Language,
discourse, and cognition:
Retrospects
and prospects1
Robert
de Beaugrande
1.
Language and cognition as a (non)topic in linguistics
The
status of cognition as a concept in the science of linguistics remains
uncertain. I recently re-scanned my survey of major linguistic discourses (in
Beaugrande 1991) and I could not find the term ‘cognition’ in the writings
such avowed mentalists as Saussure and Sapir, nor (less surprisingly) in the
writings of such avowed behaviourists as Bloomfield and Firth. The first
occurrence of the term ‘cognition’, or ‘cognitive faculties’ to be
exact, was found in the writings of Noam Chomsky, which at the time signified a
programmatic renewal of mentalism, as we shall see.
The
long absence of the term ‘cognition’ in linguistics seems interesting in
view of the well-established term ‘cognitive linguistics’, and merits some
scrutiny. For the early mentalists, the contents of cognition, variously called
‘knowledge’, ‘ideas’, ‘thoughts’ and so on, seemed under-determined.
For Saussure (1966 [1916]: 111), ‘thought’ was ‘only a shapeless and
indistinct mass’, an ‘indefinite plane of jumbled ideas’. For Hjelmslev
(1969 [1943]: 52), ‘purport’ [Danish ‘mening’] was ‘an unanalysed
entity’, an ‘amorphous thought-mass’. These linguists surmised that
language imposes determinacy: ‘thought’ ‘becomes ordered in the process of
its decomposition’ as ‘language takes shape’ and enables ‘clear-cut
consistent distinctions between two ideas’ (Saussure); ‘linguistic form’
‘lays arbitrary boundaries on a purport-continuum’ that ‘depends
exclusively on this structure’ (Hjelmslev). So mentalist linguistics might
assume that the study of cognition would be left with nothing.
For
the behaviourists, in contrast, the contents of cognition seemed over-determined,
thanks to the notion that ‘mental images, feelings, thoughts, concepts’, and
‘ideas’ ‘are merely popular names for various bodily movements’ (Bloomfield 1933: 142). For Bloomfield, ‘the meaning of a linguistic form’ is not a ‘mental
event’, but ‘the situation in which the speaker utters it and the response
which it calls forth in the hearer’; then he wondered if ‘the situations
which prompt people to utter speech include every object and happening in their
universe’ (1933: 142, 139f). If so, ‘the study of speakers’ situations and
hearers’ responses’ might be ‘equivalent to the sum total of all human
knowledge’ (1933: 74). And ‘to give a scientifically accurate definition of
meaning for every form in the language’ would require ‘scientifically
accurate knowledge of everything in the speaker’s world’. So
behaviourist linguistics might assume that the study of cognition would be left
with everything.
At
all events, these mentalists and behaviourists did leave cognition out of their
accounts, and concentrated on the units of sound (phonemes) and the units of
form (morphemes), which plainly constitute orderly systems of ‘distinctive
features’. And when linguistics and semantics eventually embarked on the study
of meaning, they predictably reapplied a conception developed in phonology and
morphology, namely the system of basic units. Parallel to the
‘phonemes’ and ‘morphemes’, these units were variously termed
‘semes’, ‘sememes’, ‘semantemes’, ‘semantic features’, or
‘semantic primitives’ (e.g. Hjelmslev 1957; Katz and Fodor 1963; Greimas
1966). However, ‘phonemes’ can be securely defined by articulation, e.g., a
‘voiced dental stop’; and ‘morphemes’ can be defined, though less
securely, by the processes of word-construction, e.g., for the Tenses or Aspects
of Verbs. What could be the basis for defining the units in semantics?
Several
possibilities were aired, such as:
(a)
the ‘linguistic image of properties, relations, and objects in the real
world’ (Albrecht 1967: 179), as in Pottier’s (1963) taxonomy of chairs;
(b)
the distinctive elements arising from the ‘apperceptive constitution’ of
‘human beings in regard to their environment’ (Bierwisch 1966: 98), a
well-known example being colours (cf. Heider 1972);
(c)
as conceptual elements into which a ‘reading" decomposes a ‘sense’
(Katz 1966);
(d)
as elements for constructing a semantic theory (Katz & Fodor 1963);
(e)
as constituents of a metalanguage for discussing meaning (Greimas 1966).
