CHAPTER VI

METAPHOR

 

 

Introduction.

 

A standard definition of metaphor is a "a figure of speech in which a word or phrase denoting one kind of object or idea is used in place of another in order to suggest similarity between them." It, always, involves comparison between two things. A version of this is the simile theory - that metaphors are abbreviations of similes. Example, "My lover is a treasure." Another version is the semantic interaction theory - that two semantic components, one intended literally and one intended non-literally, interact with each other. For example, referring to two arrogant people who came to an agreement, someone might say, "Prima donnas embraced."

 

In semiotic terms a metaphor is a dynamic, as opposed to stable, sign, and this follows the etymology of the word, which suggests a transfer or displacement of names. We can begin then by describing metaphor as the act or process of denoting one concept (the tenor) with a sign conventionally tied to another (the vehicle). The purposes of doing this are: (i) to emphasize certain associations of the tenor over others (my dentist is a barbarian); (ii) to enrich the conceptual structure of the tenor by analogy with another domain (e.g. the CPU is the brain of the computer); (iii) to convey some aspect of the tenor which defies conventional lexicalization (e.g. the leg of the chair, the neck of the bottle), a process which is called catachresis by Black (1962).

 

Donald Davidson, in his "What Metaphors Mean," elaborates the semantic treatment of metaphor which is consistent with his general theory of meaning. It is critical to the traditional views, simile theory and Max Black's interaction theory.

 

In his own words: " The central mistake against which I shall be inveighing is the idea that a metaphor has, in addition to its literal sense or meaning, another sense or meaning. This idea is common to many who have written about metaphor; it is found in the works of literary critics like Richards, Empson, and Winters; philosophers from Aristotle to Max Black; psychologist from Freud and earlier, to Skinner and later; linguists from Plato to Weinreich and George Lakoff." (Philosophy of Language, p. 416)

 

For Davidson, the concept of metaphor as primarily a vehicle for conveying ideas is as wring as the parent idea that a metaphor has a special meaning. He agrees that metaphors cannot be paraphrased, but this is not because metaphors say something novel for literal expression, but because there is nothing there to paraphrase. Metaphor is a legitimate device not only in literature but in science, philosophy, and the law. It is effective in praise and abuse, prayer and promotion, description and prescription. Metaphor belongs exclusively to the domain of use. It is something brought off by the imaginative employment of words and sentences and depends entirely on the ordinary meanings of those words and hence on the ordinary meanings of the sentences they comprise.

 

Therefore, for him, the theorist who attempts to explicate metaphor by appealing to a hidden message, like a critic who tries to state the message, is fundamentally confused. There is no explanation that can be expected because there is no message that exists at all.

 

A.P. Martinich, in his "A Theory of Metaphor", elaborates his treatment on metaphor within Grice's pragmatic theory of conversation and Searle's theory of speech acts. Metaphors are similar to indirect speech acts in that both are conversationally implied due to the non-fulfillment of conversational maxims. Without abandoning a pragmatic approach, Martinich agrees with Davidson that "understanding a metaphor is as much a creative endeavor as making a metaphor, and as little guided by rules." An audience interprets what a speaker means by trying to incorporate an understanding of the speaker's beliefs and behavior (linguistic and nonlinguistic) in the simplest, most plausible, least disruptive way. In other words, the audience attempts to achieve a cognitive equilibrium when confronted with any situation.

 

A       -        TWO MAIN FIELDS IN METAPHOR RESEARCH:

(LITERAL & FIGURATIVE)

 

I The Literalists

The literalists are holders of the classical view of metaphor.

 

(i)                 Rhetoric Claim: Metaphor is the rhetorical icing on the cake of language, an expressive device that adds colour and emotive tension to one's writing, while supporting concise expression and clever forms of emphasis and focus. However, while metaphor may be thought-provoking to the reader, it is nevertheless a second-class tool of language. Informative metaphors may change our contingent knowledge about particular concepts (e.g., billboards are ugly, surgeons are over-paid) but cannot fundamentally alter the way we structure certain concepts (e.g., argument as war, science as magic, sex as politics etc.) and thereby structure new experiences (in the sense that a Sex-as-Politics metaphor might actually change the way we view our secretary's actions).

 

 (ii) Peripherality Claim: That the act of metaphor interpretation and generation is not central to language or cognition. It does not arise naturally out of a coherent model of language, rather it is oblique to the natural processes of language and modelled by a peripheral process of mind. A peripheral theory of metaphor describes how a figurative utterance is effectively decomposed and spirited away to achieve a literal rendition of the meaning contained therein. No account is given of how metaphor can alter the underlying conceptual representation of the hearer. This thesis is commonly referred to the Substitution View of metaphor (from Black 1962), inasmuch as it claims that metaphor interpretation is a process whereby a figurative statement is replaced by its literal counterpart before true comprehension occurs.

 

 (iii) Canonical Meaning Claim: Often referred to as the Language of Thought hypothesis (see Fodor 1975), this is a reductionist thesis which claims there exists a symbolic and atomistic level of mental representation into which all possible speaker meanings may be encoded. Furthermore, it is the literalist position that not only is this level of representation essentially literal, it is necessarily so, for this is the gold standard against which all meaning is valued. That is to say, the atomic elements of the canonical substrate stand as proxies for certain external meanings.

 

 (iv) Principle Of Expressibility Claim: As a corollary to the hypothesis of canonical meaning, this thesis claims that any meaning which is expressible in such a mental representation is also expressible in a full natural language. That is to say, one does not necessarily have to resort to tropes such as metaphor to express certain meanings, though it may be convenient to do so for the reasons of rhetoric discussed earlier. This thesis thus removes any cognitive necessity for the metaphor phenomenon.

 

 (v) Falsehood Claim: If we accept the principle of expressibility, the thesis that a speaker does not have to resort to metaphor for expressive purposes but may use literal language to achieve the same descriptive ends, it follows that a speaker that employs metaphor is engaging in a socially-licensed lie (or less pejoratively, exaggeration). In this view, figurality operates much like other deceptive tropes such as irony and sarcasm, inasmuch as the speaker deliberately says one thing while meaning something else. The metaphor interpretation process of the hearer must thus recognize when the speaker is engaging in such linguistic deception, and reconstruct, puzzle-like, his intended meaning.

 

 (vi) Anomaly Claim: The use of metaphor in language is signalled by a violation of the constraints (or preferences) of semantic orthodoxy. A literal reading of a figurative statement produces an anomalous meaning structure, which in turn triggers a special metaphor processing module to take over the comprehension of the utterance. The interpretation of metaphors is thus a two stage process of initial anomaly detection and subsequent repair, whereby an alternative meaning representation is found to express the content of the metaphor. This divergence of processing demands between literal and figurative utterances thus predicts that figurative statements should require more comprehension time than their literal counterparts. However, empirical testing of this prediction supports no such divergence of processing, thus discrediting the anomaly view as a cognitively plausible model of metaphor (Gerrig 1989; Hoffman & Kemper 1987).

 

 (vii) Classical Categorisation Claim: This thesis claims that classical/logical models of concept structure, such as the intensional models of type hierarchies and necessary and sufficient truth conditions, and the extensional set-based models of Fregean language analysis, adequately meet the representational demands of the metaphor phenomenon. In particular, this thesis claims that the type hierarchy assumes a central role in the analysis of metaphor, moreso than another form of knowledge, such as associational, connotational, affective, reflective and collocative meaning (see Leech 1990 for a discussion of these various types of meaning).

 

 (viii) Dictionary Versus Encyclopaedia Claim: How much knowledge is required to adequately interpret a metaphor, and where does this knowledge reside? It is a hallmark of many literalist approaches that word meaning (i.e., lexical semantics) provides sufficient knowledge for metaphor comprehension, while recourse to contingent world knowledge (such as `who is currently president of America?', and `What military tendencies does Russia exhibit?') is deemed unnecessary. In this view, metaphor operates in the domain of analytic semantic definitions rather than in the realm of semantic memory. It follows that legitimate sources of word meaning, such as the dictionary, become the knowledge substrate of these approaches and the touchstone of literal meaning. The use of `dictionary' must be distinguished here from `encyclopaedia', which contains much more than definitional criteria (such as chairs are made for sitting), actually specifying contingent (synthetic) facts regarding the usage and typical contextual settings of a concept (e.g., chairs are usually made of wood, almost always have four legs and are sometimes upholstered in leather). The semiotic relevance of this distinction is discussed at length in Eco (1984).

 

II The Figuralists

 

(i)                 Naturalness Claim: Far from representing a semantic anomaly, or a deviant meaning which proves unpalatable to some deep-seated literal substrate, metaphor is a natural element of language, which calls for no special treatment or dedicated cognitive process. Rather, the processes which are involved in the generation and comprehension of metaphor arise naturally out of the workings of the cognitive system as a whole. A bolder expression of this thesis goes further to claim that metaphor is the very currency of thought, and that literal meaning, far from being a solid bedrock of absolute meaning, is merely a synchronic artefact of a changing metaphoric system.

(ii)               Centrality Claim: An extension of the naturalness position, this thesis claims that metaphor is central to the workings of both our language and general cognitive faculties. Metaphor is at the root of our creative powers, serving a cognitive function that is irreducible and irreplaceable. This position thus represents the antithesis of the substitution view of metaphor discussed earlier under Literalist Peripherality, denying that metaphor can be spirited away and shoe-horned into a literal meaning representation. Furthermore, this position entails that metaphor is one of the fundamental conceptual combinators that mediates in the construction of new conceptual structures.

