Two models of metaphoricity and three dilemmas of metaphor research

1 Introduction

According to the currently prevalent conception of metaphor, i.e., the Conceptual Metaphor Theory (CMT), metaphor consists in a cross-domain conceptual mapping allowing for “understanding and experiencing one kind of thing in terms of another” (Lakoff and Johnson 1980a: 5). Whatever the advantages of this definition, it amounts to a considerable impoverishment of the classical notion of metaphor as first defined, as far as we know, by Aristotle, and maintained through more than two thousand years of rhetorical tradition in Europe, most recently being expounded in the works of Max Black (1962) and Paul Ricœur (1975; but also see Wolff and Gentner 2011). In terms of CMT, a metaphor, according to what we will henceforth call the Tension model, is a mapping from one domain to another in which one domain is understood in terms of another; this mapping is experienced not only as being innovative, but as permitting the discovery of features of the object in the first domain which would not have been highlighted if the object of the second domain had not been brought to bear on it. Thus, the specificity of metaphors does not only consist in the crossing of domains (in some sense or other) but in maintaining the tension between them (Black 1962; Ricœur 1975; Sonesson 1989; 2003; 2015a). As Black (1981: 208) observes in his review of Lakoff and Johnson (1980a), from this point of view, CMT is concerned with “shallow or trite metaphors, if they deserve the name.” This is not to say (although this is no doubt what Black meant) that the kind of study that Lakoff and Johnson have initiated is uninteresting, but it really concerns the opposite of what has until recently been considered to be the nature of metaphor, and actually forms the ground against which the metaphor stands out as a figure, that is, the world taken for granted — the Lifeworld, in the sense of phenomenology. Nevertheless, we will see that the classical tradition allows for another model, that we will call the Overlap model, which has not been clearly differentiated from the Tension model.

2 First dilemma: CMT and the Tension model

Different theories, from Aristotle to Black and Groupe µ, have conceived of the relation between the two contents involved in metaphors as being one of tension and/or of union resulting from an overlap (see Section 3). Still, this tradition has retained certain constants, which are opposed to Lakoff and Johnson’s conception: (1) there is no fixed set of “cross-domain mappings,” which is in some way primary to other metaphors; (2) while metaphors are certainly cognitive in some sense, they are not implanted directly in the brain but rather often the result of socio-cultural presuppositions; (3) there is only metaphor (or any other rhetorical figure) to the extent that the term is felt to be “improper,” that is, if there is a sense of the term being misapplied (which is not true of such cross-domain mappings as, for instance, ARGUMENT IS WAR); (4) it is a relation between signs of which one is substituted for the other, not merely between free-floating contents.

In the complex classifications of figures in the French 17th and eighteenth century treatises of rhetoric, metaphor is, like metonymy, a trope (applying to words or single signs rather than to sentences or sign complexes) and a substitution (involving the exchange of one elements for another, rather than the suppression or addition of an element or the permutation of the order of several elements). What differentiates metaphor from metonymy and synecdoche is the nature of the relationship between the two elements entering into the substitutions. Whereas the tenor and its vehicle (the content and the expression of a figure) are joined by similarity in metaphor, metonymy connects them by means of a contiguity, relating as part to whole in synecdoche. Thus, in classical semiotic terms, metaphor is an iconic sign and metonymy and synecdoche are varieties of indexical signs.

At least from Aristotle onwards, metaphor has been thought of as a discovery procedure, the means for finding out, and even creating, similarities never observed before — which is the sense of metaphor that Paul Ricœur (1975) has tried to restore. According to one well-known theory of metaphor, usually attributed to Black (1962; cf. Goodman 1968: 68; Ricœur 1975 : 109), the two terms brought together by metaphor, “vehicle” and “tenor” (called so in the terminology of I. A. Richards employed by Black), must persist in a state of “tension” and “resistance,” which will naturally be the case if one sign is substituted for another sign that is less befitting to the context; indeed, another way or formulating this relationship, which I have employed elsewhere (see Sonesson 2015a), may be to say that at the root of the construction there is a willful miss-classification of something into a category to which it does not belong — or, more specifically, that an object is assigned to a prototype of another category than its own. This would seem to accord perfectly with Black’s theory, according to which the two terms brought together by the construction, the “vehicle” and the “tenor,” must persist in a state of “tension” and “resistance;” for, indeed, what could occasion more resistance than the classification of a term under a heading to which it does not belong? There is some empirical evidence for this theory of metaphor: small children, Gardner (1982: 91, 158) discovered, simply overgeneralize their terms and only later come to use real metaphors, where “such tension is sensed, and then overridden” (ibid.: 99; cf. Winner et al. 1979).