Either
the units are given in advance by the ‘real world’ or the ‘human
environment’; or they are generated by the processes of human comprehension;
or again they are purely theoretical constructs of linguistics and semantics. In
all these interpretations, the units are presumably cognitive rather than
strictly linguistic; they cannot be simply a set of words used for writing
definitions in the manner of a dictionary. Moreover, the catalogue of units
would have to be universally applicable for all the meanings of a language, or
even the meanings of all languages. And the units would constitute a system
analogous to ‘phonemes’ and
‘morphemes’ only if a given unit is applied repeatedly and consistently to
many meanings. In the most strictly orthodox view, the units would determine the
‘sense properties’ whose ‘sum’ is the ‘sense of any expression’ —
‘its indispensable hard core of meaning’, ‘deliberately excluding any
influence of context or situation of utterance’ (Hurford and Heasley 1983:
91).
An immediate problem was how to label these
cognitive units of meaning. If we use
words, then the words can have meanings of their own beyond the basic units they
are supposed to label. Let us consider a seeming straightforward example I
recall hearing discussed at the time: constructing the meaning of ‘kill’
from CAUSE + DIE (Chomsky 1971 [1968]: 188f). In authentic discourse, this
construction works for some instances like [1], and breaks down for others, like
[2].
[1] ‘It would always be a heavy
thought to me, if I’d caused his death,
even in a just cause.’ ‘Yes’, said Phineas, ‘killing
is an ugly operation…’ (Uncle Tom’s
Cabin)
[2]
Is’t I that chase thee from the country and expose
Those
tender limbs of thine to the event
Of
the none-sparing war? […]
Whoever
shoots at him, I set him there.
Whoever
charges on his forward breast
I
am the caitiff that do hold him to’t.
And
though I kill him not, I am the cause
His
death was so effected (All’s
Well that Ends Well III, 2, 105-108, 115-119)
The
problem here lies not in the correct choice of labels for ‘semantic
features’ or ‘sense properties’, but in the orthodox aspiration to
describe an ‘indispensable hard core of meaning’ free of all ‘context’.
Whilst
these discussions were in progress, the so-called ‘cognitive revolution’ set
about transforming the human sciences from the 1960s onwards. The most
‘revolutionary’ idea, in my view, was that cognition is a unifying central
faculty encompassing a wide range of human capabilities and processes that had
hitherto been studied in isolation and with narrow methods — intelligence,
learning, memory, skills, problem-solving, and, ultimately, language and meaning
(cf. overview in Newell 1990).
The
response of linguistics was a return not merely to the mentalism of its early
years (e.g. Saussure, Hjelmslev) but to the idealist and rationalist philosophy
of 17th and 18th centuries (e.g. Descartes, Leibnitz).
Early idealism held the contents of cognition to be determinate, but proposed to
explain this factor at a single stroke by invoking a ‘universal,
language-independent’ ‘system of possible concepts’ (Chomsky 1965: 160).
The order of meanings in language would be pre-determined by the construction of
the human mind; each language makes its own selection out of the universal
order. For philosophers like Descartes, the ultimate source and creator of this
order would be the mind of God. In modern linguistics, God was dispensed with by
linking cognition directly to biology (as
if to outflank behaviourism): ‘there is a highly determinate, very definite
structure of concepts and of meaning that is intrinsic to our nature, and as we
acquire language or other cognitive systems these things just kind of grow in
our minds, the same way we grow arms and legs’ (Chomsky 1991: 66).
This
renewed mentalism in effect set out to study cognition instead of language,
aspiring to construct ‘a theory of linguistic structure’ without
‘reference to particular languages’ (Chomsky 1957: 11). Eventually, real
language — now called ‘externalised language’ (or ‘e-language’ for
short) — was declared to be a mere ‘epiphenomenon’ (Chomsky 1986: 25). In
these ‘radically different theories’, ‘there are no constructions; there
are no rules’ (Chomsky 1991: 81). This version of cognitive linguistics is
concerned solely with ‘the structure of mental representations’
and expressly excludes ‘the relationship between’ ‘mental
representations’ ‘and things in the world’ (Chomsky 1991: 93). The
exclusion carries ironic (and probably unintentional) echoes of Bloomfield’s
much earlier exclusion, quoted above, upon confronting ‘the sum total of all
human knowledge’.
