points for exploring clause
combining. If we shunt along the dimensions, we can develop a well-rounded picture of
clause combining.
The dimensions of metafunction and stratification are intersected in Table 1-1. The
howrizontal axis shows the different metafunctions, each one being represented by a column;
and the vertical axis presents the different strata, each one being represented by a row or (in
the case of lexicogrammar) by a set of rows. The cells in the table represent the points of
origin of major linguistic systems such as TRANSITIVITY, THEME, MOOD and CLAUSE
COMPLEXING (names of systems are given in capitals). Thus CLAUSE COMPLEXING is
located stratally within the lexicogrammatical stratum — more specifically, at the rank of
clause along the local dimension of rank (see Section 2.1 below); and it is located
metafunctionally within the logical mode of the ideational metafunction: it is the resource for
construing sequences of processes by linking clauses into complexes that are constructed out
of logico-semantic relations, where the clauses are treated as having either equal status
(parataxis) or unequal status (hypotaxis). The dimensions of stratification and metafunction
define not only the location of clause combining itself but also its “neighbourhood” in the
total system. Looked at “from above”, its stratal neighbourhood includes rhetorical
organization — clause combining realizes patterns of rhetorical relations in text (cf.
Matthiessen & Thompson, 1988), and looked at “from below”, its stratal neighbourhood
includes tone sequence — clause combining is realized by certain sequences of tones (see
Halliday, 1967a, 1985/94: 306-307; Chafe, 1988; Mithun, 1988). Looked at “from around”,
its metafunctional neighbourhood includes (i) the textual systems of conjunction, which
provides a cohesive, non-structural alternative to clause combining (cf. Halliday & Hasan,
1976: 227-229; Halliday, 1985/94: 328-329; 398-403), and of theme, which manages the
flow of information in combinations of clauses (cf. Thompson, 1984; Thompson & Longacre,
1985), (ii) the experiential system of transitivity, which complements clause combining in the
construal of sequences by construing the organization of the processes linked in a sequence
and which competes with clause combining in the construal of circumstantial meanings of
time, cause, condition etc. (cf. Halliday & Matthiessen, 1999: Chapters 3,4 and 6), and (iii)
the interpersonal system of mood, which gives each clause linked in a clause complex a
“dialogic” of negotiatory status of varying degrees of arguability (cf. Givón, 1982; Halliday,
1985/94: 239-241; König & van der Auwera, 1988).
5
The table thus shows us how clause combining is related to everything else in the
system, by one or more degrees of separation. It helps us identify patterns of agnation beyond
clause combining itself — patterns that involve e.g. CONJUNCTION (textual) and
circumstantial TRANSITIVITY (experiential): see Section 2.3. It also helps us locate sources
of markers of clause combining within agnate systemic domains: see Figure 1-2. This figure
shows several of the sources of conditional structural conjunctions and markers, based on
Traugott’s (1985) account of the evolution of markers of conditional clauses. Traugott
presents a “cognitive map” of conditionals in relation to other categories, showing “a set of
potential, not necessary, routes for change, and a set of potential, but not always necessary,
components of meaning associated with any one category”. Such a map can, as illustrated for
condition in Figure 1-2, be outlined against the background of the general map of the overall
system shown in Table 1-1. This illustrates the value of the table as a resource in exploring
clause combining ecologically and it also illustrates the close connection between patterns of
agnation in the current system and paths of change.
conjunction: transitivity:
‘lexical” conjunctions circumstantial phrases
in (the) case (of ...)
cause
mood & modality:
interrogative;
subjunctive
textual status:
thematicity (topicality); condition had it rained,
givenness if, whether it would ...
transitivity:
mental & verbal
given that clauses of projection
supposing, assuming;
time say
whenever
Figure 1-2: Markers of ‘condition’ and patterns of agnation within the function-rank
matrix
Table 1-1 represents a partial map of the general systemic potential of English —
i.e., of the system pole of the cline of instantiation; but all the particular systems such as
clause complexing and transitivity covered by the map are instantiated in text. These systems
are nothing but generalizations across innumerable text instances. Thus patterns of clause
complexing that emerge in a particular text instantiate the system of clause complexing; but
at the same time they “re-create” (part of) that system, maintaining and changing it, because
7
the system is nothing more than the distillation or echo of innumerable instances of clause
complexing in text. We have to avoid the trap of re-ifying systems such as clause complexing
by treating them as part of a code or competence that is phenomenally distinct from text in
context: system and text are not different phenomena; they are only different “phases” of one
and the same phenomenon, just as weather and climate are just different “phases” of
meteorological phenomena (cf. Halliday, 1973: 43-46; 1992). This is particularly important if
we are to understand the nature of clause combining because of its property of emergence
mentioned above. To understand the relationship between system and text, we must observe
relative frequency of instantiation in text of systemic terms such as ‘hypotaxis’ vs.
‘parataxis’, and then, shifting our perspective along the cline of instantiation from instance
towards potential, interpret these frequencies as probabilities in registers and in the overall
system, as was done in a pioneering study by Nesbitt & Plum (1988). These quantitative
patterns can then be related to qualitative properties of the system. This will be one important
theme in my discussion of clause combining.
Metafunction will be discussed in Section 2 and stratification will be deal with in
Section 3. Instantiation will not be discussed separately but it will be highlighted throughout
the discussion.
2.1 Logical
When it is interpreted metafunctionally, clause combining turns out to be engendered by the
ideational metafunction — more specifically, by the logical mode of the ideational
metafunction (rather than by the experiential mode). This means that clause combining is
organized univariately as opposed to multivariately1 (see Halliday, 1965; 1978: 129-131;
1
Halliday (1965/81: 31) characterizes this contrast in mode of structure as follows: “we can say that language
exhibits two basic types of structure, the “multivariate” and the “univariate”. A multivariate structure is one
involving more than one variable; a univariate structure is one involving only one variable. The elements of a
multivariate structure are thus different variables each occurring once only … The elements of a univariate
structure are repetitions of the same variable …”. Transitivity structure is multivariate, as in Actor + Process +
8
1979) and that it is thus one form of complexing in language: it is complexing at clause rank.
Clauses are linked to one another through logico-semantic relations such as temporal or
causal sequence, condition, addition and elaboration and these linked clauses form logico-
semantic clause complexes. To bring out this logico-semantic, relational nature of clause
combining, we can call it clause complexing (following Halliday, 1985/ 94: Ch. 7; see also
Halliday, 1965; Hudson, 1968, in Huddleston et al, 1968).
Clause complexing is like complexing at other ranks in the grammar — complexes of
groups and phrases, of words, and of morphemes; and although complexes at different ranks
may be marked by different sets of conjunctive particles in certain languages, it is very
common for them to draw on the same set, as is the case in English, and conjunctive markers
may generalize across environments along the rank scale (Mithun, 1988: 349-351).
Complexes at clause rank cover the widest range of logico-semantic relations, so they also
involve the most extensive set of conjunctions; for example, while and, or, but are common
across all environments of complexing, then, so; after, before, because and many others are
not (cf. Matthiessen, 1995a: 173-176).
All complexes are structured as series of related elements: each relation represents a
new expansion of the complex. This serial structure was exemplified in Figure 1-1 above, but
let’s now consider a couple of simpler examples. Figure 2-1 shows a clause complex that
consists of a series of three clauses; the relationship is one of temporal sequence of elements
that are of equal status: 1 ‘and then’ 2 ‘and then’ 3. The complex represents a subprocedure
in the procedural part of a recipe. Figure 2-3 shows a similar complex at group rank; the
relationship is one of adding elements that are of equal status: 1 ‘and also’ 2 ‘and also’ 3 ‘and
also’ 4 ‘and also’ 5. The complexes are developed link by link; each pair of linked elements
is called a nexus (Halliday, 1994: 218). In other words, a nexus is a minimal subcomplex in
the unfolding complex. In these examples, the complexes are developed out of nexuses where
the linked elements are of equal status: they are related through parataxis. But nexuses may
also involve elements that are given unequal status in the relationship, where one is dominant
and the other is dependent: these elements are related through hypotaxis. Figure 2-2 shows a
clause complex that is built up from a series of hypotactic nexuses. Here each new clause
depends on the previous one in the series.
Remove the casserole from add 2 peeled sliced avocados and return to the oven for a
the oven in a line down the centre of further 5 minutes
the dish
1 → ×2 → ×3
clause clause clause
Figure 2-1: Example of serial structure — paratactic clause complex (recipe)
Goal: he + added + milk, cornflour salt, cream and sherry; co-ordination structure is univariate, as in 1 + 2 + 3
+ 4 + 5: milk, cornflour salt, cream and sherry.
9
Calypso knew that her aunt knew she knew how unwelcome Richard would
be in Enderby Street.
α → ‘β → ‘γ → ‘δ
clause: mental clause: mental clause: mental clause
Figure 2-2: Example of serial structure — hypotactic clause complex (Mary Wesley,
The chamomile lawn)
The clause complex in Figure 1-1 above is more intricate than those in Figure 2-1 and
Figure 2-2. All of the complexes are developed through linear serialization; but the complex
in Figure 1-1 also involves layering or internal nesting: several of the elements linked within
the nexuses are themselves nexuses rather than simple clauses. I will exemplify and discuss
such complexes further in Section 3.2. What is important here is to emphasize the open,
dynamic nature of the serial structure of clause complexes; this structure is, as I put it above,
emergent. There has been a strong tendency in 20th century linguistics to view linguistic
structure in terms of “building blocks”, but this conception of structure crumbles when we
begin to analyse casual, spontaneous, unselfconscious spoken discourse — or in fact, any
discourse viewed as a process of unfolding rather than only as the product of that unfolding.
This is brought out very clearly in Ochs, Schegloff & Thompson (1996). Halliday (1961/
1976: 56-59) stressed the dynamic nature of linguistic structure many years ago:
Language is patterned activity. At the formal level, the patterns are patterns of meaningful organization:
certain regularities are exhibited over certain stretches of language activity. (p. 56) The category set up to
account for the stretches that carry grammatical patterns is the ‘unit’. (p. 57) The unit being the category of
pattern-carrier, what is the nature of the patterns it carries? In terms once again of language as activity, and
therefore in linear progression, the patterns take the form of repetition of like events. Likeness, at whatever
degree of abstraction, is of course a cline, ranging from ‘having everything in common’ to ‘having nothing
in common’. ... In grammar the category set up to account for likeness between events in successivity is the
‘structure’. If the relation between events in successivity is ‘syntagmatic’, the structure is the highest
abstraction of patterns of syntagmatic relations. (p. 59)
All structure is to be interpreted as patterned activity; we should see it in terms of
“likeness between events in successivity” and we should not re-ify it as some kind
constructional form imposed on words. The serial structure of clause complexes foregrounds
the view of structure as patterned activity; and it represents the open-ended nature of the
activity of meaning — any of the complexes presented here could easily have been expanded
10
further: unlike clauses, they are not constructional units. It also foregrounds the locally
managed quality of clause complexing — “locally managed” in the sense Ford & Thompson
(1996: 134) develop when they describe spontaneous talk:
In the past two decades, conversation analysts have uncovered patterns and principles of interaction,
particularly in the areas of turn-taking and the sequential organization of talk. The picture of spontaneous
interaction that emerges very clearly from this research depicts a complex and intricately monitored human
practice that is maximally sensitive to moment-by-moment input by all parties to a conversation, and is,
therefore, characterized by an organization that is locally managed.
Thus in the following extract from casual conversation during a tea break at a
workplace (UTS/Macquarie corpus, text due to Di Slade), speaker M is producing a narrative
to substantiate the claim that a co-worker, Joanne, is an annoying person (she gets really
pushy). The narrative is sustained across three of M’s turns, as shown by the underlining:
||| M: I’m about to throw Joanne out the window. ||| F: Joanne who? ||| M: Lattimer. |||
F: Why? ||| M: She gets really pushy. ||| I’m looking for a file for Adam, || Kerry
handed me three others || and I was in the middle of finding the third one for her. ||| A:
Kerry gave you three, did she? ||| M: Yeah, || you know, they had to be done. ||| Ë
And Joanne came up || and she said, || "Oh, can you do this?" || and I said, || "Look
you're at the end of a very long line: || be prepared to wait || and she said, || "Well,
she's at the Oncology clinic right now." || and I said, || "But these have to be done as
well; || I can't help || and sort of smiled all the way through it || and she went,... || I
said, || "Look, it's three minutes to three; || these should be done in a minute || if you
want to wait till then || and she went || (sigh) ahhh. || then she went away || and I
thought || “Oh yeah, * end of story” ||| A: * She gets very worried. ||| M: Ë And then
she came back in again || and um she said, || "Are those files there; || did Kerry give
you those files there?" || and I knew || what she was going to say next || and I said, ||
"Ah, among others," || and um, she went, || "Oh, oh they can wait until after this one, ||
'cause they're not needed, okay." || and I said * ... ||| F: * Why couldn't she grab
someone else? ||| M: Because Liz and I are the only ones doing them || and they don't
know that Ann can do them. ||| Ë But um, I said, || "Look Liz is going” || I said, ||
"Look, you know it's nearly three o'clock now; || Liz should be back any second
now." || I said, || "Anyway, I've got afternoon tea now || and I've got to go to taping."
|||
The narrative gets suspended three times: the first time, A asks M a question as a reaction to
information from the narrative, and M answers the question before she returns to the
narrative; the second time, A makes a statement (which overlaps with M’s end of story),
commenting on Joanne’s behaviour, but M does not reply to this and continues her narrative;
and the third time, F asks M a question arising out of the narrative, which M answers before
picking up the narrative thread again. This kind of interleaving is typical of conversational
narratives (see Ochs, 1997). These exchanges are possible partly because the serial nature of
the structure of clause complexing. On the one hand, the series can be suspended — and even
aborted — when local conditions so demand. On the other hand, the series can be resumed
after each suspension: notice that each time M has suspended, she picks up again with either
and or but (shown in bold italics and indicated by an arrow). In fact, the whole narrative
11
sequence can be interpreted as a single complex, extended serially across three turns.2 Part of
this “megaplex” is shown diagrammatically in Figure 2-5. The figure shows how the
megaplex is developed nexus by nexus, through both linear progression (vertical dimension)
and internal layering (horizontal dimension). Each move in the development is very simple,
constructed out of a logico-semantic relation, and the complexity emerges out of these locally
simple moves. Clause complexes may equally be constructed collaboratively across turns by
more than one interactant (cf. Ono & Thompson, 1996), as in the example from casual
conversation in Figure 2-4. Here Hermine has been telling an anecdote about going to
physiotherapy. After being asked how much the therapy costs, she goes on to say that it’s a
bargain before giving the actual amount. She is starting out on the reason why it’s a bargain
when Penny jumps in like a runner in a relay-race and takes over Hermine’s unfolding clause
complex, supplying another line of development.
