CONTEXT AND COGNITION: KNOWLEDGE FRAMES AND SPEECH ACT COMPREHENSION *

CONTEXT AND COGNITION: KNOWLEDGE FRAMES AND SPEECH ACT
COMPREHENSION *
TEUN A. VAN DIJK
This paper is about the cognitive foundations of pragmatic theories. Besides the fact that the
usual appropriateness conditions for speech acts, which are given in cognitive terms, such as S
knows/believes/wants ... (that) p , require empirical investigation, a sound theory of prag-
matics must also explain how certain utterances in certain contexts are actually understood as
certain speech acts. Speech act comprehension is based on rules and strategies for so-called
context analysis , in which (epistemic) frames play an important role in the analysis of social
context, social frames, and interaction type. Results of context analysis are then matched with
those of pragmatic sentence analysis, viz. the illocutionary act indicating devices. Finally, some
results from the cognitive analysis of discourse processing are applied in a brief account of the
comprehension of speech act sequences and macro-speech acts.
1. The foundations of pragmatics
The philosophical and linguistic study of pragmatics requires an analysis of its
foundations. This basis of pragmatic theories is on the one hand conceptual, e.g. in
the analysis of action and interaction, and on the other hand empirical, viz. in the
investigation of the psychological and social properties of language processing in
communicative interaction.
One of the crucial components in an integrated theory of communicative inter-
action by natural language is constituted by a cognitive theory of language use.
Such a cognitive theory will not only provide insight into the processes and
structures involved in the actual production, comprehension, storage, reproduction
* This paper is intended as an introductory survey of some problems in the interdisciplinary
domain of pragmatics and cognitive psychology. Instead of elaborating the details of our ob-
servations, the discussion presented here remains more or less general. Some of the remarks
have briefly been touched upon during a lecture in the Department of Psychology at the Uni-
versity of Colorado at Boulder. I am indebted to Walter Kintsch for some helpful suggestions
concerning this paper and for his permanent advice and collaboration in our work on discourse
processing in general. Finally I am indebted to Sandro Ferrara for long discussions about the
numerous problems involved in the relations between pragmatics, the theory of discourse and
cognitive psychology.

---

T. A. van Dijk / Context and cognition
212
and other kinds of processing of sentences and discourses, but also in the ways
speech acts are planned, executed and understood.
In particular such a cognitive theory of pragmatics will have to elucidate what
the relations are between various cognitive (conceptual) systems, and the conditions
for the appropriateness of speech acts in given contexts.
Besides the systems of beliefs, wants, wishes, preferences, norms and values, that
of conventional knowledge plays an important role. It is so to speak the basic, and
at the same time the social condition for the operation of the other systems in com-
munication. Such systems are studied, e.g. in actual artificial intelligence, under the
label of frames. A cognitive theory of pragmatics has as one of its tasks to specify
how we are able to perform and understand acts of language, and how we are able
to act upon such understanding as it is related to cognitive frames.
2. The cognitive nature of pragmatic conditions
The appropriateness conditions of classical speech act theory are usually of a
cognitive nature, and include such conditions as:
(i) speaker knows that p
(ii) speaker believes that p
(iii) speaker wants p
(iv) speaker finds it good that p etc.
The concepts involved in such conditions are treated as primitives in pragmatics:
they are not further analyzed, neither conceptually
e.g. in terms of epistemic
logic nor empirically. It is clear however that if a theory of pragmatics should
claim to be empirically relevant, these various concepts should also be assessed at
the proper psychological level, i.e. in terms of experimental or simulative (artificial)
findings.
More specifically, it should be investigated which cognitive processes are under-
lying the assignment of appropriateness in communicative contexts. In which
respect are the actually formulated conditions idealizations , i.e. how far are they
from the actual acceptance/comprehension of speech acts?
Although pragmatic conditions have a cognitive basis, it should be borne in mind
that the ultimate rationale for a pragmatic theory of language is to bridge the gap
between utterances (and hence grammar) on the one hand, and interaction (and
hence the social sciences) on the other hand. This may mean, for instance, that
whatever a speaker really knows, thinks or wants during the accomplishment of a
speech act, is irrelevant as long as his behaviour may be interpreted as, and hence
socially counts as, exhibiting these various internal states.

