Complementation as interpersonal grammar.pdf

Full Terms & Conditions of access and use can be found at
https://www.tandfonline.com/action/journalInformation?journalCode=rwrd20
Word
ISSN: 0043-7956 (Print) 2373-5112 (Online) Journal homepage: https://www.tandfonline.com/loi/rwrd20
Complementation as interpersonal grammar
William B. McGregor
To cite this article: William B. McGregor (2008) Complementation as interpersonal grammar,
Word, 59:1-2, 25-53, DOI: 10.1080/00437956.2008.11432580
To link to this article: https://doi.org/10.1080/00437956.2008.11432580
Published online: 15 May 2015.
Submit your article to this journal
Article views: 479
View related articles
Citing articles: 2 View citing articles

WILL/AM B. McGREGOR
Complementation
as interpersonal grammar
Abstract. This paper is concerned with a range of complex sentence con-
structions including They said to shave, They knew that I had shaved, They
forced me to shave which are generally regarded as "object comple-
ments". It is proposed that this analysis is mistaken: complement clauses
do not serve as objects. Rather, I will argue that the majority of these con-
structions are characterized by interclausal grammatical relations of the
type referred to in McGregor (1997) as conjugational-whole-to-whole
relations rather than part-to-whole relations (as in the relation borne by a
genuine object). These grammatical relations construe interpersonal
meaning, and can be divided into two independent subtypes: framing and
scope, which relate to fundamentally different modes of signifying from
the interpersonal perspective, respectively demonstration and description.
1. lntroduction1• This paper is concerned with the grammatical struc-
tures of so-called "object complements", such as illustrated by exam-
ples (la)-(lg).
(la) I tried/wanted to shave
(1b) They made/forced/wanted me shave
(le) They waited for me to shave
(ld) They remember me shaving
(le) They suggested that I shave
(lf) They knew that I had shaved
(lg) They said to me "You should shave"
A comprehensive treatment of the domain must address at least the
following questions: (a) What grammatical relations are involved? (b)
What types of unit are involved in these relations? (c) What type of
structure results-e.g. is it a separate grammatical sign, i.e. construc-
tion? (d) What are the meanings of the various elements involved, and
to what extent is the meaning of the entire structure compositional? (e)
To what extent are the relations, units, structures, and meanings com-
mon/different across all complement constructions in a language, and
across languages?
25

26 WORD, VOLUME 59, NUMBER 1 (APRIL, 2008)
This paper focuses on (a), the grammatical relations in object com-
plement constructions in English, although occasional mention will be
made of corresponding constructions in other languages. The other
questions are dealt with in passing, being raised where they throw light
on (a); they are not dealt with systematically or exhaustively.
The received view-accepted by most grammarians, formalists
(e.g. Rosenbaum 1967, Bresnan 1979, Haegeman 1991) and function-
alists(e.g.Giv6n 1980,Noonan 1985,Bolkestein 1990,Dik 1997:95)-
is that the complement clause is embedded as a constituent of the
"matrix" clause.2 Moreover, it typically serves in an argument or par-
ticipant role, usually either subject or object.
My purpose is to present some evidence against these analyses, and
propose an alternative analysis for at least some ofthe constructions gen-
erally treated as complements. Specifically, I propose that many com-
plement clauses are not embedded as constituents of the matrix clause.
In these cases, the complement clause does not represent a part of the
whole construction that serves a grammatical function in it; although it
belongs to the whole construction, it does not discharge a role in that
structure.3 Rather, grammatical relations of fundamentally different
types are involved. These are relations ofone ofthe following two types:
(i) scope, whereby the matrix clause holds the complement in its
scope, modifying its propositional content (much as a negative par-
ticle holds a proposition in its scope) or speech-act status (like an
illocutionary particle such as an interrogative particle); or
(ii) framing, in which the matrix clause frames the complement
clause, indicating how it is to be taken interactively-effectively as
a representation of another utterance (much as a frame around a
painting indicates it is to be taken as an artistic object, as an iconic
representation ofthe world rather than as part ofthe world as such).
Putting things in slightly different terms, this amounts to a rejection
of the notion that the event denoted by the complement clause invari-
ably plays a substantive role in the matrix event, comparable with the
role played by the referent of an NP, as in the case of genuine clause
embedding. An example is (2). Here the non-finite clause smoking five
packets ofcigarettes a day is embedded in the matrix clause in the gram-
matical relation of Actor and Agent; it can be replaced by NP such as
John and the intense heat, which serve the same grammatical roles. The
sentence construes the habitual and excessive smoking event as agen-
tive in the killing.

McGREGOR: COMPLEMENTAT!ON AS INTERPERSONAL GRAMMAR 27
(2) Smoking five packets of cigarettes a day finally killed him
In (1g), by contrast, the projected future event of my shaving does
not serve the same sort of relation to the matrix event of speaking as a
set of abstract entities such as words does in the event denoted by They
said words to me. The latter clause construes a speech event directed
towards the speaker that comes to realization through the production of
words; the referent abstract entities play an integral role in the speech
event. In (lg) the referent event of the clause you should shave plays
no such role in the speech event. Rather, it is an utterance correspond-
ing to the complement clause-an utterance that could be enacted by a
clause like it-that is realized through the act of speaking. The matrix
clause indicates how the complement clause is to be "taken" interac-
tively: not as a representation of an event, but rather as a demonstration
or performance of a possible utterance (Clark and Gerrig 1990, McGre-
gor 1994).
The relation between the matrix and complement clause in (lg) is
interpersonal rather than experiential: it is concerned with the construal
of the complement utterance in the interactive context, and not with the
referent event of the complement clause as a representational compo-
nent of a larger situation.
This thesis is developed in the body of the paper, where it is pro-
posed that it applies not just to direct quotes, but, with modifications, to
many other object complement constructions. The basic intuition is that
where one clause falls within the scope of another, or is framed by
another, the interpersonal character ofthe former is modified by the lat-
ter. By contrast, where one clause serves as a constituent of another, as
in (2), a relation is construed between the referent situations in some
possible world. The two manners of clause combination construe dif-
ferent grammatical semiotics, experiential in the case of constituency,
interpersonal in the case of conjugational relations.
These observations are special cases of a more general hypothesis
that the grammars of human languages are organized around four pri-
mary semiotic components or types (McGregor 1997:73-80):
• interpersonal, which is concerned with the representation ofinter-
active phenomena in the speech event;
• experiential, which concerns the construal of the world of experi-
ence, possible worlds;
• logical, concerning the logical relations ('if', 'and', 'or', etc.) per-
.ceived among experiential phenomena; and