 

(iii)             All Language is Metaphoric Claim: In denying the validity of the literalist position, both in holding that there is no absolute substrate of literal meaning in the mind, and in allowing metaphor to occur at all levels of our conceptual representations, are we in effect claiming that all language is metaphorical? For to do so would effectively deny the possibility of ever uttering a truly literal statement. But clearly we can -- the sentence `The box is on the table' is as literal a statement of fact as one is ever likely to express. And so, a simple empirical demonstration would appear to terminate this line of inquiry.

 

Or does it; can we be absolutely certain that this statement is entirely literal? The component words do seem to be used in their truest and most ordinary senses -- `box' signifies just a box, `table' just a table. However, just what kind of boxes and tables do we conceptualize when understanding this sentence? I myself think of a cardboard cubic box about a foot long in all directions (that is, an easily carried box), and a varnished wooden kitchen table with flowers in the centre, and surrounded by chairs. More precisely, I imagine a table big enough and strong enough to hold the box I am conceptualizing. Additionally, this instantiation establishes a kitchen setting, which prompts me to conceptualize the room in which the box and table are located. In effect, we bring our own prototypical notions of boxes and tables to the interpretation of the utterance. But we may never have sensorially experienced such a prototypical box sitting on such a prototypical table; nevertheless, we can infer that the situation is physically plausible by remembering actual experiences of something of similar weight and size sitting on a similar table. And so, even the comprehension of an overtly literal statement involves a form of conceptual re-interpretation, in which a novel scenario is viewed through the lens of past experience. The mechanics of interpretation involve an initial construction phase wherein prototypes are evoked and assembled according to the dictates of the utterance; this instantiated scenario is then compared to an analogous situation from actual experience to determine whether it is a feasible interpretation. By this reckoning, even absolutely literal statements require some form of deep analogical reasoning, the same reasoning processes which, when more explicitly required, are the trademark of figurality.

 

The ability of a word to specify a particular concept, when used in its `ordinary' sense, should not therefore be taken as a basis of absolute literal meaning (as MacCormac 1985 advocates). For such concepts are in turn modelled via specific prototypes, drawn from the hearer's subjective experience of the world. No matter how true a statement is to the ordinary consensus meaning of a word (as defined in a dictionary, for instance), there will always exist an element of subjectivity which is only overcome by grounding in past experience. If one accepts that the mechanics of this grounding are the same as the grounding required during the interpretation of metaphor (wherein one domain becomes grounded in another), then one must conclude that all language is metaphoric.

 

Note that this thesis is quite different from the claim that all knowledge is metaphoric, discussed in Indurkhya (1992). Clearly, the theory outlined above is dependent on some component of our knowledge --- that component which models our past experience --- to be non-metaphorical, if it is to serve as a grounding for higher-level metaphors. Nevertheless, this should not be taken as a sop to the literalist position, because the inherently subjective nature of personal experience denies that such a grounding could ever provide an absolute representation of literal meaning. This point is discussed further under the thesis of Figural Empiricism.

 

(iv)             Figural Rationalism Claim: Rationalism is the philosophical enterprise which questions not how our knowledge conforms to the objects and events of the real world, but rather how these events and objects conform to our existing patterns of knowledge. From a rationalist perspective then, metaphor is not only a vital element of human communication, serving those conceptual domains in which our lexicons prove impoverished, it is also an architect of conceptual structure, organizing our memories, shaping our experiences and interpreting our senses. That is to say, each metaphor is an active cognitive agency rather than a once-off burst of linguistic ingenuity. For instance, Lakoff & Johnson (1980) claim that the Argument-is-War metaphor is deeply ingrained in western culture, and actively colours the way we act with other people. In the words of Kant, our metaphors comprise the conceptual spectacles through which we view the world.

 

(v)               Figural Empiricism Claim: Metaphor allows us to view one concept through the lens of another, and thereby structure and understand one domain in terms of another. However, if we deny the existence of a literal substrate, and allow metaphor to operate at all levels of conceptual representation, some metaphors must therefore be grounded external to this conceptual system. That is, if we to avoid infinite regress (metaphors all the way down), some metaphors must relate conceptual structures to sensory experience of the world. Though this philosophical enterprise of empiricism is antithetical to that of rationalism, inasmuch as it stresses the role of external reality and our experiences thereof in the shaping of our conceptual structures, the two enterprises are wed harmoniously in the phenomenon of metaphor. Metaphor drives a dynamic cognitive system, in which sensory experience is interpreted via existing metaphors (a rationalist process), and novel metaphors are created to meet the representational demands of new sensory experience (its empirical counterpart). Unsurprisingly, space -- as one of the dominant empiricist forces in our development -- claims the lion's share of empirically-driven metaphors. The localist thesis (see Lyons 1977) claims that much of our understanding of metaphysical abstractions such as time, emotion, and inter-personal relationships are grounded in our metaphors of space. For instance, Lakoff & Johnson (1980) and Veale & Keane (1992a.b) outline a variety of highly productive spatial metaphors which are shown to underlie a host of abstractions, such as health, marriage, divorce, kinship terms and corporate relations.

 

Harnad (1980) explains this experiential dependency without ever having to leave the confines of mental representation, by exploiting the essential duality of such representations. He posits (uncontroversially) two main levels of representation in the mind, the continuous and the discrete, the former comprising the distributed sense impressions imparted by physical experience, the latter being the inducted, symbolic structures which model this experience at a higher level. Metaphor is argued to be the bridge between these two representational forms; for the purposes of a symbolic account of metaphor, the continuous substrate can be viewed as a mental proxy of external reality, as it impinges upon the nervous system of a situated cognitive agent. In a similar vein, Indurkhya (1991) exploits a duality between Sensory Data Sets (SMDs) and the concept network to achieve a comparable explanation of metaphor grounding. Nevertheless, an admission that metaphors are ultimately grounded in a lower-level representation should offer little comfort to proponents of the literalist view. For this continuous representation, forming as it does a bridge between the high-level symbolic structures of the mind and the low-level sensory stimuli of the nervous system, is by no means an absolute or objective substrate of meaning. Indeed, we should expect that it is a highly subjective internalisation of an agent's own external experience, for is it not the posited role of the higher-level symbolic representation to discretize and rationalise these continuities in a manner which can be objectively communicated by language?

 

 (vi) Encyclopaedia versus Dictionary Claim: Literal meaning, by its very role as an absolute measure of semantic correctness, is inherently definitional and inflexible. To allow any flexibility would be to allow varying degrees of literality, thereby advocating a continuous rather than discrete division between literal and figurative expression, and denying language a solid bedrock of meaning. If literality is modelled on a continuum, not only does the situation arise wherein some literal statements are more literal than others, but the complementary situation is also possible, wherein some literal statements are more figurative than others. Recalling our earlier discussion of the literalist canonicalist position, such a move would seem to undermine the epistemological integrity and philosophical motivation for positing a literal substrate in the first place.

What this means, of course, is that only a rigidly definitional source of knowledge may act as a suitable basis, or guarantor, for this literal substrate. The dictionary, for instance, informs us about what each word means; if there is an ambiguity as to the meaning of a particular word, this polysem is itself divided into a number of contextually unambiguous readings. In this sense the contents of the dictionary are analytic, for unlike an encyclopaedia, the ideal dictionary explains in terms of necessary and sufficient conditions, rather than communicating by example. For instance, the dictionary entry of `bachelor' might define the concept in terms of the concepts Man and Unmarried. In contrast, the encyclopaedia should offer a wealth of diverse information on the topic, providing examples (e.g., Hugh Heffner in the seventies, Richard Branson now, etc.), while situating the concept in its correct social setting (western society, dinner parties, singles bars, etc.). The latter is essentially a representation of a collective semantic memory, the former a sterile distillation of semantic abstractions. The literalist dictionary vs. encyclopaedia camp (typified by MacCormac 1985) contend the former is sufficient for metaphor interpretation, while the figuralist encyclopaedia vs. dictionary camp hold that the latter is essential.

Major Proponents: Umberto Eco

 

(vii) Systematicity Claim: A metaphor is not simply an isolated burst of linguistic ingenuity, the colourful exclamation of a novel perspective, but a glimpse into the underlying conceptual structure of a domain. Metaphors are deep-seated in our cognitive systems, and thus, exploiting the full generative capacity of language (as championed by Chomsky 1957) manifest themselves at the lexical surface in a myriad of different, but coherent and systematic, guises. A metaphor is a system of concepts, a many-faceted productive schema which offers a host of related perspectives on the same domain. The Sex-as- Politics metaphor system, for instance, not only views gender relations as a form of political arena in which sexuality sets the agenda for discussion, but additionally views the genders as opposing political parties, with different social manifestos and terms of power. Within this system, to hold an opinion on sexuality is to express a voter preference, while the traditional rhetoric of politics becomes the language of gender awareness: women are seen as the oppressed masses, and are called upon to let their voices be heard and claim their rightful place in society. In this way, a single metaphor can motivate an entire social movement, creating a host of different analog bindings (such as gender: party, women: oppressed masses). The metaphor is so entrenched that many who operate within it may not believe it to be a metaphor at all, but a literal description of fact (women are oppressed by men!!). This additionally demonstrates the subjective (and therefore, hardly absolute) nature of the literal/figurative divide.