---

T. A. van Dijk / Context and cognition
213
This methodological caveat does not imply however that the cognitive analysis
of pragmatic concepts is irrelevant. On the contrary, actual communication does
involve real knowledge, beliefs and wants. There is even a general condition viz.
that of sincerity which requires a direct link between what is thought, etc. and
what is shown.
More generally it may even be said that the proper social conditions involved in
the formulation of pragmatic rules, such as authority, power, role and politeness
relations, operate on a cognitive basis: i.e. they are relevant only inasfar the speech
participants know these rules, are able to use them, and are able to relate their
interpretations of what is going on in communication with these social properties
of the context. 1
3. Pragmatic comprehension
A pragmatic theory provides rules for pragmatic interpretation. That is, given
certain utterances of natural language, it specifies the rules assigning a particular
speech act or illocutionary force to this utterance, given a particular structure of
the pragmatic context.
Now, this task is more a programme, at the moment, than actually carried out.
We know something about various speech acts and their contextual conditions, but
little about the systematic relationships with the (grammatical or other) structures
of the utterance. These relationships are most conspicuous in the use of explicit
performatives, certain particles, the syntactic form of sentences (indicative, inter-
rogative and imperative), and in the propositional content of sentences. 2
Since we know not much more, we have little to offer for a more general theory
of pragmatic comprehension which is part of a cognitive theory of information
processing. Pragmatic comprehension is the series of processes during which lan-
guage users assign particular conventional acts, i.e. illocutionary forces, to each
other s utterances. The problem thus is: how do hearer actually know that when a
speaker utters such or such a sentence, that the speaker thereby makes a promise or
a threat? What information must be available to the hearer in order to be able to
make such assignments?
Obviously this information may come from various sources and through various
channels :
1 This kind of cognitive relativism does not imply an idealistic approach to language and
interaction. As we will briefly mention below, conventions are involved, where conventions are
taken as strictly objective properties of social structure. Methodological issues related to this
issue will however be left undiscussed in this paper.
2 The basic pragmatic concepts used in this paper have been discussed elsewhere; for a recent
summary and for an analysis of the relations between the semantics and the pragmatics of dis-
course, see van Dijk 1977a, also for further references.

---

T. A. van Dijk / Context and cognition
214
A. properties of the structure of the utterance (as assigned on the basis of gram-
matical rules);
B.
para-linguistic properties, such as speed, stress, intonation, pitch, etc. on the
one hand, and gestures, facial expression, bodily movements, etc. on the
other hand;
C. actual observation/perception of the communicative context (presence and

properties of objects, other persons, etc.);
D.
knowledge/beliefs in memory about the speaker and his properties, or about
other properties of the actual situation;
E. more in particular: knowledge/beliefs with respect to the type of interaction
going on, and the structures of preceding contexts of interaction;
F.
knowledge/beliefs derived from previous speech acts c.q. previous parts of the
discourse, both at the micro (or local) level and on the macro (or global) level;
G. general semantic, in particular conventional, knowledge about (inter)action,
rules, etc. especially those of pragmatics;
H. other kinds of general world-knowledge (frames).
The very fact that all these components may be involved in pragmatic comprehen-
sion justifies the well-known insight that often we are unable to assign a definite
illocutionary force on the basis of a (semantic) comprehension of the utterance
(sentence) alone:
(1) I ll come tomorrow!
may function as a promise, a threat, an announcement/assertion, etc. A specific
force or function may be assigned only if the communicative context yields infor-
mation about whether the speaker has certain obligations, the hearer certain wishes,
the action a beneficiary role for the hearer, etc. All this information must however
be detected in complex processes of comprehension of previous acts and utter-
ances, observation, stored or inferred assumptions, etc.
The question is, thus, how exactly are all these complex processes related?
4. Frames and speech acts
The actual comprehension of utterances as certain speech acts is based on a com-
plex process involving the use of the various kinds of information mentioned above.
The comprehension of particular observable indices, however, should be given in
terms of more general knowledge: understanding involves general concepts, catego-
ries, rules and strategies.
This general knowledge is not amorphous but organized in conceptual systems.