28 WORD, VOLUME 59, NUMBER I (APRIL, 2008)
• textural, which is concerned with texturing the utterance, unifying
its components with one another, with other utterances, and with
the non-linguistic context.
These semiotic types are characterized by syntagmatic relations of dis-
tinct types, respectively: conjugational (whole-whole relations, cover-
ing scope and framing, and obtaining between linguistic items capable
of representing whole utterances), constituency (part-whole relations),
dependency (part-part relations), and linking (indexical, non structure-
bound).
Thus syntagmatic relations are linguistic signs in the Saussurean
sense: they are constituted by mutually defining correspondences of
forms with meanings. Their establishment requires specification of both
aspects. It is beyond the scope ofthe present investigation to identify the
semiotically significant interclausal syntagmatic relations involved in
all object complement construction types in English (or any other lan-
guage). Instead, I discuss some significant characteristics of the gram-
matical relations in a selection ofthe constructions, and attempt to show
some ofthe advantages oftreating their form as conjugational, and their
meaning as interpersonal.
The discussion is organized as follows. I begin in section 2 by out-
lining an etic catalogue of the main types of complement constructions
in the world's languages, almost all of which can be found in English.
Section 3 turns to the various types of object complement constructions
represented by English, and presents evidence in favor of the proposal
that the interclausal relationships are in most instances of the interper-
sonal type. Section 4 comments on a few further theoretical issues. Sec-
tion 5 winds up the paper with a summary and reflections on the pre-
sumptions of the enterprise engaged in. The tentative nature of the
proposed analyses is acknowledged; my purpose is to present a general
framework for the analysis of a range of phenomena, and as a conse-
quence, many details are glossed over.
2. Range ofobject complements. Object complements have been clas-
sified in a variety of ways, morphological (e.g. according to mood ofthe
complement, presence of a complementizer, etc.), syntactic (e.g. raising
vs. equi, nature ofunits involved and their grammatical relations), prag-
matic (is the complement proposition presupposed (as for regret) or not
(as for believe)), and semantic (e.g. type of matrix event). The type of
classification I would ultimately aim for is a semiotic one that combines

McGREGOR: COMPLEMENTATION AS INTERPERSONAL GRAMMAR 29
syntagmatic relations with their semantics. Such a classification is, un-
fortunately, not yet available for any language.
For the present paper, with its focus on the semiotic aspects ofcom-
plement constructions, it is useful to begin with an admittedly etic
semantic classification of the domain of complement constructions to
give some notion of their extent. Such a categorization has been sug-
gested by Michael Noonan, who identifies fourteen "classes of comple-
ment-taking predicates" (Noonan 1985:110-133). This, he suggests, is
relatively exhaustive: it characterizes the semantic domains likely to be
covered by complement constructions in any language. There is no
implication that the semantic types identified will necessarily be repre-
sented by distinct complement construction types in a given language,
or indeed even by complement constructions. The fourteen types are
(where I construe them as applicable to the matrix clause, rather than
specifically to its verb):
• Utterance clauses, which concern the transfer of information via
the speech channel. These have verbs like say, tell, promise, ask,
inform, report, allege, state, acknowledge, certify, decree, declare,
proclaim, specify, stipulate, and so forth.
• Propositional attitude clauses, which express an attitude towards
the validity of the proposition encoded by the complement clause.
This may be a positive attitude, as with believe, think, suppose,
assume, or a negative one, as with doubt, deny.
• Pretense clauses, which put forward the complement situation as
a hypothetical or unrealized event. Verbs include imagine, pretend,
make believe, trick (someone into believing), and the like.
• Commentative clauses, commenting on the situation described
by the complement clause, evaluating it or reacting emotionally to
it, as with regret, be sad, be sorry, or a judgment on it, e.g. be
important.
• Knowledge and acquisition ofknowledge clauses,which describe
a state of knowledge or its acquisition. These employ verbs of cog-
nition (know, discover, realize,find out) and perception (see, hear).
• Fearing clauses, which describe a fear or concern that the comple-
ment has occurred, or will occur. Verbs include fear, be afraid,
worry, be anxious, and the like.
• Desiderative clauses, which express a desire that the complement
situation occurs, and employ verbs like want, wish, hope, desire,
like, etc.

• Manipulative clauses, which describe means of getting someone
to do something. Included are causatives and permissives (make,
force, have, cause, get, let, permit), illocutionary acts (tell, com-
mand, order, request, ask, cajole, beg), and propositional manipu-
lation, i.e. getting someone to adopt an attitude to a proposition
rather than perform an action (convince, persuade, and teach) (Dik
1997:107).
• Modal clauses, expressing either an epistemic (be possible, be
likely) or deontic (be obliged, ought, be necessary) evaluation of
the occurrence of the complement clause.
• Achievement clauses, indicating successful (manage, chance,
dare, remember, happen) or unsuccessful performance of the com-
plement event (try,fail, avoid, refrain, abstain,forget).
• Phasal or aspectual clauses, that specify a phase in the comple-
ment situation. Verbs include begin, continue, keep on,.finish, stop,
and cease.
• Sensory perception (Noonan's "immediate perception") clauses,
describing the sensory mode by which the complement situation
was perceived. The matrix clause has verbs like see, watch,
observe, hear, smell, ignore,feel, and imagine.
• Negative clauses, like English it is not the case that, and perhaps
negative predicates in languages such as sega ni in Fijian.
• Conjunctive clauses, which involve a conjunctive predicate such
as 'and then', as in Lango tt 'and then', which inflects as a verb,
albeit with a reduced inflectional paradigm.
Noonan's fourteen types can be reduced to a more manageable
scheme by extracting four recurrent features: (a) communicative,
grouping together matrix situations involving speech and other semiotic
systems; (b) cognitive, including human cognitive processes of all
types; (c) status, concerning the truth value of the complement clause;
and (d) Aktionsart, concerning the aspectual character of the comple-
ment clause. Table 1 shows how the types can be categorized according
to these features.

Table 1. Grouping of the major etic-semantic types of matrix clauses
Communicative Cognitive Status Aktionsart
Utterance Knowledge Negative Manipulative,
causative
Manipulative, Propositional Modal Achievement
illocutionary (epistemic)
Pretense Modal Phasal
Commentative (deontic) Conjunctive
Fearing
Desiderative
Sensory
perception
In the next section I take a selection of subtypes from each of the
four major etic categories in turn, and examine how they can be con-
strued in interpersonal terms. Doing this will involve a certain amount
of reorganization of the etic categories, and the exclusion of some.
Along the way, we will see that identification of the above domain of
complex sentence constructions in terms of an experiential grammar of
complementation, focusing as it must on modes of participating in
events (in particular, as Agent, Undergoer, etc.), is flawed. The domi-
nant semiotic models are the interpersonal ones of framing and scope.
3. Grammatical structures of the major complement types
3.1. Communicative complements. In section 1, I outlined a few
semantic considerations that argue against treating communicative com-
plements as embedded in, or as dependent on, the matrix clause of
speech. It was suggested that the grammatical relation between the
clauses is instead one of framing. The idea is that the clause of speech
encloses the quoted utterance, specifying its status as a demonstration of
an utterance, rather than-or at least only indirectly-as a construal of
an event. The clause of speech establishes a frame within which the
quoted clause is to be interpreted and evaluated (McGregor 1997:252ff).
In the case of direct quotation, as in (lg), the clause of speech and
the quoted clause each have their own independent deictic centre from
which deictic categories such as tense, person, and spatial and temporal
proximity are specified. They also have independent specification ofthe