Clive: How much does this ’cost every every ... time?
Hermine: Well it’s quite …
α Y’know I think
‘β α I’m getting a bargain
‘β ×β => be{laughs}cause —
Penny: ‘β ×β 1 —’cos I love being tortured
×β +
‘β 2 and he does it’s quite cheap really.
Figure 2-4: Clause complex constructed collaboratively across turns (UTS/Macquarie
corpus)
2
In the analysis on which the counts given in this chapter are based, I have been more conservative and not
extended any clause complexes across turns.
12
M:
1 I'm looking for a file for Adam,
linear expansion
x3 and I was in the middle of finding the third one for her.
x6 1 and I said:
x2 be prepared to wait!”
x8 1 and I said:
=2 I can’t help.”
x11 1 I said,
“2 “(sigh) ahhh”
1 and I thought
x14
Figure 2-5: Narrative unfolding as clause complex in casual conversation (extract from
gossip text “Joanne”)
Whether we interpret the whole narrative sequence produced by M above and shown
in Figure 2-5 as a single clause complex or not will of course depend on various
considerations — including both rhetorical-semantic considerations and phonological ones:
13
see further Section 3 below. But regardless of where we draw the line, we have to show that
behind the serial structure of complexing is the system of CLAUSE COMPLEXING — the
strategies open to speakers in developing clause complexes and to listeners in tracking and
reconstructing clause complexes. A description of this system is set out in Figure 2-6. The
system characterizes the potential of one nexus in a developing clause complex — the nature
of the relation by which the nexus is formed (the systems of TAXIS: ‘hypotaxis’ vs.
‘parataxis’ and LOGICO-SEMANTIC TYPE: ‘projection’ vs. ‘expansion’) and the option of
stopping or of expanding the complex further by opening up another nexus (the system of
[systemic] RECURSION: ‘’stop’ vs. go on’).
The nature of the relation is determined by TAXIS and LOGICO-SEMANTIC TYPE.
TAXIS is the contrast between ‘hypotaxis’ and ‘parataxis’ — between “unequal” and “equal”
status in the relationship or, to put it in terms of Rhetorical Structure Theory, between a
nucleus-satellite relation and a multi-nuclear relation (see further Section 3.2 below). The
system of LOGICO-SEMANTIC TYPE is concerned with the logic of the relation forming
the nexus — it is either projection (reporting and quoting) or expansion (which includes the
traditional categories of apposition, co-ordination and “adverbial clauses”). These two
systems are simultaneous, so they intersect to define the semiotic space shown in Table 2-1.
This table illustrates typical patterns of realization and specifies the labels used in the
analysis of the serial structure of clause complexes.
14
hypotaxis
TAXIS
parataxis
idea
clause projection
LOGICO- locution
SEMANTIC
elaborating
TYPE
expansion extending
enhancing
stop
RECURSION
go on
parataxis hypotaxis
[1 2 3 …] [α β γ …]
projection idea 1 Calypso knew Ë α Calypso knew Ë
[‘] ‘2 “My aunt knows”. ‘β that her aunt knew.
locution 1 Calypso said Ë α Calypso said Ë
[“ ] “2 “My aunt knows”. “β that her aunt knew.
expansion elaboration 1 Calypso knew; Ë α Calypso knew, Ë
= =
[=] 2 she was fully aware of it. β being fully aware of it.
extension 1 Calypso knew Ë α Calypso knew Ë
+ +
[+] 2 and her aunt knew. β as well as her aunt knowing.
enhancement 1 Calypso knew Ë ×β Because Calypso knew Á
[×] ×2 so her aunt knew. α her aunt knew.
The systemic options set out in Figure 2-6 are not equally likely to be instantiated in
text; and while certain combinations are favoured, others are disfavoured. The systemic
probabilities associated with the options and the combinations of options can be estimated by
counting relative frequency in text. The larger the sample of texts and the more registerially
balanced it is, the more accurate the estimates are likely to be. Table 2-2 presents counts of
TAXIS and LOGICO-SEMANTIC TYPE from a small sample of spoken and written
Australian, British and American texts, taken from my ongoing project of lexicogrammatical
analysis of text.3 The counts must be taken as a tentative indication of the probabilistic profile
of the system of complexing. The profile may change as the sample of texts is enlarged and
extended to cover a more balanced range of registers; but there are certain features that stand
out very clearly, as is brought out when the frequencies are visualized as the graph in Figure
2-7:
The two terms in the system of TAXIS, ‘hypotaxis’ and ‘parataxis’, are roughly
equiprobable: while the division of labour between them varies across the
different logico-semantic types, they are balanced overall.
The most favoured combination is that of hypotactic enhancement; the least favoured
combinations are those of paratactic projection of ideas (i.e., quoting thoughts)
and of hypotactic extension (i.e., some form of addition or disjunction as a
3
For the analysis of clause complexes, I used a sample of around 52,000 words — around 6,100 clauses. These
clauses form 1,600 clause complexes with two or more clauses, constructed out of around 2,900 clause nexuses.
The sample is registerially diverse, including written and spoken narratives, written news reports, written
scientific expositions, written procedures, written persuasive text, spoken interviews, spoken gossip, spoken
anecdotes, spoken banter among work mates, and casual conversation in the family. However, the sample is
“opportunistic”: I have not attempted to make it registerially balanced; for example, 55% of the sample is
spoken whereas 45% is written discourse. (In contrast, Nesbitt & Plum’s, 1988, study was based entirely on
socio-linguistic interviews with dog breeders, covering narratives, anecdotes, recounts, exemplums and
observation/comment.) All the texts in the sample are complete as examples of the registers they instantiate.
16
The system of RECURSION is also subject to probabilistic patterning: see Figure 2-8.
The figure shows counts from the sample of spoken and written texts already mentioned. As
the systemic path of recursion, shown on the left-hand side of the diagram, increases, the
number of instances decreases: clause complexes with two clauses are more common than
complexes with three clauses, which in turn are more common than complexes with four
clauses, and so on until we reach more intricate complexes of 7-8 clauses, where the descent
in the number of instances levels for a while, as it does again for 9-11 clauses. But for
complexes with 9 or more clauses, there are so few instances that we can’t draw any
conclusions. In the present sample, there are only 6 complexes with more than 11 clauses;
these are scattered between 12 and 22 clauses. Top begin to see a pattern here, we would
need to analyse a sample on the order of 500K or 5 million words. This is a daunting task for
18
manual analysis! But it is in samples of the order of 20 to 200 million words that many
fundamental grammatical principles are still hidden from us. In the present sample, all the
most intricate complexes come from spoken texts; see further Section 3.2 below.
stop
go on
stop
stop
go on
stop go on
stop
go on stop
stop go on
stop go on
go on stop
stop stop
go on
stop
go on
go on stop
go on
go on stop
go on stop
go on
go on
The system of RECURSION interacts with the other two systems, those of TAXIS
and LOGICO-SEMANTIC TYPE: as the degree of systemic recursion increases,
probabilities change within both of these other two systems. I will just give the counts for
TAXIS here: see Figure 2-9. Here “level 1” represents the most local nexuses and “level 7”
the most global ones. As the graph indicates, there is a clear tendency for ‘parataxis’ to
increase as the level increases but for ‘hypotaxis’ to decrease as the level increases. This
picture may reflect the semogenetic relationship between parataxis and hypotaxis. It has often
been assumed that hypotaxis evolves out of parataxis (cf. Hopper & Traugott, 1993: Ch. 7)
19
— what Harris & Campbell (1995: 282) call the “parataxis hypothesis” (a hypothesis which
they themselves in fact view critically). And since grammaticalization is overwhelmingly a
downwards movement in the grammar — downwards in terms of complexing and rank (cf.
Matthiessen, 1995a: 49-50), it seems plausible that the increase in hypotaxis as we move
towards “level 1” reflects the semogenetic drift downwards.4 At any rate, we will see later
that the division of labour between hypotaxis and parataxis in relation to the levels is part of a
more general picture that also includes circumstantiation within the clause and cohesive
conjunctions across clause complexes.
Clause complexing varies considerably across registers: the frequencies shown so far
simply represent an indication of the systemic generalizations that are distilled out of the vast
range of different registers, which in turn simply represent registerial generalizations distilled
out of the vast range of texts instantiating the different registers. To get a sense of this
variation, we can start at the instance pole of the cline of examination, returning to the gossip
phase of casual conversation from which I presented a narrative extract above. The analysis
of this extract was shown in Figure 2-5 and it suggests the way in which the text unfolds. The
patterning out of which the narrative passage is created can be seen very clearly if we
examine the way in which the whole gossip text unfolds: see Figure 2-10. Overall, clause
complexes in “Joanne” are dominated by relations of enhancement and projection; and within
4
This view is given further support when we consider differences between spoken and written discourse: see
Section 3.3 below.
20
these, temporal succession and projection of speech (locution) are the favourite subtypes. As
we have seen in Figure 2-5, temporal succession helps construct the flow of events in
narrative passages and projection of speech constructs dialogic passages within the narrative.
Therefor enhancement and projection are correlated within the unfolding of the conversation
in Figure 2-10; together they constitute a narrative strategy in this conversation — an in
casual conversation in general.
The patterns that we see in a single text, or a phase within a single text, may or may
not be representative of something more general. When the patterns of a single text tend to
recur in other texts, we can often recognize these texts as constructing a “family” of texts —
a text type or register. To group the texts in this way will make sense if we can discern a
common kind of context — a context type that selects for a particular register (remembering
that registers, like everything else in language, are just zones of relative similarity against the
background of relative difference). To illustrate the nature of register variation in clause
complexing, I have selected seven different registers5 and displayed the relative frequencies
of the terms in the systems LOGICO-SEMANTIC TYPE and TAXIS in Figure 2-11. Since
the systems are simultaneous (cf. Figure 2-6 above), each system is counted separately.
In TAXIS, hypotaxis is less common than parataxis in the two registers of casual
conversation (chat — around 35%; and gossip — around 30%); but it is more common than
parataxis in news reports (hard news — around 65%), scientific reports (around 60%) and
topographic procedures (walking and driving tours in guide books — around 60%). In stories
for children hypotaxis is also less common — around 40%, which is closer to the frequency
in conversation than that of the registers of writing in the selection (news reports, scientific
reports and topographic procedures). In interviews with writers, hypotaxis and parataxis are
more or less equally balanced. These interviews are, of course, less spontaneous than casual
conversation and have no doubt been edited after they were transcribed; but they are less
planned than any of the written registers. I shall return to TAXIS and the difference in mode
between writing and speaking in Section 3.3. But it seems likely that the equal balance
between hypotaxis and parataxis in the overall system of English represents a “compromise”
between spontaneous spoken discourse and planned written discourse.
5
Here it is not possible to review the criteria for recognizing registers — or registerial zones — within register
variation. As always in language, we have to look for criteria from all angles — “from above” (contextual
considerations), “from around” (semantic properties) and “from below” (lexicogrammatical, and perhaps also
phonological, realizational characteristics). In identifying the registers in the selection, I used contextual
considerations in the first instance, before examining lexicogrammatical patterns in clause complexing or any
other grammatical systems. For example, gossip texts meet the contextual criteria discussed in Eggins & Slade
(1997) and Slade (1996).
21
enhancement
9 10 11 12 13 14 15 16 17 18 19 20 21
elaboration
0
0
0
0
projection
extension
projection
enhancement
elaboration
0
0
0
0
extension
2
0
1
1
0
0
1
0
9 10 11 12 13 14 15 16 17 18 19 20 21
0
0
0
0
0
0
0
0
0
0
0
0
projection: ‘I said ...’; ‘she said ...’
3
0
0
0
enhancement: ‘and then ...’