Such a procedure could obviously give rise to surrealist metaphors, which, in the limiting case, may only produce an effect of bizarreness or, as the Russian formalist said, estrangement. Comparing metaphors with scientific models, however, Black is rather interested in figures that make you see new sides of familiar objects or discover new facts by pointing to unsuspected similarities. Thus, faced with the Shakespearean metaphor “Shall I compare thee to a summer’s say,” you have to search the metaphorical space offered for the possible coincidences between the properties of tenor and vehicle. Clearly, such a search of metaphor space does not necessarily have to be as demanding as in Shakespeare’s case. To say that the lion is the king of the jungle is still felt as something of a metaphor, although one producing only a slight amount of tension. The statement that the king fought as a lion could perhaps engender a little more suspense.

CMT famously redefined metaphor as a cross-domain conceptual mapping allowing for “understanding and experiencing one kind of thing in terms of another” (Lakoff and Johnson 1980a: 5). This does obviously not mean that metaphors, in the classical sense of the term, are excluded — but they are not singled out. I will have more to say about why this is a problem later (in Section 4). For the moment, let as ponder the notion of domain. Following the inspiration of Roman Jakobson, CMT only recognizes one other figure, metonymy, which is characterised as within-domain association. To begin with, it is not easy to see what defines a domain here. Take the celebrated case of “LIFE IS A JOURNEY.” In what sense, one may ask, is a journey a different domain from life? Rather, it would seem to be a possible episode of the whole, which is life, which would make this a kind of synecdoche. Or take the case of “ARGUMENTATION IS WAR.” As Black (1981) observes in his review, not all of the examples given really concern war. However, I, unlike Black, think they can all be subsumed by a notion such as “(reciprocal) aggression:” “ARGUMENTATION IS AGGRESSION.” Aggression is no doubt manifested in war and many other occasions of life, including argumentation, but it remains unclear in which sense argumentation, aggression, and/or war are different domains. Let’s admit, however, that these are domains defined ad hoc. After all, we could ask the same question concerning the relation between “a summer’s day” and the “you” of the Shakespearean address — only that, in this case, the domains are defined by the tension experienced between them. There is clearly no tension between the “domains” of CMT. Or, at least, there is not the degree of tension taken for granted in classical metaphor theory. This is the moment to remember Peirce’s definition of the three kinds of hypoicons:

Hypoicons may be roughly divided according to the mode of Firstness of which they partake. Those which partake of simple qualities, or First Firstnesses, are images; those which represent the relations, mainly dyadic, or so regarded, of the parts of one thing by analogous relations in their own parts, are diagrams; those which represent the representative character of a representamen by representing a parallelism in something else, are metaphors. (CP 2.277, EP 2: 274)

Elsewhere, I have suggested that the metaphors of CMT are really diagrams in the sense of Peirce (see Sonesson 2015a). That is, they simply serve to rely two domains (to the extent that they really are domains^[1]) without producing any modification of the relation between them, which would have the function of highlighting formerly unobserved properties of the second domain brought to bear on it by the first domain.^[2] Although the Peircean diagram is a much broader category that that of ordinary language, it includes diagrams in the everyday language sense; the lines of the population curve on the paper go up, just as the population does. This is, at least from one point of view, simply an equivalence between two two-place relations. The terms need to be reinterpreted, but the relation itself is one and the same. There is thus no tension. On the contrary, in the case of a real metaphor, I suggest the relation itself has to be reinterpreted. This explains our feeling that metaphors transgress borders. Whether Peirce’s definition really serves to define metaphors in the classical sense, of whether it singles out a particular kind of metaphor, is something which we will have to consider later (see Section 7 below).

3 A second classical model of metaphoricity: The overlap model

Another theory of metaphor is suggested (perhaps unintentionally, since it has never been returned to) by Groupe µ (1970: 107): the metaphor, we are told, “extrapole, elle se base sur une identité réelle manifestée après l’intersection de deux termes pour affirmer l’identité des termes entiers. Elle étend à la réunion des deux termes une propriété qui n’appartient qu’à leur intersection.”^[3] Thus, where Black’s theory supposes a non-coincidence of the two categories brought together, this conception appears to posit a partial identity that is transmuted into a total coalescence. Here, Groupe µ seems to anticipate the blending theory of Fauconnier and Turner (2002), though illustrating it with two overlapping circles (which seems to me to be more intuitively satisfying) rather than by positing four different spaces. Indeed, according to Fauconnier and Turner (2003: 57), “The essence of the operation is to construct a partial match between two input mental spaces, to project selectively from those inputs into a novel ‘blended’ mental space, which then dynamically develops emergent structure.” The difference, however, is that Fauconnier and Turner are not interested in defining metaphor, but in pinpointing a very general linguistic (or possibly semiotic) operation which brings different spaces or frames together, which includes not only counterfactuals (as in Fauconnier’s original book), the sign, complex grammatical structures, fictive motion, but also a boat race, a ritual of a new-born child, and a Buddhist riddle (see Fauconnier 2001; Fauconnier and Turner 1998, Fauconnier and Turner 2002, Fauconnier and Turner 2003) — and no doubt also the arts, the sciences, and humour, which is what Arthur Koestler (1964), from whom they claim to take their inspiration (Fauconnier and Turner 2002: 37), was primarily concerned with (cf. Sonesson 1989: 78f). While all these things may conceivable have something in common, one may think that the differences between them are at least as important (see Sonesson 2001).