The
wheel seems to have turned a long ways. Cognition has passed from being first an
‘indistinct mass’, then a stupendous catalogue of ‘every object and
happening in the universe’, then a determinate system of ‘hard-core
meanings’, and finally a ‘universal system of
possible concepts’. As far as I can tell, none of these conceptions adequately
defines the relation between cognition and language; and I
shall suggest an account that is markedly different in theory (what cognition
and meaning are) and in practice (how we go about describing them through data
from language and discourse).
2.
Language and discourse as theory and practice
Applying our terms
broadly, we can define a language as a theory of human
knowledge and experience (what speakers know how to talk about), and discourse
as its practice (what speakers do talk about). Language and discourse would
thus interact in a
dialectical cycle, with each side informing and guiding the other,
as modelled in Figure 1. Language specifies the standing constraints (e.g., what the
words usually mean), whereas discourse manifests emergent constraints
(e.g., what the words mean in this particular stretch of discourse).
The
practices are heavily ‘theory-driven’ in the sense that discourse compels
its participants to ‘theorise’ about what words mean, what makes sense, what
people intend, and so on. Yet the theory is also heavily ‘practice-driven’
in the sense that discourse continually tests and adjusts implicit theories
about meanings, senses, or intentions,
In this account, discourse is
probably the most theoretical practice humans can perform, particularly
in respect to meaning. In return, language is the most practical theory
humans can devise, offering us the resources to shape and guide almost any of
our practical activities. Yet the ‘theoreticalness’ of language is extremely
well-hidden from most speakers who practice it. They would regard discourse as a
‘practical matter’; they would be surprised if we told them they possess a
‘theory of their language’ that makes them ‘theoreticians’.
All
human practices produce knowledge, which is stored as the contents of cognition.
Language must have originated and evolved as a refined means for delimiting,
organising and stabilising those contents into
meanings, and for sharing the meanings among the community.
The content of cognition and the content or meaning of language are thus
phenomena of the same order, differing in degree more than in kind. By talking
about what we know and sharing it with other people, we also build or change our
knowledge in ways that could hardly be done without the use of language.
Cognition
and language would thus also interact in a dialectical cycle, with each side
informing and guiding the other, as modelled in Figure 2.
Most
essentially, cognition generates meanings, whereas language determines
meanings. But here also, the difference is only in degree, and these
dialectical processes are inseparable.
A
helpful analogy here might be cognition as the landscape
and language as the map (compare the ‘cognitive maps’ in Neisser
1976). We could live in the landscape and move about after a fashion with no
map, but language makes our lives and our movements much more organised and
productive. Similarly, each of us can have a ‘mental map’ of the landscape,
but without the language-map, we could not compare and improve the goodness of
fit. The meanings of language are in turn not just names of points on the map,
but directions about how to navigate with the aid of the map. And discourse is
the activity of actually walking through the landscape by means of the map.
How
accurate the map might be in any literal cartographic sense could never be
established, because both the landscape and the map are continually evolving.
Indeed, just using the map can change the landscape or the map, or both. We can
only try to notice how helpful the map proves to be in practice, and hope that
we will get where we want to go and not be led too far astray.
Nor
again could we establish which language is a better map of human cognition; or
whether cognition is the same for all humans whilst only our maps differ from
one language to another. Such questions are irresolvable because we do not have
any language-independent modality for accessing cognition beyond the limits of
what we happen to perceive with our senses. And even what we see, hear, and so
on is partly pre-organised by language, so its potential for testing language
remains limited in principle.
The
analogy to the map is inappropriate in that a map is deliberately drawn by an
explorer or cartographer, whereas language evolves both spontaneously and
communally. Somehow, the contents of cognition get organised as meanings of
language with remarkable precision and delicacy even though speakers are not
following an explicit plan or design. The organisation is ‘delivered for
free’, so to speak, and one might empathise with the 17th century
idealists who imagined it all to be the work of God.