 

 (viii) Learning Claim: Metaphor is more than an expressive device for communicating facts about an obscure domain, but a cognitive tool whereby our conceptual structures are developed and consolidated. That is, metaphor allows us not only to talk about one domain in the vocabulary of another, but also to understand the causal mechanics of a weakly modelled domain in terms of those of another, more fleshed-out domain. Given this hypothesis, it is fruitful to consider the workings of metaphor at the conceptual level independently of its use as a linguistic device.

A learning theory of metaphor must demonstrate how the comprehension of a figurative statement (or concept combination, in a non- linguistic setting) can alter the representation of a domain in a deep and enriching manner. For instance, Ortony's theory of salience imbalance (Ortony 1979) might be described as a learning theory, inasmuch as it allows for a process of predicate promotion, whereby the salience of a predication in the metaphorized subject is raised as a result of the metaphor. For instance, the metaphor `Motorways are snakes' might raise the salience of the properties Twisting and Dangerous in the Motorway schema. However, it is questionable as to whether this promotion, which modifies existing predications rather than adding new ones, has achieved true learning and enriched the domain of Motorway transport. The re- appreciation of previous knowledge in new ways (`I never really thought of motorways as deadly') may be qualitatively different to the acquisition of new knowledge (e.g., the metaphor `Motorways are railroads for cars' conveys that motorists are seriously limited about where they can travel on a motorway, at what speeds they may drive, and at what points it is permissible to get on and off). In contrast, the analogical domain transfer theories of Winston, Carbonell, Gentner, Holyoak & Thagard provide mechanisms by which new conceptual material may be projected into novel domains.

 

 (ix) Non-Classical Categorisation Claim: A figuralist re-appraisal of the nature of the literal meaning, one which decides that no such absolute representation is feasible, effectively challenges the ability of classical models of category/concept structure to describe the metaphor phenomenon. The umbrella term `classical model' is used here to designate those representational strategies traditionally employed by the Literalist camp, such as the omnicompetent type hierarchies of Aristotle and Way, the necessary and sufficient truth conditionals of Searle, and the markerese of Russell and Aarts & Calbert. Two literalist exceptions to this classical trend are MacCormac (1985), who proposes a continuum model of metaphor based on the mathematics of fuzzy logic (while nevertheless quantizing this continuum into well defined areas of true literality and figurality), and Wilks (1978), who weakens the notion of hard semantic constraints to that of soft semantic preferences which can accommodate anomalous metaphoric constructs.

This thesis of non-classical categorisation is a correlate of the Encyclopaedia vs. Dictionary debate, which claims that the traditional dictionary structure, modelled as it is upon classical models of categorisation which limit its world view, is insufficient for the purposes of metaphor analysis, which needs full access to all the contingent nuances of everyday life. However, non-classical models of category structure such as those championed by Lakoff (1987) gel nicely with the representational demands of an encyclopaedia - categories are both organized relative to prototypes -- examples par excellence - of those categories, and specified relative to their idealised contextual settings. This allows for a more natural and altogether more flexible model of category structure, one that is required if the essential action of metaphor -- the stretching of established conventions - is to be adequately modelled.

 

 

 

 

B       -        THE PHILOSOPHERS ON METAPHORS:

 

1.         ARISTOTLE.

 

The field of metaphor research is one that has its roots in antiquity, springing from the work of the philosopher-scientist Aristotle (circa 300 B.C.), who offers us the first theory on the workings of metaphor in The Poetics:

"Metaphor is the application to one thing of the name belonging to another. We may apply (a) the name of a genus to one of its species, or (b) the name of one species to its genus, or (c) the name of one species to another of the same genus, or (d) the transfer may be based on a proportion." (this translation from Hutton 1992)

 

It is a concise definition. It attempts to tie all the strands of metaphor into a coherent whole. The use of the words genus and species notifies us immediately that we are about to encounter a theory which is strongly rooted in the type hierarchy. This tradition which is still very much alive today in the computational treatment of metaphor. Metaphor is seen here as a conceptual phenomenon arising from the manipulation of the type hierarchy: in generating a metaphor, which is defined here as the transfer of one term onto another, the speaker may (a) move up the hierarchy to a find more general term; (b) move down the hierarchy to find a more specific term; (c) move across the hierarchy to find a sibling term of the same specificity; or (d), use the classic proportional analogy A/C = B/D, in which the term combination A:D is allowed to stand for term C, and the combination B:C for the term A.

 

For each of the above cases of metaphor which Aristotle characterises, he offers the following examples:

(a) genus to species: `Here stands my ship', in which `to be at anchor' is a specialisation (species) of the more generic term `to stand still'.

(b) species to genus: `Truly ten thousand noble deeds hath Odysseus done', in which `ten thousand deeds' is a specialisation of the more generic term `large number'.

(c) species to species: `Drawing off the life with bronze', and `Cutting off the water with unwearied bronze', in which `to draw off' and `to cut off' are each a sibling species of the genus term `to take away'.

(d) proportional analogy: `The wine cup is to Dionysus as the shield is to Ares', in which the analogy allows the term combination `cup of Ares' to replace `shield', and the combination `shield of Dionysus' to replace `cup'.

 

There are a number of obvious problems with Aristotle's treatment of metaphor. Firstly, Aristotle places a strong emphasis on the role of the type hierarchy in metaphor production and interpretation, which results in altogether unconvincing metaphor derivations that appear overly convenient and contrived. In effect, Aristotle shifts the burden of metaphor interpretation onto an omnicompetent type hierarchy, where he expects that every metaphor can be resolved in terms of a single ontological traversal. This, however, adheres to the classicist perspective on the representation of knowledge, and Aristotle cannot be unduly faulted for an approach that has survived in philosophy for two millennia, and which fails to incorporate the recent upheavals in categorisation theory as heralded in the work of Rosch & Mervis (1975), and Lakoff (1987).

 

Secondly, Aristotle's theory essentially advocates a substitutionist view of metaphor, inasmuch as one term stands for, or replaces another. Metaphor is not seen in his framework to be an essential feature of human communication, rather it is described in the Poetics as a formula for achieving more colourful expression; given a particular term which one wishes to use, Aristotle's definition above provides four different ways in which that term can be replaced with another.

 

Thirdly, because metaphors are resolved as ontological manipulations, they have no representational status in themselves, and thus they cannot assume the role of active conceptual agents in the structuring of other concepts; for instance, there is no room in Aristotle's theory for the systematic families of metaphors (such as the spatial metaphors) which underlie much of our common-sense understanding of the world (see Lakoff & Johnson 1980; Veale & Keane 1992a,b).

 

Fourthly, because metaphors have no explicit representation in Aristotle's theory - they essentially exist in the rules of knowledge use, rather than in the knowledge itself - many of the ontological relations which comprise the type hierarchy will themselves be metaphorical. For example, just as `to stand at anchor' is a species of the genus term `to stand still', then `to feel ill' and `to feel unhappy' must each be a species of the genus term `to feel down', which of course is an instance of the orientation metaphor. Thus Aristotle's theory is fundamentally confused in the division it enforces between knowledge representation and knowledge use.

 

However, Eco (1981) acknowledges that Aristotle should be praised for treating metaphor as a necessary cognitive function; in this respect, his is a precursor to the romantic view of Rousseau (see De Man, 1973), and Nietzsche. He elucidates this position in his Rhetoric:

`Accordingly, it is metaphor that is in the highest degree instructive [...]. It follows, then, that for style and reasoning alike, that in order to be lively they must give us rapid information. [...] What we like are those that convey information as fast as they are stated - so long as we did not have the knowledge in advance - or that our minds lag only a little behind. With the latter two kinds there is some process of learning.' (translation from Eco 1981).

 

This description elevates metaphor beyond the level of sterile comparison, to which it is seemingly relegated in the Poetics, and into the realm of learning, where it can be seen to serve a necessary cognitive function. Thus we obtain a somewhat conflicting view of metaphor from the Poetics and the Rhetoric. On one hand, Aristotle's grants high cognitive status to metaphor in claiming that it is not an empty word game, but a vital (lively) process of communication and learning, but on the other, he undermines this elevated status by describing the metaphoric process in overly simplistic ontological terms.

 

2.                FRIEDRICK NIETZSCHE.

 

What does Nietzsche mean by the term metaphor? Here, Nietzsche considers all conceptualizing to be metaphoric -- it is an approximation, inexact, a convenient lie. In fact, Nietzsche writes that "we believe that we know something about the things themselves when we talk of trees, colors, snow, and flowers; and yet we possess nothing but metaphors for things -- metaphors which correspond in no way to the original entities."

 

Thus, the concept (i.e. a concept of a tree) is found to be an abstraction three powers away from the original sense data, or nerve stimulus. This fundamental metaphysical position can be described via the three metamorphoses of the original stimulus: first, the nerve stimulus is transformed into image (thus leaving dreams as originary writing?). Second, the images become sounds, or words. Language, which is so pervasive in our human existence, is thus the second level of metaphor. Finally, there is the transformation from the sonic realm back to the conscious, as the sound/word becomes the concept. Thus, there is no proper concept, or direct correspondence, but only figurative, metaphoric conceptualizing.