Oddly enough, after presenting this model, Groupe µ (1970: 108) goes on to claim that this is the same thing as to say that a metaphor is made up of two synecdoches, one of which is generalizing, e.g., from birch to flexibility, while the other is particularizing, e.g., from flexibility to girl. Of course, this is a quite different conception: it amounts to forming a single class made up of all things that possess in common the property of being flexible, neglecting all further attributes. The theory therefore only accounts for the intersection, not for the union. Curiously, I have never seen the Overlap model being explicitly addressed elsewhere (except by Wolff and Gentner 2011, to which I will return below), but it often seems to be taken for granted in the discussion of metaphors.

Implicitly, it would seem, Amos Tversky (1977) presupposes such an Overlap model of metaphor, but with a difference. If a union of the two sets of semantic features is to result, the metaphorical connection has to be symmetrical, but this is clearly unsatisfactory in most cases. According to Tversky (ibid.: 328), metaphors, like other similarity judgements, are asymmetric. Thus we say “Turks fight like tigers,” not the reverse; and “My love is as deep as the ocean,” not the reverse; and to say that “A man is like a tree” is not at all the same as to say that “A tree is like a man,” for in the first instance, we attribute roots to the man, and in the second instance we give a life history to the tree. In this last example, Tversky is, I believe, entirely right (though many more features may actually be transferred in each instance, as we shall suggest below), but is also seems to me that the two other metaphors, and indeed all metaphors, also possess their inversions, which have a different meaning.^[4] Of course, in examples such as those above, the inversions would probably only satisfy a surrealist. Tversky (ibid.: 349) observes that in the judgement of similarity, “a particular feature space” is given beforehand, but in metaphor, this space has to be sought out, but this is, as we shall see, not the only difference. Rather, it is an effect of something else: that metaphorical similarities have to be created anew, to a large extent, while judgements of similarity rely on similarities taken for granted.

It will be noted that ordinary similarity judgments are also asymmetric, as shown experimentally by Tversky and Gati (1978). Thus, people find the similarity judgment “North Korea is similar to China” much more acceptable than its inversion, “China is similar to North Korea.” Or at least this was the case at the moment of the experimental study. I employed this discovery in my criticism of the symmetry argument against iconicity, which was advanced by Bierman (1963) and Goodman (1968), who both wanted to deny the possibility of iconic signs on the grounds of the symmetry of the equivalence relation of logics (see Sonesson 1989, Sonesson 1998, Sonesson 2001). Nevertheless, similarity judgments are clearly different from metaphors. As I have described them elsewhere, they start out from a standard of comparison, something which is more prominent, which is used to evaluate the other item. Prominence may be due to different reasons: in a case such as China, the standard may be more well-known, more historically and/or politically important, a bigger country, more powerful, etc. In the case of one identical twin, the one we have got to know first becomes the standard of comparison. However, it would be quite misleading to say that properties are transferred from one of the members of a similarity judgment to the other. The point of such as judgment is precisely to keep the similarity circumscribed.

We can now try a synthesis. In Tversky’s terms, metaphors emphasize certain properties of the subject of comparison (corresponding to the tenor) — comparatively low on its feature hierarchy — that are then also sought out in the reference point (corresponding to the vehicle). The whole feature set of the reference point is then projected onto the subject of comparison. The result is no mere identity, however, as that part of the subject of comparison that does not intersect with the reference point remains in awareness and resists integration in the new whole; thus, there may be tension between the new category created and the received categories. While I formulated this model long ago (vaguely in Sonesson 1989, and more explicitly in, 2003), its psychological reality has more recently been experimentally proven (quite independently of me, but not of Tversky) by Wolff and Gentner (2011).

As in ordinary judgements of similarity, the non-intersecting features of the subject of comparison are heavily weighted in metaphors. In this case, however, this part of the feature set not only contains numerous features but also features directly opposed to those of the new ecumenical category. Indeed, this is exactly what makes us perceive a comparison as a metaphor rather than an ordinary judgement of similarity. In contrast, there is, of course, no resistance to be overridden in an ordinary judgement of similarity. This, I submit, is also the case with much of what Lakoff and Johnson call metaphors.