Perhaps
a different analogy than a map might be helpful here. The termite hill on the
African veldt is an impressively complex structure built by an immense number of
individual events and agents, each of these eminently simple in itself. A single
termite certainly possesses no knowledge of the total design of the termite
hill; such knowledge could not be justly attributed even to the entire termite
colony. Instead, the design is generated by a dynamic convergence among millions
of movements of termite moving tiny grains of sand or earth.
An
intriguing fact about the architecture of termite hills was demonstrated in
computer simulation by Peter Kugler when I was at the Institute for Medical
Engineering of the University of California, Los Angeles. He was able to show
that the main supports of the structure correspond to the points where the
trajectories of the several termites most frequently intersect. The structure
thus emerged from the crossing of pathways — like ‘order out of chaos’
(Prigogine and Stengers 1984).
In
an analogous manner, the complex architecture of cognition could arise from the
intersections among the trajectories in networks of knowledge representing the
relations among classes of perceptions, sensations, and so on, each one of which
is simple in itself. What something ‘means’ or ‘represents’ is a
function of what it is connected to — the pathways leading toward or away from
it in the network. When some item of knowledge gets put into practice, a point
in the network, or more likely a cluster of points, gets ‘activated’; and
this activation automatically spreads out along the contiguous pathways without
the intervention of executive control (cf. Collins and Loftus 1975). The
relevant pathways retain and increase their activation, whereas the irrelevant
pathways lose their activation in about half a second — a mode of processing
cognitive psychology has been able to demonstrate under the label of the
‘construction-integration model’ (Kintsch 1988, 1989). A strangely
counter-intuitive consequence is that all meanings of a word are initially
activated, and not just the one relevant to the context, as ‘relevance
theory’ complacently asserts without regard for the findings of experimental
psychology. The relevant one is the one whose activation level is sustained or
raised.
What is
conventionally called a ‘meaning’, concept’, ‘thought’, ‘idea’ and
the like is not a standing unit like an entry in a dictionary or a message in a
pigeonhole, but an activation pattern. Each instance of activation may
strengthen its links, but may also adjust or alter them, depending on other
areas of parallel activation, which together constitute the ‘context’.
In this account,
the contents of cognition and the meaning of language are products of massive
parallel distributed processing (Rumelhart, McClelland, et al. 1986). They
are not so much ‘stored’ as ‘prefigured’ in the cognitive architecture
of human knowledge networks, which would be the psychological and neurological
instantiation of language as ‘theory’; their activations would be the
instantiations of discourse as ‘practice’. So language would have naturally
emerged as a ‘user-friendly front end’ of cognition, which itself possesses
language-like qualities. Language is a product and process of ‘higher-order
consciousness’ that reconciles the individual and the social factors of
cognition (cf. Edelman 1992).
Now,
this account could explain the failure of projects in linguistics and semantics
for describing meaning in basic units that constitute the ‘hard core of
meaning independent of context or situation’. Meanings are not units
at all, but events in a dialectical process which always has a context as
its cognitive architecture. If semanticists purport to be analysing as meaning
out of context, they are in fact just replacing natural contexts with artificial
ones, I cannot see why the results could claim to be generally valid or
representative. I would go on to point out, as a logical corollary, that
meanings will not hold still or remain constant while we ‘analyse’ them. On
the contrary, analysis necessarily entails an increased or specialised
activation, the meaning will keep on spreading and thus gaining in richness and
complexity.
We
might draw here upon the tradition of ‘gestalt psychology (e.g. Koffka 1935)
by modelling cognition as the ground, and a language as the figure.
In a typical gestalt, the figure is generated by receiving the central focus of
attention or consciousness, whereas the ground receives only peripheral
attention. This model would be recursive in the sense that the meaning of
a discourse is also the figure set off by focused attention against the ground
the meaning in the language, which also receives peripheral attention. The
participants in the discourse understand the current meanings with the aid of
the usual meanings being, in the commonplace metaphor, in the ‘back of the
mind’. This model too indicates the impossibility of describing meaning by
‘indispensable hard-core senses’; a meaning does not have such a core.