modal categories, including illocutionary mood. By framing the repre-
sented utterance as a direct quote, it is specified that the deictic and
modal categories are shifted from the center in the present speech situ-
ation to the center in the reported speech situation.
Figure 1 provides a diagrammatic representation of the proposed
framing relation. The two systems of coordinate axes are intended as
schematic representations of the centers for deictic (x and y axes) and
modal categories (z axis) in the speech situation (for the clause of
speech) and reported speech situation (for the quote).
I am not, of course, the first to have invoked the notion of framing
in describing directly represented speech. But whereas previous inves-
tigators apparently take it as merely a useful metaphor, I believe that it
goes deeper than this, and framing may be taken as a label for the gram-
matical relation between the quoting clause of speech and the quoted
utterance. The most salient aspects of the grammatical relation are
firstly that the clause of speech serves to enclose the quoted utterance,
and secondly, it specifies how the latter is to be construed interactively.
This is a rather loose relation, that is not restricted to single clauses, and
may extend to sentences and paragraphs. At the opposite extreme, it
may involve non-words such as boo-hoo, brr, tsk-tsk. These character-
istics are not observed in constructions involving embedding or depen-
dency relations. Other features that appear more consistent with the
framing analysis than the alternatives include:
• Repetition and omission ofclauses of speech is permissible, indeed
not uncommon, in many languages.
• Usually a special marked prosody can indicate the status ofa clause
as a quote: prosody is apparently in paradigmatic alternation with a
Figure 1 Framing analysis of example (lg)

clause of speech. The marked prosody often defines the extent of
the quoted material, so that it can also be regarded as establishing
an interpretative frame.
• In many languages, including English, the clause of speech can be
interpolated within the boundaries of the reported utterance.
• In some languages reduced forms ofclauses of speech occur in rep-
resented speech. This happens in English, Warrwa, and Miriwoong,
where the clause of speech can be reduced to a single NP repre-
senting the speaker (marked by the ergative in Warrwa, by the abla-
tive in Miriwoong).
English distinguishes between direct and indirect speech represen-
tation. Whereas in direct representation the deictic centre is that of the
reported speech situation, in indirect representation it is the deictic cen-
tre of the reporting speech situation. Moreover, modal categories are no
longer independent; mood is represented primarily from the point of
view of the present speech situation. Indirect quotations are, like direct
quotes, demonstrations; the two types differ in terms of the perspective
from which the demonstration is enacted: primarily the here-now of the
present speech situation, rather than the here-now of the referent speech
situation. Figure 2 shows the proposed grammatical structure of an indi-
rect speech representation. The single system of axes indicates that both
clauses adopt the present speech situation as deictic centre and modal
reference point. The grey color of the inner box represents the more per-
meable boundary between the clauses.
In indirect speech representation in English the deictic centre of the
reported speech situation may also be employed. A temporal adverbial
such as now in an indirectly represented clause can be evaluated from
the deictic centre of the reported utterance, as in Then they told me that
now I should go.
Interestingly, differences exist among languages in terms of their
treatment of deictic and modal categories in indirect speech representa-
tion (Li 1986, Roeck 1994). In Modern Georgian, for instance, person
categories are evaluated according to the present speech situation, tense
is typically according to the reported speech situation, but may alterna-
tively be according to the present speech situation (Boeder 2002:10).
Evidential categories in Modern Georgian also admit these two alterna-
tive centers (Boeder 2002: 11).
The framing analysis provides an intuitively plausible explanation
for the two major contrasting possibilities in speech representation,
direct and indirect. To situate something in a framed picture one could

They said that I should shave
rx
Figure 2 Framing analysis of an indirect quotation version of example (lg)
choose to specify it in terms of the axes defined by the frame (the cor-
relate of direct speech) or in terms ofthe viewer (indirect speech). It also
explains why other possibilities exist in some languages, intermediate
between these two, involving mixed perspectives. The category of free
indirect speech, as in Was he certain, Henry asked, involves representa-
tion of the mood of the quote from the perspective of the reported
speech situation, and deictic categories of tense and person from the
actual speech situation. Modals such as probably, really, surprisingly,
frankly, and the like also represent the reported speaker's modal evalu-
ation, rather than that of the actual speaker (Vandelanotte 2004a:495).4
Temporal adverbials also employ the deictic centre of the referent
speech situation: the temporal deictic tomorrow in She would go tomor-
row, Mary said situates the event of the quoted clause with respect to
Mary's saying, not the speaker's.
The structure of a free indirect speech quote is illustrated in Figure
3. Here I have attempted to represent by grey scale the intermediate sta-
tus of free indirect speech: the integration of the two clauses lies
between the integration for direct speech (least) and indirect speech
(maximal).
Vandelanotte has argued (Vandelanotte 2000, 2004a, 2004b) that in
addition to free indirect speech representation, another grammatically
distinct type of speech representation is identifiable in English, which
he dubs distancing indirect speech. Examples this category are (3a)
and (3b).
(3a) John will be late, he said. (Reinhart 1975:136)
(3b)He mailed you earlier today, he said, so please do answer him.
(Genuine e-mail example, cited in Vandelanotte 2004a:496)
Vandelanotte refers to this type ofdistancing indirect speech as sco-
pa1.5 This is, in my opinion, correct. The clause of speech seems to be
added as a type of disclaimer to the initial clause, a type of parenthetical

distancing the speaker from the proposition it expresses (Reinhart 1975,
Ifantidou 2001, Vandelanotte 2004a, 2004b). The clause of speech, in
examples such as (3a) and (3b), appears to serve an evidential function,
modifying the proposition expressed by the first clause in regard to its
evidential basis. In other cases, a different type ofmodal modification is
invoked, as in parenthetical use of I suggest/claim/propose/submit in
academic discourse. These serve an illocutionary function, specifying
how the modalized clause is to be taken; simultaneously the speaker dis-
tances themself from the proposition by admitting that the evidence for
it is not compelling. Correspondingly, the quoted clause serves as a
description of a situation rather than as a demonstration or enactment of
an(other) utterance, in the terminology of Clark and Gerrig (1990).
Certain peculiarities attest to the grammatical distinctiveness of
(3a) and (3b), and to the notion that the interclausal relation is scope
rather than framing-see Figure 4. First, the clause of speech is obliga-
tory: without it, no modification ofany type ofthe first clause is implied.
It cannot be inferred, in the absence of the clause of speech (Vandelan-
otte 2004a:505), as is typical for absent framing clauses. Second, the
clause of speech has characteristic properties of a parenthetical inser-
tion: it occurs either within the modalized clause, or following it, and on
a separate, reduced intonation contour. Third, the clause of speech is in
paradigmatic contrast with other scoping parenthetical insertions such
as tags (McGregor 1997:244-249). These cannot eo-occur as paren-
theticals modalizing the same clause. In He mailed you earlier today,
didn't he, he said, the final clause of speech, he said, must, it seems to
me, be interpreted as a frame for the preceding tagged clause. Lastly,
only clauses expressing propositions can occur as quotes in this type of
distancing indirect speech construction: imperatives and exclamatives
are precluded, as are minor clause types such as interjections.
Was he certain
If H•my~k•d
-?L/ X
y
Figure 3 Framing analysis of a free indirect speech quote

36 WORD, VOLUME 59,NUMBER 1 (APRIL, 2008)
To wind up this somewhat cursory treatment of quotatives, it is
remarked that there are at least two other circumstances in which a com-
municative complement serves a modalizing function. One is perfor-
matives such as (4a) and (4b), where the clause of speech indicates
explicitly the illocutionary value of the quoted clause. The other, exem-
plified by (4c), expresses a type of intensification, that the speaker holds
a strong attitude to the matter of the quoted clause. These are what I
have elsewhere referred to as illocutionary and attitudinal modaliza-
tions (McGregor 1997:66).
(4a) I order you to keep quiet
(4b) I promise that I'll be there
(4c) I'll say he's a nuisance
As in the case of distancing indirect speech, the clause of speech appar-
ently has scope over, rather than frames the quote. Moreover, these con-
structions show grammatical peculiarities that distinguish them from
corresponding framing constructions such as I ordered him to keep
quiet, I promised that I would be there, and I said that he's a nuisance.
Limitations of space preclude discussion of these types of quotative
construction.
John will be late he said
z
y
Figure 4 Scopal analysis of a distancing indirect speech quote
3.2. Cognitive complements. Just as utterance complements involve
the mimetic representation of potentially spoken utterances, cognitive
complements involve mimetic representation of potentially cerebrated
thoughts, phrased linguistically. Examples of the seven subtypes listed
in Table 1 are given in (5a)-(5g).