3
1
0
3
1
0
1
1
0
0
0
0
4
1
0
6
0
0
0
0
7
3
0
8
8
0
1
0
0
7
0
0
0
0
6
8
6 7
0
0
0
2
5
0
0
0
0
4
4 5
0
0
0
0
3
0
0
0
0
2
3
0
0
0
enhancement 0
1
2
1
elaboration
projection
extension
0
1
2
3
4
5
6
7
8
& Mulac, 1991; Halliday, 1985/94: 354-363). Elaboration is quite frequent in casual chat; at
around 35%, it is the most common type of relation used to link clauses. This means that a
good deal of chat develops by restatement, clarification, exemplification and the like. In
contrast, elaboration is quite infrequent in stories (for children); at around 5%, it contrasts
sharply with the dominant type of relation — enhancement, at around 50%. This makes very
good sense: stories develop mainly through sequence in time, which is one type of
enhancement, whereas elaboration is used only to provide some characterization, description
or background. Stories for children contrast in interesting ways with news reports. In news
reports, enhancement is only a little over 30% but elaboration is more than three times as
common as in stories, no doubt reflecting the fact that news reports are, generically speaking,
no longer stories — they are no longer a kind of narrative, but rather a kind of report (cf.
Nanri, 1993; Iedema, Feez & White, 1994). News reports stand out among all the registers in
the sample in the way that the projection of locutions (quoted and reported speech) has been
foregrounded at around 30%. The modern news report is just as much about what people say
as about what they do. Very often the news is a semiotic event, like a statement or the
publication of a document; but even when it is concerned with a material event like a disaster
or a death, locution will play a prominent part in the form of comments by politicians, experts
and eye-witnesses. In this way, it is possible to review all the quantitative patterns that
emerge in Figure 2-11 and to interpret them in qualitative terms. There is no point in trying to
comment on all the patterns that emerge since the present selection is merely illustrative.
However, it is worth noting that certain tendencies seem to be pervasive. For example,
enhancement will always play a major role whereas hypotactic extension will always play a
minor one. The former never goes below around 30%, whereas the latter never goes above
3%. Such differences in orders of magnitude are likely to be manifested all along the cline of
instantiation.
23
Like all other lexicogrammatical systems, the system of clause complexing is thus
inherently probabilistic in nature. This is one kind of indeterminacy inherent in language;
but like all other systems, the system is also systemically indeterminate: the systemic terms
24
are all “prototypes” with clear, distinct centres but with peripheries that shade into one
another. We can think of the system as defining a multidimensional semantic space where the
systemic terms correspond to regions that are more or less close to one another and which
may shade into one another: this is a topological perspective that complements the
typological one of the system network (see Martin & Matthiessen, 1991). At the same time,
we can also think of the terms in the system network as representing fuzzy sets (rather than
crisp ones) in Zadeh’s (1987) sense, with varying degrees of membership (see Matthiessen,
1995b). The indeterminacy of the terms in the systems of clause complexing becomes very
clear when one analyses large volumes of text; but they are also apparent in the system itself,
for example from the range of meanings of a number of structural conjunctions (cf.
Matthiessen, 1995a: 148-149) and from the emergence, and paths of grammaticalization of,
new uses of such conjunctions over time (e.g. Hopper & Traugott, 1993: 177-184; Traugott,
1985), as illustrated in Figure 1-2 above. Table 2-3 shows how elaboration, extension and
enhancement are related through multiple meanings of structural conjunctions (and the non-
finite clause type ∅ v-ing). It is very common for languages to display this kind of spread in
the senses of conjunctions. For example, in Modern Standard Arabic, like and in English, wa
has both an extending sense of addition, contrasting with /aw and /am ‘or’, and an enhancing
sense of sequence in time, contrasting with fa and Tumma (Holes, 1995: 217-220).
Table 2-3: Multiple meanings of structural conjunctions
expansion projection
elaboration extension enhancement
and ‘and also’[additive] ‘and then’
[temporal], ‘and so’
[causal]
but ‘and on the other ‘and yet, although’
hand’[adversative] [concessive]
while ‘and also’ [additive]; ‘meanwhile’
‘and on the other [temporal]
hand’ [adversative]
or ‘in other words’ ‘alternative’ ‘otherwise’
[expositive] [alternative] [conditional:
negative]
∅ v-ing ‘specifically’ ‘while, after’
[clarifying] [temporal]
since [i] ‘because’
[causal]
[ii] ‘from then on’
[temporal]
as [i] ‘because’
[causal]
[ii] ‘when’
25
[temporal]
[iii] ‘like’ [manner:
comparison]
if, ‘in case’ [condition] [reported yes/no
whether question]
Finite ^ conditional clause [direct yes/no
Subject with interrogative question]
sequence Finite ^
Subject
2.2.1 Textual
The textual system of THEME is a resource for organizing the “flow of information” within
the clause as a message in the unfolding text. In a ‘declarative’ clause, the unmarked Theme
is the Subject; this often remains constant as a clause complex unfolds and this continuity
may be indicated by ellipsis:
I have to go || and [∅] find the scissors. [UTS-Macquarie Corpus, 76_146]
The lbo saw the Europeans and the power [[they had]] || and [∅] listened to [[what they
preached‘; || they put two and two together, || and [∅] said, you know, || these people
are so powerful, there must be something in [[what they believe]]. [Interview with
Chinua Achebe, 16_29]
This illustrates how a clause complex can serve as the domain of local thematic progression.
In her work on Japanese, Thomson (1998) treats the clause complex as a thematic domain.
The significance of the clause complex as a textual domain also includes referential
strategies, as in languages with so-called switch-reference systems (e.g. Longacre, 1985;
Whaley, 1997: 276-279).
A marked Theme may set a circumstantial frame of reference in the first clause of a
clause complex that continues to hold for the remainder of the clause complex. For example,
27
in the following clause complex, in Ibo belief is the domain for both the initiating clause and
the continuing clause:
In Ibo belief, every man and woman is created separately by a unique god-agent
[[called a chi]], || and this chi is virtually responsible for success in life. [Interview with
Chinua Achebe, 16_32]
The operational domain of the system of THEME in fact extends to the clause complex
(Matthiessen, 1995a: 154-158): when alternative sequences of clauses within a nexus are
possible, as they typically are in a hypotactic nexus, the contrast is a thematic one. This is the
reason why the conditional dependent clause appears before its dominant clause in the
complex in Figure 1-1: the clause sets up the local context of conditionality in which the
speaker intends the complex to be interpreted. Thompson (1984) shows how this applies to
hypotactically related clauses of purpose: initial purpose clauses have thematic status but
final ones do not; and thematic purpose clauses may specify the theme for a passage of text
that extends beyond the clause complex of which they are part (cf. Section 3.2). Thompson &
Longacre (1985: Part II) show how initial hypotactically related clauses (“adverbial clauses”)
often serve to distill information from the preceding discourse to present it as the point of
departure for the next move. This is again a thematic choice, relating to the method of
development of the discourse (cf. Fries, 1981), and the same principle is in fact in operation
at the level of semantics within rhetorical paragraphs and whole texts (cf. Martin’s Chapter
11 in Halliday & Martin, 1993, and Halliday’s, 1987b, notion of “hierarchy of periodicity” in
discourse).
While the system of THEME is a resource for assigning textual prominence to
elements within the clause, thus adapting it to its textual environment within and beyond the
clause complex, the textual system of CONJUNCTION is concerned with transitions in text
(cf. Bateman & Matthiessen, 1993; Matthiessen, 1992). This system complements clause
complexing: it provides the resources for indicating rhetorical relations regardless of whether
they hold within a clause complex or across clause complexes (Halliday & Hasan, 1976: Ch.
5; Martin, 1992: Ch. 4). The system covers the same range of relations as the structural
conjunctions that indicate the structural relations of hypotaxis and parataxis within clause
complexes (Halliday & Hasan, 1976: 227-230; Halliday, 1984/94: 328-329, 398-403;
Halliday & Matthiessen, 1999: 222-226), but unlike structural conjunctions, cohesive ones do
not indicate taxis (in terms of Rhetorical Structure Theory, they do not differentiate between
nucleus-satellite relations and multi-nuclear ones): see Table 2-4 for examples (the
“circumstantiation” column will be referred to below). Semantic patterns beyond the clause
complex will be discussed in Section 3.2 below.
28
2.2.2 Experiential
While the logical system of CLAUSE COMPLEXING is a resource for construing our
experience of the “flow of events” in the world as sequences of “quanta of change”, the
experiential system of TRANSITIVITY is the resource for construing each quantum of
change as a configuration of a process, participants directly involved in this process, and
more indirectly involved circumstances, as illustrated in Figure 2-13. This example is a nexus
where one clause of ‘material’ process type is linked to another of ‘behavioural’ process type
through a relation of paratactic enhancement. In general, different types of relation favour
clauses of different kinds of process. For example, paratactic temporal sequence often
involves one ‘material’ clause followed by another ‘material’ one. But the interpretation of a
clause nexus frequently depends on more delicately differentiated process types. Thus the
following paratactic nexuses marked with and are most likely ‘extending’ (‘and also’):
Kukul crouched low to the ground || and moved slowly.
His skin became feathers || and his hair, a gorgeous crest.
The first nexus involves two ‘material’ clauses. They are both concerned with movement in
space; but the movements, crouching and moving, are interpretable as perspectives on the
same movement rather than as two sequenced movements. The second nexus links two
inchoative intensive ascriptive relational clauses and they are again interpretable as different
perspectives on one and the same transformation rather than as two sequenced
transformations. In contrast, the following paratactic nexuses with and are most likely
‘enhancing’ (‘and then’). They link ‘material’ clauses that represent conventional activity
sequences (cf. Martin, 1992: 321-325, for the notion of activity sequence):
29
In pain, Kukul pulled out the arrow || and headed for the river || to wash his wound.
He fell onto a sea of emerald grass || and there he died.
They carved its image into stone || and placed it on their temples and palaces.
The limiting case of constraints on process type in a logico-semantic relationship is
projection, where the projecting clause is always either ‘mental’ or ‘verbal’ (or ‘behavioural’
pressed into verbal service to quote a clause (complex)).
The interpretation of the relationship between the clauses that enter into a nexus
within a clause complex thus depends on transitivity factors; but other experiential systems
also come into play — particularly, the system of tense/aspect. For example, in (Modern
Standard) Arabic, the common conjunction wa ‘and’ may, as already mentioned above, be
interpreted as either extending (‘and also’) or enhancing, temporal — sequence (‘then’) or
simultaneity (‘while’), and the aspectual selections in the clauses linked by wa help
determine the interpretation, as illustrated by the following two examples taken from Holes
(1995: 218):
experiential logical
circumstantiation hypotaxis parataxis
projection 24 — 3.7% 438 — 30.70% 185— 12.50%
expansion elaboration 11 — 1.7% 218— 15.30% 295 — 19.90%
extension 27 — 4.2% 20 — 1.40% 560 — 37.80%
enhancement 579 — 90.3% 748 — 52.50% 441 — 29.80%
641 — 100% 1424 — 100% 1064 — 100%
32
2.2.3 Interpersonal
The interpersonal system of MOOD is the central resource of the clause as a quantum of
dialogic interaction; it is concerned with the arguability status or negotiatory value of the
clause. The system of MOOD is open to ‘free’ clauses (such as we would all be dead in
Figure 1-1 above) but not to ‘bound’ ones (such as If there weren’t trees on the earth, um
earth in Figure 1-1); the contrast between ‘free’ and ‘bound’ is thus a contrast in
interpersonal status in the development of dialogue. Free clauses select freely for different
mood options, each one thus being capable of making an independent contribution to the
development of dialogue: ‘indicative’ clauses exchange information, either giving
(‘declarative’) or demanding (‘interrogative’) it, whereas ‘imperative’ clauses typically
demand goods-&-services. For example, the following passage from pre-dinner interaction
between a mother and her 7-year-old son, whom we have already met in the example in
Figure 1-1 above, is developed entirely through ‘free’ clauses.
dominant clause in a hypotactic relationship may be either ‘free’ or ‘bound’; just as with
parataxis, the status of the clause is not influenced by the nature of the role in the
relationship.) In the development of dialogue, hypotaxis often plays an important role since it
assigns unequal status to the elements of a complex: the dominant element is the proposition
or proposal developing the argument, whereas the dependent elements only provide support.
For example, the following passage is taken from a later stage in the pre-dinner interaction
already quoted above. After moving on to other issues (such as whether Dano should have a
shower before dinner), the interactants return to lasagna:
Mother free: indicative: declarative Well, a hungry man will even eat
lasagna.
Dano free: indicative: declarative [crying] Well, I won’t;
free: indicative: declarative I hate it,
free: indicative: declarative I just hate it.
Mother free: indicative: declarative Lasagna?
Dano free: indicative: declarative Yeah; I just hate it;
free: indicative: declarative I just loathe it.
Mother α free: indicative: declarative But you might like this one
x
β α bound: finite ‘cos I remember
x
β ’β1 bound: finite you’ve eaten it in the <Dano: What?>
past
x
β ’β =2 bound: finite and you’ve actually quite liked it.
Dano free: indicative: declarative No, I won’t.
