Department of linguistics, Gothenburg University, 1995, Bilyana Martinovski

Three cognitive theories: major differences and similarities - Talmy, Langacker, Jackendoff

1. Introduction

The modern direction of language study towards cognitivism gets more and more clear results, in form of theories and models. They seem to be similar and distinct in various ways. In this paper I will try to outline what is common and what is different in the cognitive theories of language presented by Langacker 1988-94, Talmy 1988 and Jackendoff 1983. My comparison will be based on the following aspects of their theories:

i) ontologies, epistemologies;

ii) relations between syntax and semantics;

iii) relations between grammar and lexicon;

iv) relations between lexicon and encyclopedia;

v) kinds of cognitive categories and operations;

vi) opinions on feature analysis;

vii) opinions on the idea of prototypes;

viii) ideas on the determination of the meaning of a linguistic expression.


2. Basic assumptions

The basic assumptions of cognitive theories of language are related to the ontology and the epistemology of human language. It is not the case that each of the mentioned authors express clearly the ontological and epistemological notions they are working with. In fact, only Jackendoff gives clear expression of the ontology behind his approach. The reason is that, for example, for Langacker cognitive theory of language is analyzing meaning only on conceptual level. One may interpret his first basic claim that "meaning reduces to conceptualization (mental experiences)" in the sense that perception is part of the process of conceptualization and if so then there are no clear boundaries between perception and interpretation. But if perception is incorporated in the conceptualization process and if we accept that what we perceive is not always exactly what is going on in the external world (f.ex. our perception of light from a lamp, or perception of colors) than one may say that for Langacker there is no need of ontology, but just of epistemology, since his theory is concentrated only on the process of conceptualization: we cannot say anything about how the world really is but how we conceptualize it.

Talmy is not explicit on that point either. However, his basic concern is the hypothesis that we conceptualize and concentrate on different cognitive structures and notions. These are only cognitive, which does not mean that the world must or is organized according to these structures and notions.

Jackendoff distinguishes between a real world and a projected world. We have conscious access only to the projected world, which is "the world as unconsciously organized by the mind" . Hence, for Jackendoff, there is also a clear difference between real reality and conceptual reality:

"...we can talk about things only insofar as they have achieved

mental representation through this processed of organization.

Hence the information conveyed by language must be about the projected world."

(in Semantics and Cognition)

His major ontological categories are identified by features like thing, place, event, action, manner, amount, direction, sound, smell, time. These features are called basic domains in Langacker and basic cognitive structures and notions in Talmy, but there is no claim, in the latter two authors, that these notions identify ontological categories. Maybe, we have to understand them as such, but they are not explicitly defined so.

We can already state some basic assumptions common to all those authors:

1. Meaning is conceptualization.

2. There is difference between real world and conceptualized world.

3. There is no direct correspondence between these two worlds.

4. The cognitive theory of language describes only the organization of this conceptualized world.

What follows from these assumptions is that these theories are working only with epistemological categories and not with ontological categories.

Another common view is that there are special cognitive processes and operations of conceptualization which are used of human beings not only for organization of linguistic but also for non-linguistic information.

5. The cognitive operations used of humans to organize and structure linguistic information are the same as those used to structure non-linguistic information.

6. Human beings have inborn capacity for such internal organization of information which is expressed by these operations.

3. On the relations between levels of linguistic description

A fundamental part of a theory of language are the claims concerning the relations between the levels of linguistic description such as semantics, syntax, pragmatics etc. Here I will include the problem of the distinction between lexical and encyclopedic meaning, which for Langacker and Jackendoff is directly related to the distinction between semantics and pragmatics.

However, it is not the case that all authors express opinion on all these matters. Jackendoff is typically most concerned with syntax and semantics; Langacker is most explicit on the semantics/pragmatics question; Talmy gives a special version of the relation between grammatical and lexical notions; he is not explicit about the other two relational pairs.