---

T. A. van Dijk / Context and cognition
215
One of the ways to account for this organization is in terms of frames. 3 Frames are
not arbitrary chunks of knowledge. First of all they are knowledge units organized
around a certain concept. But, unlike a set of associations such units contain the
essential, the typical and the possible information associated with such a concept.
Finally, frames seem to have a more or less conventional nature, and thus should
specify what in a certain culture is characteristic or typical . This criterion seems
to single out especially certain episodes of social interaction, such as going to the
movies, travelling by train, or eating in a restaurant. Although we might also call a
frame the set of epistemic units we have about books, balls and bananas, such units
do not as such organize our understanding of the world in a way a conceptual
frame as we define it does, viz. by also organizing our behaviour with respect
to the world, and the ways we interpret other s behaviour, as in the frames we have
about cashing a check or shopping. Although the distinction between mere con-
cepts and the frame-like organization of conceptual knowledge is still vague there
may be fuzzy boundaries in the theory - we provisionally keep ourselves to the
more restricted interpretation of the notion of frame.
The question which arises with respect to pragmatic theory is then: in what
sense may we consider speech acts as frames . Clearly, speech acts are acts and they
also have a conventional nature. We know, intuitively, how to promise something,
how to congratulate somebody, and this knowledge is clearly part of our world-
knowledge. In which respect, however, would our promise -frame organize our
knowledge of the world in a different way than our knowledge of bananas or bark-
ing, i.e. in the sense that we know one when we see or hear one? Could we say
that there is a whole episode during which a speaker is going about promising
e.g. by accomplishing all kinds of preparatory, component and auxiliary acts, as is
the case in taking a train?
Although the appropriate accomplishment of speech act involves a number of
conditions, and although there are ways of complex promising or threatening, we
would at first sight see no reason why a simple speech act would be a frame (and
not, for instance, laughing or hitting). The only organizing principle involved is
that relating certain purposes, intentions and certain doings (utterances with certain
properties) to contextual states and events.
3 The current notion of frame has been discussed mainly in artificial intelligence and cognitive
psychology after Minksy s influential paper (Minsky 1975). For further elaboration of this con-
cept, see the contributions in Bobrow and Collins, eds. 1975 also for further references. The
notion, of which variants have become known as scripts , scenarios or schemata , already
appeared in artificial intelligence in Charniak s dissertation (1972), where the term demon was
used in order to denote conventional knowledge structures used in the interpretation of dis-
course. In fact, much of the actual discussion about knowledge representation has its roots in
Bartlett s work on remembering (1932) in which the notion of a schema played an important role.
For a critical discussion about the nature of frames and their function in discourse comprehension,
see van Dijk 1977b.

---

T. A. van Dijk / Context and cognition
216
Yet, speech acts may nevertheless be connected with frames. First of all, we have
typical speech act sequences 4 of which the structure has a more or less conven-
tional or ritual character, such as giving lectures, preaching, making everyday
conversation, or writing love letters. In such cases we clearly have a number of dif-
ferent (speech) acts, of which each may have a characteristic function in the per-
formance of the episode: opening, introducing, greeting, giving arguments, defend-
ing, closing, etc. In such cases we may have different strategies for fully accom-
plishing our goals. Moreover, unlike (most) speech acts, they may be culture
dependent.
Secondly, speech acts are interpreted on the basis of frame-like world knowl-
edge, e.g. because they are part of such frames. Especially the institutional speech
acts, such as baptizing, marrying, convicting or firing (taken as speech acts) are part
of often highly conventionalized episodes. Without such frame-knowledge I would
for instance be unable to differentiate the utterance I sentence you to ten years of
prison when spoken to me by a judge, in a courtroom, at the end of a trial, etc.
or as spoken by my friend being angry against me. We would know that the first
speech act counts and the second not, because only the first is part of an institu-
tional frame.
Thirdly, the interpretation of speech acts requires knowledge of what might be
called meta-frames: we know the general conditions under which actions are
accomplished, when they are successful, etc. Thus, if somebody in his sleep tells me
Can you open the window? , I would hardly do so, because I have the general
knowledge that only controlled, conscious and purposefully intended doings count
as actions. 5
Finally, the interpretation of speech acts involves world knowledge more in gen-
eral. Speech acts often pertain to past or future activity of the speaker or the
hearer: they are essentially functioning as expedient ways in which such activities
are planned, controlled, commented upon, etc., or they are intended with the
purpose to provide information for such actions. Hence, they basically require
knowledge about what is necessary, plausible or possible in the real world. If some-
body would tell us: I just jumped from the Eiffel Tower , we would hardly take
him seriously.
Similarly, when I congratulate somebody I should assume that something
pleasant occurred to him, but our more general world knowledge will have to tell us
what is pleasant, for whom in what circumstances. Pragmatics itself will not make
explicit the latter conditions
which belong to a representation of our cognitive
semantics.
In other words, whether the necessary conditions for the appropriateness of
4 For an analysis of speech act sequences and their super-ordinate macro-speech acts , see van
Dijk 1977a.
5 For these general notions from the theory of action and their relevance for pragmatic theory,
see van Dijk 1977a.