Although the finite bound clauses look like declaratives in that Subject and Finite
appear in the declarative sequence Subject ^ Finite, they are not true statements (cf. Hooper
& Thompson, 1973; Harris & Campbell, 1995: Ch. 10): for example, they cannot generally
be tagged. Thus while we can say he was leaving, wasn’t he, so she left we cannot say
because he was leaving, wasn’t he, she left (and tagging a non-finite bound clause is
completely impossible). This makes sense since the tag is a device for eliciting a negotiatory
evaluation from the listener of a proposition or proposal presented as arguable: ‘tell me
whether you agree with my statement or not’. In many languages, there are modal contrasts
that are (largely) specific to bound clauses such as the contrast between ‘subjunctive’
(irrealis) and ‘indicative’ (realis). The mood value of bound clauses is determined by their
structural function: they are either hypotactically dependent within a clause complex, as in
the example above, or embedded to serve as a constituent within a clause or a group.
Table 2-6 sets out the relationship between MOOD and TAXIS. This is a relationship
between the interpersonal and the logical: as illustrated in the table, this is a link between two
clines — the interpersonal cline of arguability (how arguable is the proposition / proposal
enacted by the clause?) and the cline of interdependence (how interdependent is the clause
35
structurally on other clauses).6 The second cline has two outer values, both of which represent
the lack of a tactic relationship in the grammar: cohesive interdependence and embedding.
The link between the interpersonal and the logical is also manifested in the area of tone.
Contrasts in tone (falling vs. rising vs. level vs. falling-rising vs. rising-falling and more
delicate distinctions) realize interpersonal contrasts; but the contrast between falling (tone 1)
on the one hand and level (tone 3) vs. falling-rising (tone 4) on the other is also significant in
the area of taxis: in a sequence of tones, tone 1 indicates ‘no further tactic relationship’, tone
3 indicates ‘paratactic link’ (to the following clause) and tone 4 indicates ‘hypotactic link’ (to
the following clause); see Halliday (1967a: 34-36), Halliday & Greaves (in prep.: Chapter 2)
for the account.
Table 2-6: MOOD (cline of arguability) and TAXIS (cline of interdependence)
MOOD cohesion TAXIS [clause embedding
[non-structural] complex] [clause]
paratactically hypotactically (downranking)
interdependent dependent
free finite √ √ — —
[He was leaving [he was leaving
(wasn’t he). (wasn’t he), so
Therefore she she left]
left.]
bound finite — √ √
[because he was [the reason for
leaving (*wasn’t her departure
he), she left] was that he was
leaving (*wasn’t
he)]
non- — √ √
finite [because of him [the reason for
leaving, she left] her departure
was him leaving]
6
In this systemic functional interpretation, Givón’s (e.g. 1980, 1995) binding hierarchy is thus composed of a
number of functionally motivated clines, including the two clines shown in Table 2-6 but also the scale of rank
and the parameter of logico-semantic type.
36
118), illustrated by examples taken from a retelling in English for children of a Guatemalan
legend in Table 2-7. This cline is iconic: the more integrated the grammatical construal is, the
more unified the quantum of change is; and the more separate the grammatical construals are,
the more distinct the quanta of change are. As we have already seen from the ideational part
of the cline, different regions along the cline favour different kinds of projection and
expansion (see Table 2-5 above).
Table 2-7: The cline of integration (examples taken from The Hummingbird King)
structural ideational: circumstantiation One day,
experiential Kukul was hunting in the forest.
As a young boy,
he spent long hours with his father.
ideational: hypotaxis As he came to a thicket, ||
he heard the faint rustling of
logical leaves.
parataxis He heard the rustling of leaves ||
and [∅] raised his bow
and arrow.
In describing the cline in terms of the degree of distinctness between events in the
construal of change along the flow of events, I am of course foregrounding the ideational
perspective since it is the ideational metafunction that is concerned with our construal of
experience. However, the other two metafunctions also play a role in defining the cline:
[i] Ideationally, the cline is one of experiential distinctness between processes and
salience of individual process. At one pole of the cline, there is just one process, which is
expanded by a circumstance without a process in its own right (e.g. one day). At the other
pole of the cline, there are two processes, each salient in its own right and each construed by
a grammatically separate clause: the listener or reader has to infer the relationship between
the clauses. Intermediate between these two poles is the zone of clause complexing, shading
into circumstantiation on the one hand and cohesion on the other.
[ii] Interpersonally, the cline is one of arguability (as already indicated in Table 2-6
above). At one pole of the cline, there is just one proposition or proposal; the circumstantial
element is an Adjunct within this proposition / proposal, but it is not assigned propositional
status in its own right — it cannot be negotiated separately in dialogue. At the other pole of
the cline, there are two independently arguable propositions / proposals, each with the full
array of interpersonal features of negotiation. Intermediate between these two poles is the
zone of clause complexing. Parataxis means that the two propositions / proposals are still
independent of one another, but one of them may be subject to structural ellipsis of e.g. the
Subject, Finite or Predicator (as in He heard the rustling of leaves and [∅] raised his bow
37
and arrow or His skin became feathers and his hair, [∅] a gorgeous crest). Hypotaxis means
that the dependent clause does not have the status of a prototypical proposition / proposal; it
is removed from the mainline of dialogic negotiation and is not variable in mood, being either
non-finite or finite (structurally, like a declarative clause). Thus when a speaker says β When
you look at it, α you think “Oh yes!”, don’t you? (Svartvik & Quirk, 1980: 202), the mood
tag don’t you can only relate to the α clause — that is, you think, don’t you, but not you look,
don’t you.7 (This is one reason why examples of projection such as I think people just have
to, don’t they?(op cit., p. 205) must be interpreted as metaphorical: ‘people surely just have
to, don’t they?’.)
[iii] Textually, the cline is one of informational prominence and density. At one pole
of the cline, there is just one message, so there is just one assignment of thematic and
rhematic status. The circumstantial element may be thematic or rhematic. If it is thematic, it
will be marked Theme and is often given the status of a unit of information on its own, with
the remainder of the clause being a second unit of information; the information unit that
includes the marked Theme is then likely to be spoken on tone 4 (falling-rising), just like a
hypotactically dependent clause (Halliday, 1967a: 33; see Section 3.1 below); for example
(Theme underlined): // 4 ^ In / this job / Anne we’re // 1 working with / silver //(Halliday,
1994: 368). At the other pole of the cline, there are two messages, so there are two
assignments of thematic and rhematic statuses, one for each independent clause. This is also
true of the intermediate region of the cline — that of clause complexing; but here there is
often thematic continuity from one clause to another (as noted above in Section 2.2.1). In the
case of hypotaxis, the Theme is typically unmarked in the dependent clause (as in β: Because
they have only one institution, α: they don’t have different ah you know levels of um attention
that the males do, you know. [UTS/Macquarie Corpus, Text 10]) and usually elided when the
clause is non-finite (as in α: You brought some β: after [∅] having bought some.
[UTS/Macquarie Corpus, Text 11]); but as noted above in Section 2.2.1, the dependent clause
may itself be thematic within the clause complex.
Each region located along the cline illustrated in Table 2-7 is itself extended along the
cline. Thus some circumstances are closer to being part of the clause nucleus whereas others
are in a sense closer in status to (non-finite) dependent clauses in hypotactic clause nexuses;
dependent clauses in turn range from non-finite ones that presume various features from the
main clause to finite ones that could, in principle, stand on their own (Halliday, 1985/1994:
240-241); some of these finite dependent clauses lie on the border of paratactically related
clauses; and some paratactically related clauses in turn lie on the border of cohesively related
clauses. (And through the process of grammatical metaphor, additional points are added
along the cline; in particular, circumstances with nominalizations are intermediate between
ordinary circumstances and hypotactically dependent clauses: see Section 3.3 below.)
In a study of (what is here called) hypotactic projection of ideas, Thompson & Mulac
(1991) show that projecting cognitive clauses with a first person Senser/ Subject are in
various ways more closely integrated with the projected proposition than are clauses with a
7
It is important to emphasize that we are talking about a cline of arguability. Clauses that look like (finite)
dependent ones can still be tagged under certain conditions, as in That’s the one I should have if I had any, I
think, because it’s jolly, isn’t it? (Svartvik & Quirk, 1980: 210). Under certain conditions even embedded
clauses can be tagged, as in It’s this sort of thing that makes an absolute fool of Mallet, doesn’t it?, because this
is one of the … this is the line that you’ve been plugging (op cit.: 62).
38
third person Senser/ Subject; for example, they are much more likely to occur without the
binder (“complementizer”) that: see Figure 2-15. This reflects the status of clauses such as I
think, I reckon, I suppose in the grammatical system: they have come to serve not only as
projecting clauses in clause complexes but also as interpersonal Adjuncts of modality within
the projected clause (cf. also Halliday, 1985/1994: 355-356 and Matthiessen, 1991). This
integration into the projected clause is reflected in the strong tendency to favour clauses
without that, as Thompson & Mulac (1991) show. This tendency to occur without that is in
fact part of a long-term transformation in the status of clause nexuses of this kind. They have
developed out of what I think can be interpreted as paratactic nexuses of elaboration in Old
English on the model of they told me this: that it was too late, as suggested by Hopper &
Traugott’s (1993: 185-189) account (on the connection between elaboration and projection,
cf. Matthiessen, 1991). Here OE πæt served both as a pronoun and as an emerging marker of
projected clauses. It came to serve exclusively as a binder marking projected clauses but
Rissanen (1991) shows that that decreased steadily from 1350 to 1710 with (certain verbs of)
both mental and verbal projection: see Figure 2-16.
1000
800
600 Subject: oher
400 Subject: you
200
0 Subject: I
unmarked 0
marked that
Figure 2-15: Marking of projection in modern conversation (AE), from Thompson &
Mulac (1991) — SUBJECT PERSON
39
200
180
mental: know: th
160
mental: think: th
140 vebal: tell: 0
verbal: say: th
verbal: say: 0
120 vebal: tell: th
mental: think: 0
100 mental: know: 0
mental: know: 0
80 mental: think: 0
vebal: tell: th verbal: say: 0
60
verbal: say: th vebal: tell: 0
40
20 mental: think: th
0
mental: know: th
1350-1420
1420-1500
1500-70
1570-1640
1640-1710
Figure 2-16: The decline of that as a binder marking certain mentally and verbally
projected clauses (from Rissanen, 1991: 279)
To interpret the phenomenon that Thompson & Mulac (1991) draw attention to, we
have to add an interpersonal branch to the cline set out in Table 2-7, as shown in Figure 2-17.
The interpersonal branch is concerned with the manifestation of projection as interpersonal
assessment within the clause. This includes both modality — I think : probably; I urge you :
you must — and evidentiality — they say : reportedly. In either case, the meaning of
projection is manifested structurally within the clause as (i) an interpersonal Adjunct, realized
either by an adverbial group (probably), a prepositional phrase (in all probability) or a clause
of the type discussed by Thompson & Mulac (I think); or (ii) a modal Finite, realized by a
modal auxiliary (must). For further discussion of this interpersonal branch in English,
Japanese and other languages, see Matthiessen & Teruya (in prep.).
The cline is based on the fact that there are certain patterns of meanings that
“permeate” the grammatical system (Halliday & Hasan, 1976: 227-230; Halliday, 1985/94:
328-329) — the patterns of projection and expansion. Projection and expansion permeate the
grammatical system in the sense that they are patterns of meaning that are manifested in a
range of grammatical environments — the environments forming the cline shown in Figure
2-17 (and other environments as well). In relation to these different grammatical
environments, projection and expansion are fractal patterns of meaning (Halliday &
Matthiessen, 1999: 222-226; Matthiessen, 1995a: 91, 97): they are recurrent patterns of
meaning manifested in different grammatical environments, varying somewhat from one
40
environment to another but reflecting the same basic pattern. When we move from the
lexicogrammatical stratum to the semantic one, we find that these meanings are in fact also
manifested within the semantic domain of the text as rhetorical relations (see Section 3.2).
textual:
cohesive
# ##
COHESION
>1 clause
logical: 1 2
TAXIS: parataxis
logical: !
TAXIS: hypotaxis "
structural
experiential: interpersonal:
TRANSITIVITY ASSESSMENT
1 clause
Figure 2-17: Cline of integration with both experiential and interpersonal branches
the vantage point of semantics — will make it possible to relate discourse-semantic patterns
of rhetorical complexing to the lexicogrammatical cline discussed here.
x
β falling-rising // 4 By the time the Great Central was built //
α falling // 1 the trains could manage the gradients much more easily
and the //
x
β falling-rising // 4 ^ Well to / start with
α falling it // 1 means being / free of / worries and re/sponsi/bilities //
42
falling // 1 ^ I / know //
falling // 1 Then / all of a / sudden she / started wearing / make-up //
falling //1 ^ It was a / real / classic //
And a sequence of two or more instances of the same tone — tone concord — indicates that
the combining relation is one of elaboration (Halliday, 1967a: 36; 1985/94: 306); for
example:
As always in language, the marked departures from the unmarked case are motivated
and meaningful (as Halliday, 1967a: 36, puts it, “as regularly with intonation choices, there is
a probabilistic correlation but the choice remains”).
semantics: lexicogrammar:
RST complexing
x
satellite: 1 β By the time the Great Central was built,
circumst.
nucleus: α the trains could manage the gradients much more easily
x
2 1 α and the Great Central line usually went across the valleys
+
β instead of[going] round them like the earlier railways
x
2 1 so the distances were shorter
8
In our discussion, we focussed on hypotactic enhancement; but the proposal applies to all types of relation.
44
+
nucleus: 2 and you got better view.
satellite: The whole of the way between Nottingham and London there
elaborate was nowhere where another railway line crossed overhead.