 

This is a non-Aristotelian notion of what a concept is, whereas for Aristotle metaphor refers to the concept itself. In this view, metaphor is dependent on the concept for material to utilize in the metaphoric event, i.e. the word-play depending on the concept for its basis. Nietzsche reverses this by insisting that the concept refers to metaphor -- seeing as all we have are metaphors in this human existence, there is no escaping this conclusion.

 

Therefore, Aristotle's definition of metaphor could not be retained as such by Nietzsche since it is based on a division of the world into well-defined genera and species corresponding to essences. Whereas for Nietzsche, the essence of things is enigmatic, so genera and species are themselves but human, all-too-human metaphors.

 

However, one must not make the mistake of privileging metaphor at the expense of concept. To recognize metaphor solely falls into the same sorts of traps that does the privileging of concept -- one forgets that metaphor is only metaphor, abstraction, a convenient lie, and metaphysical posturing returns. One must realize that metaphor is encased within concept, and that we have forgotten that concepts consist of varying metaphoric constructs.

 

So, in discarding traditional, correspondence epistemology and by uttering the now-famous "God is dead,' Nietzsche plunges us headlong into a infinite system of metaphor. There is no truth per se, only illusion; there is no knowledge, but instead the knowledge that a concept has no greater value than a metaphor and is itself a condensate of metaphors. Yet we still believe in the fictions that concepts create, i.e. the self, will, the 'thing.' How does/has this happened? Here we find ourselves moving into perhaps Nietzsche's greatest contribution to the Western philosophic tradition (in terms of sheer utility) -- the genealogical method, and his historical analysis of how metaphors come to be 'mummified' into concepts, thus enforcing the forgetting of the true nature of the metaphor.

 

The treatment of metaphor offered by Nietzsche is itself rich in metaphor. We must therefore look to Nietzsche as a source of philosophical insight rather than computational constraints on metaphoric interpretation. The most useful of such insights is to be found in his treatment of the literal/figurative divide, where he discusses the role and origins of metaphor in language:

`There is no `real' expression and no real knowing apart from metaphor. but deception on this point remains [...] The most accustomed metaphors, the usual ones, now pass for truths and as standards for measuring the rarer ones. The only intrinsic difference here is the difference between custom and novelty, frequency and rarity. Knowing is nothing but working with the favourite metaphors, and imitating which is no longer felt to be an imitation.'

 

Therefore, Nietzsche is much bolder than Aristotle in his claims of cognitive centrality for metaphor - metaphor is seen here not just as one facet of the human learning process, but as the basis of all knowledge, of all cognition. Human knowledge is ultimately metaphorical, a truth about which we deceive ourselves by holding to such false ideals as literal meaning. The difference between the literal and the figurative is not as that between true and false, but between old and new.

 

Language is pictured by Nietzsche as forming a constant flux of conceptual accommodation, whereby novel metaphors enter the language and gradually become established, worn to the extremes of conventionality; but in doing, old metaphors provide a framework around which new metaphors can be elaborated. This view is again expounded by Nietzsche using the metaphor of monetary currency:

`What is truth? A moving army of metaphors, metonymies and anthropomorphisms, in short a summa of human relationships that are being poetically and rhetorically sublimated, transposed, and beautified until, after long and repeated use, a people considers them as solid, canonical, and unavoidable. Truths are illusions whose illusionary nature has been forgotten, metaphors that have been used up and have lost their imprint and that now operate as mere metal, no longer as coins.' (from `Uber Wahrheit und Luge im aussermoralischen Sinn', translated by Culler 1981).

 

This notion of metaphor as linguistic coinage is much more than colourful rhetoric, it is particularly apt, from the perspective of the Sapir- Whorf view of language forwarded a number of years after Nietzsche (see Whorf 1950). In this view, language shapes the thought processes of a culture, comprising a social contract by which speakers agree to adhere, in effect holding the members of a society to use the same words and to make the same conceptual discriminations. A powerful metaphor carries more value in this linguistic economy, as value is ultimately measured in terms of descriptive utility.

 

If Nietzsche does not offer a treatment of metaphor amenable to algorithmic expression, what then does he offer? It might be said that Nietzsche has a vested interest in exalting the role of metaphor in thought, acting as he does as a philologist, as a philosopher, but foremost as a rhetorical stylist. However, Nietzsche recognises the central role of metaphor in thought, and the necessity of understanding the roots of a philosopher's metaphors before one can truly understand his arguments. To use Kant's metaphor, one must always be aware of the conceptual spectacles worn by an opponent, and the attendant possibility of conceptual parallax that can confound an argument. Thus a theory of metaphor is essential to a theory of argumentation, and to philosophy itself.

 

3.                JOHN SEARLE.

 

Searle (1979) offers a treatment of metaphor from the perspective of speech act theory, in which a statement possesses both a Speaker Utterance Meaning (SUM) and a Literal Sentence Meaning (LSM). In this model, the SUM of a statement is that meaning the speaker wishes to convey to the hearer, while the LSM is that meaning to be found by analysing the truth conditions of the sentence, independent of the speaker's intentionality. A statement is thus `literal' whenever the SUM and LSM are the same, i.e., the speaker both means what he says and says what he means. A metaphorical utterance, however, as with other rhetorical devices such as irony and hyperbolae, exhibits a rift between SUM and LSM, which is reconciled by the hearer via the construction of a figurative interpretation.

 

When adhering to the social contract that underlies a conversation, the speaker will normally endeavour to minimise the conceptual distance between LSM and SUM, inasmuch as he wishes to be understood by the hearer (this consideration is embodied in the Gricean `Principle of Co- operation', discussed in Coulthard 1985). From this speech act perspective, literal truth can be viewed as a form of semantic `honesty' or `plainness', that is, the hearer believes the speaker to mean what he says, and thus to say what he means in as co-operative a fashion as possible. Consider, for example, the statement `My friend is walking on air', which supports both a literal and metaphoric interpretation, the latter being preferable in a null context. Note, however, the statement contains no semantic anomaly that would give preference to a figurative reading - one could contrive a technological context in which one's friend really was walking on air, or a fictional context, in which one's friend was Superman, Peter Pan or Tinkerbell. In fact, it is not so much a case of a figurative reading being triggered, in a sense, by some overtly recognisable aspect of the statement - it simply happens that the metaphoric interpretation makes fewer demands upon the reader/hearer's credulity, while the literal interpretation demands that the hearer construct a complex model of explanation founded upon tenuous and implausible hypotheses. In effect, it is the application of Occam's Razor, rather than the detection of a semantic anomaly, that ear-marks the statement as figurative.

 

As this example demonstrates, valid models may be constructed for both a metaphoric and a literal interpretation, and therefore the classification of an utterance as either literal or figurative is more than a matter of truth conditions and semantic anomaly; it is essentially a matter of hearer credulity. The hearer brings to bear both consensus knowledge, and personal experience as a situated agent in the world, to decide upon that interpretation that best fits the conversational (or narrative) context.

 

Often the speaker may anticipate an unwillingness on the part of the hearer to accept a particular intended interpretation , due to the excessive demands it makes upon his world model, and may therefore qualify his utterance with an adverbial modifier such as `honestly', `actually', `really' or `literally'. This qualification serves to bolster the hearer's readiness to arrive at an otherwise tenuous interpretation. As Way (1991) has observed, this usage of `actually' or `literally' as a conversational strategy is the predominant, consensus meaning of the word `literal'. Literality in this sense is not simply a matter of semantic correctness or orthodoxy, but a measure of semantic credibility. When explicitly invoked in an utterance, a literal qualifier often serves to overturn the most plausible interpretation in favour of one less likely, providing, in Searle's model, the hearer with an overt marker that the SUM and LSM of the utterance coincide.

 

The speaker may also feel it necessary to explicitly qualify an utterance in this manner even in cases where there is no competing metaphoric interpretation, or for that matter, no more plausible literal interpretation. As Searle observes, for example, the utterance `The Cat is on the ceiling' is a literal statement, but one that is nevertheless difficult to reconcile with mundane world experience. The speaker may thus qualify his statement in recognition of the difficult interpretation task it presents to the hearer. This phenomenon suggests a intuitionist method of determining the literal content, or measure, of an utterance. The more figurative the intended interpretation of an utterance is, the more its meaning will change when qualified with a literal modifier (such as `actually' or `literally'). The more literal the intended interpretation is, however, the more redundant and unnatural such a qualification will seem. Thus the statement `Bill is on the golf course' is literal due to the gross redundancy of using `literally', while `Bill is on the phone' is metaphoric (albeit conventionally so), as attested by the change of meaning incurred by specifying `literally'. Likewise, the statement `The cat is on the mat' is seen to be grossly literal. Interestingly, however, the statement `The cat is on the ceiling' yields far more readily to literal qualification, and not only does the modifier `literally' not change the intended meaning of the utterance, it serves to make it all the more palatable. In this sense then, the statement `The cat is on the ceiling' is seen to be somehow less literal than the statement `The cat is on the mat'. This would suggest, contrary to Searle's strict division in terms of the satisfaction (or otherwise) of truth conditions, that literality and figurality comprise a continuum of many shades, rather than a black-and-white dichotomy.