---

T. A. van Dijk / Context and cognition
217
speech acts are actually satisfied must be decided by our knowledge of the world
and its frame-like mental organization.
5. Context analysis
Before language users are able to match incoming information against the more
general linguistic and other knowledge in memory, they must analyse the context
with respect to which a certain speech act is performed.
One of the methodological principles which should be kept in mind is that the
notion of context is both a theoretical and a cognitive abstraction, viz. from the
actual physical-biological-etc. situation. That is, a great number of features of the
situation are not relevant for the correct comprehension of the illocutionary force
of utterances. It will seldom be the case that, whether my speech participant has
red hair or not, I will understand his utterances in different ways (with the possible
exception of those cases where such situational details are thematized). Hence a
speech understander will focus attention on specific properties of the situation
which might be relevant for correct interpretation of both meaning/reference and
pragmatic intentions/purposes. A next methodological point is that unlike prag-
matics and (the rest of) grammar, a cognitive theory does not (only) have rules and
concepts, but strategies and schemata, i.e. devices for a fast and functional proces-
sing of information.
Strategies and schemata are the basis of the normal processes of hypothetical
interpretation: given a certain textual and contextual structure they allow fast
assumptions about probable meaning and intention
even if the rules at a later
moment may lead to a rejection of the hypotheses.
One of the most obvious examples of such cognitive processing devices is based
on the typical structure of the sentence: if an interrogative structure is given, we
may provisionally conclude that a question or request is made.
Similarly in the analysis of context. If a complete stranger is heading for us on
the street, we may be pretty sure that (globally) he will ask a question or make a
request and not make some assertion about his love life, or a threat.
It seems to follow that in pragmatic comprehension we not only establish a
context out of the situation, but also have schemata for the analysis of such con-
texts. That is, if a context obviously satisfies a set of (ordered) key features it will
be taken as characteristic for a specific set of possible speech acts.
It is clear that the analysis of context is a necessary but usually not a sufficient
condition for the pragmatic comprehension of utterances. That is, a language user
will merely have a certain set towards the possible speech acts which may follow.
The definite assignment of a speech act takes place, of course, after comprehension
of the utterance itself, and after matching of the pragmatically relevant information
from the utterance with the information from the context analysis. In this respect,
pragmatic comprehension parallels the relative process of semantic comprehension,

---

T. A. van Dijk / Context and cognition
218
in which previous discourse and knowledge of (semantic) context is important in
the interpretation of individual sentences. Parallel to the notion of presupposition,
then, we may introduce the notion of pragmatic precondition, defined as a neces-
sary contextual property.
Pragmatic comprehension schemata pertain to the initial context of the verbal
communication process, viz. the state which changes by the performance of an act
of speech. Such an initial context is not only characterized by the events/actions
immediately preceding the speech act, but possibly also by cumulated information
from earlier previous states and events.
Since however not all details from previous interaction states can be stored and
retrieved, permanent procedures of relevance assignment must be at work to sort
out the information which hypothetically will be important in further (inter-)action
production/comprehension. It may be assumed that the procedures involved here
are similar to those based on macro-rules in the processes of semantic comprehen-
sion of complex information (e.g. of discourse). 6 In other words, the initial con-
text with respect to which a speech act is to be interpreted contains three kinds of
information:
(i)
general semantic information (memory, frames);
(ii) final state information from immediately preceding events/acts;
(iii) global (macro -)information about the whole previous interaction structures/
processes.
Since (ii) and (iii) are relevant only for the processing of the actual context,
these kinds of information are of the episodic kind.
From the remarks above we may conclude that pragmatic contexts are struc-
tured. More in particular we assume that fast cognitive processing requires that con-
texts are hierarchially structured - as is also the case for semantic (macro-)struc-
tures of discourse. This hierarchy is defined in terms of social structure: speech acts
are integral part of social interaction (we do not usually perform them when being
alone).
The hierachical structure of society allows us to determine which units (e.g.
institutions, roles, actions) and relations are determined by those of higher level. In
order to be able to determine whether a speech act is appropriate, we thus should
first of all be aware of the most general social setting in which the interaction takes
place, and then about the more specific or ad hoc particulars of this setting, e.g.
actual properties of the speech participants.
Although the social context of speech acts is not the topic of this paper, it
should be recalled that the relevant social structure should, as mentally represented,
6 The relevance of macro-structures and macro-rules in cognitive processing of complex infor-
mation has been shown, both theoretically and experimentally, in van Dijk 1975, 1976, 1977a,
van Dijk and Kintsch 1977 and Kintsch and van Dijk 1977.