While Halliday’s account of clause complexing had been worked out in the 1960s
(see e,g, Halliday, 1965, 1967a: 34-36, 1977; Berry, 1975), Rhetorical Structure Theory was
developed fairly independently of it, starting around 1983 with the work by Bill Mann, Sandy
Thompson and myself. It was thus quite striking when it turned out that both accounts posited
the same kind of relational organization — even though this point in fact emerges clearly
from the approaches to discourse developed earlier by Grimes (1975) and Longacre (1970,
1976).9 As already hinted at, this can be explained if we assume that clause complexing
grammaticalizes rhetorical complexing within the semantics, as suggested by Matthiessen &
Thompson (1988).
The general principle is very straightforward. At the semantic stratum, texts are
organized as rhetorical complexes — passages are linked through rhetorical relations of
projection and expansion and there may be internal nesting: the passages that are linked may
themselves consists of passages linked by rhetorical relations. This gives us the kind of text
organization brought out by Rhetorical Structure Theory and by other accounts that interpret
text in terms of relational organization — accounts such as Grimes (1975), Longacre (1976),
Beekman, Callow & Kopesec (1981) E. Pike (1992) and Martin (1992).10 Rhetorical
complexes are realized lexicogrammatically (and then, in turn, via the lexicogrammar,
phonologically). The lexicogrammatical opportunities for realization fall within the cline set
out in Table 2-7 and in Figure 2-17 above; they are either textual or logical but not
experiential (except through metaphor; see below):
textual: [i] The rhetorical relation of the rhetorical complex is marked conjunctively
by means of a cohesive conjunction (Halliday & Hasan, 1976: Ch. 5) such as for
example, moreover, however, nevertheless, therefore, meanwhile or by a
structural conjunction functioning cohesively at the beginning of a clause
complex (And, Or, But). [ii] Alternatively, the rhetorical relation of the rhetorical
complex is not marked explicitly but is left implicit and has to be inferred from
other lexicogrammatical patterns such as the selection of theme and lexical
cohesion (cf. Martin, 1992, on implicit conjunctive relations).
logical: The rhetorical complex is realized by a clause complex. This means that not
only is the rhetorical relation marked (as with cohesive conjunctions) but the
tactic nature of relation is also marked, as are the scopes or domains of the
passages being related. In principle, nucleus-satellite relations are realized by
hypotactic links and multinuclear ones are realized by paratactic links (see
Matthiessen & Thompson, 1988: 308).
9
The point can only be seen if hypotaxis is differentiated from “embedding”; the notion of “subordination” fails
to make this important distinctions (see Matthiessen & Thompson, 1988).
10
RST is more like Grimes’ (1975), Longacre’s (1976) and E. Pike’s (1992) accounts in at least one respect:
like Grimes but unlike Martin (1992), we operate with a tactic-like distinction between hypotactic (nucleus-
satellite) relations and paratactic (multinuclear) ones.
45
Father: But you’ll be home from school at three o’clock, er, you, you’ll be able to do plenty
of things in the afternoon, but at now, now in the aftern— <Dano: Plenty?>, in the
morning, you won’t need to do half an hour homework.
Dano: Well, do I have to do more in the afternoon?
Father: No, probably not. Just do half an hour now.
Dano: Don’t do it in the afternoon?
Father: Dano. [sighs]
Dano: Okay, I’ll do it now [2 seconds] if you desperately want me to.
Father: Sorry?
Dano: If you desperately want me to I’ll do it now.
Father: All right, er, if you do half an hour now. … Come on, my darling boy … you want to
do it here or, no let’s do it in the kitchen.
However, Dano could also have realized the rhetorical complex of condition by either a
cohesive sequence of clauses not grouped into a clause complex — You desperately want me
to. In that case, I’ll do it now, or by a paratactic complex — You desperately want me to, and
in that case, I’ll do it now. All the possibilities are represented diagrammatically in Figure
3-1 — we could even add an additional variant, Do you desperately want me to? I’ll do it
now, where there is no cohesive conjunction to mark the relation of conditionality, but the
condition is realized as an interrogative clause (Figure 1-2). Each lexicogrammatical strategy
in turns has a default realization in the phonology — a distinct sequence of tone groups,
discussed the preceding subsection.
SEMANTICS
RHETORICAL
RELATIONS
condition
LEXICOGRAMMAR
PHONOLOGY
textual:
cohesive
# ##
COHESION
TONE
SEQUENCE
You desperately want me to. In that case, I’ll do it now.
tone 1 ^ tone 1
>1 clause
logical: 1 2
TAXIS: parataxis
46
tone 3 ^ tone 1
You desperately want me to, and in that case, I’ll do it now.
patterns
1 x2
logical: !
TAXIS: hypotaxis "
tone 4 ^ tone 1
x" !
structural
tone 1 ^ tone 1
experiential: interpersonal:
TRANSITIVITY ASSESSMENT
(circumstances) (mood Adjuncts)
tone 3 ^ tone 1
1 clause
PHONOLOGY
TONE
tone 4 ^ tone 1 SEQUENCE
! CONJUNCTION:
!: Therefore I
[you] vote!
Therefore,
urge you
enhancing
! ="
we
endorsing ...
!: I don’t
‘": that
! CONJUNCTION: enhancing
think
But
open government
I
let’s
...
Rheme
! COMPLEX
’"x"
we
if
Rather,
!’"
other players
we
! x"
where
! COMPLEX
!!
we
When
x"
we
! COMPLEX
we
!
tempting
x"
endorsing ...
!: I don’t
believe
‘": that
Figure 3-2: Rhetorical analysis of the “California Common Cause” text (Mann &
Thompson, 1985), with lexicogrammatical realizations superimposed
In any given text, it is very likely that all the different lexicogrammatical strategies
will be used. This is illustrated by means of one of the short written texts discussed in the
RST literature by Mann & Thompson (1985, 1986), analysed lexicogrammatically by
Matthiessen (1995a: 827-842) and used as a basis for comparison of RST and conjunctive
48
analysis by Martin (1992: 244-261) — the California Common Cause text, set out in Table
3-1 together with clause complex analysis. The rhetorical analysis presented and discussed by
Mann & Thompson is shown in Figure 3-2 together with aspects of the lexicogrammatical
realization. Rhetorical complexes that are realized by complexes are shown in boxes together
with the structure of the complex (Matthiessen, 1995a: 829) and structural conjunctions are
shown in italics (for examples of other texts, see op cit.:168-173). Rhetorical relations that
are marked by cohesive conjunctions are annotated by Ê CONJUNCTION: [type of
conjunction] and the conjunction is shown in bold italics where they occur in the text. These
are direct realizations of rhetorical relations. In addition, I have exemplified how the textual
system of THEME contributes to the realization of the rhetorical development of the text (see
Matthiessen, 1995a: 839-840, for the full thematic analysis): the clause that realizes rhetorical
unit [9] in the text has a Theme (Open government, campaign finance reform, and fighting
the special interests and big money) that identifies the point of elaboration in relation to the
previous text (most immediately, the Rheme of the clause that realizes unit [8] (stick to those
issues of governmental structure and process, broadly defined, that have formed the core of
our agenda for years). (For a more extended example of theme selection in relation to
rhetorical organization, see Matthiessen, 1995a: 581.)
Table 3-1: California Common Cause text (from Mann & Thompson, 1985)
[1] I don’t believe
α:
that endorsing the Nuclear Freeze initiative is the right step for California CC.
‘β:
[2] x Tempting as it may be,
β:
[3] we shouldn’t embrace every popular issue that comes along.
α:
[4] x When we do so
β:
[5] we use precious limited resources
α α:
[6] where other players with superior resources are already doing an adequate job.
α xβ:
[7] Rather, I think
α:
we will be stronger and more effective
‘βα:
[8] if we stick to those issues of governmental structure and process, broadly defined, that
‘β xβ:
have formed the core of our agenda for years.
[9] Open government, campaign finance reform, and fighting the special interests and big
money — these are our kinds of issues.
[10] Let’s be clear:
[11] I personally favour the initiative and ardently support disarmament negotiations to
reduce the risk of war.
[12] But I don’t think
α:
‘β: endorsing a specific freeze proposal is appropriate for CCC.
[13] We should limit our involvement in defense and weaponry to matters of process,
α:
=
β: such as exposing the weapons industry’s influence on the political process.
[14] Therefore, I urge you
α:
“β: to vote against a CCC endorsement of the nuclear freeze initiative.
(Signed) Michael Asimow, California Common Cause Vice-Chair and UCLA Law
Professor
49
11
Concession may be interpreted either internally — ‘I concede that x, but I still hold that y’ — or externally —
frustrated cause. According to the relation definitions given in Mann & Thompson (1987), concession is treated
as an internal relation (cf. Mann & Matthiessen, 1991). However, here I will use the external interpretation (for
example, in the counts presented below), treating it as frustrated cause. In general, it behaves more like external
enhancing relations than like internal ones.
50
12
I have left out all occurrences of the justify relation since in Stuart-Smith’s analysis they involve projection to
a large extent — Smith et al have claimed that … and the like. I have also left out all uncertain cases. As already
mentioned, concession is interpreted as external rather than internal. For elaborating relations, I have not
differentiated between internal and external: all instances have been given under “external”.
51
elaboration are in a sense the reverse of one another: while enhancement decreases steadily as
we move from circumstantial realization within the simple clause to cohesive realization
within text beyond the domain of clause complexes, elaboration increases. Like elaboration,
extension increases throughout the cline from clause to text, but it decreases at two points —
hypotaxis and cohesion. The second decrease is more significant; but it is quite likely to
reflect the rhetorical nature of the specific sample of undergraduate psychology essays. Thus
while the counts for clauses and clause complexes based on samples of registerially diverse
texts (both spoken and written), the count for patterns beyond clause complexes is based on a
registerially homogeneous sample (written psychology essays, as mentioned above), so this
count is more likely to show the effects of registerial skewing. The tendencies are fairly clear;
but the picture is quite tentative: future research will have to explore a much larger,
registerially varied sample, and it will have to provide counts for all the points of realization
along the cline from clause to text from the same sample.
Figure 3-3: Expansion manifested structurally within clause complexes and cohesively
between clause complexes (summary of Stuart-Smith, in prep.: Ch. 9, Table 2)
52
Texts differ individually in what rhetorical relations they draw on, just as they differ
in what types of clause nexuses they favour and the two are obviously related: rhetorical
variation is realized in clause-complexal variation. But such variation is not confined to
53
individual texts; it is characteristic of the text types or registers (genres) that individual texts
instantiate (cf. Kamyab, 1997: 261-271). For example, the pervasiveness of internal
enhancing relations in the global organization of the California Common Cause text is, as
already noted, characteristic of persuasive registers — as illustrated by the analyses of letters
of appeal in Mann, Matthiessen & Thompson (1992) and Abelen, Redeker & Thompson
(1993) and of advertisements in Mann & Thompson (1987) and Fries (1992). In contrast,
expository texts such as encyclopedic entries are dominated by relations of elaboration (cf.
Mann, 1988; Matthiessen, 1995a: 579-584), whereas procedural and narrative texts involve
relations of sequence in time. Similarly, different registers have distinct profiles of clause
complexing.
However, texts and registers vary not only in which patterns of rhetorical complexing
and clause complexing they favour and disfavour, but they also vary in how they divide the
labour of lexicogrammatical realization. The picture shown in Figure 3-4 represents
something of a registerial patchwork: as already noted, the counts for clausal
circumstantiation and clause complexes are based on registerially mixed samples whereas the
counts for cohesive patterns come from Stuart-Smith’s (in prep.) study of written psychology
essays. If we had similar counts for texts characteristic of prototypically spoken English and
texts characteristic of prototypically written English, I predict that we would find interesting
differences in the way that the labour of realization is divided.