Speakers
know what words mean, and how those meanings can be shared, through the
convergence of partially similar events. The meaning of a given word evolves in
a dialectical cycle between what it might mean by itself and what other sorts of
words it is likely to appear with in collocations, i.e., typical
combinations of lexical choices (Sinclair 1991; Hunston and Francis 1999). In
cognitive terms, the meaning takes on its characteristic content by virtue of
multiple activations in networks with other meanings.
A
grand challenge for cognitive linguistics, I would suggest, is to analyse
meanings by examining large sets of such events as provided by very large
corpora of authentic text and discourse (Beaugrande 2000). Let us return to
examples of ‘kill’ and ‘die’. Using my general corpus of about ten
million words of British and American writers on the WordPilot programme (Milton
1999), I found 442 occurrences of ‘killed’. In the large majority, the thing
‘killed’ was indeed a person or an animal (especially a snake) being caused
to die. But this was not an ‘indispensable core’. One can histrionically
speak of being ‘killed’ by sadness [3] or discomfort [4]. Or, one can use
the term to mean ‘put a sudden and complete end to’ something that is
neither a person nor an animal [5-6].
[3]
How should I ever have borne it! It would have killed me never to come to
Hartfield any more! (Emma)
[4]
The heat was excessive; he had never suffered any thing like it--almost wished
he had staid at home--nothing killed him like heat (Emma)
[5]
I’m in love with change and I’ve killed my conscience (This Side
of Paradise)
[6]
you have killed my love. You used to stir my imagination. Now you don't
even stir my curiosity. You simply produce no effect. (Dorian Gray)
My
corpus of Shakespeare plays returned some highly creative meanings, such as
cease an emotion [7], cancel one vow with another [8], outshine a light [9],
close one’s eyes [10], or torment a person with misplaced kindnesses [11].
This last has become a standing collocation in English.
Hath
kill’d the flock of all affections else
That
live in her (Twelfth Night)
[8] You do advance your cunning more and more.
When
truth kills truth, O devilish-holy fray! (Midsummer
Night’s Dream)
[9] Arise, fair sun, and kill
the envious moon,
Who
is already sick and pale with grief
That
thou her maid art far more fair than she. (Romeo and Juliet)
[10] To bed, to bed! Sleep kill
those pretty eyes,
And
give as soft of all thought! (Troilus and Cressida)
[11] This is a way to kill
a wife with kindness,
And
thus I’ll curb her mad and headstrong humour (Taming of the Shrew)
[12]
He opened his lips to speak, but his accents died
away ere they were formed. (Wieland)
[13]
the illumination died away in a few seconds, and my heart died
away within me as it went. (Arthur Gordon Pym)
[14]
His face was ruggedly formed, but it looked like ashes—like something from
which all the warmth and light had died out.
(My Antonia)
[15]
Next year you return to the same point of view, perhaps to find that the glory
has departed, as if the mountain had died (Steep
Trails)
I
was surprised to find from my Shakespeare corpus that only people were said to
have ‘died’. No other uses of the term occurred.
It
is customary in studies of language or literature to distinguish between ‘literal
meaning’ and ‘metaphoric meaning’. The implication is that a
listener or reader first supplies the literal meaning and, noticing this does
not fit, casts about for a suitable metaphoric meaning. But the ‘spreading
activation’ or ‘construction-integration’ models of cognitive processing
suggest that the suitable meaning could by produced without such a detour
through the literal meanings. Certainly such could be expected if the meaning is
familiar, e.g. ‘sound died away’ covering a range of other things that
‘died away’ in my corpus data, such as ‘voice’, ‘whisper’,
‘footsteps’, ‘trampling’, ‘music’, ‘rhythm’, and ‘cadence’
In
the purview of text linguistics, the collocations we can identify in very large
corpus data represent a special phenomenon: they are more specific than the
language, yet more general than the text, like some category of ‘missing
links’ in between (Beaugrande 2000). In terms of cognition, the account I have
proposed would explain collocations as interactive priming loops among
expressions whose probability of being chosen together is significantly higher
than the normal base state prior to activation. The links among, say, ‘sleep
– close – eyes’ would be closer to activation, whereas no such linkage
would be primed among ‘sleep – kill – eyes’.