 

In summary, Searle strives for a treatment of metaphor that is compatible with that of other pragmatic phenomena, such as irony and hyperbolae. The opposition of Literal Sentence Meaning (LSM) and Speaker Utterance Meaning (SUM) implies that metaphor is a form of conversational sleight-of-hand, in essence a lie, that allows the speaker to say one thing (the LSM) but mean another (the SUM). Certainly there is often a multiplicity of meanings inherent in a metaphor, but this arises out of the applicability of different conceptual models to the interpretation of the metaphor, and not from a desire on the part of the speaker to communicate by misdirection. Returning to Searle's example of the "The cat is on the ceiling", the multiplicity is a product of at least two different conceptual models of "ONness", each of which is surely valid in its own context. The first of these models is that employed in "The fly is on the ceiling", and the second is that which underlies the interpretation of "The cat is on the floor". Within the idealised conceptual model of a flying insect (to use the terminology of Lakoff 1987), a fly is considered to be free from the dictates of gravity, and thus ceilings, walls and floors become interchangeable. Once the speaker has satisfied himself that the cat is indeed physically attached to the ceiling (by glue, for instance), the effects of gravity upon the cat are also seen to be nullified, and the speaker is thus free to employ the fly/ceiling model to communicate this situation truthfully to the hearer. Within the cat/floor model, however, certain truth-conditions are violated; clearly, the cat is not pressing down upon the ceiling, rather he is pulling upon it, and the ceiling is not pushing back with the expected normal-reaction. If the hearer is restricted to this particular model as a definition of ONness, the speaker would indeed have to say one thing to mean another.

 

In the final analysis, however, the limitations of Searle's theory emerge from a fundamental belief in the homogeneity of literal meaning, a belief which represents a denial of the fuzzy rather than absolute nature of the literal/metaphor classification, and a refusal to accept that so-called literal truth is in fact a patchwork of overlapping conceptual models, each with specific contextual limitations on its applicability. These limitations are a product of what might be called an overly Fregean, or early- Wittgentein, style of rigid analyis, with its emphasis on truth conditions and questions of reference, rather than upon the context-dependent relationship between words and concepts, as considered by the later- Wittgenstein in his theories of family resemblances and language games.

 

4.                IVOR RICHARD AND MAX BLACK.

 

A perspective upon metaphor which is currently in vogue, and deservedly so, among both philosophers and computationalists, is the Interaction View, developed by the philosopher Max Black as an elaboration and progression of the work of Ivor Richards (1936). At a general level (and if this theory has a serious flaw, it is this generality and vagueness), the Interaction view claims that metaphor is a cognitively irreducible phenomenon that works not at the level of word combination, but much deeper, arising out of the interactions between the conceptual structures underlying words. But what does it mean to refer to the interaction of conceptual structures - surely any model which acknowledges, even in an oblique manner, the conceptual level of representation must involve some degree of interaction?

 

It is therefore necessary to distinguish between the process of conceptual combination, and that of conceptual interaction. Combination is a compositional operation which results in an aggregate conceptual structure, one that contains the meanings of the concepts involved. Thus the concepts Red and Box may validly combine to form the aggregate structure Red-Box, whose representation is a compositional arrangement of the concepts Red and Box. However, the existence of the combination Red-Box does not change the meaning of either the concept Red or the concept Box. The combination Red-Box is thus a purely compositional arrangement of concepts, in which the aggregate concept Red-Box derives its meaning from that of Red and Box, but through a relationship which preserves the independence of these component concepts. In contrast, metaphor, as argued by Richards and Black, is a process that transcends simple combination. Consider Black's metaphor `Man is a wolf'; this certainly involves some level of combination, resulting in a new conceptual metaphoric structure, the Man-As-Wolf schema. However, the very existence of this schema changes the meaning of both the Man and Wolf concepts, allowing certain predications normally applied to one to be applied (with perhaps a change in meaning) to the other. This interactionist schema Man-as-Wolf is more than a simple combination, inasmuch as it causes its component concepts to move conceptually closer together, with the effect that people are to some extent dehumanised, and wolves are to a complementary extent anthropomorphized. Because all motion (even conceptual motion) is relative, the metaphor affects not only the way we view people, it must also affect our views on wolves. So while combination is simply an interlocking of two conceptual structures, interaction involves an interchange of predications/ associations between the two. Richards refers to this interchange between domains as `fundamentally a borrowing between and intercourse of thoughts'.

 

Lacking a computational paradigm in which to express his theories, and eschewing the logical formalism that so often accompanies a substitution view of metaphor, Black chooses metaphors themselves as a framework in which to couch his model of metaphor. My interpretation of the interaction process described above, as involving the movement together of, and subsequent interchange among, two conceptual structures, arises not only from a specific (and not untendentious) reading of Black's descriptive analogies, but also from a natural desire to find in Black's work support for my own particular model of metaphor. In the end, this is the ultimate limitation of Black's metaphor writings, for while they provide us with an interesting and colourful vocabulary to describe the metaphoric process, their vagueness lends them all the mystique and hermeneutic charm of the cabala. If only we could look beyond his metaphors, to understand the significance of his signs, we begin to think, and some deep insight into the workings of this most creative of mental phenomena would be ours. Inevitably, however, Black's writings possess no more than the sum of their parts, and like the cabala, are open to numerous conflicting interpretations (see Indurkhya 1992 for a similar reconstructive interpretation, and a good analysis of the contradictions inherent in Black's theory). Rather than serve as a magnifying glass on the mechanics of metaphor, the interaction view as it is described instead serves more readily as a looking glass, in which various researchers (and I don't exclude myself here) see reflected their own theories. This fact is evidenced by the variety of different computational approaches to metaphor which claim some form of intellectual indebtedness to the interaction view, for instance, Way (1991), Indurkhya (1992), and this current work. In the end, Black's view is probably best taken not as a model of metaphor, but as perhaps a meta-theory, an agenda of how a real model should be created.

 

As an example of polymorphous nature of the interaction view, consider the following metaphor from Black (1962):

"Suppose I look at the night sky through a piece of heavily smoked glass on which certain lines have been left clear. Then I shall see only the stars that can be made to lie on the lines previously prepared on the screen, and the stars I do see will be seen as organized by the screen's structure. We can think of the metaphor as such as screen and the system of 'associated commonplaces' of the focal world as the network of lines upon the screen."

 

Indurkhya considers this analogy to highlight the essential "similarity- creating" powers of metaphor, claiming that the pattern imposed by the screen is a new (rather than previously inherent) view of the tenor (the night sky). But how exactly are we to understand the model conveyed by the metaphor? Do the stars represent entities, and the lines relations among these entities? If so, the analogy can be seen as supporting the Structure Mapping view of Gentner (1983), in which a structure from the vehicle domain is imposed onto the structurally-impoverished tenor domain, thereby organising the entities of the tenor. However, if the stars represent attributes, then the lines provide emphasis to these attributes, and the analogy is seen to support the Salience Imbalance view of Ortony (1979). Gentner for one considers the structure mapping model to be in competition with Ortony's model (see Gentner & Bowdle 1994). Likewise, from within the framework of his own theory of metaphor, Indurkhya deems malapropos a further elaboration of the night sky analogy by Black, in which the observation that a metaphor often uncovers new aspects of a tenor is likened to the way a telescope discovers new stars that are beyond the range of the naked eye. However, this elaboration might be taken as further support of the structure mapping approach, claiming that the transfer of candidate inference structure into the tenor domain might be responsible for the creation of new entities in the tenor.

 

To summarise then, Black's interaction theory is to be praised for the active role it assigns to metaphor as a cognitive (rather than a purely rhetorical) device. Alternately, it is to be both praised and damned in the same breath for providing such a clear demonstration of the intrinsic unsoundness of theories which are neither grounded in a mathematical (formal) or computational (algorithmic) framework.

 

 

 

 

 

5.                DONALD DAVIDSON.

 

Davidson (1979) provides an analysis of metaphor that reaches much the same conclusions as Searle, namely: (i) that metaphoricity is defined relative to a strong notion of literality; (ii) that literality is somehow an inherent property of words and not of concepts and world models; and (iii) that the meaning of metaphoric utterance must be comprehended both in terms of what it literally means (Searle's LSM), and the meaning that it causes the hearer to infer (Searle's SUM). Davidson's main contribution, however, over and above Searle, harks back to the writing of the later- Wittgenstein, in which he claims that

"I depend on the distinction between what words mean and what they are used to do. I think metaphor belongs exclusively to the domain of use." [Davidson, 1979]

 

Referring back to Searle, then, the LSM (Literal Sentence Meaning) belongs to the domain of meaning, while the SUM (Sentence Utterance Meaning) belongs to the domain of use. As an extension to Searle, however, Davidson offers a caution to the reader regarding the nature of SUM, claiming that the SUM is not actually carried by the metaphor (as a secondary meaning), rather it is created by the hearer in response to the metaphor:

"Metaphors mean what the words, in their most literal interpretation, mean, and nothing more. [...] The central mistake [...] is the idea that a metaphor has, in addition to its literal sense or meaning, another sense or meaning." [ibid]

 

This interpretation of Davidson's theory is also supported by Culler (1981), when he paraphrases Davidson's thesis thus

"It [Interpretation] is not a matter of structure, but of effect, and the study of metaphor should be a study of response" [Culler, 1981].