In prototypically spoken English, clause complexes may reach great intricacy (see
Halliday, 1985, 1987a), as already illustrated by the example in Figure 2-5 above. What this
means is that clause complexes can extend beyond local rhetorical patterning in text to realize
extended rhetorical complexes. One way of exploring this characteristic is to track the
unfolding of spoken text, simply measuring the number of clauses per clause complex. One
of the many texts I have analysed in this way is an interview with Dirk Bogarde, published on
the web. The pattern of unfolding in this text is quite typical — a pattern of peaks and troughs
in the intricacy of clause complexes: see Figure 3-5. The highest peak is part of a narrative
sequence in one of Bogarde’s turns in the interview:
||| In another instance, a great friend of mine from California, a cousin, was staying
with me in France, || we were raking in the hay; || I said || "I'm getting old || I'm going
to have to sit down in the shade," || and she said, || "Thank God, || at least we can do it.
|||
||| Wouldn't it be awful || if..." |||
||| And that's exactly [[what happened to her]]. |||
||| She had a massive stroke || while she was dressing for dinner in San Francisco, ||
and that was the end of it. |||
||| People with [[whom she was dining]] couldn't find her. |||
||| When she didn't arrive, || they went to her flat || and found her on the floor || and got
her to the hospital. || but she was a total vegetable. |||
||| Now, she had told me || — and she'd told my sister and one or two other people-||
that if such a thing ever did happen to her || she didn't want any help || and could we
get it over to her daughter || not to be put through it. |||
||| As it happens, || Margo was an extremely rich woman. |||
54
||| Her daughter wanted "the best" for Mommy-|| so Mommy got the best [[that money
can buy]]. || and a lot of people profited || in sustaining her hopeless condition. |||
||| She had day and night nurses of all kinds, shapes and sizes. |||
||| They stripped her clean of every bit of jewellery [[she ever had]], || though that's
neither here nor there. |||
||| The doctors charged exorbitant fees; || the hospital bills mounted month by month.
|||
Ë ||| One day her daughter Penny went || to see her || and called me afterwards; || she
said: || "I must tell you, || something extraordinary's happened, || I must tell you || what
it is, || it's simply incredible, || she cried." || and I said || what do you mean || she cried?
|| so she said: || "Well we were in there || and suddenly she must have heard my voice
|| because there was one long tear [[that started to trickle down]]." |||
||| Now, come on, || that was Margo saying || "For Christ's sake, let me go! || I'm
[[where I promised never to be]]!" |||
||| It took a year for her to die. |||
Such passages are very common in spontaneous spoken English. Rhetorical relations
play a central role in the ongoing development of casual conversation: they allow the text to
develop fairly locally — not as a pre-planned, constructional whole but as a spontaneously
unfolding journey of exploration; at each point in this journey, it is possible to expand the
text further by projecting or expanding to include a new passage. From Eggins’ (1990) study
of casual dinner table conversation among close friends, we can see how the rhetorical-
relational resources enable interactants to go on for several hours — provided they have
something that fuels the conversation such as the need, among close friends, to explore
difference against the background of shared values and experiences. And clause complexing
is precisely the lexicogrammatical resource that can congruently choreograph such rhetorical
patterns: as I pointed out at the beginning of this chapter, the clause complex is an open and
dynamic resource, constructed out of relations rather than out of the constituents of a
constructional whole. The clauses combined into such complexes are internally fairly simple:
they tend not to be greatly expanded by circumstantial elements and they tend not to involve
many embedded clauses. Consequently, the density of information per unit of information is
kept manageable for real-time processing.
18
1! One day her daughter Penny went 1x" to see her x21 and called me afterwards; |x2=21 she said: x2=2”21! "I must
tell you, x2=2”21”" something extraordinary's happened, x2=2”2=2! I must tell you x2=2”2=2”"11 what it is,
16 x2=2”2=2”"1=2it's simply incredible, x2=2”2=2”" =2 she cried." x31 and I said x3”21 what do you mean x3”2’2 she
cried? x41 so she said: x4”21 "Well we were in there x4”2x2! and suddenly-she must have heard my voice- x4”2x2x"
because there was one long tear
14
12
10
1
6
11
16
21
26
31
36
41
46
51
56
61
66
71
76
81
86
91
96
101
106
111
116
121
126
131
136
141
146
151
156
161
166
Figure 3-5: The unfolding of clause complexes in an interview with Dirk Bogarde
struck me when I have analysed such passages of spoken English is how similar the analysis
seems to be to various characterizations of languages without a (long) tradition of writing.
Thus in describing certain languages of New Guinea and South America, Longacre (1985:
282-283) discusses “the problem of the ‘endless’ sentence”:
In respect to Foré and certain other languages of New Guinea we noted above that the medial-final chain is
equivalent to the paragraph rather than to the sentence and that shorter levels of chaining characterize
sentences within the paragraph. We also challenged the assumption that the medial-final chain is
necessarily a sentence in structure. It is a shock to realize, however, that in some languages, both in New
Guinea and South America, we sometimes find chaining carried to such (by our standards) excessive
lengths that the chain is plausibly neither a sentence nor a paragraph — unless we consider that the body of
a discourse consists of but one sentence or one paragraph. While the latter is not impossible and, in fact,
characterizes some shorter discourses, it is hard to believe that a text of seven or eight pages reduces to the
structure of one sentence or even that the body of a text is simply one sentence. It seems here that we must
take stock and realize again that chaining is a surface-structure phenomenon which is capable of being
plugged into various functions. As such,. we can expect it to confine itself neither to the sentence nor to the
paragraph in all languages.
Thus, Millie Larson (1978) presents an Aguaruna discourse in which, while the aperture of the discourse,
closure, and finis are discrete and consist of smaller chains, the entire intervening body of the discourse
which contains the episodes of the story is one long run-on chain of over sixty clauses. Furthermore, this
chain is very evidently not a simple linear sequence. It may itself be divided into paragraphs and sentences
by various defensible criteria. In such an instance, very plainly the surface structure of chaining is no longer
marking sentence or paragraph, but is really co-extensive with the entire body of the discourse.
A similar situation probably exists for Waffa in New Guinea. In Waffa, in at least one discourse type (the
legend narrative discourse), Hotz and Stringer (1970) report that there are long chains which clearly can be
broken up into weakly delineated paragraphs and weakly delineated sentences. Furthermore, there
obviously are groups of paragraphs distinguished from each other within such a long chain. These groups
seem to correspond to weakly delineated embedded discourses. (Footnotes omitted.)
What Longacre calls a “sentence” corresponds to our clause complex. A sixty-clause clause
complex is clearly significantly more intricate than Bogarde’s 17-clause complex or even the
clause “megaplex” in Figure 2-5 above; but the principle behind them is, I believe, very
similar — both emerge in a locally managed way, engendered by the logical metafunction.
In contrast, registers characteristic of prototypical writing in English do not tend
towards highly elaborated clause complexes. The California Common Cause text is not
characteristic of prototypical written discourse, but even so the most intricate complex
involves only three clauses. Prototypical writing includes academic books and articles and
administrative, bureaucratic and legal discourses. The following extract comes from a text
book in geology (Robert J. Foster. 1971. Physical Geology. Columbus, Ohio: Merrill.):
||| Dolostone (dolomite) is composed of dolomite. ||| It is recognized by effervescence
after scratching (to produce powder) with dilute hydrochloric acid; || it will also react
(without scratching) with concentrated or with warm dilute hydrochloric acid. |||
Dolostone is generally formed by replacement of calcite, presumably soon after
burial. ||| The reduction in volume in this replacement may produce irregular voids ||
and generally obliterate fossils. |||
57
[Geology exposition] —
[Bogarde interview] —
maximum: 17 clauses
average: 2.6 clauses
clauses / complex
8
0
20
18
16
14
12
10
Figure 3-6: Comparison of the unfolding of spoken text (light grey) and written text
(dark grey)
The extract is characteristic of the geology text: clause complexes tend not to be very
intricate but there are many nominalizations (underlined in the passage above). To get a sense
of how this written text unfolds through clause complexes, we can compare it with a spoken
58
passage such as the interview with Bogarde:13 Figure 3-6 shows 93 clause complexes from
the geology text and the Bogarde interview. The difference between the two texts is very
striking. On the one hand, the average number of clauses per clause complex is considerably
higher for the spoken text (2.6 as opposed to 1.6) and on the other hand, the potential for
intricacy is much greater: the maximum number of clauses per complex is 17 in the spoken
text but only 4 in the written one.
The picture that emerges from the comparison of the two texts in Figure 3-6 does
seem to be representative of the general contrast between spoken and written mode. This can
be seen when we compare them in terms of the relative number of nexuses at different
“levels” of layering: see Figure 3-7. Both modes show a steady decline from level 1 (the most
local nexuses) to level 7 (the most global nexuses). (I have used a logarithmic scale so that
the lower numbers at the higher levels of layering will still show up clearly as part of the
overall trend.) However, the written mode occupies fewer levels: there are no complexes with
nexuses above level 5 and fully 70% of all nexuses occur at level 1. The spoken mode starts
lower (around 60% at level 1) extends to level 7. It is these tendencies that are significant
because the point to potentials for the two modes. Thus if we compare the two modes in
terms of the average number of clauses per complex, they are likely not to be hugely
different. In my sample, the average is 3.0 for spoken and 2.6 for written: see Table 3-3.
However, as the table shows, prototypical examples of spoken and written discourse are
further apart. Casual conversation (represented by gossip) has an average of 4.1 clauses per
clause complex, whereas scientific writing (represented by a chapter from a conference
proceedings on climate change) has an average of 2.5 clauses per complex. Here it is the
spoken variety that differs most from the average for the total. In addition, the table also
gives counts of the average number of words per clause — 5.8 for casual conversation and 19
for scientific writing (the latter being very high, even for written discourse).14 This count
gives an indication of how “compressed” the discourse is in terms of the density of
information per clause.
Table 3-3: Spoken and written mode — counts for total sample and for prototypical
examples
spoken: gossip spoken: total written: total written:
scientific report
# of clauses 341 3870 2197 120
# of complexes 64 932 653 30
(of ≥ 2 clauses)
# clauses / 4.1 3.0 2.6 2.5
complex (of ≥ 2)
# words / clause 5.8 7.6 10.2 19
13
The ideal would be to compare pairs of text where the only variable is the mode — (prototypical) spoken vs.
(prototypical) written.
14
These counts are simply based on orthographic words. It would be more revealing to count lexical items,
excluding grammatical ones, since this would directly measure the lexical density (cf. Ure, 1971; Halliday,
1985). When grammatical items are excluded, the difference between spoken and written mode emerges more
clearly.
59
Figure 3-7: Spoken and written mode and percentage of nexuses at different levels of
layering (logarithmic scale)
It is certainly possible to find spoken discourse where the potential for intricacy is not
taken up — for example, arguments with short turns or banter during a coffee break. And it is
also possible to find written discourse where clause complexes are more intricate — for
example, procedural discourse. Thus the following algorithmic passage could be analysed as
a single clause complex of 23 clauses (in spite of the punctuation): see Figure 3-8.
1. Put the start node on a list, OPEN, of unexpanded nodes.
2. Remove the first node, n, from OPEN.
3. Expand node n — generating all its immediate successors and, for each successor
m, if m represents a set of more than one subproblem, generating successors of m
corresponding to the individual subproblem. Attach, to each newly generated node, a
pointer back to its immediate predecessor. Place all the new nodes that do not yet
have descendants at the end of OPEN.
4. If no successors were generated in (3), then
a. Label node n. unsolvable.
60
x x
2 β 5. Otherwise, if any terminal nodes were generated in (3), then
α 1 a. Label these terminal nodes solved.
+ x
2 β b. If the solution of these terminal nodes makes any of their
ancestors solved,
α label these ancestors solved.
+ x
3 β c. If the start node is labeled solved,
α exit with success.
+
4 d. Remove from OPEN any nodes “that are labeled solved or
that have a solved ancestor‘.
x
5 6. Go to (2).
Figure 3-8: Algorithm interpreted as (largely enhancing) clause complex (total of 23
ranking clauses)
However, pushing the clause complex analysis this far is probably unhelpful and the
algorithmic passage is in any case not like prototypical writing; it could, in principle, be a
sequence of spoken instructions. In contrast, the geology text is closer to prototypical writing
— though it is not an extreme example of it (cf. Figure 3-6 above). Both spoken and written
discourse is organized at the level of semantics as complexes of rhetorical relations; but they
differ in how these complexes are realized lexicogrammatically.
In spoken discourse, clause complexes may extend quite far in realizing such
rhetorical complexes, covering not only short local passages but also less local ones such as a
whole episode in a narrative; but in prototypical writing such as the geology text, clause
complexes tend to extend much less far in realizing clause complexes. The two modes thus
differ in how they use the cline set out in Table 2-7 and Figure 2-17 above. Interestingly, they
also differ within the logical zone of the cline with respect to which type of taxis they favour.
While spoken text favours parataxis over hypotaxis, written text favours hypotaxis over
parataxis. This is shown by a count of the 2,870 clause nexuses in the sample of spoken and
written text used earlier: see Figure 3-9 (cf. with the comparable picture for all the texts in the
sample given earlier in Figure 2-9). Overall, the ratio of hypotaxis to parataxis is roughly
30% to 70% in speech and roughly 50% to 50% in writing. Speech thus shows a clear skew
towards parataxis whereas writing shows only a very slight preference for hypotaxis (51.4%).
If we consider these differences in relation to “level”, they are even more striking. (Here level
1 represents the least intricate complexes and level 7 the most intricate ones found in the
sample of texts.) As the graph shows, in speech, there is a steady increase in parataxis as we
move towards higher levels, but there is a steady decrease in hypotaxis as we move towards
higher levels. Thus while hypotaxis dominates slightly locally, parataxis dominates globally.
In writing, hypotaxis dominates significantly locally; but as the level increases, the difference
between hypotaxis and parataxis becomes less marked.
We can generalize the situation as follows:
Hypotaxis dominates over parataxis in local organization; that is, the closer we are to
the experiential organization of the single clause in Table 2-7, the more dominant
is hypotaxis.
62
Parataxis dominates over hypotaxis in global organization; that is, the closer we are
to the textual organization of sequences of clause complexes in Table 2-7, the
more dominant is parataxis.
Speech makes use of global organization whereas writing does not.
This provides evidence for the view that has often been expressed that speech is “looser”,
more paratactic-like; but it shows us very clearly that the division of labour between parataxis
and hypotaxis in speech has to be seen in relation to “level”. The increasing dominance of
parataxis at higher levels can be seen as related to Eggins’ (1990) finding that casual
conversation can go on for hours since the interactants are able to keep expanding the
discourse relationally. In contrast to speech, writing clearly favours hypotaxis locally and
disfavours global, or higher-level, organization within the clause complex. This can be
interpreted as a shift towards the experiential pole of the cline in Table 2-7.