Looking
at multiple occurrences of an expression in corpus data is a special means of
distributing our attention between figure and ground. Instead of just
determining the meaning of an expression in a single discourse, we are seeking
to assess the similarities and differences among its meanings in multiple
discourses. Some of those meanings will match our intuitions, e.g., how
‘voices died away’, whilst others may not match, e.g., how a ‘mountain
died’. Usually the match is a matter of degrees rather than a yes-or-no
division.
Still,
our knowledge of collocations is part of the cognitive and linguistic ground for
the meanings that figure in any single text. Quite plausibly, the greater
challenge for a ‘cognitive text linguistics’ would be to account not just
for the figures in the focus of processing, i.e., the expressed meanings of a
text, but for the ground at the peripheries of processing, i.e., the (possibly)
implied meanings. If activation works itself out through the cognitive
architecture of knowledge and meaning without executive control, then
determining its full breath or scope may prove difficult indeed.
Consider
this brief report about someone being ‘killed’ in a road accident, taken
from the US news media (Cullingford
1978):
[16]
A New Jersey man was killed Friday evening when a car swerved off Route 69 and
struck a tree. David Hall, 27, was pronounced dead at Milford Hospital. The
driver, Frank Miller, was treated and released. No charges were filed, according
to investigating officer Robert Onofrio.
In
exchange, we seem to know all manner of things implied but not stated. We know
that the hapless Mr. Hall was ‘killed’ by the impact of the car against the
tree, rather than being bumped off just then by the New Jersey Mafia. Miller
lost control rather than deliberately heading for the tree with the plan of
knocking it down and carting it home for firewood. The ‘tree’ was a large
sturdy outdoor tree rather than a dwarf bonsai on a window sill or a Christmas
tree in a shopping mall. What was left of Hall got taken to ‘hospital’ by an
ambulance and not by canoe, camel, or skateboard. Miller was ‘treated’ by
dressing his injuries and not by giving him a fancy-dress dinner. The
‘charges’ would be documents for criminal proceedings and not swift advances
of mounted cavalry. These excluded alternatives may sound rather zany or
scurrilous, but how and why we know they are excluded is by no means a trivial
question. The psychological evidence I have cited for the ‘construction-integration
model’ indicates
that large amounts of irrelevant knowledge are indeed initially activated, such
as these alternate meanings of ‘tree’, ‘treat’, and ‘charge’.
Now
contrast this passage from a news item from the Ipswich
Journal (Jan. 12, 1878):
From a purely linguistic standpoint, the Pronouns
‘he’ and ‘him’ could be ambiguous about who pronounced whom dead. But
from a cognitive standpoint, no doubt can arise. The actions of ‘staggering’
and ‘falling’ (to say nothing of not wanting good brandy) are nicely apropos
prior to becoming ‘dead’; and a ‘surgeon’ is just the person who has the
authority to make such ‘pronouncements’.
Notice
that we are not actually told that the solicitor has died. But just to introduce
a modest margin of doubt would require some macabre ingenuity, e.g.:
Whether
we elect to call such explorations by the term of ‘cognitive text
linguistics’ is a purely academic matter. In the past, linguistics has been
rather too liberal at coining labels in advance of substantial achievements.
What counts will be the vast dimensions and richness of the terrain, for which
at the moment we shall for some time still be struggling to attain a shared
cognitive map.
Note
[1]
This paper is a fully revised version of a video lecture recorded in Botswana
and presented at the Conference of the International Association of Cognitive
Linguistics in Stockholm in July 1999. I am deeply indebted to Ibolya Maricic of
Vaxjö University for the transcription.
Albrecht,
E. 1967. Sprache und Erkenntnis.
Berlin: Deutscher Verlag der Wissenschaften.
Beaugrande,
R. de. 1980. Text,
Discourse, and Process. Norwood, NJ:
Ablex.
Beaugrande,
R. de. 1991. Linguistic
Theory: The Discourse of Fundamental Works.
London: Longman.
Beaugrande,
R. de. 1997. New
Foundations for a Science of Text and Discourse.
Stamford, CT: Ablex.
Beaugrande,
R. de. 2000. Text linguistics at the
millennium: Corpus data and missing links. Text 20.