3.1. The relation between syntax and semantics

There are different opinions on the relation between syntax and semantics. For Langacker, Talmy and Jackendoff semantic structures are treated as a special case of the conceptual structure. But for Langacker syntactic structures are dissolved in, expressed by semantic structures and the semantic structures are characterized relative to cognitive domains, called cognitive structures in Jackendoff (they are called differently and they consist of different elements, although their general function in both theoretical bodies are compatible). For Langacker there is a need of only two levels for description of a linguistic expression, a semantic one and a phonological one. He describes syntactic categories in terms of his basic cognitive notions profile/base, figure/ground and trajector/landmark. Thus a subject is said be a nominal expression that corresponds to the trajector of a clausal head, and a direct object - to its primary landmark. For example, in the construction "long snake" the trajector of the adjective ( a thing ) is identified to the noun, but the noun adds more specific information to it; the landmark is a salient participant other than the trajector, in the case of "long" it is a region along the length scale. In the same manner

"...verbs, adjectives, adverbs and prepositions are all attributed trajectors and landmarks, regardless of whether they function as clausal heads."

Consequently, for Langacker the so called syntactic categories and syntactic relations are part of the internal structure of the concepts; the word order is a reaction on these internal properties and structures.

Jackendoff, on the other hand, distinguishes between phonetic representation, syntactic structures, semantic structures and conceptual structures. He adopts the Conceptual Structure Hypothesis which in his case "proposes the existence of a single level of mental representation onto which and from which all peripheral information is mapped. This level is characterized by an innate system of conceptual well-formedness rules. ...the concerns of semantic theory with the nature of meaning and with mapping between meaning and syntax translate into the goals of describing the conceptual well-formedness rules and the correspondence rules, respectively." Semantic properties are not sufficient for Jackendoff to explain how the syntactic form of language reflects the nature of thought. To do that one needs Grammatical Constraints which are part of his cognitive theory and explain the relation between syntax and lexicon. But "syntax is formally unlearnable unless the learner makes use of information from the underlying structure of the sentence, which they take to be derivable from the meaning" . None the less, the Grammatical Constraint is a mystical notion since it lacks explicit definition. According to it, several grammatical constructions characteristic of reference to #thing# (thing in the projected world) find close parallels in constructions that refer to other ontological categories. It seems that Jackendoff is applying the Chomsky-inspired syntactic theories to conceptualization processes, using the formal feature representation as in Bresnan's Lexical-Functional Grammar. There is, however, a distinction between lexical and grammatical categories, which is not different from the traditional view on that point.

3.2. The relation between lexicon and grammar

Talmy's distinction between lexical and grammatical notions is special. He puts the so called open-class categories in the set of lexical categories, that is: verbs, substantives, adjectives; everything else is counted as close-class categories or grammatical notions: inflection, prepositions, conjunctions, determiners, particles, derivations, grammatical relations and categories, "zero" forms, word order (he avoids the use of the term syntax), intonation pattern. Obviously, Talmy is reinterpreting the distinction between categorematic and syncategormatic terms known from Aristotle's and logica moderna from the Middle ages (representatives of which are William of Sherwood, Peter of Spain, Lambert of Auxerre etc.) by omitting some elements from the former and adding a lot more to the latter. In logica moderna most nouns, pronouns, verbs, participles and adjectives were considered to be categorematic words, that is, words capable to serve as subjects of predicates; and the syncategoremtic words were those which can occur in a statement only together with categorematic words . The latter had the properties signification, supposition, copulation and appellation, whereas the former did not have these properties. Talmy's grounds for this distinction are different. As a conceptualist and cognitivist he shares the idea that all categories and notions in language have meaning (signification), but thinks that the lexical and the grammatical elements specify different kinds of notions (f.ex. grammatical elements in no human language specify for kind of material, color, rate etc.). 3.3. The relation between lexical and encyclopedic meaning, semantics and pragmatics There are different views on the status of syntax in relation to semantics but very similar opinions on the relation between encyclopedic and lexical meaning. Talmy is not discussing the problem but judging from his treatment of linguistic meaning one can say that it is very similar to that of Langacker. The latter was reluctant to accept the Chomskyan distinction between syntax and semantics, now he is rejecting the assumed significance in the distinction between semantics and pragmatics. This opinion of his is based on his impossibility to distinguish between lexical and encyclopedic meaning, since according to his basic assumptions the conceptualization processes and structures are relevant for all kinds of knowledge, linguistic and non-linguistic. Both for Langacker and Jackendoff this distinction has only methodological grounds, but assuming it in the framework of a cognitivistic theory of language will damage the entire enterprise. "I see no a priori reason to accept the reality of the semantics/pragmatics dichotomy. Instead, gradation of centrality in the specifications constituting our encyclopedic knowledge of an entity." (in A View of Linguistic Semantics) This statement is very similar to Jackendoff's position: "...There is not a form of mental representation devoted to a strictly semantic level of word meaning, distinct from the level at which linguistic and nonlinguistic information are compatible. This means that if, as it is often claimed, a distinction exists between dictionary and encyclopedic lexical information, it is not a distinction of level; these kinds of information are cut from the same cloth." (pp. 110 in Semantics and Cognition)