 

Like Searle, Davidson clearly subscribes to the view that metaphors communicate by misdirection, as evidenced by his claim that

"Generally, it is only when a sentence is taken to be false that we accept it as a metaphor and start to hunt out the hidden implication. It is probably for this reason that most metaphorical statements are patently false, [...] Absurdity or contradiction in a metaphorical sentence guarantees we won't believe it and invites us, under proper circumstances, to take the sentence metaphorically." [ibid]

 

The determination of falsehood, then, is a key step in the full comprehension of a metaphor, if its "hidden implications" are to be inferred. And how should falsehood be determined if not against the yardstick of literal orthodoxy? Davidson provides the following basis for literal meaning:

"Literal meaning and literal truth-condition can be assigned to words and sentences apart from particular contexts of use". [ibid]

 

Essentially, Davidson claims that literal meaning resides in "the ordinary meaning of a word". This view is quite an old-fashioned one, and one which is forever being refuted in the fields of cognitive science and cognitive linguistics. Word meaning is always dependent on context, even if that context is so obvious as to be transparent. Words only have meaning inasmuch as there exists some contextually-determined cognitive model that ties them to the correct concepts (see Lakoff 1987 for a discussion on these issues). Presumably, when Davidson employs such terms as ordinary, and patently false, he is attempting to evoke some notion of a default, or prototypical, word sense. Inevitably, then, it is the limitations of the philosopher's logical vocabulary, and a basic desire to avoid the vocabulary of the psychologist/cognitive scientist, that allows Davidson to place unwarranted faith in a notion as thoroughly problematic as literal meaning.

 

 

6.                STEVAN HARNAD.

 

A most intriguing perspective on metaphor, one that exploits a duality of representation in the mind, is proposed by Stevan Harnad (see Harnad 1982), a philosopher/psychologist who is perhaps better known for his views on Searle's Chinese Room argument, and for introducing the term 'symbol grounding' into the computational literature.

 

The duality central to Harnad's theory is that which dichotomises mental representation into the discrete and the continuous, essentially the bounded and the unbounded, or as the anthropologist Brenda Beck(1978) terms the divide, the semantic and the analogic. In Harnad's view, which is significantly echoed (via a computational expression) in the work of Indurkhya (1992), mental representation can take the form of either sensorially-grounded continuous impressions, or logically founded conceptual abstractions which are formed by induction from these sensorial bases. By their very nature, these abstractions discretize their underlying continuous (fuzzy) sources, leading to the creation of symbols and symbolic structures, which presumably capture some useful generalities and thereby form a discriminatory basis for the categorisation of future stimuli. However, in abstracting to a discretized representation that permits distal access and high-level conceptual combination, fringe elements at the continuous level that are deemed inductively irrelevant will necessarily be lost to the higher-level of representation (inasmuch as abstraction necessarily entails reduction). The kernel of Harnad's argument is that conceptual combinations (or appositions) which do not apparently yield sensible interpretation at the discrete or bounded symbolic level may nevertheless bear fruitful results if performed at the continuous level of processing. Essentially, underlying fringe elements in the continuous domain, which were original deemed irrelevant at the time of induction and thus reduced out of the derived symbolic representation, might be later reinterpreted as germane when in combination with continuities underlying other concepts. In this view, metaphor is a cognitive device through which the continuous, analogic (and sensorily-based) grounds that drive our categorisation (i.e., symbol- making) processes are re-evaluated in new conceptual contexts. This recombination often leads to the creation of new concepts, novel combinations which are more that the sum of their symbolic parts, inasmuch as these concepts emerge from continuities not present at the symbolic level.

 

In Harnad's terms, sensory experiences are recorded as mental impressions which constitute continuous or unbounded engrams, mental traces which arise from sensory experience and which form the ultimate psychological basis of memory. Derived from these continuous traces, by induction and other abstraction processes, are bounded engrams, which quantize the underlying unbounded impressions into useful discretized mental objects or symbols. Given an unbounded engram, it is often possible to evoke the original unbounded source, and in some sense re- experience (relive?) the original primary impression. In this way, high- level conceptual combinations may trigger a low-level combination of impressions, which in turn may lead to a re-experiencing of the original stimuli and a subsequent recategorisation of this composite impression into a new conceptual category.

 

This view is at once both enervating and depressing. In one light, Harnad is reaffirming the cognitive irreducibility of metaphor, and reformulating in specific psychological terms the rather nebulous interaction theory of Black (1962). In another light, however, Harnad convincingly denies the possibility of a wholly symbolic account of the metaphor phenomenon. Rather than argue for the bankruptcy of the symbolic enterprise, though, Harnad instead emphasises that a competent theory of metaphor must incorporate both discrete and continuous levels of representation, or in computer science terms, such a theory must successfully integrate symbolic and connectionist modes of processing.

 

B      -       THE LINGUISTS ON METAPHORS.

 

1.                  GEORGE LAKOFF.

 

If the work of Aarts & Calbert demonstrates the essential wrongheadedness of the linguistic approach to metaphor, then the writings of George Lakoff, with his various collaborators Mark Johnson, Claudia Brugman, and Mark Turner, represent the conceptual antithesis to this view, stressing both the cognitive role of metaphor and its roots in the human conceptual architecture.

When criticising the mainstream linguistic community (as typified by the Chomskyan transformational model of language, with its enforced separation of competence and performance) for its sterile views on metaphor, Lakoff speaks from an insider's position of knowledge. As one of the early generative semanticists, Lakoff questioned both the validity and primacy of a syntactic deep structure in formal models of language, overturning the Chomskyan Standard Theory view that semantics emerges from an interpretative reading of syntax, and not vice versa (see Leech 1990). The new generative position instead posited that a deep semantic structure was at the core of language comprehension, effectively merging deep syntactic and semantic structures into a unified whole. To formalists, who consider the central concern of linguistics to be the study of abstract language competence, rather than actual performance, this reversal is of little empirical consequence. Nevertheless, the generative position does clear the way, ever so slightly, for a new view of metaphor in linguistics: no longer must meaning be subservient to syntax, and no more is metaphor to be considered a superficial phenomenon of language.

 

Lakoff's writings call for a reification of the status of metaphor, from that a superficial rhetorical device that decorates our speech, to the status of a deep, cognitively-realised agency that organises our thoughts, shapes our judgements, and structures our language. Metaphor is shown to pervade (indeed, saturate) our everyday, non-rhetorical language to such a degree that a deviance view of the trope would leave little, or no scope for orthodoxy. Lakoff cites as empirical evidence of this cognitive role of metaphor the immense systematicity that the phenomenon exhibits: individual metaphor systems, such as Argument-is-War or Love-is-a- Voyage, seem to possess a vast generative capacity, inasmuch as they manifest themselves in a variety of guises, each guise portraying a different facet of the metaphor. (see Lakoff & Johnson 1987).

 

Lakoff's central thesis is that metaphors facilitate thought by providing an experiential framework in which newly acquired, abstract concepts may be accommodated. The network of metaphors that underlie thought in this way form a cognitive map, a web of concepts organized in terms which serves to ground abstract concepts in the cognitive agent's physical experiences, and in the agent's relation to the external world. A major component of the human cognitive map is what Lakoff terms a cognitive topology, essentially "a mechanism by which we impose structure on space, in a way to give rise to spatial inferences" (see Lakoff 1988). The cognitive agent is itself an essential player in this organisation - abstract thoughts are not structured in terms of objective spatiophysical properties of the world, but subjective, egocentric properties that the agent projects onto the world via his cognitive map. Consider a concept such as "In Front of"; this is clearly a subjective relation, as it relies on the perspective of the cognitive agent performing the observation. In our observer-centric culture, one entitiy (the figure) is considered to be in front of another (the ground) if the figure lies between the observer and the ground. In another, less egocentric, culture, however, "In front of" might be defined relative to the ground rather than the observer; in such a culture, the observer would have to project himself onto the ground (in other to determine the correct perspective) and thus "in front of" would imply that the ground lies between the figure and the observer. Similarly, misinterpretations often arise in the comprehension of temporal phrases such as "let's push the meeting back a week". Considering that the western view of time is that of a line advancing forward from past to future, such a request is easily interpreted as "let's schedule the meeting one week earlier", in which the meeting is moved backwards one week along an objective time-line. However, an egocentric analysis results in the interpretation "let's schedule the meeting one week later", in which the observer, standing at a particular point on the time-line, pushes the meeting away (and thus forward) in time. These examples demonstrate that even the same metaphor of time can produce different interpretations, depending on the relative position of the observer within his cognitive topology.

 

In the Lakoffian view, a metaphor is a schema in the original Kantian sense, that is, a unifying framework that links a conceptual representation to its sensory and experiential ground. Garnder (1987) offers a most concise explanation of Kant's notion of a schema:

"A schema serves as a mediating representation which is intellectual in one sense, sensible in another. Thus, a schema is directly activated in terms of sensory experience and can plausibly thought to provide an interpretation of that experience. As a cognitive scientist might put it today, Kant had entered the world of 'mental representation'". [Gardner 84]

 

Lakoff (1988) conjectures that metaphors project the cognitive map of a source domain (i.e., the vehicle) onto a target domain (the tenor), thereby causing the target to become grounded in spatiophysical experience via the source. The result is that the schemas which mediate between conceptual and sensory levels in the source become active also in the target. In this view, a metaphoric schema is a mental representation that grounds the conceptual (intellectual) structure of an abstract domain in the sensory (sensible) basis of another, more physical, domain. This schematic view of metaphor is similarly expressed by Indurkhya (1992).