(5a) John realized that Mary had been talking to him
(5b) I believe you have eyes in the back of your head
(5c) Watson imagined that he was Holmes
(5d) The management regrets that the performance
is cancelled
(5e) Watson feared that Holmes was dying
(5f) Everyone wants their child to be successful
(5g) Holmes saw Watson throw down his paper
Knowledge
Propositional
Pretense
Commentative
Fearing
Desiderative
Sensory
perception
The same range of reporting types are available for representing
thoughts as representing speech: direct, indirect, free indirect, and dis-
tancing indirect. Not all are available or equally likely for each of the
subtypes. For instance, sensory perception complements are not nor-
mally direct quotes-usually if a verb ofperception occurs with a direct
quote it is a knowledge complement. Thus Holmes saw "Watson has
thrown down his paper" indicates Holmes' acquisition of the belief that
Watson threw down his paper (e.g. inferred from seeing it on the floor
when it was previously in Watson's hands) rather than perception of the
actual event. Free indirect and distancing indirect subtypes are illus-
trated in the following two examples, respectively:
(6a) Did she mean before he met me, I wondered (Corris 2003:5)6
(6b) It might snow, I think
The four representational subtypes can be analyzed in exactly the
same way as the corresponding subtypes of communication comple-
ments. Thus communication and cognitive complements could well
have been treated together in the previous section. I treat them sepa-
rately not because I intend to provide a detailed discussion of commu-
nication complements, much less motivations for the framing analysis
(which I take to be fairly evident), but rather because two general
remarks are worth drawing out explicitly.
The first observation is that, as has been widely noted, indirect rep-
resentation ofthoughts predominates over direct representation not only
in English but also in many other languages. Assuming that the contrast
between direct and indirect speech is that the former represents the
wording of the reported utterance while the latter represents its mean-
ing, this has sometimes been put down to the fact that knowledge of
another's thoughts can never be certain (e.g. Halliday 1985:230). There
are difficulties with this explanation. For one thing, knowledge of

another's exact words and form of expression can never be certain
either, at least in the absence of modem technologies of recording.
Memory of the lexical items and grammatical structures used in an
utterance fades more rapidly than memory of meaning, though this does
not hinder speakers from habitually quoting directly.
This explanation also wrongly predicts that cognitive complements
with first person thinkers that are set in present time should be directly
quoted. Surely our most certain knowledge is of what we are thinking at
the moment: so why are not I want "I will go now", and I believe "he is
mistaken" the norm, rather than I want to go now, and I believe him to
be mistaken?
Except for the final two types in the list above, i.e. desiderative and
sensory perception complements (which we will return to later), the
most natural and frequent mode of expression in English usually
involves the ordinary indirect quotation construction (Figure 2). My
explanation concerns the pragmatics of use, and follows from Levin-
san's Q-principle (Levinson 2000) or Grice's First Maxim of Quantity.
One frames a clause as a thought-rather than simply states it directly
as a proposition-when, for one reason or another, the speaker wishes
to distance themself from the proposition expressed. Often this is
because of some sort of conflict or discontinuity with the knowledge
system of the actual speech situation. The best way of representing the
thought is therefore indirectly, from the perspective of the actual speech
situation. Direct representation presents things from the perspective of
the referent situation of cognition, and the contrast with the present sys-
tem of knowledge is not brought out so forcefully.
My second observation is related to the last point, and concerns
cognitive complements such as (7a) and (7b) that involve I (don't) think,
I (don't) believe, and the like. Here the clause of cognition typically
serves as an expression of the speaker's evaluation of the probability of
the proposition expressed by the complement, that it is probable (I
think/believe) or improbable (I don't think/believe).
(7a) I don't think you locked the door
(7b) I believe we will eventually know the truth of the matter
This construction is reminiscent of the constructions involving
verbs of speech discussed at the end of the previous section. The point
is not to mark the complement clause as a demonstration, but rather to
express an evaluation of the proposition it expresses. The clause of cog-
nition thus appears to have the other clause in its scope: it can be

replaced by a modal particle, and a tag will apply to the complement
clause. Other types of modalization are possible with clauses of cogni-
tion: I guess you locked the door and I wonder whether you noticed the
bad smell apparently express illocutionary modification, in these cases,
interrogative. For none of these constructions does it seem necessary, or
even useful, to invoke the notion of grammatical metaphor, as per Hal-
liday (1985:332ff).
3.3. Status complements. Status complements clearly exhibit both the
formal and semantic features of scoping constructions. Little needs to
be said about them here beyond the observations that: (a) they express
interpersonal slants on the complement clause-rhetorical modification
in the terminology of McGregor (1997:224ff)-and (b) the matrix
clause can normally be replaced by either a scoping particle (not,prob-
ably, or necessarily) or auxiliary verb or verb sequence such as should
have, must have, and the like. The three types identified in Table 1 are
illustrated in (8).
(8a) It is not the case that they followed his dripping blood
until nightfall Negative
(8b) It is possible that they followed his dripping blood
until nightfall Epistemic
(8c) It was necessary for them to follow his dripping blood
until nightfall. Deontic
3.4. Aktionsart complements. Aktionsart complements would seem to
be the least amenable to interpersonal analysis. What have causation,
achievement, phase, or conjunction got to do with interpersonal semio-
sis? The short answer is nothing. But this is not the issue: our question
is not about the semantics of causation (or any of the other relations) as
such, but rather the semantics of the relation between the clauses. Is
there any evidence that it expresses interpersonal meaning? I would
argue that there is. Due to limitations of space, I will confine the dis-
cussion to the causative subtype, mentioning the achievement and phase
types briefly at the end of the section. The conjunctive type is left out
entirely, since it is not represented in English.
Causative manipulative complements, as understood here, include
causatives with matrix verbs such as force, get, induce, and have and
permissives with matrix verbs like let, permit, and allow. Examples are
given in (9a)-(9f).