4. Criteria for characterization of meaning

4.1. Conditional criteria

All the discussed authors are concerned with the characterization of linguistic meaning. In this context, one may formulate another basic assumption which seems to be shared by the described theories as follows:

7. We begin constructing our mental universe of experience registered in basic domains (or primitives, or basic cognitive categories), arriving at ever higher level of conceptual organization by means of innately specified cognitive operations.

Langacker criticizes both the description of meaning by lexical primitives or prototypes and by feature analysis. The primitives approach is not relevant mainly because for Langacker the cognitive domains are open-ended, that is are not fixed. However, as I already mentioned, he accepts that there are some basic cognitive domains: "It is however necessary to assume some inborn capacity for mental experience, i.e. a set of cognitively irreducible representational spaces or fields of conceptual potential. ... Among these basic domains are the experience of time and the ability to conceptualize configurations in 2- and 3-dimensional space...color space, the ability to perceive a particular range of pitches, domains defining possible sensations of taste and smell, and so on." (pp.54-55 in View of Linguistic Semantics) This definition strongly reminds of Jackendoff's features identifying the major ontological categories, which I mentioned above and Talmy's topological and non-topological categories of cognitive notions, which are :

topological (relativistic)

linera extent/plurality point/singularity same/locatedness within/different region/"adjacency" of points side/one-to-one correspondence partition/pattern of distribution

non-topological (absolute)

material space time motion medium entity currently indicated quantified magnitude shape/contour of line rate kind of material sensorimotor character color

This list is, of course, reducible to the more general notions of Langacker and Jackendoff. What is important is that all this authors assume that there are some basic irreducible representational fields of conceptual potential, but for some reason they don't see them as fixed and they avoid to call them primitives. One fundamental reason that could partly explain this position is that all these three students of language speak of conceptualization while emphasizing the subjective nature of linguistic meaning, which is one of the reason for their assumption that even the basic cognitive fields are not fixed, although subjectivity does not presuppose dynamics and the opposite. It is however different when we come to cognitive operations which are also inborn capacities of conceptualization. (More about that follows below.)

Langacker criticizes also the feature analysis, not by arguing but by presenting an alternative view on that point.

" ...a cognitive domain is an integrated conceptualization in its own right, not a feature bundle." (pp.54 in View of Linguistic Semantics)

This sounds promising, but when he describes the main categorizing relationships - schematicity and extension - it appears that the former one is a relation where the more specified concept has a domain which adds some new non-conflictual features to the more abstract concepts (ex. circular object-> circular piece of jewelry) and the latter one add new and conflictual information (ex. circular object-->arena, since there are rectangular arenas). It does not help much that this new information is presented by other domains, the fact is that it is properties or features that are added or omitted.

Jackendoff represents one branch of the decomposition school because he believes in the necessity of decomposition of meaning but not by binary or n-ary features. Like Langacker he opposes the position (expressed by Katz 1966, generative semanticists like Lakoff 1971, Johnson-Laird's procedural semantics etc.) which major premise is as follows:

"The meaning of a word can be exhaustively decomposed into finite set of conditions that are collectively necessary and sufficient to determine the reference of the word."