4 Lifeworlds we live in

According to Max Black (1981: 208), as quoted earlier, Lakoff and Johnson (1980a) is concerned with “shallow or trite metaphors, if they deserve the name.” In the classical treatises of rhetoric, this is the residue figure called catachresis or, in everyday language, dead metaphors. But perhaps such metaphors may be more or less close to death and there may be more or less possibility of bringing them back to life. In this sense, the term “dormant metaphors,” suggested by Perelman and Olbrechts-Tyteca 1958/2000: 405), appears to be more fitting — unless, of course, we want to bring the very metaphor contained in these terms back to life, suggesting that what happens is more like a resurrection than simply waking up from sleep. No doubt these different construals may be apt for different cases.^[5]

Let’s start with a classical example of a dead metaphor: the foot of a mountain. We are concerned with something that occupies the same position on a mountain as the feet do in relation to the human body, that is, the part that is closest to the ground.^[6] Although the body is certainly, in a sense, a part of the landscape — and thus a synecdoche in relation to it — it is easy to see the figurative aspect here justifying the division of domains, since, for a human being, nothing comes more naturally than to project his/her own body to the environment — which is part of the half assimilated lesson of phenomenology that supposedly inspired this “philosophy in the flesh” (Lakoff and Johnson 1999).

It certainly makes sense to think of this as a Peircean diagram, projecting the domain of the human body to another domain, that of a part of the landscape. Now suppose there is a small band of forest close to the top of the mountain, or perhaps close the very foot of the mountain. In the first case, I could talk about “the beard of the mountain.” In the second case, it would be possible to use the phrase “the stockings of the mountain.” None of these expressions may be particularly enlightening, but they certainly serve to bring the dead metaphor to life. That is, they transform a diagram into a real metaphor. Dead metaphors like “the foot of the mountain” are of course special in that there is no literal expression to describe the fact, at least not in English and not in any other Germanic or Romance language. But this is not the kind of example central to CMT.

Nevertheless, so-called conceptual metaphors such as “LIFE IS A JOURNEY,” favoured by CMT, can be brought to life in a similar fashion. For instance, you might say: “Starting school is like unpacking the travel bag of life.” Once again, I am not claiming that this is a great metaphor, but it is certainly a metaphor in the classical sense, because it offers a new perspective on what it means to go to school. In other words, it introduces an element of tension, which serves to highlight properties of the domain of going to school that are not immediately obvious in that domain but that are more familiar in the domain of travelling. The point is not that poets may start out from such cross-domain mappings and invent really creative metaphors, as Lakoff and Johnson also recognize — it is that already as speakers of the language we are provided with the virtual capacity to awaken or resurrect, as the case may be, such petrified syntagma.

What Lakoff and Johnson (1980a, 1980b) call “orientational metaphors” are, I submit, something much more fundamental. As they go on to suggest, most of them:

have to do with spatial orientation: UP-DOWN, FRONT-BACK, IN-OUT, ON-OFF, DEEP-SHALLOW, CENTRAL-PERIPHERAL. These spatial orientations arise from the facts that we have bodies of the sort we have and that they function as they do in our physical environment. Orientational metaphors give a concept a spatial orientation, for example, HAPPY IS UP. The fact that the concept HAPPY is oriented UP leads to English expressions like ‘I’m feeling up today’ (Lakoff and Johnson 1980b: 461f)

and in the following pages they give a long list of “metaphors” based on the first dimension, that is, UP-DOWN:

Happy is up; sad is down. […] Conscious is up; unconscious is down. […] Health and life are up; sickness and death are down. […] Having control or force is up; being subject to control or force is down. […] More is up; less is down. […] Foreseeable future events are up and ahead. […] High status is up; low status is down. […] Good is up; bad is down. […] Virtue is up; depravity is down. […] Rational is up; emotional is down. (Ibid.: 462ff)

For each of these cases, they add some “physical basis” which sounds more or less plausible but is clearly based on our human way of being in the world; thus, for instance, happy is up and sad is down because “dropping posture typically goes along with sadness and depression, erect posture with a positive emotional state” (ibid.: 462). Further light is thrown on this idea by Johnson's (1987:74) observation, that, in learning to stand up, “the baby becomes a little homo erectus,” that is, I take it, an incarnation of the human being as being (or becoming) different from other animals, something which is epitomized by the erect posture. Taking the lead from such passages, we can understand embodiment, not as some kind of neural structure as Lakoff and Johnson sometimes do, for which there is no serious proof (See Forest 2015; Guillaume et al. 2013; Fuchs 2018), but as originally conceived by Edmund Husserl and then explicated by Maurice Merleau-Ponty, as the presence of the human being in the world and, more specifically, in the Lifeworld, the world taken for granted. If so, the human body is part of the whole world experienced, and what is up on the human body is simply part of what is up in the world. In this sense, we are really dealing here with a synecdoche — or, more precisely, since this is not necessarily a relation between signs, with a part-whole relationship (which I have elsewhere termed factoriality: see Sonesson 1989). Of course, to a human being, the upper part of his/her own body is not just any part of the upper sphere. It is a marked part of a whole. And it is marked because, while at the same time as it is a part of that whole, it constitutes the point of view from which the whole is conceived. ^[7]

There is no contradiction in claiming that these “metaphors” are really, on one hand, diagrams because they do project properties of the “domain” of the human body to the “domain” of the landscape, and claiming, on the other, that the relation is one of factoriality because the body is a part of the human environment. Even though signs may cumulate iconicity (as in diagrams) and factoriality (as in indexicality), this is not even an issue here, because we are involved with basically perceptual Lifeworld meanings (See Figure 1). It is doubly misleading to talk in this context about metaphors because metaphors are signs substituting for other signs — in this case, we are really involved with the direct experience of the human Lifeworld.