(9a) The police forced the trackers to follow his dripping blood
until nightfall
(9b) The police got the trackers to follow his dripping blood
until nightfall
(9c) The police induced the trackers to follow his dripping blood
until nightfall
(9d) The police had the trackers follow his dripping blood
until nightfall
(9e) The police let the trackers follow his dripping blood
until nightfall
(9f) The police allowed the trackers to follow his dripping blood
until nightfall
What I want to propose is that causative complements involving
these matrix verbs are amenable to an interpersonal analysis whereby
the complement clause represents an event that falls within the scope of
the matrix clause. The matrix clause provides interpersonal modifica-
tion of the rhetorical type on the complement clause: it integrates the
complement clause into the framework of knowledge and expectations
of the discourse. These causative complements express more than mere
causation, such as is expressed in causative-transitive clauses like I
closed the door and the guards marched the prisoners. Causative com-
plements inherently convey information about expectations concerning
the occurrence of the complement event.
The basis for this suggestion is brought out especially clearly in the
semantic explications suggested by Anna Wierzbicka (Wierzbicka
1988) for to complements ofjorce, get, and induce in English, as shown
in (10)-(12):
(10) X forced Y to doZ.~
Y thought this: I don't want this: I will do Z
X did something to Y
one could think at that time: Y will do Z because of that
Y thought this because of that: I cannot not do Z
Y did Z because of that (Wierzbicka 1988:43; my emphasis)
(ll)X gotYto doZ.~
X thought this: I want this: Y will do Z
X did something to Y because of that
because of this, Y thought this: I want this: I will do Z
one could think this: Y will do Z because of that

(12)X induced Y to doZ.~
X thought this: I want this: Y will do Z
X knew this: Y will think this: I don't want this
X did something to Y because of that
because of this, Y thought this: I want this: I will do Z
one could think this: Y will do Z because of that
It is not my intention here to argue either for (or against) the style
of semantic representation adopted by Wierzbicka, or for the validity of
the explications she proposes for the three complement types. I merely
remark that it seems to me that her analysis is insightful, and on the right
track. What is particularly notable for our purposes is the observation
that each of the three complement types involves numerous instances of
the event primitives think and want, as well as one instance of know.
These have as thinkers/wanters participants in the sentences (the object
offorce, the subject of get and induce), as well as a generic one. Very
clearly these causative complement constructions are deeply rooted in
the human or personal domain in ways that the causative-transitive con-
structions involving verbs like open or close are not.
More significantly, in each explication there are three instances of
a modalized version of the to-complement clause 'Y [to] do Z'. These
represent in each case something that was thought or desired by X, Y, or
a non-specified generic one. Thus, if Wierzbicka is correct, the comple-
ment clause represents the content of at least one thought-event, as well
as at least one desire-event (which, of course, includes desire to the con-
trary). Thus semantically the complement clause may be considered to
lie within the scope or frame of an event of thought and an event of
desire.
What is desired, as well as who desires it, differs across the three
complement types, and thus cannot be associated with causative com-
plements as such (presuming them to represent a single over-arching
construction type). However, what is asserted as thought by the generic
one (the bolded line in the explications) remains virtually constant across
the three explications. It is reasonable to interpret this thought as the
specification of an expectation that the complement event occurred, or
would occur, only by virtue ofthe unspecified action ofX on Y, that oth-
erwise it would not have occurred. The expectation is of course some-
thing that someone would think will occur, given the relevant conditions.
These conditions are invoked by the specification because ofthat in the
bolded components of explications (10)-(12). As phrased in (10) it
would appear to be an expectation of any individual situated in the ref-

42 WORD, VOLUME 59,NUMBER 1 (APRIL, 2008)
erent world, whereas in (11) and (12) it appears to be an expectation of
anyone in the actual speech situation.lt is not clear thatforce differs from
get and induce in this way, and Wierzbicka (1988) does not adduce argu-
ments that it does. Regardless, the main point is that the bolded clauses
express (in the style of natural language metalanguage) what can rea-
sonably be regarded as expectations about the occurrence of the com-
plement event that are not specifically attributed to anyone in particu-
lar-they are thoughts or expectations that anyone might entertain.
If the story adumbrated in the previous paragraph is correct, it is
apparent that each of the three causative complements includes as a part
of its inherent meaning an evaluative component that links the comple-
ment event to the framework of expectations and knowledge engen-
dered in and by the discourse. The expectation, it should be noted, is not
associated with the discourse environment as such: it is not suggested
that these causative complements are used only in discourse environ-
ments in which expectations of non-occurrence of the complement
event are invoked. Rather, the expectations are created by the causative
construction itself, much in the way they are by scoping adverbs such as
unexpectedly. This interpersonal component of meaning is, I submit, the
semantic invariant associated with the interclausal relation.
Not to make too fine a point of it, examples (9a)-(9f) above are in
paradigmatic opposition with (13).
(13) The trackers followed his dripping blood until nightfall
What the matrix clause in the causative complement constructions does
is modalize the complement event by specifying that otherwise it would
not be expected to have happened. The responsibility for its occurrence,
that is to say, lies squarely in the matrix event-in the unspecified action
X performed on Y. Y is evaluated as not ultimately responsible for per-
forming Z, despite actually doing it.
Two formal grammatical features of these causative constructions
should be singled out as consistent with the proposed analysis. I refer
first to the fact that the complement is in non-finite form, and second to
the use of to as complementizer.
The non-finite form ofthe complement clause indicates that it is the
event itself, 'Y do Z', that is modalized as unexpected, not the corre-
sponding proposition.7
This is consistent with the bolded components of
Wierzbicka's explications, which concern (thoughts about) the perfor-
mance of events, not the truth value of propositions. This is clear from

the because clause, which can only be interpreted as specifying the rea-
son for the occurrence of the event, not reasons for entertaining the
thought.
As to use of the complementizer to, again Wierzbicka (1988) is
enlightening. She comments (Wierzbicka 1988:165) on the future ori-
entation of most to complements in English, and indeed suggests that
this is perhaps a component of the semantic invariant of these comple-
ment constructions. Again, the future orientation is prominent in the
bolded components of explications (10)-(12): the complement event is
not only specified as a thought, but modalized by will.
At this point it is well to return to Noonan's manipulative comple-
ments, which include both causatives and permissives, and illocution-
ary acts (Noonan 1985:125-127). It seems to me that the same sort of
expectation modification is invoked in illocutionary complements as in
causative complements. Thus in (14a)-(14d) it seems reasonable to pre-
sume that an expectation is invoked that otherwise-were it not for the
illocutionary act ofthe matrix clause-the complement event would not
occur. In (14b), for instance, the presumption is invoked that mathe-
matics department would not offer the course to liberal arts students
were it not for the request, presumably from the administration.
(14a) The police commanded the trackers to follow his dripping
blood until nightfall.
(14b) The math department has been requested to offer an "Apprecia-
tion of Math" course for liberal arts students beginning next
semester.8
(14c) Your wife begged me with tears in her eyes to go easy on you,
and I don't mind admitting I was deeply moved.
(14d) All the while they must keep an eye on the Soviet Union to the
east, insuring stability at home to ward off a possible Russian
intervention, while cajoling Western banks and governments to
aid their fledgling democracy.
Of course, there is an important difference between these illocu-
tionary complements and causatives and permissives: the latter are fac-
tive, whereas the former are unspecified for factivity. I suspect that
Wierzbicka is correct that factivity is a part of the inherent meaning of
causatives and permissives, as per the final component of explications
(10)-(12). If factivity was a pragmatic inference, it ought to be defeasi-
ble, which seems not to be the case. Rather, it appears to be invariably