His argument is:

"But once the marker COLOR is removed from the reading of "red", what is left to decompose further? How can one make sense of redness minus coloration?" (pp.113 in Semantics and Cognition)

Thus, Jackendoff also ends up with basic irreducible not-decomposable cognitive fields. He argues that there are different necessary condition for the field, for the thing in reality and for the projection of this thing. For example, spatial continuity is not a necessary condition for connecting 4 points in a rectangle but #spatial continuity" is a necessary condition for the projection of the #thing# stimuli. His criticism of the feature-based traditional decomposition approach analyzing word-meaning with necessary and sufficient conditions results in a modification of the theory expressed in new types of conditions:

a. necessary conditions - in a hierarchical structure of meaning determination of the superordinate concept is a necessary condition for the subordinate one. Ex.: COLOR is a necessary condition for determining the meaning of "red".

b. typicality conditions - these are conditions which are typical but subject to exceptions and the latter are discrete, not continuous as the centrality conditions. Ex.: There are green leaves but there are also leaves which are not green. Or, typical for Swedes is that they have fair hair, but there are Swedes with red and brown hair etc.

c. centrality conditions - they specify a central value for a continuously variable attribute. Ex.: An argument and an example here may be Berlin and Kay's finding that leave-green is the prototypical green hue of color, which obviously satisfies certain centrality conditions of color (light) intensity.

d. intentional conditions - they are neither necessary nor sufficient. For example, in defining the conditions on the projected (represented in the mind) notion #thing# the intentional conditions are qualities like size, brightness, contrast (when the input is a visual stimuli).

There are also graded judgments which are kind of categorizing judgment and their main characteristic is that they define the categorization of a thing in relation to the context in which it appears. Jackendoff uses the graded judgments in the formulation of the centrality conditions.

The notion of the centrality conditions reminds us of Langacker's "gradation of centrality" (despite Langacker's claim that his theory has nothing to do with that of Jackendoff). I will repeat the quotation in it's new context:

"I see no a priori reason to accept the reality of the semantics/pragmatics dichotomy. Instead, ... gradation of centrality in the specifications constituting our encyclopedic knowledge of an entity.... I adopt an encyclopedic conception of linguistic semantics." (pp.57-58 in A View of Linguistic Semantics)

Another obvious similarity between Jackendoff's and Langacker's ways of determining the meaning of linguistic expressions is that both emphasize the hierarchical order of cognitive semantic structures. For Jackendoff's superordinate concept is a necessary condition for the subordinate one. The same idea is expressed by Langacker and his concept of base and matrix.

"The base of a predication is nothing more than its matrix (or more precisely, those portions of such domains which the predication actually invokes and requires."

(pp.58-59 in A view of Linguistic Semantics)

The base is the knowledge which is pressuposed for the determination of a concepts meaning, and this knowledge is organized hierarchically as the superodinate nodes of a network. "Right triangle" is a superordinate domain of "hypotenuse" and without it is impossible to understand the meaning of the concepts "hypotenuse". In that sense, Langacker's base or matrix is identical to Jackendoff's necessary condition for determination of meaning.

Jackendoff's typicality conditions are expressed in Langacker by his two categorizing relationships of schematicity (which involves modification of information) and extension (which involves change of information).

Jackendoff's intentional conditions are equal to Talmy's categories of notions which are typically not specified by grammatical elements.

However, these are incorporated in the description of the meaning of what Talmy calls grammatical elements in both Jackendoff and Langacker, mainly because they don't make this distinction and because they use interrelated conceptual hierarchies.

4.2. Operational criteria

All the three cognitivists' common premise is that there are universal cognitive operations used for the structuring of knowledge, also linguistic knowledge. It is interesting to see if they end up with the same, similar or different operations. This structuring operations are essentially related to what I will call the creativity hypothesis of human mind which all cognitivist share. Langacker calls this sum of operations and cognitive structures "the Dimension of Imagery", which are the following:

- profiling a profile on a base of a term, ex. "line segment" is the profile on the base "right triangle" of the term "hypotenuse";

- level of specificity, ex. animal->reptile->snake->rettlesnake;

- background assumptions and expectations, the distinction between given and new information;

- secondary activation, f.ex. in creating of metaphors;

- scale and scope of predication, ex. *armnail is impossible because it is not composed according to the expected scope of predication;

- relative salience of a predication's substructures, ex. the salience of "compute" in "computer" lies at the margins of awareness;

- perspective (orientation, vantage point, directionality, objective construction).

