A Cognitive Theory of Language. Semantic Theory and Analysis.

List of contents

Introduction

PART ONE

1. Linguistic Categorisation
1.1 John R. Taylor: Prototype Theory and Basic Level Terms
1.2 George Lakoff: Radial Categories, Metonymic Models

2. Lakoff’s (et al.) Cognitive Theory of Language
2.1 ICM: Image-Schema, Metonymy and Metaphor

3. Implications for Semantic Analysis

PART TWO

4. Case Study
4.1 Analysis
4.2 The Semantics of Metaphor

5. Conclusion

References

Introduction

This paper will be concerned with cognitive-linguistic approaches to semantics and their practical application to semantic analysis. Part one will review and discuss semantic theory: Cognitive Semanticists base their theories on the assumption that language is part of general cognitive processes rather than an autonomous module independent of general cognition. This assumption entails an approach towards meaning which takes into account world knowledge and language use (what is otherwise distinguished as ”pragmatics”). Evidence for this approach as given for the question of linguistic categorisation by John R. Taylor^[1] will be reviewed, and discussed with reference to George Lakoff’s (et al.)^[2] Cognitive Theory of language. The notion of Conceptual Metaphors as structuring everyday language is developed into a theory of basic metaphoricity of thought, as reflected by its surface expressions in language. In part two I will apply Lakoff and Johnson’s theory of Conceptual Metaphoricity to a case study: I want to discuss and assess the potential, and the limits, of this approach in the analysis of short poetic text.

PART ONE

1. Linguistic Categorisation

”Classical” approaches to categorisation, as called by Taylor (1989, 21 ff.) and Lakoff (1987, 6 ff.), can be characterised by an understanding of linguistic categories as clearly bounded sets of members which have certain essential features in common. These categories are taken to reflect existing categories in the world which are defined by, as Aristotle called it, necessary and sufficient features. These features are binary, i.e. they determine or rule out membership and thereby establish clear boundaries. This understanding of categorisation is manifested in traditional philosophical and scientific thought as well as in ”common sense” theories about language and the world. Lakoff calls this approach ”objectivism” and identifies the following assumptions:

‘OBJECTIVIST METAPHYSICS: All of reality consists of entities, which have fixed properties and relations holding among them at any instant. [...]

OBJECTIVIST ESSENTIALISM: Among the properties that things have, some are essential; that is, they are those properties that make a thing what it is, and without which it would not be that kind of thing. Other properties are accidental – that is, they are properties that things happen to have, not properties that capture the essence of the thing. [...]

THE DOCTRINE OF OBJECTIVE CATEGORIES: The entities in the world form objectively existing categories based on their shared objective properties.’

(Lakoff 1987, 158-161; emphasis in the original)

Cognitive scientists have set out to prove these assumptions wrong. Categorisation, in their view, is an operation of human cognition which is determined by experiences of physiological and physical characteristics of the human body and bodily interactions with the environment. Categorisations and their realisations, however, can provide the basis for human experience once they are firmly established, so that conceptual categories can work back upon human perception. Language categories, they maintain, are reflections of conceptual categories. Thus, they disclaim the autonomy of language which is presupposed by most formalist and generative linguistic theories (cf. Lakoff 1987, ch. 9), as well as a metaphysical notion of reality which is reflected in human perception and categorisation. Rather, reality is made meaning of through these cognitive processes, which in turn are predominantly structured by bodily experiences and interactions. Therefore, cognitive semantics take empirical evidence about language as the basis for understanding human cognition. Prototype theory has developed from empirical studies of linguistic categorisation, and it is taken as evidence for processes of conceptual categorisation.

1.1 John R. Taylor: Prototype Theory and Basic Level Terms

Prototype theory has emerged from experimental findings which suggest that human language categories are not clear cut sets but rather have fuzzy boundaries, showing varying degrees of typicality. Taylor (1989, 40 ff.) introduces the use of ”attributes” instead of (binary) features as defining the properties of entities. Attributes are, very often continuous, characteristics of appearance or function, relating to real-world properties readily accessible through experience:

‘Ultimately, the attributes have to do, not with inherent properties of the object itself, but with the role of the object within a particular culture...’

(Taylor 1989, 41)

They can be functional, interactional, or image-schematic (gestalt perception). Categories are structured by networks of attributes, each member of the category sharing some of them while no attributes have to be shared by all members. Eleanor Rosch^[3] established the notion of degree of category membership, observing that each category has (one or more) prototypical examples which serve as reference points for this category. Those entities only sharing very few of the prototypical attributes are classified as more marginal members of the category, sometimes being ambiguous in their categorisation and blending into other categories. The prototypicality of entities also shows up as a variable in psycholinguistic experiments, with e.g. more prototypical members of a category being processed faster in recognition-tasks (Taylor 1989, 45). Prototypical entities can be identified by their high salience in people’s minds, as shown in their language use. Thus, when people are asked to name ‘a bird’, speakers of English most often say robin, which is one of the most prototypical members of the category. Thus, prototypicality can be shown to have psychological reality. Prototype categories have the evolutionary advantage of being able to flexibly accommodate new, unfamiliar data, and thus to adapt themselves to the ”flux of experience”, an ability classical categories do not have to the same extent.