associated with the construction, and thus represents a portion of the
inherent semantics. This final component would merely be missing
from the explication of to illocutionary complements.
This difference can be captured in the framework I am proposing
under the presumption that it is not directly related to the interclausal
relationship. Instead, it can be accounted for under the notion of vari-
able scope of the tense of the matrix clause: for causatives and permis-
sives tense has scope over the non-finite clause, whereas for illocution-
ary complements it does not. My guess is that this is a consequence of
a more fundamental difference: the interclausal relation is scopal for
causatives and permissives, but framing for illocutionary complements.
It must be stipulated that my proposals concern English causative
and permissive constructions involving to-complements. They are not
presumed to apply to other putatively similar constructions, such as
make causatives, which do not take complements marked by to. As
Wierzbicka (1988:43-44) observes, "unlike force, make refers neither
to the causee's thoughts or intentions, nor to anybody else's expecta-
tions regarding the causee's future actions" (my emphasis).9 Nor are
they presumed to apply toforce causatives with into as complementizer,
as in The police forced the trackers into following his dripping blood
until nightfall. For these also, according to Wierzbicka (1988:43), the
expectation component is also absent.
Might not the complement clauses in these constructions be candi-
dates for genuine objects? Perhaps, though I do not think there is com-
pelling evidence in favor ofthe possibility. In addition, there are at least
two other equally plausible alternatives. First, the complement clause
could be a secondary predicate on the object NP. This would be consis-
tent with (15c) as an agnate for (15a) and (15b).
(15a) The police made the trackers follow his dripping blood until
nightfall.
(15b) The police forced the trackers into following his dripping blood
until nightfall.
(15c) When the police did something to the trackers, they followed his
dripping blood until nightfall.
In favor of this analysis, one might compare (16a) and (16b) with the
more prototypical secondary predicates:
(16a) The police made the trackers tired.
(16b) The police forced the trackers into the truck.10

A second possible analysis, at least for make, is that the construc-
tion is monoclausal and make is a causative marker.
Regardless of the analysis of these two exceptional complement
types, I have, I hope, established in this section that causative and per-
missive to-complements are amenable to an analysis in which the
matrix clause has scope over the complement, expounding expectation
modification on the event denoted by the non-finite clause.
That similar cases could be mounted for achievement and phase
complements emerges from the explications Wierzbicka proposes for
the complements of some key verbs in these domains. In the following
two explications the crucial components indicating that expectation
modification is inherent to the construction are again shown in bold:
(17) He failed to hit the target. ~
he thought this: I want this: I will hit the target
he did something because of that
one could think this at the time:
it will happen because of that
it didn't happen (Wierzbicka 1988:35)
(18)At timet, the rain ceased.~
before t, it was raining
one could think then: more of it will happen after now
at t, one could think this:
more of it will not happen after now
after t, more of it didn't happen (Wierzbicka 1988:81)
4. Some further theoretical issues. In the previous section it was
argued that much of the domain traditionally included under the rubric
"object complementation" falls under interpersonal grammar. Many
complement clauses are linked to their matrix clause by conjugational
relations of framing or scope, grammatical relations that express inter-
personal semantics. I have attempted to give some notion of the range
of complex sentence constructions in English that fall into the domain,
and to identify some that are excluded. Determination of the bound-
aries, however, is a more distant goal.
There are of course many types of complementation I have not
touched on in this paper, and thus have not situated in relation to inter-
personal grammar. Subject complements (examples (19a) and (19b))
have been ignored entirely, and fact-type object complements (as in
examples (19c) and (19d)) are mentioned only in passing.

(19a) That they followed his dripping blood until nightfall cannot be
doubted.
(19b) For them to get divorced would be unexpected.
(19c) Tom regrets that they're getting divorced.
(19d) Tom resents that he has to wait in line.
As to subject complements, I content myself with two remarks.
First, a constituency analysis is viable only if the complement clause
serves in an experiential role in the matrix clause. This is not the case
for subject complements generally, the matrix clauses typically being
relational in structure rather than situational-as is clearly the case for
the matrix clauses in (19a) and (19b).11
Second, the complement clause
does indeed appear to serve in the grammatical role subject in the matrix
clause, as indicated by the form of tags. If Halliday is correct in his con-
strual of this role as an interpersonal one (Halliday 1985:76; see also
McGregor 1997:282), subject complements would be situated firmly in
interpersonal grammar, their relation to the matrix clause being an inter-
personal one-subject.12
Fact complements are more difficult to situate semiotically.
Davidse (1994) adduces compelling arguments that they show different
grammatical structures from cognitive complements of the proposi-
tional type involving verbs like think. She argues further that the inter-
clausal grammatical relation is embedding: the complement is embed-
ded in the matrix clause, in which it serves an experiential role-for
instance, the complement would be an Undergoer in (19c) and (19d).
Thus, fact complements would represent genuine complement clauses,
to be understood in terms of experiential grammar. What this analysis
fails to account for is the fact that, as Halliday (1985:243) puts it, facts
come "as it were ready packaged in projected [i.e. represented] form".
One way of capturing this observation, as well as the essence of
Davidse's proposed analysis, is to assume that fact complements are
framed by either nominals such as fact, or NPs such as the fact, which
may, like framing clauses of speech, be omitted. What is embedded as
Undergoer in the matrix clause then is not the complement clause, but a
complex NP involving a framed complement clause.
Aside from the problem of the delimitation of the boundaries of
interpersonal complementation, there are a number of formal-grammat-
ical issues in complementation, prominent in the complementation lit-
erature, that I have not dealt with systematically in this paper. These
include: (a) "raising"/"control" and "equi"; (b) "long distance depen-
dencies"; and (c) the form of the complement and complementizer. Any

grammar purporting adequacy must account somehow for formal facts
such as these. In the remainder of this section I sketch out some sug-
gestions for their treatment.
Beginning with (c),Anna Wierzbicka has adduced coherent seman-
tic explications of the various formal types of complement clause in
English (Wierzbicka 1988:162-168). She demonstrates convincingly
that each is semantically significant, and has an associated invariant of
meaning. Complementation is not a rag-bag of meaningless grammati-
cal forms, a domain that is amenable to exclusively formal analysis, as
has sometime been presumed.
(a) and (b), by contrast, appear to represent purely formal facts of
English grammar, that have no semiotic value. They must presumably
be accounted for by grammatical rules specifying admissible structures.
Equi-which is represented by examples such as (9a)-(9f)-can be
accounted for under a grammatical rule of obligatory ellipsis of the sub-
ject of the complement clause, as proposed by Dik (1997:148).
Both "raising"/"control" (examples (20a)-(20c)) and "long dis-
tance dependencies" (examples (21a)-(21d)) cause analytical difficul-
ties because they exhibit linguistic units where they ostensibly should
not be.
(20a) Howard believes the trackers to have followed his dripping blood
until nightfall.
(20b) Last night I heard my neighbor belting his children.
(20c) We smelt the farmer spreading pig shit.
(21a) For how long does Howard believe the trackers followed his
dripping blood?
(21b) Do you think they followed his dripping blood until nightfall?
(21c) I don't think they followed his dripping blood until nightfall.
(21d) Who do you reckon followed his dripping blood until nightfall?
Thus, in the examples under (20) the apparent objects (or Under-
goers) of the matrix verbs are not genuine objects!Undergoers: (20a)
does not imply that Howard believes the trackers; (20b) does not imply
that I heard my neighbor (e.g. I might have heard only children crying);
and (20c) does not imply that we smelt the farmer (more likely we smelt
what he was spreading). This reasoning, however, is spurious, as
observed by Langacker (1999:321ff). That the post-verbal NPs in these
examples do not correspond with Undergoers in the corresponding
monoclausal constructions has no bearing on their grammatical roles in