---

T. A. van Dijk / Context and cognition
219
be taken into account when analysing the processes of pragmatic comprehension.
The important methodological correlate of this assumption is of course that it is, at
this level of analysis, not so much the social situation itself, but the interpretation/
construction by the social members of that situation which counts in actual com-
munication. Clearly, this kind of cognitive relativism does not imply that those
interpretations have no objective basis. On the contrary, successful interaction
requires that the interpretations of social structure are conventionalized.
Note that the social context is also an abstract construct with respect to actual
social situations. In the first place all those properties are socially irrelevant which
do not somehow condition the interaction of the social members: it is irrelevant
what I actually think if I do not show my thought in my behaviour; nor is it rele-
vant what I carry in my car if it has no interactional meaning . Similarly for those
interactions which are particular to one situation, such as sneezing or smiling to
someone during a public lecture or a court session: they do not essentially deter-
mine the typical (inter-)actions defining such social settings in general.
Thus, social context analysis begins at the level of general social context. This
general social context may be characterized by the following categories:
(i) private
(ii) public
(iii) institutional/formal
(iv) informal
The precise definition of these concepts cannot be given here. Important is only
that they define different kinds of social contexts, e.g. public institutions such as
courts, traffic, or hospitals, public informal places such as restaurants or buses,
private institutions such as families and private informal settings such as making
love or beating somebody.
The different social contexts thus globally characterized are in turn defined by
the following properties:
(i) positions (e.g. roles, status, ect.)
(ii) properties (e.g. sex, age, etc.)
(iii) relations (e.g. dominance, authority)
(iv) functions (e.g. father , waitress , judge , etc.)
These properties of social contexts, and their members, are systematically
related. They define the possible actions of the social members in the respective
contexts. Social contexts may themselves be organized, e.g. by a certain structure
of (social) frames. 7 Thus, within the general institutional context of a court, there
7 We make a provisional difference between epistemic (cognitive) and social frames, as notions from
cognitive psychology and sociology, respectively. Clearly, there are relations between the two notions,
because the organization of social interaction has its influence on the organization of our knowledge
about social structures. An analysis of social frames has been given by Goffman (1974).

---

T. A. van Dijk / Context and cognition
220
are several frames, which are e.g. chronologically ordered, such as the charge-frame,
the defense-frame and the judgement/conviction-frame. In these frames members
are assigned specific functions/positions/properties and relations. More in particular
these frames regulate which kinds of acts may be performed. In the private institu-
tion of a family the parent-function, being associated with a set of properties and
relations (power, authority), defines a set of possible social actions, e.g. sending the
children to bed, whereas the child function, conventionally, does not allow that a
child sends one of his parents to bed. Besides the four defining categories of social
contexts, then, we must have a set of conventions (rules, laws, principles, norms,
values) defining which sets of actions are associated with which positions, func-
tions, etc.
Thus, a convention will determine when in an informal public place a member
with a certain property, relation, and function may greet another member. I may
not greet anybody in a crowded street, but may do so on a lonely mountain path,
whereas I may greet any acquaintance or even people I have just interacted with in
specific ways, etc. The complexity of these conventions and their constraints can-
not be gone into here.
For our discussion about the contextual analysis taking place during pragmatic
comprehension the (very fragmentary and informal) analysis of social context given
above suggests that each language user should take into account the following
information about this general social context: its specific type, the frame of the
context now being relevant, the properties/relations of social positions, functions
and the members filling these categories, as well as the conventions (rules, laws, prin-
ciples, norms, values) determining the socially possible actions of the members
involved.
Note that the analysis of a particular context in terms of the concepts men-
tioned above is possible only with respect to general knowledge of social structure.
It is within this broader framework of the analysis of the social context in which
the specific properties and relations (e.g. actions, interactions) of the speaker may
be analyzed, viz. his previous behaviour (doings, actions, etc.) - e.g. the specific
things he said before - as well as the inferences we make about the internal struc-
ture of the speaker, in terms of:
(i) knowledge, beliefs
(ii) wants, desires, preferences
(iii) attitudes
(iv) feelings, emotions
both at he particular as well as on the more general level (norms, values, beliefs the
speaker also exhibits in other situations). Note that part of the acts involved are
instances of the conventional acts belonging to the action sets of some informal or
institutional social context and its properties: in the institutional public context of
traffic, a person with the function of policeman, has the right to give me the signal

---