Since Langacker does not distinguish between close-class elements and open-class elements, his description of the cognitive image and cognitive operations.But from Talmy's point of view Langacker is describing only lexical-item-senses. In fact, LangackerŐs examples are lexical and morphological in character, that is roots, affixations, compounds, inflections, prepositions, adverbs, particles, nouns, verbs, adjectives, idioms, grammatical categories. However different Langacker's and Talmy's perspectives and frameworks are, one may notice that the phenomena they are describing are very similar. First of all, Talmy organizes all cognitive categories, operations and structures in four Imaging Systems. It is not only the name of this systems that is identical to LangackerŐs dimensions. Exactly as Langacker, Talmy defines those systems as "great complexes in language that organize the structuring and "viewing" of conceptual material" (pp.. 194, in The Relation of Grammar to Cognition)

These Systems are:

1. Structural schematization: this system involves all forms of conceptualization of quantity or relations between quantities, in dimensions like time, space etc. The categories listed here are: dimension, plexity, state of boundedness, state of dividedness, degree of extension, pattern of distribution, partitioning of space and time, axiality, scene-division, geometrical schematization.

2. Deployment of perspective: this system examines how one places one's "mental eye" to look out upon a scene. The categories which belong here are: perspectival mode and degree of extension.

3. Distribution of attention: this system examines the allocation of attention which can be directed differentially over the aspects of the scene. The categories included in this system are: level of synthesis, level of exemplarity, global vs. local scope attention, figure/ground, plus discourse concepts like focus, topic, comment, given and new.

4. Force dynamics: this system involves "the forces that the elements of the scene exert on each other". The categories involved here are not discussed in this article but are said to be: force, resistance to force, overcoming of such resistance, blockage to the exertion of force and the removal of such blockage.

Before we start a more detailed comparison, one must say that the phenomena included in Talmy's last cognitive system are not described by Langacker. Talmy's concept of schematization is not the same in Langacker, but this difference is not relevant. Lets me first describe briefly Talmy's categories and operations for structuring of a linguistic expression's meaning:

1. Structural schematization:

Category

Notions Example Cognitive operations
dimension matter/objects call->give a callrefection
action/events hail->hailed actionalizing
plexity matter

action

bird->birds

breathe->take a breath

multiplexing

uniplexing

state of boundedness bounded

unbounded

(mass/count)

(perfective/

imperfective)

She slept in 8 hrs.

She dressed in 8 hrs.

shrub->shrubbery water->body of water

sleep-> sleep for 1hr.





debounding

bounding

bounding

degree of extension

(aspect)

point

bound

extent

unbounded

speck

climb a ladder in 5m.->-"- at midday

climb->keep climbing


reduction

magnification

pattern of distribution

(aspect)

non-resetable

resetable

full cycle

multiplex

steady-state

gradient

died->is dying (gradient)

fall

flash

breathe

sleep

widen

extension
partitioning (no description)
axiality

directed shift

end-point

sick

well

scene-division

of a lexical item

of a gram.form

serve

(Act, item, guest-host personation)

Subj+Refl.Obj.






monadic personation
geometrical

schematization

(no description available)

2. Deployment of perspective:

Category

perspective
mode
(related to the events basic pattern of distribution)
degree of (no description available)
extension

Notions

steady-state

moving mode

Examples

There are houses at various points.
There is a house every now and then.

3. Distribution of attention:


CategoryNotions Example Cognitive operations
level of synthesis Gestalt (np1)

multiplexity(np2)

a cluster of trees

(np1) (np2)

iceberg->2 halves of iceberg

bounding

form specification

Gestalt formation

componentiation

level of exemplarity

(for multiplexity)

full complement

exemplar

oysters have siphon(s)

an oyster has siphon(s)

global vs. local scope attention (related to the perspective mode))

figure/ground the cluster of trees

the trees in that cluster

(discourse concepts like focus, topic, comment, given and new)

4. Force dynamics:

force, resistance to force, overcoming of such resistance, blockage to the exertion of force the removal of such blockage.