In prototypical writing, the division of realizational labour seems to be shifted away
from the logical strategy of clause complexing towards the experiential domain of the clause
(cf. Halliday & Matthiessen, 1999: 264): see Figure 3-10. In relation to spoken discourse, this
shift is one of de-complexing and down-ranking. Complexes are compressed into clauses and
clauses are compressed into nominal groups, so the informational density is higher in writing
than in speech (cf. the measure of words per clause presented in Table 3-3 above). A central
feature of this shift is nominalization — e.g. reduce Ë reduction, replace Ë replacement;
but there are other concomitant features — e.g. conjunctions are replaced by prepositions, as
with because Ë in, or by verbs (in relational clauses), as with so Ë produce. We can see the
difference between the written mode and the spoken mode, if we reword some of the
examples from the geology text in a more spoken-like mode:
† ||| You can recognize it || by boiling it || after scratching it <<(to produce powder)>>
with dilute hydrochloric acid || …
† ||| Dolostone is generally formed || by calcite being replaced, || presumably soon
after it has been buried. |||
† ||| The volume is reduced || because [calcite] is replaced, || so it may become
irregularly empty || and fossils are generally obliterated. |||
63
In writing, the clausal system of CIRCUMSTANTIATION (cf. the cline set out in
Table 2-7 and Figure 2-17 above) is brought into service to realize parts of rhetorical
complexes: the rhetorical relation is realized by a preposition (rather than by a structural
64
conjunction) and the rhetorical unit by a nominal group with a nominalization as Head (rather
than by a clause; the nominalization is typically a process nominalization corresponding to
the Process of the clause, but other types corresponding to certain other clausal elements also
occur: see Matthiessen, 1995a: 678-682). In addition to the system of circumstantiation, the
metaphorical mode of realizing rhetorical relations also involves the system of PROCESS
TYPE: in ‘relational’ clauses, the rhetorical relation is realized by a verb — as if it were a
process; for example: The reduction in volume in this replacement may produce irregular
voids and generally obliterate fossils. This also happens with internal relations, e.g. Large
amounts of feldspar in a sandstone may imply rapid deposition and burial — ‘we know that
[material] may have been deposited rapidly and buried because feldspar is abundant in
sandstone’. In one further step, the rhetorical relation is itself nominalized as cause, manner,
condition and the like or the projecting process is nominalized (claim, requirement,
suggestion, proposal etc.) and this nominalized relation or process is construed as a
participant in a relational clause of identity, as in the following examples (from Global
Change and the Changing Atmosphere by William Clark):
the only way [[we will ever get out of playing "crisis response" to the degree [[that we
have been doing of late (be it in ozone depletion, acid deposition, or some other
"problem of the month")]] ]] is [[to get the necessary broad-based basic research
going]]
The problem [[that has preoccupied scientists and administrators over the last several
years]] is [[how such a goal can be approached in practicable, doable steps [[that, at a
minimum, do no harm to scientific research already under way]]]] — steps [[that, in a
more optimistic vein, promote some of the new long-term research, observations, and
synthesis [[that are necessary to turn the notion of connections into a real revolution in
our understanding of, and ability to cope with, global change in the geosphere-
biosphere system]]]].
Again, a first requirement is [[to do no harm to organizational frameworks [[that,
through years of evolution, are finally at the stage [[where they are supporting
programs [[that are actually helping us to get on with the business of increasing
understanding]]]]]]]].
As the examples illustrate, relational clauses of this kind may serve as hosts to quite a few
embedded clauses: the clause structure is very simple — participant = participant; but the
participants are construed as experientially very “heavy” through nominalization and
embedding.
The shift in the division of labour in the realization of rhetorical complexes is in the
first instance a shift within the ideational metafunction from the logical mode of construal to
the experiential one and it has significant consequences for the way in which experience can
be construed — as flow of events or as classes of entities. But at the same time there are also
textual consequences affecting the organization of the flow of information (see Halliday &
Martin, 1993) and interpersonal consequences affecting the way ideational meanings are
propositionalized since a nominalization group is even less arguable than a non-finite clause.
In terms of the cline in Table 2-7, it seems very clear that writing relies more on the
textual resource of explicit cohesive CONJUNCTION than does speech. This is illustrated by
the following passage of scientific writing (cohesive conjunctions in bold):
65
||| The data showed [[that this theory is completely and utterly wrong]]. ||| The oxides of
nitrogen were measured || as being unusually low. ||| Some other theories [[that require
an increase in nitrogen compounds]] are likewise incorrect. ||| The fluorocarbon-halon
theory suggests [[that there should be a change in the partitioning of chlorine from the
inactive forms of chlorine, namely hydrochloric acid and chlorine nitrate, into the
active forms of chlorine, namely chlorine atoms and chlorine oxide radicals]]. |||
Therefore, Anderson's tests were critical || for determining || whether the chlorine
oxide (and bromine oxide) abundances were enhanced. ||| Another theory,
<<advanced by K.-K. Thong,>> requires a change from downward to upward
motions over Antarctica in association with other circulation changes. ||| If this is
correct, || one should see enhanced levels of tropospheric trace gases such as nitrous
oxide and methane in the lower and middle stratosphere. ||| Therefore, the Heidt
measurements were critical for this purpose. ||| The ER-2 aircraft could not climb
higher than 18.5 km because of the very cold, dense atmosphere and the need [[to
carry a lot of fuel for safety reasons]]. ||| Also, it did not range farther south than 72oS
latitude, about midway down the Palmer Peninsula. ||| Therefore, many of the
measurements were made close to the inside edge of the polar vortex. ||| It would have
been scientifically desirable [[to have flown higher and farther southward in the
vortex]]. ||| In any event, our flights from Puntas Arenas to 72oS and back were useful
|| for comparing conditions just inside the vortex to those outside. ||| One of our first
flights was made on August 23, 1987. ||| Water vapor dropped from about 3 ppm
outside the vortex to about half this value inside, || indicating [[that the atmosphere
inside the vortex was dehydrated]]. ||| Ozone changes were only slight across the
vortex boundary. ||| However, the abundance of the chlorine monoxide radical (ClO)
increased from about 10 parts per trillion by volume to about 500 parts per trillion. |||
[From Stratospheric Ozone Depletion: Antarctic Processes by Robert T. Watson.]
The cohesive conjunctions help the reader construct the reasoning by which the passage is
developed. A number of internal rhetorical relations are, however, realized metaphorically by
verbs in relational clauses — show, suggest and indicate. In this respect, writing thus relies
more on the outer, non-logical poles of the cline in Table 2-7 and Figure 2-17, whereas
speech relies more on the logical zone between these two poles.
The shift that has evolved in the division of labour in the realization of rhetorical
relations is additive within the total linguistic system in the sense that the range of possible
realizations in the lexicogrammar has gradually been expanded over the centuries as part of
the evolution of the registers of scientific English, Italian, etc. This evolution involves
patterns of meaning that can be traced back to Ancient Greek; but scientific English itself has
evolved together with modern science over the last 500 years or so (Halliday, 1988, 1998;
Halliday & Martin, 1993). The patterns of reasoning that are characteristic of modern science
may have precursors in the evolution of legal discourse (cf. Olson, 1994: 50, on the evolution
of “the forms of sceptical argument” in “judicial and political contexts” already in Greece).
Hoffman (1993: Ch. 4) suggests that the “papal revolution” (approximately 1072-1122) was
an important legal precursor to the development of modern science in the West: it involved
“working out a new legal system” (p. 125) — the canon law, and “it is this great legal
transformation … that laid the foundations for the rise and autonomous development of
modern science” (p. 119); the birth of the new science of law involved three separate
elements: a body of legal materials to work with; a new method of analysis; and a place in
66
which to carry on these new legal studies, the universities” (p. 124). These developments
were important both socially in establishing universities as legally independent bodies of
scholarship and semiotically in engendering an extensive range of new registers — registers
that are now associated with standard languages in modern nation states. These registers are
to a large extent written registers and they are produced and interpreted in institutions of the
modern state — institutions that are foreign to many (tribal) communities with a purely oral
tradition such as those of the languages mentioned by Longacre (1985) in the passage quoted
above. If the expansion of the range of realizations of rhetorical relations is part of the
evolution of these registers, then it follows that the division of realizational labour is a major
area of typological variation.
Central to this variation is the complementarity of the logical and the experiential as
modes of construing human experience. Both modes are in operation in all languages; but the
division of labour between them would appear to be variable across languages — just as it is
across the registers within a single language such as Chinese, English or Italian, with the
experiential taking over in the realization of rhetorical relations in registers of prototypical
writing. This suggests that grammatical metaphor of the ideational kind is quite a recent
development in the evolution of human languages — and one that has only taken place in
certain languages, viz. those that have evolved the range of registers referred to above (cf.
Rose, 1998: 432, on the relative absence of grammatical metaphor of the ideational kind in
Western Desert in General: “Western Desert has the same realisational potential for
transcategorisation as Greek, Latin and English, but ideational metaphor remains a minor
motif in the Western Desert code.”). It also suggests that there will be languages that favour
the logical mode over the experiential one, perhaps even in contexts where casual, spoken
registers of English would draw on the experiential mode. This area needs considerable work
since typological research has not, on the whole, been concerned with the division of labour
between the logical and the experiential and with the phenomenon of grammatical metaphor
of the ideational kind.
However, Pawley’s (1987: 335-336) account of Kalam, a language spoken in the
Highlands of Papua New Guinea, reveals a very rich use of the logical mode together with
constraints on the experiential mode:
We turn now to the comparison of English and Kalam. Such a comparison is of interest, I believe, in that it
may indicate roughly the outer limits of variation among languages in resources and conventions encoding
event-like phenomena. In these respects Kalam may be as different from English as any language on earth.
… English speakers who are learning Kalam, or translating Kalam discourse into English, find many Kalam
accounts of happenings extraordinarily explicit and long-winded. When describing an activity Kalam
speakers will routinely single out for mention certain aspects or components of the activity that English
speakers normally leave out or fuse together with other components. Kalam speakers often found my
descriptions of events to be cryptically or ambiguously telegraphic, relying heavily on assumed knowledge
and inference. (p. 335) … one important difference between English and Kalam is that Kalam has few event
expressions that are very high on the scale of conceptual complexity. Kalam clauses (if they are clearly
single clauses) usually denote simple events. English speakers, on the other hand, freely use single clauses
to represent episodes and other complex conceptual events. (p. 336)
Thus the movement of an object along a path can be construed as a simple event by a
single clause in English — for example, The man threw a stick over the fence into the
garden; but “in Kalam such an ‘event’ must be encoded as an episode, a sequence of four
conceptual events: (1) the man takes hold of the stick, (2) the stick is thrown, (3) it flies over
the fence, (4) it falls into the garden.” (op cit., 353-354). Kalam thus construes the motion
67
along the path as a logical sequence rather than as an experiential configuration. Here we
would assume that Kalam routinely uses rhetorical sequences in the development of text
where English might only have a single rhetorical unit. This also relates to the nature of the
experiential system of process type in Kalam — for a discussion of this and of languages
with “serial verb constructions” (complexing at group rank), see Halliday & Matthiessen
(1999: 316-318) and Matthiessen (forthc.: Section 9.3.1).
SEMANTICS
RHETORICAL
RELATIONS
condition
LEXICOGRAMMAR
PHONOLOGY
textual:
cohesive
# ##
COHESION
TONE
SEQUENCE
You desperately want me to. In that case, I’ll do it now.
tone 1 ^ tone 1
68
>1 clause
logical: 1 2
TAXIS: parataxis
expansion
tone 3 ^ tone 1
You desperately want me to, and in that case, I’ll do it now.
metaphorical
realization 1 x2
logical: !
TAXIS: hypotaxis "
tone 4 ^ tone 1
x" !
structural
tone 1 ^ tone 1
experiential: interpersonal:
TRANSITIVITY ASSESSMENT
(circumstances) (mood Adjuncts)
tone 3 ^ tone 1
1 clause
PHONOLOGY
My performance of it depends on your desparate desire for it
TONE
tone 4 ^ tone 1 SEQUENCE
4 Conclusion
In this chapter, I have tried to locate clause complexing as system within the total meaning-
making resources of language. This means locating it along the dimensions of instantiation,
stratification, metafunction and rank — both in order to map out its ecological niche and to
show how clause complexing shades into other strategies. The location of the system of
clause complexing is shown diagrammatically in Figure 4-1.
hypotaxis
METAFUNCTION
TAXIS
parataxis
idea
system
projection
clause
LOGICO- locution
SEMANTIC
TYPE
elaborating
expansion extending
enhancing
ACTIVITY
stop
RECURSION
go on
SEQUENCE
RHETORICAL
STRATIFICATION
COMPEXING
phylogenesis —
evolution of system
CLAUSE TRANSIT- MOOD THEME;
COMPEXING IVITY COJUNCT.