Figure 1:

Orientational (or primary) “metaphors” as (a) diagrams and (b) factoriality.

I am not the first to note the essential difference between these “orientational metaphors” and the others. Grady (1997) calls them “primary metaphors,” and this term is even taken up later by Lakoff and Johnson (1999: 59), who suggest that they “are a consequence of the nature of our brains, our bodies, and the world we inhabit.”^[8] This is clearly something quite different even from dead metaphors. Actually, it is unclear how you would be able to bring such a “metaphor” to life — something which might perhaps be suggested as a kind of operational definition of dead or dormant metaphors. No matter how you vary the description of the upwards movement (in the example above, but equivalently in all other cases), you are still within the same space shared by the human body and the environment.

Instead, I maintain, we are here in the domain of what Husserl called the Lifeworld and what James Gibson termed the world of ecological physics. The idea of a common-sense world has reappeared numerous times in philosophy as well as in the social sciences, no doubt suggested independently by different scholars. Husserl posits the Lifeworld to explain the foundation on which the models of the natural sciences are constructed, both serving as the primary objects studied and transformed by the model and as the common sense world in which the scientists are accomplishing their work. Indeed, you cannot treat the accelerator permitting you to study the electrons as being at the same time a bundle of electrons itself. Students of Husserl such as Aron Gurwitsch, Alfred Schütz, Maurice Merleau-Ponty, and Herbert Marcuse considerably extended not the meaning but the function of the concept of Lifeworld, using it to explain social reality itself. We owe to Schütz, in particular, the characterization of the Lifeworld as “the world taken for granted.” The “commens” characterised by Peirce (1998: 478) would seem to be a similar domain of shared assumptions. Greimas certainly took the idea of a semiotics of the natural world from Husserl via Merleau-Ponty. Coming from a very different tradition, Jakob von Uexküll introduced the notion of Umwelt to serve as some kind of world taken for granted by the animals — although, of course, in a deeper sense, the tick and his kin do not have the choice of taking anything for granted at all. When the psychologist Gibson (1982: 217ff) postulated the world of “ecological physics” to explain the possibility of immediate perception where the older school of constructionists had to suppose complex calculations, he did not refer to Husserl explicitly anywhere in his writings, but he often uses the same phrases and examples. Many of the “laws of ecological physics,” formulated by Gibson and which are defied by magic (that is, our expectancies held in the common sense world, not anything going against any contemporary accepted theory about the world in the natural sciences), are also such “regularities [that] are implicitly known,” as, for instance, that no substantial object can come into existence except from another substance; that a substantial detached object must come to rest on a horizontal surface of support; that a solid object cannot penetrate another solid surface without breaking it, etc.

The Lakoff and Johnson type of “metaphors” should be understood in this framework: as diagrams, but not as just any diagrams: as diagrams founded on the invariants of the common human Lifeworld.^[9] Some of these pseudo-metaphors, however, should rather pertain to one of the instances of what Aron Gurwitsch (1974) called the particular sociocultural lifeworlds characterizing a society or a group, which may be added to the common structures of the Lifeworld, which we all share as human beings. A case in point may be “TIME IS MONEY,” which, according to Black (1981: 201), “admits of no direct equivalent in French, and other ‘Western’ languages. (Larousse calls it an English proverb.)” Since Black (and Larousse) wrote this, however, the globalization of Anglo-Saxon culture may well have proved Lakoff and Johnson right in this respect. In any case, whether or not “time is money” is (or was) an “English proverb,” it is probable that also the French and many other nations have an idea about time being a limited resource, although they may have expressed this notion in other terms. The problem is rather whether, thus generalised, this can be considered an invariant of the human condition, and, if so, how far back in time we can suppose this to have been the case.

Be that as it may, it might well be argued that dead metaphors are (at least potential) metaphors. But this is certainly not the case with “orientational” or “primary metaphors.” They are rather literal or, if you like, “physiognomic” properties of the Lifeworld we inhabit (See Sonesson 2013, Sonesson 2015).