 

The extent to which particular spatial metaphors have pervaded our language is representative of the way our cognitive topologies dissect the world. For instance, up/down orientation metaphors (e.g., Up is good, Down is bad, Top is best, Bottom is worst, High is happy, Low is sad) are considerably more productive than front/back metaphors (e.g., forward is future, back is past), which in turn are more productive than left/right metaphors (e.g., right is good, left is bad, unnatural or sinister). From an objective stance, each spatial dimension possesses the same descriptive power. From a subjective position, however, we note that the paucity of left/right metaphors reflects the general symmetry of the human body in that direction; because left and right are so topologically undifferentiated, this dimension is utilized less in a egocentric model of the world. However, this symmetry does not exist in either the front/back or up/down dimensions. The fact that vision operates to the front of the body and not to the back is reason enough for this dimension to be differentiated, and thus descriptively useful, while the presence of an active force, gravity, acting in the vertical dimension justifies the significance we attach to the concepts of up and down (see Lyons 1977).

 

The cognitive maps that ground our metaphors and provide the conceptual substrate for our most abstract thoughts are thus very much influenced by our bodily experiences of the world. Such a view denies an objectivist account of the world, as the properties we perceive are thus contingent upon our physical makeup (i.e., the way our bodies interact with the world) and cultural biases (i.e., the metaphors we choose to model these interactions). In light of this contingency, Lakoff argues for a subjective psychology that rejects the classic model of categorisation (the Fregean model so beloved by philosophers such as Searle), and proposes instead a non-reductionist model of natural categorisation, one that replaces such staples of objectivism as necessary and sufficient conditions with prototypes, hedges, radial categories and ICMs - idealised cognitive models (see Lakoff 1987).

 

The basic notion underlying a radial category is that some members of a category will be more representative than others (i.e., members par excellence); taken together, the members of a category thus form a radial structure, the most representative, or prototypical, members located at the centre, with less representative outliers clustered around this hub. For instance, a robin is a bird par excellence, while a penguin is quite a poor example of the class. Membership of a category is therefore a gradated, rather than absolute black and white, notion. Hedges are a form of category membership that correspond to linguistic judgements such as "kind of" or "sort of"; for instance, a whale is like a fish (sort of), and a moped is like a motorbike (kind of), inasmuch as many of the same common-sense inferences hold about each.

 

Lakoff suggests that radial categories and prototypical effects arise due to a tendency in humans to conceptually the world through partial and idealised cognitive models (or ICMs). Each category is defined relative to such a partial model, which captures the expectations that should hold whenever the category finds valid application. The degree to which the background ICM of a category thus fits a real world entity or situation is a measure of the category representativeness of this entity/situation. Prototypical category members are therefore those members which make a perfect fit with the background ICM, i.e., those members which satisfy all of the background assumptions and conventions associated with the category. For instance, the classic definition of the category Bachelor, in truth-conditional terms, stipulates that a member should be an unmarried man. However, such a definition grants full and equal membership status to such dubious representatives as priests, homosexuals, geriatrics, young teenagers, shipwrecked sailors, children raised by wild animals, men in long-term relationships who are not free to see other woman, and all other males prohibited from marriage either by law, common-sense, or inclination. Quite simply, the concept of bachelorhood is defined relative to an ICM that expects category members to be unmarried but of a marriageable age, to be of heterosexual disposition, and exist in a human society that both supports marriage and that provides enough eligible males and females for the practice to survive. Thus, Humphrey Bogart in "Casablanca" is a prototypical member of the class Bachelor, while Tarzan and Quentin Crisp are very poor examples. Note that as ICMs provide an idealised context that licenses the use of a category, should the real world context of a category member shift, that member's position within the radial category structure may also be modified. For instance, in moving between the jungle and the city, Tarzan becomes a highly representative and eligible bachelor (indeed, this is the plot to at least one Tarzan cinema feature); likewise, after his rescue and subsequent return to civilisation, no doubt Robinson Crusoe also fell prey to the marriage pressures imposed by society.

 

This reformulation of categorization, in terms of ICMs, prototypes and radial categories, yields a much more plastic account of world knowledge, one I believe is necessary if metaphor is to be viewed not as a deviant phenomenon that violates truth conditions, but as a cognitive device that stretches conventions and establishes new perspectives. Referring back to out previous treatment of Searle, the thoroughly troublesome problem of literal meaning is considerably simplified when expressed within this new repertoire of cognitive notions. For not only does a psychology of prototypes and idealized cognitive models considerably clarify (and resolve) the issues fuelling the literal meaning debate, such a psychology provides a clear cognitive motivation for the existence of metaphor in the first place. When traditional philosophers and formalists, such as Searle, Davidson and MacCormac, stake their claims of literality on the meaning of "words used in their ordinary senses", I suspect that they are in fact groping to express the notion of "prototypical category members used in their idealized setting", but in the end are too constrained by an objective logical vocabulary that seeks to avoid any psychological postulation. The raison d'etre of metaphor in such a cognitive setting is thus the expansion of existing categories; metaphor allows us to reorganise our conceptual models by reclassifying concepts under new categories, thereby testing the conventions inherent in an ICM and broadening the scope of our existing categories. The tension inherent in a novel metaphor is a result of the initial unrepresentativeness of the new categorization, but as the metaphoric reading becomes more representative, possibly through the modification of the associated ICMs, the metaphor is seen to lose its tension and become more literal (i.e., more prototypical).

 

2.                  UMBERTO ECO.

 

A semiotic perspective on the interpretation of metaphor, which views the phenomenon as an unstable or dynamic sign, is offered by the philosopher, historian and semiotician Umberto Eco in his 1984 book 'Semiotics and the Philosophy of Language'. Eco's treatment is interesting not only for the semiotic lexicon he brings to the discussion of metaphor, but also for the depth to which he plumbs Aristotle's original theory, in doing so demonstrating a clear link between many modern theories of metaphor and the classical theory which was proposed by Aristotle over two thousand years ago. In particular, Eco's analysis is a significant parallel between Aristotle's theory and the dynamic type hierarchy (DTH) model of Eileen Cornell Way, a modern computational theory which shall be considered in some depth later. In general, Eco makes a somewhat uncharitable claim that mirrors, in its dismissive sweep, Whitehead's famous reduction of the European philosophical tradition to 'a series of footnotes to Plato':

"... of the thousands and thousands of pages written about the metaphor, few add anything of substance to the first two or three fundamental concepts stated by Aristotle. In effect, very little has been said about a phenomenon concerning which, it seems, there is everything to say." [Eco'84]

 

In Eco's opinion, then, the basic metaphor research agenda was initially founded by Aristotle, and most thought that has since followed on the subject has remained largely faithful to the tenets established therein. The idea that a well-structured concept hierarchy in itself provides a suitable basis for metaphor interpretation (what Eco might himself call the Porphrian Fallacy) has, as a result, become somewhat of a mind-set to researchers in the field, as can be witnessed by the number of different theories which depend, ultimately, on either an explicit or implicit taxonomy of world knowledge, against which the operations of metaphoric analysis can be conducted.

 

However, if most metaphor researchers are caught in the grasp of an idae fixe, Eco's own perspective is not itself entirely untrammelled by this fixation of vision. This is apparent from the five semiotic rules he provides to sketch a basic interpretation process (note: Eco seems to describe here an interpretative process for dealing with metaphors in which the tenor is not explicitly provided, but has to be eked out of the narrative context; for instance, the "house of the birds" refers to, but does not mention, the sky):

 

1. Derive a componential representation of the vehicle, focusing on those components (markerese features, slots, attributes, predicates, etc.) which are deemed germane in the current co-text (narrative context).

2. Look abductively in the encyclopaedia (world knowledge-base) for some other sememe (concept symbol) that possibly shares some of the focused properties of the vehicle, but exhibits other, interestingly different properties of its own (presumably these interesting differences supply the emotive tension of the metaphor). The sememe found in this way is a plausible candidate for the tenor of the metaphor.

3. Try to link the opposing (interestingly different) properties of tenor and vehicle via a shared superordinate in the knowledge-base taxonomy. Presumably, such a link legitimises the opposition and thus the metaphor.

4. Evaluate the metaphoric reading. The higher up the knowledge-base taxonomy these oppositions are united, the more interesting the metaphor is deemed. Presumably, this is a criterion for establishing metaphoric quality, inasmuch as the more obscure the opposition, the more novel the metaphor will seem.

5. Establish new predications/attributions/relations in the tenor as a result of the metaphor, 'so as to enrich the cognitive power of the trope'.

 

On the whole, Eco's five rules put forward a liberal agenda for metaphor interpretation, one that acknowledges its cognitive centrality and role in acquiring new conceptual structure. The difficulty with his formulation, as suggested previously, is that it is not entirely free of the taxonomic mind-set which can be considered a failing in other models of metaphor. Steps three and four of his scheme should (and can) be generalised so as to avoid unnecessary commitment to omni-competent concept hierarchies. Determination of a shared superordinate is not the only plausible manner in which an opposition of interestingly different properties in the tenor and vehicle may be reconciled; they may instead share some common attributions, partake in the same predications in the same contexts, etc., or indeed, the commonality may not be explicable at the symbolic level at all before the metaphor is interpreted, if the perspective of Harnad is the correct one to adopt.

 

C         -           SOME CONCLUSIONS.