TONE
SEQUENCE
ACTIVITY
registers/
SEQUENCE
text types
RHETORICAL
COMPEXING
TONE
SEQUENCE
systemicization of
instantial, emergent
patterns
ACTIVITY
texts
SEQUENCE
RHETORICAL
COMPEXING
logogenesis —
CLAUSE
COMPEXING
TRANSIT-IVITYMOOD THEME;
COJUNCT.
unfolding of text
TONE
SEQUENCE
18
16
14
12
10
0
11
16
21
26
31
36
41
46
51
56
61
66
71
76
81
86
91
96
1
101
106
111
116
121
126
131
136
141
146
151
156
161
166
Figure 4-1: The system of CLAUSE COMPLEXING within the total linguistic system
The point of this fairly holistic exploration of clause complexing is not to draw sharp
boundaries around it as a grammatical system, insulating it from other systems. Rather it is
precisely the opposite — to interpret clause complexing as a grammatical zone with
permeable boundaries that shades into (i) cohesive sequences of clauses that are not
structurally related and (ii) circumstantial expansions of the clause nucleus. This grammatical
zone is located stratally between rhetorical complexing (semantics) and tone group
70
sequencing (phonology); but it is not insulated from these systems. We can only understand
clause complexing fully if we explore how it realizes rhetorical complexes and how it is
realized by sequences of tones. The value of such a cross-stratal view is demonstrated very
clearly by Ford & Thompson (1996) in their study of interactional units in conversation. In
the unmarked case, semantic, lexicogrammatical and phonological patterns correlate; but
against the norm of the unmarked case, marked mappings emerge as meaningful alternatives.
Clause complexing is thus highly indeterminate — as is every other system in
language: a zone that shades into other zones. It is also indeterminate in the sense that it is an
inherently variable system that is constantly changing. We can look at this process of change
close up, focussing on individual texts — as is illustrated in Figure 3-5 above. This change in
the course of the unfolding of text is logogenesis (the lower vertical arrow in Figure 4-1). It is
through the cumulative effect of logogenesis that the system of clause complexing changes
— the evolution of patterns of clause complexing as part of the phylogenesis of language (the
higher vertical arrow in Figure 4-1). The bridge between logogenesis and phylogenesis is the
evolution of registers — like the evolution of scientific registers referred to above in relation
to the changing division of labour between the logical and experiential modes of the
ideational metafunction. For example, the evolution of news reporting over the last 250 years
or so has involved the transformation of the traditional chronologically organized news story
— essentially a single enhancing paratactic-like sequence — into today’s news report that is
not organized chronologically but rather as a multi-perspectival cycling through the events,
with early prominence given to those events that are most likely to grab the reader’s attention
— a nucleus that is then elaborated by multiple, hypotactically related satellites (cf. Nanri,
1993; Iedema, Feez & White, 1994). These are changes in the patterns of rhetorical
complexing; but the changes have of course also been lexicogrammatical. One aspect of the
modern news report is the way it defines the division of labour among the different
combinations of TAXIS and LOGICO-SEMANTIC TYPE in clause complexing — in
particular, how it marshals the resources of reporting and quoting. One common pattern in
reports on disasters is a move from reports (hypotactic projection) attributed to officials and
experts to quotes (paratactic projection) attributed to eyewitnesses. Such changes in
registerial patterns are part of the evolution of the general system of clause complexing.
It is of course of interest to try to see further back in time to explore how clause
complexing might have evolved in languages. It has been noted for various languages with an
exclusively oral tradition that they tend to develop explicit markers, often through borrowing,
where they previously used asyndetic combinations when they come into contact with
(colonial) languages with more elaborate repertoires of conjunctive markers (Mithun, 1988).
This is an interesting development and is likely to reflect the registerial difference between
speech and writing in the use of conjunctive markers (cf. Chafe, 1988: 24). However, this is
in the first instance an evolution in the patterns of realization — not a change in the system of
clause complexing itself. I would think clause complexing is a very ancient grammatical
strategy — one that predates documentary evidence by many millennia. (As already
suggested above, the shift towards the experiential through grammatical metaphor of the
ideational kind is, in contrast, a much more recent development.) If ontogenesis can be taken
as a guide to phylogenesis, then it is clear that it is an early strategy: young children have
problems processing intricate clause complexes. The ontogenetic evidence suggests strategies
into clause complexing, such as the way into generalized mental projection via first-person
projection (I think etc.) that also has an interpersonal sense of assessment (Painter, 1993) or
71
6 References
Abelen, Eric, Gisele Redeker & Sandra A. Thompson. 1993. “The rhetorical structure of US-
American and Dutch fund-raising letters.” Text 13.3: 323-350.
Bateman, John. 1989. “Dynamic systemic-functional grammar: a new frontier.” Word 40.1-2:
263-87.
Bateman, John A. & Christian M.I.M. Matthiessen. 1993. “The text base in generation.”
Hermann Bluhme & Renzhi Li Keqi Hao (ed.), Proceedings of the international
conference on texts and language research, Xi’an, 29-31 March 1989. Xi’an: Xi’an
Jiaotong University Press.
Beekman, John, John Callow & M. Kopesec. 1981. The semantic structure of written
communication. Dallas: Summer Institute of Linguistics.
Berry, Margaret. 1975. Introduction to systemic linguistics: I structures and systems. London:
Batsford.
Capra, Fritjof. 1996. The web of life: a new synthesis of mind and matter. London:
HarperCollins.
Chafe, Wallace. 1988. “Linking intonation units in spoken English.” In Haiman &
Thompson (eds.). 1-27.
Cumming, Susanna & Tsuyoshi Ono. 1997. “Discourse and grammar.” In van Dijk (ed.).
112-137.
72
Halliday, M.A.K. 1979. “Modes of meaning and modes of expression: types of grammatical
structure and their determination by different semantic functions.” D.J. Allerton et al
(ed.), Function and context in linguistic analysis. Cambridge: Cambridge University
Press. 57-79.
Halliday, M.A.K. 1985. Spoken and written Language. Geelong, Vic.: Deakin University
Press.
Halliday, M.A.K. 1987a. “Spoken and written modes of meaning.” Rosalind Horowitz and S.
Jay Samuels (eds.), Comprehending oral and written Language. New York:
Academic Press. 55-82.
Halliday, M.A.K. 1987b. “An interview with Michael Halliday.” Conducted by Paul
Thibault. 601-627. In Ross Steele & Terry Threadgold (ed.), Language topics. Essays
in honour of Michael Halliday. Volume II. Amsterdam: Benjamins.
Halliday, M.A.K. 1988. “On the language of physical science.” Mohsen Ghadessy (ed.),
Registers of Written English: situational factors and linguistic features. London &
New York: Pinter Publishers. 162-178.
Halliday, M.A.K. 1985/1994. An introduction to functional grammar. 2nd ed. 1994. London:
Edward Arnold.
Halliday, M.A.K. 1992. “The notion of 'context' in language education.” Thao Le & Mike
McCausland (ed.), Interaction and development: proceedings of the international
conference, Vietnam, 30 March — 1 April 1992. University of Tasmania: Language
Education.
Halliday, M.A.K. 1993. “Towards a language-based theory of learning.” Linguistics and
Education 5.2: 93-116.
Halliday, M.A.K. 1998. “Things and relations: regrammaticising experience as technical
knowledge.” In J.R. Martin & Robert Veel (eds.), Reading science: critical and
functional perspectives on discourses of science. London: Routledge. 185-235.
Halliday, M.A.K. & William S. Greaves. forthc. Intonation in the grammar of English.
London: Equinox.
Halliday, M.A.K. & Ruqaiya Hasan. 1976. Cohesion in English. London: Longman.
Halliday, M.A.K. & James R. Martin. 1993. Writing science: literacy and discursive power.
London: Falmer.
Halliday, M.A.K. & Christian M.I.M. Matthiessen. 1999. Construing experience through
meaning: a language-based approach to cognition. London: Cassell.
Harris, Alice B. & Lyle Campbell. 1995. Historical syntax in cross-linguistic perspective.
Cambridge: Cambridge University Press.
Hasan, Ruqaiya. 1992. “Rationality in everyday talk: from process to system.” Jan Svartvik
(ed.), Directions in corpus linguistics. Berlin: de Gruyter.
Hjelmslev, Louis. 1943. Omkring sprogteoriens grundlaeggelse. Copenhagen: Akademisk
Forlag. (English translation: Hjelmslev, L. 1961. Prolegomena to a theory of
language. Madison, Wisconsin: University of Wisconsin Press.)
74
Holes, Clive. 1995. Modern Arabic: structures, functions and varieties. London & New
York: Longman.
Hooper, Joan B. & Sandra A. Thompson. 1973. “On the applicability of root
transformations.” Linguistic Inquiry 4: 465-497.
Hopper, Paul J., & Sandra A. Thompson. 1980. “Transitivity in grammar and discourse.”
Language 56.2:251-299.
Hopper, Paul J. & Sandra A. Thompson. 1984. “The discourse basis for lexical categories in
Universal Grammar.” Language 60.4:703-752.
Hopper, Paul J. & Elizabeth Closs Traugott. 1993. Grammaticalization. Cambridge:
Cambridge University Press.
Hotz, J. & M. Stringer. 1970. Waffa sentence, paragraph and discourse. Unpublished
manuscript, Ukarumpa, New Guinea, Summer Institute of Linguistics. Ref. in
Longacre (1985).
Huddleston, Rodney D. , Richard A. Hudson, Eugene Winter & A. Henrici. 1968. Sentence
and clause in Scientific English: final report of O.S.T.I. Programme. University
College London: Communication Research Centre.
Huff, Toby E. 1993. The rise of early modern science: Islam, China and the West.
Cambridge: Cambridge University Press.
Iedema, Rick, S. Feez & P. White. 1994. Media literacy. (Write it right industry research
report no. 2.) Sydney: NSW, Department of Education, Disadvantaged Schools
Program Metropolitan East.
Kamyab, Ghodratollah. 1997. Rhetorical structure analysis and popularised medical review
articles. Monash University: Ph.D. thesis.
König, Ekkehard & Johan van der Auwera. 1988. “Clause integration in German and Dutch
conditionals, concessive conditionals, and concessives. In Haiman & Thompson
(eds.). 101-133.
Larson, Millie. 1978. The functions of reported speech in discourse. Dallas: Summer Institute
of Linguistics. (Publications in Linguistics, 59.)
Lehmann, Christian. 1988. “Towards a typology of clause linkage.” In Haiman & Thompson
(eds.). 181-225.
Longacre, Robert E. 1970. “Sentence structure as a statement calculus.” Language 46: 783-
815.
Longacre, Robert E. 1976. An anatomy of speech notions. Lisse: de Ridder.
Longacre, Robert E. 1985. “Sentences as combinations of clauses.” In Timothy Shopen (ed.),
Language typology and syntactic description III: complex constructions. 235-286.
Mann, William C. 1988. “Text generation: the problem of text structure.” In David D.
McDonald & Leonard Bloc (eds.), Natural language generation systems. Berlin:
Springer-Verlag. 47-68.
Mann, William C. & Christian M.I.M. Matthiessen. 1991. “Functions of language in two
frameworks.” Word 42.3.
75
Mann, William C., Christian M.I.M. Matthiessen, and Sandra A Thompson. 1992.
“Rhetorical structure theory and text analysis.” In William C. Mann & Sandra A.
Thompson (eds.), Discourse description: Diverse analyses of a fund raising text.
Amsterdam: Benjamins.
Mann, William C. & Sandra A. Thompson. 1985. “Assertions from discourse.” Proceedings
of the Eleventh Annual Meeting of the Berkeley Linguistic Society. Berkeley: Berkeley
Linguistics Society.
Mann, William C. & Sandra A. Thompson. 1986. “Relational propositions in discourse.”
Discourse Processes 9.1: 57-90.
Mann, William C. & Sandra A. Thompson. 1987. Rhetorical Structure Theory: a framework
for the analysis of texts. Technical Report, ISI/RS-87-185. Marina del Rey, CA: USC/
Information Sciences Institute.
Martin, James R. 1988. “Hypotactic recursive systems in English: towards a functional
interpretation.” James D. Benson & William S. Greaves (ed.), Systemic Functional
Approaches to Discourse: Selected Papers from the Twelfth International Systemic
Workshop. Norwood, NJ.: Ablex. 240-70.
Martin, James R. 1992. English text: system and structure. Amsterdam: Benjamins.
Martin, James R. & Christian M.I.M. Matthiessen. 1991. “Systemic typology and topology.”
Frances Christie (ed.), Literacy in social processes: papers from the Inaugural
Australian Systemic Functional Linguistics Conference, Deakin University, January
1990. Darwin: Centre for Studies of Language in Education, Northern Territory
University.
Matthiessen, Christian M.I.M. 1988. Representational issues in systemic functional grammar.
James D. Benson & William S. Greaves (ed.), Systemic Functional perspectives on
Discourse. Norwood, NJ: Ablex. 136-175.
Matthiessen, Christian M.I.M. 1991. “Language on language: the grammar of semiosis.”
Social Semiotics 1.2: 69-111.
Matthiessen, Christian M.I.M. 1992. “Interpreting the textual metafunction.” Martin Davies
& Louise Ravelli (eds.), Advances in systemic linguistics: recent theory and practice.
London: Pinter. 37-82.
Matthiessen, Christian M.I.M. 1993. “Register in the round: diversity in a unified theory of
register analysis.” Mohsen Ghadessy (ed.), Register analysis: theory and practice.
London: Pinter. 221-292.
Matthiessen, Christian M.I.M. 1995a. Lexicogrammatical cartography: English systems.
Tokyo: International Language Sciences Publishers.
Matthiessen, Christian M.I.M. 1995b. “Fuzziness construed in language: a linguistic
perspective.” Proceedings of FUZZ/IEEE, Yokohama, March 1995. Yokohama.
Matthiessen, Christian M.I.M. 1995c. “THEME as an enabling resource in ideational
‘knowledge’ construction.” In Mohsen Ghadessy (ed.), Thematic development in
English text. London: Pinter. 20-54.
76