Another related concept in Rosch’s theoretical framework is that of ”basic level terms”: Categories are arranged in hierarchies of sub- and superordination. Thus, Taylor identifies two axes of categorisation, the vertical axis representing hierarchical levels of categorisation, and the horizontal axis representing contrasting categories which are included in the superordinate category. On the vertical axis, there seems to be a more salient level of categorisation which people are shown to use most commonly in language. At this ”basic level” people conceptualise things as perceptual and functional gestalts (Taylor 1989, 48), e.g. they use chair more commonly than its superordinate furniture or its subordinate lounge chair. These basic level terms are usually morphologically simple, very frequent, and short.^[4] Basic level terms, it is assumed, gain their salience because they most fully and most efficiently exploit the real-world correlations of attributes. As Rosch points out, they

‘(a) maximize the number of attributes shared by members of the category; and
(b) minimize the number of attributes shared with members of other categories.’

(Taylor 1989, 51)

That makes entities on basic level those terms which we most directly manipulate and interact with in everyday experience. Basic level terms, I would conclude, are explicable as a function of language efficiency, providing the means to categorise the world in the most useful way requiring the least energy, and thus, communicating most efficiently.

1.2 George Lakoff: Radial Categories, Metonymic Models

Lakoff (1987) claims that prototype categories are based on the principle of metonymy, where central members can stand for the whole category. As important characteristics of categorisation, Lakoff exemplifies metonymic models such as social stereotypes, typical examples, ideals, paragons, generators, submodels, and salient examples (Lakoff 1987, 85 ff.). These models give rise to prototype effects in that in each of them certain members of the category can, according to different criteria, stand for the whole category. For instance, a social stereotype of the category MOTHER is the HOUSEWIFE, where several reference models for MOTHER coincide (nurturance model, birth model, legal binding model, etc.).

Some categories, according to Lakoff (1987, 153) are scalar (showing degree-of-membership effects), some are classical (having clear cut boundaries and necessary and sufficient conditions), some are metonymic (allowing one central member to stand for the category as a whole), and some are radial. Radial categories are built on the general principles of centrality (prototype categories with central and peripheral members, with basic-level objects at their centre) and chaining, i.e. the motivated extension of a category from its central to non-central members (Lakoff 1987, 95 f.). These links can be motivated by experiential domains (which may be culture specific), Idealised Cognitive Models (ICMs, which will be discussed in ch. 2), and specific knowledge overriding general knowledge. The cognitive mechanisms (Lakoff 1987, 110) involved in such categorisations are those of image-schema transformations, metonymy, and metaphorical mapping, making use of ”conventional mental images” and knowledge about these, as well as knowledge from experiential domains. The corresponding cognitive models (propositional, image-schematic, metonymic and metaphoric; Lakoff 1987, 113/ 114) will be discussed in my next point.

This understanding of categories as radial provides an account for categorisations through associative links. Category members do not need to share essential features, but are connected by motivated chains of meaning. This theory of categorisation can show how such links are motivated, however, it has no claim to predictive (generative) value for categorisation processes. In my next chapter I will discuss Lakoff’s (et al.) Cognitive Theory of Language in more detail.

2. Lakoff’s (et al.) Cognitive Theory of Language

This theory is based on the following general assumptions: Language is part of the general cognitive apparatus, it reflects the structure and functioning of cognitive models and conceptualisation. Cognitive models, in turn, are embodied, i.e. they are based either directly or indirectly on bodily or social experience. This provides, as Lakoff (1987, 154) emphasises, a non-arbitrary link between cognition and experience, but it does not mean that conceptualisations are necessarily predictable from experience. Rather, they can be shown to be motivated along systematic lines based on embodied cognitive models. This approach suggests that we perceive and make sense of the world in terms of cognitive processes, i.e. it means a total break with objectivism and with the autonomy of language.

2.1 ICM: Image-Schema, Metonymy and Metaphor

Underlying Lakoff’s (et al.) theory are three principles of conceptualisation, called ”types of cognitive models.” Lakoff uses the term ”model” for various different theoretical concepts which makes it difficult to make use of this term in a systematic way. I will, therefore, try to paraphrase these concepts without always using the original terminology. For a revised version of this theory, I will refer to Lakoff 1992, an article I found on the internet.

Lakoff (1987, 68) proposes that our knowledge is organised by means of Idealised Cognitive Models (ICMs). He draws for this theory on Fillmore’s frame semantics, Lakoff/ Johnson’s theory of metaphor and metonymy, Langacker’s cognitive grammar, as well as Fauconnier’s theory of mental spaces. Lakoff’s concept of ICMs is designed to provide reference models for concepts relative to which they can be made meaningful. For instance, the days of the week can only be understood relative to a complex cognitive model which understands TIME as defined by the movement of the sun, differentiating day and night, and puts the days in a linear sequence, the week. This model, Lakoff maintains, is idealised in that it imposes a cognitive structure on ”nature”, and such models can be culture specific. Each ICM is a complex structured whole, a gestalt, based on the following kinds of structuring principles: Image-schematic transformations, metaphoric, and metonymic mappings (cf. Lakoff 1987, 113/ 114). In his later outline of the theory Lakoff (1992) states an Invariance Principle which governs all of these operations.

‘The studies of basic-level categorization suggest that our experience is structured preconceptually at that level. We have general capacities for dealing with part-whole structure in real world objects via gestalt perception, motor movement, and the formation of rich mental images. These impose a preconceptual structure on our experience. Our basic level concepts correspond to that preconceptual structure and are understood directly in terms of it.’

(Lakoff 1987, 270; emphasis in the original)

[...]

^[1] Taylor (1989)

^[2] Lakoff/ Johnson 1980, Lakoff 1987, 1989, 1990, 1992

^[3] Cf. Lakoff 1987, 39-47

^[4] In addition, in German basic level terms are gender marked whereas superordinates are usually neuter. Subordinates (most often compounds including the basic level term) inherit the gender marking from the basic level.