the complement constructions. And indeed the grammatical evidence
clearly points to the fact that these NPs are Undergoers in the matrix
clauses. In the complement construction, the grammatical relation of
Undergoer acquires a somewhat different etic value to what it possesses
in the corresponding monoclausal construction. This being the case,
there is no reason why the same rule of obligatory NP ellipsis that
replaces "equi" cannot also replace "raising".
Examples (21a)-(21d) are considered problematic for the same
sorts of dubious reasons as the (20) examples. They are generally
accounted for under cycles of rules that move the WH and negative
words from their "real" position within the complement clause to a posi-
tion in the matrix clause. But there are other ways of accounting for the
facts (McGregor 1997:261-263), that require no movement rules. The
basic idea is represented in Figure 5, which shows the relevant aspects
of examples (21b) and (21c). What happens is that in circumstances
such as these-when the clause of cognition expresses a rhetorical
stance with respect to the complement clause-the interclausal relation
is scopal, and the modal values of the complement clause are repre-
sented, as expected, from the vantage point of the present speech situa-
tion. The complement clause lies within the scope of two rhetorical
modifiers, one of which is a (matrix) clause. All that is required is a
grammatical rule that fuses the outer scoping elements into a single
clause.
theyfollow , his dripping blood until nightfall
._INTER------------------------~
theyfollow:/., his dripping blood until nightfall
--NEG--------------------------__.
Figure 5 Proposed structure of examples (20b) and (20c)

5. Conclusions. Complementation, as it is generally construed and as
delimited by Noonan (1985), is not a unified grammatical domain. Sev-
eral fundamentally distinct interclausal grammatical relations are
involved, representing semiotic construals of fundamentally different
types. Nevertheless, a significant subset of complement constructions
can, I have suggested, be unified by virtue of the fact that the inter-
clausal relation is interpersonal-conjugational, either framing or scope.
This conjugational relation specifies how the complement clause is to
be "taken" interactively. Experiential and logical semiosis appear to
play a relatively minor role in the domain of complementation, and
probably account for just a handful of complement types.
That interpersonal meaning is prominent in at least some comple-
mentation constructions is not a new idea. Many investigators have
observed that clauses of thought often serve an evidential or modal
function (e.g. Halliday 1985; Diessel and Tomasello 2001). What is
novel about the approach adopted in this paper is that this aspect of the
meaning is considered to be central to the grammar of a non-trivial sub-
set of object complement constructions; it is not a metaphoric extension
from a more fundamental experiential meaning, namely of reporting of
speech, thoughts, desires, and so on. The latter meaning represents, in
my view, the extended meaning. It is a defeasible pragmatic inference,
available in a restricted subset of circumstances; it is not encoded by the
interclausal grammatical relation. This is consistent with the story of
acquisition reported in Diessel and Tomasello (2001), in which the
reporting interpretation emerges late, subsequent to the evidential and
performative senses.13
To wind up the paper I wish to make two general observations.
First, in my analysis of English causative and permissive complements
I have relied heavily on Anna Wierzbicka's careful and insightful
semantic analysis of these types of complements Wierzbicka (1988).
My major point of departure from her treatment concerns the locus of
the meaning components. Wierzbicka associates detailed and precise
semantic specifications with constructions defined narrowly via precise
specification of the lexical verb in the matrix clause, and choices of
grammatical categories and forms (as in the first lines of explications
(10)-(12)). The problem is that each such construction becomes a sign
in itself, whose relations to other construction-signs is specifiable only
on a metalevel, through comparison of the semantic explications. There
is no recognition of semiotic complexity in the constructions, that they
are not radically unanalyzable signs in themselves.
In my approach (the parameters of which are laid out in McGregor
1997), complements are at least partly analyzable grammatical entities.

They are complexes of signs of various types, including at least gram-
matical relations, constructions, morphemes, categories, and lexemes.
The problem for the grammarian then is not just to identify meanings
inherently associated with the narrowly specified constructions, but to
determine the loci of the various components of meaning. Thus I have
suggested that certain components in Wierzbicka's representations
belong not to the causative construction per se, but to the grammatical
relationship between the clauses; others belong to the grammatical
shape of the complement clause; still others (I hazard to guess) belong
to lexical choices (such as tell vs.force).
Second, I have dealt exclusively with the grammatical grammar of
English complement constructions. This naturally raises the question of
the wider applicability of the proposals. To what extent do complemen-
tation constructions in other languages employ interclausal relations of
scope and framing? And to what extent is complementation the concern
of interpersonal semiosis? These are questions that obviously cannot be
addressed satisfactorily in the conclusion of a paper. Nevertheless, it
does seem not unreasonable to presume that the domains of communi-
cation and cognition representation, and status modification are domains
that belong squarely in interpersonal grammar, regardless of whether
they are represented by complement constructions in a language-the
other obvious alternative being that they are given mono-clausal repre-
sentation. The Aktionsart domain is less prototypically interpersonal,
and one expects likely to be represented by interpersonal grammar only
in some languages, and then only in certain constructions. Thus, for
instance, in the Australian languages Gooniyandi and Ngarinyin causa-
tion has a variety of grammatical realizations, including a complement
construction in which the caused event is represented as a quote (McGre-
gor 2000, Rumsey 1982:160ff).
Department ofAesthetics and Communication-Linguistics,
Aarhus University,
lens Chr. Skous Vej 4, Office 1483-547,
DK-8000 Arhus C,
Denmark
linwmg@hum.au.dk
ENDNOTES
1This is a revised version of a presentation given to the conference Exploring Interpersonal
Grammar, held at the Catholic University of Leuven, November 2000. I am grateful to the audi-
ence for feedback on the presentation, to an anonymous referee for useful comments on an earlier
draft, and especially to Kristin Davidse for her forbearance and patience with a tardy contributor.