5 Second dilemma: Pictorial vs. verbal metaphors

In their seminal study of a coffee pot picture, which is also a cat picture, Groupe µ (1976: 45) asks whether there is something in pictures that can be called a metaphor.^[10] They observe that a pictorial metaphor differs from an ordinary linguistic one in that its expression plane must contain the features of both contents involved — in this case, of the cat and the coffee pot. Some features have admittedly been suppressed: in this case, there is nothing to indicate either the handle or the lid of the coffee pot, but it is more difficult to discover any part of the cat that is lacking. Both objects, in any case, may be easily recognized. It is unclear from the paper whether Groupe µ therefore wants to abandon the category of pictorial metaphor entirely, but this conclusion is borne out by consulting their later work in which quite a different approach to figurativity in pictures is suggested.

In this work, Groupe µ (1992) offers a cross-classification of the figures of visual rhetoric into those that are present or absent, and conjoint or disjoint.^[11] In the latter conception, a figure is in absentia conjoint, or a trope, if the two units involved occupy the same place in the statement, one being totally substituted for the other. It is in praesentia conjoint, an interpenetration, to the extent that the units appear in the same place with only partial substitution of one for the other. There will be a figure which is in praesentia disjoint, a pairing, if the two entities occupy different places without any substitution taking place. Finally, the figure will be in absentia disjoint, a projected trope, when only one unit is manifested, while the other is exterior to the statement, but is projected onto it. In spite of its elegance, this model is in the end problematic in several ways, as I have demonstrated elsewhere (cf. Sonesson 1996a, Sonesson 1996b, Sonesson 1997, Sonesson 2003, Sonesson 2005, Sonesson 2010a), where I have proposed a more complex model, which is however basically conceived in the same fashion. In the present context, I think it will be sufficient to point out that, given the numerous possibilities of amalgamation of the two categories involved in the picture, it is impossible to account for all the differences in terms of what is conjoint and disjoint, and what is present and absent.

The difference between the tropes and the interpenetrations, in Groupe µ’s scheme, has to do with the relations of wholes to parts, that is, factoriality: in the first case, epitomised by the panel showing Captain Haddock (a personage in Hergé’s comic strip) with bottles instead of pupils in his eyes, one sub-ordinated element of a interpretative scheme has been substituted for another in a whole that otherwise fulfils the expectations generated by the scheme. In another case, which is exemplified by the cat-coffeepot (“la chafetière”), two conflicting schemes of interpretation are actualised, and neither of the sets of predictions generated for the whole by the two schemes are entirely born out. It is in this respect that these cases are opposed to the third one, which, in terms of the theory of indexicality, is a pure case of contiguity, the pairing. Since both are some kind of indexical grounds, factoriality and contiguity may be seen as extreme cases on a scale of relative integration, of the kind that is expected to obtain in the Lifeworld. Consider the case of a picture representing a bottle of liquor inside the Coliseum as if the latter were an ice-pail (cf. hichSonesson 1989, Sonesson 1997): here two independent objects are represented, but they are not merged as in the interpenetrations; nor is (a part of) another object substituted for a part of the globally presented object, as in the tropes; nor are the objects simply put side by side in a surprising combination, as would seem to be the case with the pairings.

In the second place, as already observed by Kerbrat-Orecchioni (1977), all (or most) pictorial figures require both some amount of presence and some amount of absence: a presence, normally, of that which is not expected, and an absence of that which is expected.^[12] The joint appearance of Coliseum and the bottle (of this size and in this precise position) is unexpected, whereas there would be nothing strange in encountering a cat and a coffee pot or a bottle and Haddock side by side — the strangeness largely resides in the relations of part to whole. Yet, unlike the unexpected combinations of the pairings, the Coliseum picture does anticipate the presence of a well-defined element which is not there to be seen, that is, an ice-pail. Together with the ice cubes, the bottle requires the presence of an ice-pail, like the body scheme requires the pupils of Haddock’s eyes: yet the ice cubes, the bottle, and the ice-pail do not make up any complete whole, only a set of interconnected objects. The Coliseum picture does not break up the unity of an object as we know it in the world of our experience, but it associates things that are normally parts of different sets.

On the other hand, Kerbrat–Orecchioni (1977) is clearly mistaken in her claim that pictorial metaphors depend on a contradiction on the expression plane, for although the expression is necessarily involved, it rather serves to mediate the contradiction. The contradiction in Magritte’s “Le viol” is not between one line or point and another, but between the breasts and the eyes, the nose and the navel, the mouth and the pubis, and so on.^[13] Indeed, it is as something “seen in” the pictorial surface rather than as objects in the world that the content participates in the pictorial metaphor: for the real coffee pot and the real cat have hardly any common features of appearance, and the Coliseum and the ice-pail also have very few.