Let us now start the comparison. First of all, Talmy's description and categorization is much more detailed but it is much more difficult to defend the necessity of such categorization. By his concept of profiling, Langacker covers all Talmy's systems. Consequently, the first category "dimension" is explained in Langacker by the distinction between profile and base. What is included in plexity in Langacker is analyzed in terms of levels of specificity, scale of predication and analyzability. Talmy's operations of de/bounding is only one kind of proving and delineation of scale and scope of predication. Talmy's degrees of extension and the operations of reduction and magnification are explained in more general terms by Langackers relative salience of predication's substructures. The same goes for Talmy's patterns of distribution. Talmy's scene-division and the associated operation of monadic personation in reflexive expressions is explained by both authors as relation between trajector and landmark, where for Langacker either the trajector of the landmark is the profiled concept. This category is also described by Langacker in terms of his perspectival dimension of imagery.

Talmy's imaging systems "deployment of perspective" and "distribution of attention" deal with the same phenomena which Langacker includes in his dimension of imagery, called perspective.

The difference is that Talmy's level of synthesis and level of examplarity are also covered by LangackerŐs levels of specificity. Both authors exploit the cognitive distinction between figure and ground. Langacker defines the trajector "as the figure within a relational profile" (pp.76 in A view of Linguistic Semantics). Thus for him the figure/ground notions are much more general. Talmy mentions this distinction as a category in the system of distribution of attention, but the analysis offered by both authors of expressions like the cluster of trees/the trees in that cluster is the same. Discourse concepts like focus, topic, comment, given and new are incorporated in the system of attention distribution in Talmy. Langacker accepts only the distinction given/new as a part of the background assumptions and expectations, which influence our understanding and characterization of meaning. As a result of this brief comparation one is forced to observe that both authors are concentrating on very similar structures and cognitive operations. There are differences, but they are differences of detail and terminology. The overall theoretical body in which these imaging systems and imagery dimensions are incorporated are also different, as I already mentioned, but anyhow the structures are very much the same.

5. Conclusions

Contemporary grammatical cognitivism expressed in the theories of Langacker, Jackendoff and Talmy shares a great deal of assumptions and views concerning the cognitive organization of language.

These common assumptions are:

1. Meaning is conceptualization.

2. There is difference between real world and conceptualized world.

3. There is no direct correspondence between these two worlds.

4. The cognitive theory of language describes only the organization of this conceptualized world.

5. The cognitive operations used of humans to organize and structure linguistic information are the same used to structure non-linguistic information.

6. Human beings have inborn capacity for such internal organization of information which is expressed by these operations.

7. We begin constructing our mental universe of experience registered in basic domains (or primitives, or basic cognitive categories), arriving at ever higher level of conceptual organization by means of innately specified cognitive operations.

8. Polysemy is a fundamental way of meaning creation.

8. The distinction between pragmatics and semantics is negligible.

9. Lexical and encyclopedic meaning are unseparable.

10. There are continuous cognitive spaces and specific cognitive operations in and by which words pick out focal values.

Furthermore, we found out that the categories of cognitive notions

described in one way or another are also similar. The shared concepts are ( I will try to avoid the specific terms used by each of the authors):

boundedness degrees

scales and scopes

perspectives, vantage points, directionality

magnitude

countability

scene-arrangement

type/token

figure/ground

trajector/landmark

part-whole relations

The main difference between Jackendoff, Langacker and Talmy is their treatment of the relation between, on one hand, syntax and semantics and on the other, between grammatical and lexical notions.

References:

Jackendoff, R., What is a Concept that a Person May Grasp it?, in Mind&Language, 4(1-2):68-102,1989.

Jackendoff, R., Semantics and Cognition, MITT, 1983.

Kneale&Kneale, The Development of Logic, col. of. Norman Kretzman on History of Semantics. London:Oxford University Press, Ely House, 1962.

Langacker, R.W., A View of Linguistic Semantics, in Rudzka-Ostyn (ed.)Topics in Cognitive Linguistics, Amsterdam/Philadelphia:John Benjamins, 1988. .

Langacker, R.W., Concept, Image and Symbol, Berlin/N.Y.: Mouton de Gruyter, 1991.

Talmy, L., The Relation of Grammar to Cognition. In Rudzka-Ostyn, B. (ed.) Topics in Cognitive Linguistics.Amstredam:Benjamins. p.165-205,1988.

Talmy, L., Lexicalization patterns: semantic structure in lexical forms, in Timothy Shopen (ed.), Language Typology and Syntactic Description, vol.III, Cambridge: Cambridge UP, 1985.