2In just a few grammatical theories-including standard Systemic Functional Grammar (Hal-
liday 1985:193ff) and Word Grammar (Hudson and Van Langendonck 1991)-the relation is con-
sidered to be a dependency one, with the complement clause (or its verb) dependent on the main
clause (or its verb). There is no scope in this article to enter into detailed discussion of the depen-
dency accounts, or to compare them with one another, or with constituency accounts.
3In fact, few treatments of complementation seriously address the issue of the inter-clausal (or
inter-unit) syntagmatic relation involved, and merely presume it to be a constituency relation.
McGregor (1997:28- 31, 55) and McGregor (2003) argue that a distinction must be drawn between
the relations 'belong to' (which is a purely formal relation, roughly comparable with alienable pos-
session) and 'be a constituent of' (which is a semiotic relation, roughly inalienable possession-
'be a functioning part of').
4In indirect quotation, by contrast, modals in the quoted utterance typically represent the
speaker's intrusion on the proposition expressed by the quoted clause. This is illustrated by the fol-
lowing, from Barnard (2000: 18), The degree to which they believed that early humans really did
devise an actual social contract is difficult to access [his emphasis]. The modal really here clearly
relates to the knowledge-base constructed in the book. Although this is indirectly quoted thought,
it would work as well with They explicitly state that the early humans really did devise an actual
social contract.
5Vandelanotte distinguishes another type of distancing indirect speech which does appear to
involve framing (Vandelanotte 2004a, 2004b). To go into the complexities of this construction
would take us too far from our present concerns.
6It is clear from the context of this example that, despite the absence of quotation marks
around "before he met me", that the complement of the first clause with mean is a direct quote. It
picks up the previous directly quoted utterance of the woman, restating it in a completely different
way.
7This observation also accounts for the normal non-finite representation of desiderative com-
plements in English: it is the occurrence of events, rather than truth or falsity of propositions that
is represented as desirable, consistent with the emotions human beings typically feel.
8This and the following two examples come from the Leuven corpus.
9Unaccountably, Wierzbicka goes on to provide an explication of 'X made Y do Z' involving
the component Ydidn't want it. This does not however vitiate the story as I am constructing it, since
what is crucial to the interpersonal semantics of the interclausal relation is the generic evaluative
component, which remains absent.
10Notice that the into phrase in this example does not specify a resulting location for the situ-
ation, but rather exclusively for the Undergoer. (See McGregor 1997:155- 175 for further discus-
sion of this point.)
11True, the matrix clause can represent a situation, a passive clause such as That he had been
away was definitely said by him. This relates to the following point; these do not represent coun-
terexamples to the observation.
12There are many gaps to be filled in in this tentatively suggested analysis. Assuming the irrel-
evance of experiential-constituency grammar to the interclausal relation, it still remains possible
that a logical-dependency relation is involved as well as the interpersonal one. Second, the con-
strual of subject as an interpersonal relation can-in the absence of convincing arguments-only
be regarded as an intriguing possibility, rather than an established fact. Whether and how it might
be incorporated into the interpersonal semiotic as construed by semiotic grammar remains to be
determined.
13A further advantage of my approach is that it involves no conflict between acquisition and
historical processes of grammaticalization from a fully representational complementation con-
struction to a parenthetical construction, as emerges in the approach taken by Diessel and

Tomasello (2001). Specifically, in my story no change in interclausal relations needs to be invoked;
that the parenthetical construction (!merged at a historically later time is expected as a normal
process of formal reduction accompanying increased frequency of use (Bybee 2003).
REFERENCES
Barnard, Alan. 2000. History and theory in anthropology. Cambridge: Cambridge University
Press.
Boeder, Winfried. 2002. "Speech and thought representation in the Kartvelian (South Caucasian)
languages." Reported discourse: a meeting groundfor different linguistic domains. Eds. Tom
Giildemann and Manfred von Roncador. Amsterdam/Philadelphia: John Benjarnins. Pp. 3-48.
Bolkestein, A. Machtelt. 1990. "Sentential complements in functional grammar: embedded predi-
cations, propositions, utterances in Latin." Layers and levels of representation in language
theory: functional view. Eds. Jan Nuyts, Machtelt A. Bolkestein, and Co Vet.
Amsterdam/Philadelphia: John Benjarnins. Pp. 71-100.
Bresnan, Joan W. 1979. Theory ofcomplementation in English syntax. New York and London: Gar-
land.
Bybee, Joan. 2003. "Cognitive processes in grammaticalization." The new psychology of lan-
guage: cognitive and functional approaches to language structure. Volume 2. Ed. Michael
Tomasello. Mabwab NJ and London: Lawrence Erlbaum. Pp. 145-167.
Clark, Herbert H. and Gerrig, R. 1990. "Quotations as demonstrations." Language 66: 764-805.
Corris, Peter. 2003. Master's mates. Sydney: Alien and Unwin.
Davidse, Kristin. 1994. "Fact projection." Perspectives on English. Studies in Honour ofProfes-
sor Emma Vorlat. Eds. Keith Carlon, Kristin Davidse, and Brygida Rudzka-Ostyn. Leuven:
Peeters. Pp. 257-284.
Diessel, Holger, and Michael Tomasello. 200I. "The acquisition of finite complement clauses in
English: a corpus-based analysis." Cognitive Linguistics 12: 97-141.
Dik, Simon C. 1997. The theory offunctional grammar. Part 2: Complex and derived construc-
tions. Berlin: Mouton de Gruyter.
Giv6n, Talmy. 1980. "The binding hierarchy and the typology of complements." Studies in Lan-
guage 4: 333-377.
Haegeman, Lillian. 1991. Introduction to government and binding theory. Oxford: Basil Black-
well.
Halliday, Michael A.K. 1985. An introduction to functional grammar. London: Edward Arnold.
Hudson, Richard and Van Langendonck, Willy. 1991. "Word grammar." Linguistic theory and
grammatical description. Eds. Filip G. Droste and John E. Joseph. Amsterdam: John Ben-
jamins. Pp. 307-335.
Ifantidou, Elly. 2001. Evidentials and relevance. Amsterdam: John Benjamins.
Langacker, Ronald W. 1999. Grammar and conceptualization. Berlin and New York: Mouton de
Gruyter.
Levinson, Stephen C. 2000. Presumptive meanings: the theory of generalized conversational
implicature. Cambridge, Massachusetts and London: The MIT Press.
Li, Charles. 1986. "Direct and indirect speech: a functional study." Direct and indirect speech. Ed.
Florian Coulmas. Berlin: Mouton de Gruyter. Pp. 29-45.
McGregor, William B. 1994. "The grammar of reported speech and thought in Gooniyandi." Aus-
tralian Journal ofLinguistics 14: 63-92.
--. 1997. Semiotic grammar. Oxford: Clarendon Press.
--. 2000. Causatives in languages of north-west Australia. Unpublished manuscript.

--. 2003. "A fundamental misconception of modem linguistics." Acta Linguistica Hafniensia
35: 39-64.
Noonan, Michael. 1985. "Complementation." Language typology and syntactic description. Vol-
ume 11: Complex constructions. Ed. Timothy Shopen. Cambridge: Cambridge University
Press.Pp.42-140.
Reinhart, Tanya. 1975. "Whose main clause? (Point of view in sentences with parentheticals)."
Harvard studies of syntax and semantics: volume 1. Ed. Susumu Kuno. Cambridge MA:
Department of Linguistics, Harvard University. Pp. 127-171.
Roeck, Marijke de. 1994. "A functional typology of speech reports." Function and expression in
Functional Grammar. Eds. Elisabeth Engberg-Pedersen, L. Falster Jakobsen, and L. Schack
Rasmussen. Berlin: Mouton de Gruyter. Pp. 331-351.
Rosenbaum, P. 1967. The grammar ofEnglish complement constructions. Cambridge, Mass.: MIT
Press.
Rumsey, Alan L. 1982. An intra-sentence grammar of Ungarinjin, North-western Australia. Can-
berra: Pacific Linguistics.
Vandelanotte, Lieven. 2000. Disclaiming indirect speech and thought: a missing link in speech and
thought representation. Paper presented to conference Exploring Interpersonal Grammar,
Catholic University of Leuven.
--. 2004a. "Deixis and grounding in speech and thought representation." Journal ofPragmat-
ics 36: 489-520.
--. 2004b. "From representational to scopal 'distancing indirect speech or thought': a cline of
subjectification. Text 24: 547-585.
Wierzbicka, Anna. 1988. The semantics ofgrammar. Amsterdam: John Benjamins.

Complementation as interpersonal grammar.pdf

More Related Content

Similar to Complementation as interpersonal grammar.pdf (20)

Recently uploaded (20)

Complementation as interpersonal grammar.pdf