The great insight contained in the work of Groupe µ, which I have pursued in my own papers, is that pictorial figures tend to function in a different way from verbal ones. This is an insight lost on most scholars who have simply tried to extend the characteristics of linguistic metaphors to other semiotic resources. A typical representative of this brand of scholars is Charles Forceville (2004–2009), who has, however, recently started to attend to some of these differences (see Forceville 2008, 2017) without really disposing of the means to do so. It is important to realise that pictorial metaphors are different from verbal metaphors because pictorial signs are constituted in ways which differ from verbal signs. The key to this difference is found in the analysis of pictorial consciousness proposed by Edmund Husserl (1980), which I have extended in many of my works (see, for instance, Sonesson 1989, Sonesson 2011, Sonesson 2015c). According to this analysis, the picture sign has at least four layers: the picture thing (originally the “physical picture”), the picture object, the picture subject (“Bildding,” “Bildobjekt,” and “Bildsujet,” respectively) and (my addition) the picture referent. The picture thing is that which may hang lop-sided on the wall. The picture object it that which we can “see in,” a kind of quasi-perception of the shapes and colours arranged on the surface of the picture thing. The picture subject is that same picture object, to which is added the knowledge we have of the things depicted, which may not be rendered in the picture. The picture referent is the real three-dimensional object that may be retrieved in the world of our experience. It is the existence of these layers that allow for the intricate figurativity of the picture sign.

Nevertheless, I think Groupe µ — as well as myself as their follower — have been wrong in thinking that the term metaphor is not adequate for visual figurativity. In fact, it seems that most visual figures are really metaphors (even if they often are some other kind of figure at the same time) in the sense that they posit a similarity between one item and another. It could be argued that this only makes them into diagrams in the Peircean sense of the term. However, because of the very nature of pictorial figurativity, tension comes easily to such likenesses. In the work of Groupe µ, as well as in my own, there is a confusion of levels that results from pictures being not only generally iconic but also pictorial — that is, what we really have been studying is how the visual metaphors are brought about by means of the manipulation of the pictorial surface. And since the pictorial surface is normally very different from the objects it depicts (e.g., the real cat and the coffee pot are hardly similar), tension can usually be had for free.

To understand the kind of classification I have suggested in my earlier papers, as well as that of Groupe µ, I submit that we must attend to the difference between two kinds of indexicality (See Sonesson 2010b). We already saw above that pictorial figurativity is very much involved with mereological relations such as contiguity and factoriality (part-whole-relations), that it to say, with indexicality. As I have pointed out elsewhere (most recently in Sonesson 2010b), there are really two very different kinds of indexicality (only the first of which really corresponds to Peirce’s definition according to which it is independent of the sign relation). In the case of an abductive index, you have gained knowledge from past experiences that allow you to draw a conclusion from something you observe at present. Contiguity or factoriality is not found in the sign but rather before it. For instance, you observe something that in your earlier experience has always appeared together with another object, and you posit the other object as part of the experience of the former perception. When you see a foot print, you would normally presume an earlier presence of a foot. If you recognize an object as normally being a part of a whole, you would anticipate the presence of the whole. If a branch appears over the wall, you may reasonably expect there to be a tree at the other side. In the case of performative indexicality, on the other hand, contiguity and factoriality do not exist outside the sign but are created by it. For example, the arrow, the pointing finger, and the personal pronouns create the contiguity they signify. Here, there is no previous relationship and thus no possible knowledge about it, but the relation is created by the concurrent appearance of different objects in contiguity with each other or one as part of the whole constituted by the other.

Basically, pictorial metaphors are created performatively: one item is placed in contiguity or in any other mereological position with respect to another item. The meaning of the pictorial metaphor, on the other hand, is always partly iconic though it may also contain abductive indexicality. Thus, for instance, if the stick you normally use to hold an ice cream bar is attached to a cow,^[14] this operation on the pictorial surface is itself performative, but it also serves to suggest abductively that the ice-cream bar, which would normally be at the end of the stick, is somehow similar to a cow. This leads to a preliminary taxonomy as in Table 1.

Table 1:

A preliminary taxonomy of performative indices creating pictorial figures.

1. Table constructed by the author

6 Beyond the second dilemma: Verbal and pictorial metaphors

As we have seen, it is necessary to analyse the pictorial surface in order to understand the workings of pictorial figurativity. However, the application of a similarly detailed analysis to the expression level of verbal signs may also profit the study of verbal figures. Although the contiguity of the expression plane in verbal metaphors is constrained (more or less) by syntax, there clearly are different possibilities of expression which have not been properly investigated. Because of the nature of verbal signs, notably because of double articulation (also called duality of pattering), some kinds of metaphors (and perhaps other figures) that figure prominently in pictures may be impossible or at least difficult to bring about in language. But the opposite is no doubt also the case.