Bierwisch,
M. 1966. Strukturalismus: Geschichte, Probleme, Methoden. Kursbuch. 5,
77-152.
Bloomfield,
L. 1933. Language. Chicago: University of Chicago Press.
Chomsky,
N. 1957. Syntactic Structures. The Hague: Mouton.
Chomsky,
N. 1965. Aspects of the Theory of Syntax. Cambridge: M.I.T. Press.
Chomsky,
N. 1971 [original 1968]. Deep structure, surface structure, and semantic
interpretation. In D. Steinberg and L. Jakobovits (eds.), Semantics: An
Interdisciplinary Reader in Philosophy, Linguistics, and Psychology.
Cambridge: Cambridge University Press, 183-216.
Chomsky,
N. 1986. Knowledge of Language. New York: Praeger.
Chomsky,
N. 1991. Language, politics, and composition (with Gary Olsen and Lester
Faigley). In: G. Olsen, and I. Gales (eds.) Interviews: Cross-Disciplinary
Perspectives on Rhetoric and Literacy. Carbondale: Southern Illinois UP,
61-95.
Collins,
A. and Loftus, E. 1975. A spreading-activation theory of semantic processing. Psychological
Review 82, 407-428.
Cullingford,
R. 1978. Script Application. New Haven: Yale Univ. dissertation.
Edelman,
G. 1992. Brilliant Air, Bright Fire: On the Matter of the Mind.
New York: Basic Books.
Greimas,
A. 1966. Sémantique structurale: Recherches de méthode. Paris:
Larousse.
Heider,
E. 1972. Universals in color naming and memory. Journal
of Experimental Psychology 93, 10-20.
Hjelmslev,
L. 1969 [original 1943]. Prolegomena to a Theory of Language. Madison:
University of Wisconsin Press.
Hjelmslev,
L. 1957 Sémantique structurale. In Essais linguistiques. Copenhagen:
Sprog- og Kulturvorlag, 1970, 96-112.
Hunston
S. and Francis, G. 1999. Pattern Grammar. Amsterdam: Benjamins.
Hurford,
J. and Heasley, B. 1983. Semantics.
Cambridge: Cambridge University Press.
Katz,
J. 1966. The Philosophy of Language. New York: Harper & Row.
Katz,
J., and Fodor, J. 1963. The structure of a semantic theory. Language 39,
170-210.
Kintsch,
W. 1988. The role of knowledge in discourse comprehension: A
‘construction-integration model’. Psychological
Review 95/2, 163-82.
Kintsch,
W. 1989. The representation of knowledge and the use of knowledge in discourse
comprehension. In R. Dietrich and C. Graumann (eds.), Language
Processing in Social Context. Amsterdam: North Holland, 185-209.
Koffka,
Kurt. 1935. Principles of Gestalt Psychology. New York: Harcourt, Brace,
& Co.
Milton,
J. 1999. Lexical thickets and electronic
gateways: making text accessible by novice writers. In C. Candlin and
K. Hyland (eds.), Writing: Texts, Processes and Practices. London:
Longman, 221-243.
Neisser,
U. 1976. Cognition and Reality. San Francisco: Freeman.
Newell,
A. 1990. Unified Theories of Cognition. Cambridge, MA: Harvard University
Press.
Pottier,
B. 1963. Recherches sur l’analyse sémantique en linguistique et en
traduction méchanique. Nancy:
University of Nancy Press.
Prigogine,
I. and Stengers, I. 1984. Order out of
Chaos: Man’s Dialogue with Nature. London: Heinemann.
Rumelhart, D., J. McClelland, et al. 1986. Parallel Distributed Processing: Explorations in the Microstructures of
Cognition (2 vols.). Cambridge, MA: MIT Press.
Saussure,
F. de. 1966 [original 1916]. Course in General Linguistics (transl. Wade
Baskin). New York: McGraw-Hill.
Sinclair,
J.McH. 1991. Corpus, Concordance Collocation.
Oxford: Oxford University Press.
Virtanen,
T. 1990. Problems connected with the choice of data and the use of standardised
multipurpose corpora in linguistics. In. K. Battarbee and R. Hiltunen (eds.), Alarums
and Excursions. Turku: University of Turku Press, 163-173.