 

The mathematician/philosopher Alfred North Whitehead suggested that European philosophy is a series of footnotes to Plato. In the history of metaphor research, the semiotician Umberto Eco suggested likewise comprises little more than a procession of annotations to Aristotle. Inasmuch as the hallmark of Aristotle's theory of metaphor, essentially a reliance on taxonomic ordering and ontological twiddling as evidence in many of the approaches surveyed.

 

There are various common flaws in the theories surveyed. The main flaws are the following: i) an adherence to the Standard Pragmatic Model of interpretation, and a corresponding denial of empirical evidence to the contrary; ii) An anomaly-driven view of processing, whereby metaphors are first triggered by the violation of some semantic constraints; iii) a repair-driven view of metaphor, essentially deep substitution at the semantic/conceptual level, in which an initial primal representation needs to be mapped/transferred/projected into a more literal domain; iv) a determination that metaphor is an issue of lexical semantics only, and that recourse to contingent world knowledge is unnecessary for interpretation; v) the lack of a coherent structural binding/matching facility, such that systematicity between different metaphors/domains may be enforced; vi) an over-reliance on taxonomic ordering (the Aristotelian fallacy), whereby metaphors are interpreted relative to some omni-competent type hierarchy (or markerese).

 

The current approach to metaphor interpretation -- Conceptual Scaffolding and Sapper -- remedy these failings in a homogeneous and cognitively-plausible fashion. This approach exhibits a strong allegiance to the perspectives offered by Ortony, Harnad, Eco, and Lakoff & Johnson, and it is to these thinkers that this work owes its greatest intellectual debts.

 

D       -        REFLECTIVE POWER OF METAPHORS

 

For the last decades, philosophy has acknowledged the function of metaphor as a heuristic  means of cognition. As Nelson Goodman has pointed out, a metaphorical predication creates novel symbolic distinctions and rearranges things within a symbolic sphere in a new way. Connecting Goodman's findings with linguistics, philosophy of science, and hermeneutics, one can argue that nearly every symbolic system is based on indissoluble root metaphors and on fundamental metaphorical distinctions.

 

However, the rationality of these basic distinctions is controversial, particularly with respect to academic language and thought: one side of the dispute, represented for instance by Goodman, Hesse and Black, sees metaphor both as a non-substitutable basis of language and as a rational element with its own truth. The other side, represented by Davidson and Rorty, regards metaphors as contingent results of prevailing historical symbol systems, which implies they can neither be true nor have a meaning of their own.

 

Moreover, it remains controversial whether fundamental and background metaphor is a priori pre-reflective and irreducible. And therefore, as Jacque Derrida argues, the basis of all metaphysics ­ or whether it is possible to transcend metaphor through critical-hermeneutic reflection, so that its validity claim can actually be examined, which is the position held by Paul Ricoeur.

 

According to Nietzsche, any assertion that is taken to be true is necessarily rooted in conventional metaphors that are only contingently valid. For this reason, no rational criteria and no truth can exist for metaphors. Metaphor hereby becomes myth. However, if truth depends on the contingency of established sets of metaphors, then the truth or the untruth of metaphor can only be asserted in the mode of performative contradiction.

 

A resolution of the paradox first emerges with the ascertainment of the reflective function of metaphor, namely with its redefinition from a disruptive factor that threatens the order of literal discourse as a semantic anomaly, to a reflective factor that permits the order of language to be reflected and recategorised. Only in this reflective function can metaphor become the "linguistic kernel of a post-analogical ontology of modernity". Here, the contingency of ontological distinctions and categories becomes known and can be reflected by means of metaphor.  The difference between literal and metaphorical speech then remains a pragmatic one, which can be determined only from concrete contexts. Metaphor introduces a meta-linguistic level in opposition to the ostensibly literal level of linguistic reference to objects, which reflects this literal level in a specific way. By transgressing the boundaries of the literal and thus reflectively thematising them, metaphor can be said to have a reflective structure.

 

In addition, metaphor also has a self-reflective structure, since it proposes a new insight in the mode of the 'as-if'.  'Metaphorical redescription' can be understood as a rational anticipation of new phenomena, an anticipation that contains in itself the reflected negation of the old order and at the same time proposes a new order. If, however, metaphorical redescription is actually to be a rational anticipation and not just an arbitrary myth, then its rationality may not be evaluated only in respect of its internal aptness to the symbolic system in which it is used, for this might be mere myth or ideology. Instead, metaphor must be evaluated in terms of the epistemological paradox of the metaphorical 'as-if' predication, which simultaneously expresses both an identity between two entities and their difference.

 

Metaphorical predication always entails the 'insuperable paradox' (Ricoeur) that is built into the notion of metaphorical truth. The paradox consists in the fact that there is no other way to do justice to the notion of metaphorical truth than to include the critical incision of the (literal) 'is not' within the ontological vehemence of the (metaphorical) 'is'.

 

The usage of a metaphor with validity claims thus implies both the possibility and the necessity of metaphor reflection. Here, the difference 'literal versus metaphorical' is less decisive than the difference 'mythical versus metaphorical' insofar as myth results from taking a metaphor too literally and thus reifying language. Metaphor becomes myth when it is considered to be the one and only invariable truth, instead of a possible, fallible truth. That is, metaphor becomes myth when it is anchored so deeply in conviction that it can no longer be called into question.

 

Thus reflection upon metaphor does not entail converting metaphorical speech into literal speech. Instead it requires a determination of the difference between myth and metaphor; this means distinguishing the 'as-if' status of metaphor from the unquestionable certainty of myth ­ even where metaphor is unavoidable and irreducible, as in the case of absolute metaphor. Only metaphor reflection that is both self-referential and context-referential and that reflects the broader sense of a metaphorical expression can prove the truth of metaphor and thus rationally legitimise its use. In the same way that every actual consensus about truth remains open to revision without losing its unconditional validity claim, the metaphorical redescription has the status of an 'as-if' predication and nevertheless can assert a truth claim: As Berggren points out, metaphor is a "counterfactual statement" that possesses a "possible or counterfactual truth".

 

The rationality of the metaphorical redescription has to be measured by its adequacy. Adequacy can be determined by comparing the new metaphorical description with previous descriptions of the object and by interrogating it for new perspectives and insights. However, this reflective process of catching up with the metaphor's implications is possible only retrospectively and within argumentative discourse concerning the reflection of both metaphorical anticipation and metaphorical myth. Thus metaphor must prove itself in the language-game of truth.

 

The idea of assessing the rationality and truth of metaphor through critical reflection upon metaphor, however, is itself based on the fundamental optical metaphors of light, mirror and observation. The metaphorical roots of reflection thus deserve closer consideration. One could say that the light of truth cannot be recognised directly but needs to be refracted and mirrored from different angles. This indirect recognition is necessary not because the light of truth is too bright (as Plato believed) but because the mirror allows selection and comparison. Reflecting a metaphor from different angles means creating new views on this metaphor, and thus introducing both new distinctions into interpretation and more distance from the metaphor. Although the blind spot of observation, as such, remains insuperable, it is possible to overcome the blind spot of a particular observation by changing the position and the basic distinction of the observer.

 

Changing position and distance does not imply a reduction of the evocative and imaginative power of metaphor. In other words: metaphor reflection is not meant to be a procedure of metaphor control and surveillance but instead an operation that broadens understanding by opening up new perspectives and points of comparison. This requires that reason act not as a Foucaultian observer in the centre of a cognitive panopticon,but rather as a Socratic traveller who constantly changes position and perspective by wandering through the network of metaphorical associations and connections and by creating new relationships or abolishing old ones. Metaphor reflection is in itself a metaphorical process of transference and interference.

In conclusion, I would like to propose a systematic method of metaphor reflection, based on an extension of Shibles' 'metaphorical method', which I call reflective metaphorisation. I understand reflective metaphorisation to comprise all metaphor-critical processes, such as the creation of metaphors, their expansion, variation, exhaustion, confrontation, and historicisation. All these processes serve reflection upon a metaphorical or (ostensibly) literal expression. By setting in motion processes of reflective metaphorisation, we decontextualise the problematic expression, call it into question, and challenge its self-evident truth, thus consciously shifting it into an 'as-if' perspective. Through the systematic exhaustion of metaphorical potential and through the decontextualisation of metaphors, we can open up new ways of seeing and new connections, make visible new distinctions and differentiations, and investigate the internal and external limits of validity for each metaphor. Even paradigmatic background metaphors and absolute metaphors can be rendered accessible to reflection through this operation of reflective metaphorisation.

 

Only with a reflected consciousness of the constructive 'as-if' character of each metaphor can we avoid the mythification of metaphor and the development of essentialistic conceptions. If we bear in mind this 'as-if' reservation, we can avoid what Immanuel Kant would call 'dogmatic anthropomorphism'. We then limit ourselves to a 'symbolic anthropomorphism', which, indeed, concerns only language and not the object itself. Metaphor is necessarily anthropomorphic and as such anticipates and impinges on the world; but as reflected anthropomorphism, limited to the symbolic realm, metaphor does not resort to the myth of reifying, essentialistic ontology. Metaphor can be a rational anticipation (and thus capable of truth) only in the synthesis of anticipatory evidence and rational reflection of meaning and validity. However, metaphor possesses not only its specific reflective power but also (particularly in the sciences) an immanent tendency to become mythified, due to its suggestive power. As a result, metaphor reflection that dissolves epistemological obstacles is not only a theoretical task for metaphorology and the philosophy of science. It is also an ethical duty of academic praxis.