The intricacy of mereological relations on the pictorial surface can go well beyond contiguity. This is at least one of the reasons that account for the fact that pictorial metaphors are sometimes much more powerful than their verbal counterparts. In portmanteau words, just as in pictorial metaphors, the expression planes of two signs are merged instead of being related by means of a third sign as in a verbal metaphor. However, portmanteau words are rare phenomena that are difficult to produce because the fusion of two verbal expressions tends to hamper recognition. In contrast, no problems of this kind would seem to exist for pictorial signs — first because the picture referents of common Lifeworld things can be adapted in numerous ways in order to go together, as we saw in the case of the cat and the coffee pot, and second because pictorial signs are easily embedded in other pictorial signs with no or only small consequences for recognisability. A case in which such a portmanteau world can be created, and which was in fact created by Groupe µ (1976), is the term “chafetière” used as a designation for the hybrid cat-coffeepot. In Swedish, a similar but more simplistic word can be created as a designation of this picture, “kattekanna,” which combines “katt” (“cat”) and “kaffekanna” (“coffeepot”). In English, as in many other languages, no similar word seems to be available (see Figure 2).

Figure 2:

Typical surface mechanisms of (above) verbal metaphors and (below) pictorial metaphors.

I have suggested elsewhere (Sonesson 2015a) that, unlike verbal metaphors, pictorial metaphors tend not to possess any inversion. As I noted above in Section 3, quoting Amos Tversky (1977: 328), a metaphor such as “a man is like a tree” does not carry at all the same meaning as “a tree is like a man,” for in the first instance, we attribute roots to the man, and in the second instance we give a life history to the tree. A by now classical example of the same phenomenon is the difference between “the surgeon is a butcher” and “the butcher is a surgeon” (see Figure 3). It may be difficult even to express an iconic relation between a butcher and a surgeon in a picture using the same or even a similar operation on the level of the expression plane.^[15] It could no doubt be done, showing half a picture of a butcher joined to half a picture of a surgeon.

Figure 3:

The workings of the butcher-surgeon metaphors.

But the question remains how we would go about deciding whether the meaning of this picture corresponds to “the surgeon is a butcher” and “the butcher is a surgeon.” Indeed, I suggested elsewhere that the cat-coffeepot (Sonesson 1997, Sonesson 2005, Sonesson 2010a, see also Section 3) makes the coffee-pot more cat-like and the cat more similar to a coffee-pot, and the same thing could be said about the picture of the butcher and the surgeon. In fact, directionality is often appended to the pictorial metaphor, because of verbal labels, genre conventions and other kinds of world knowledge. Once we know that the “chafetière” was created as an advertisement for a coffee brand, we realize that, in this case, it is the cat properties which have to be transferred to the coffee pot and, indirectly, to this particular brand of coffee. The genre conventions of advertisements decide which part of the picture is the standard of comparison. And if a real tree trunk has been carved so as to resemble a human body, we would naturally understand this as a transfer of tree properties to the human body and, indirectly, to the human being as a whole. It may not be this easy to make the butcher-surgeon picture directional. It would probably be necessary to use a verbal label.

The verbal metaphor can however be conveyed in another way. Instead of saying “the surgeon is a butcher,” we may opt for the expression ”the surgeon lifted her chopper” (see Figure 4a) and instead of “the butcher is a surgeon,” we could say “the butcher applied his scalpel” (see Figure 4b). “Chopper” brings with it its “selection restrictions” (Katz) or “classemes” (Greimas), which are contrary to the semantic features (semes) of “surgeon” and contradictory to its (positive) value. Or, to use a somewhat more updated semantic theory, the word “chopper” introduces elements of the frame (script, scheme, etc.) of “slaughter” into the overall frame of “surgeon,” occasioning a lower evaluation of (some aspects of) the primary frame. Similarly, “scalpel” brings with it its “selection restrictions” (Katz) or “classemes” (Greimas) which are contrary to the semantic features (semes) of “slaughter” and contradictory to its (negative) value. Or: “scalpel” introduces elements of the frame of (script, scheme, etc.) “surgeon” into the overall frame of “slaughter,” occasioning a higher evaluation of (some aspects of) the primary frame (see Figure 5). This variety of the slaughter-butcher metaphor can be expressed in pictures with this directionality maintained. It is sufficient to adapt a picture of a surgeon by changing the scalpel in the hand into a chopper, or the to adapt a picture of a butcher by changing the chopper in the hand into a scalpel. The reason why we here obtain a true unidirectional metaphor without having to add anything in particular to the picture (except, of course, for the recognition of the roles played by the persons and objects, including their habitual evaluations) is no doubt that the human figures playing their respective professional roles spatially dominate the pictorial surface. Perhaps a scalpel touching a piece of meat without the butcher being seen could convey the idea of a butcher working like a surgeon, but it would be less unambiguous than the full picture.^[16] This suggests there are several factors involved here.

Figure 4:

Pictorial equivalents of (a) “The surgeon lifted her chopper” and (b) “The butcher applied his scalpel.” Depictions by the author.

Figure 5:

The workings of the chopper-scalpel metaphor.

Something which seems very difficult, if not impossible, is to render in pictures is a verbal metaphor of the type “Fred wallowed in self-pity” (Figure 6). The verb “wallowed” brings with it its selection restrictions (classemes), which are contrary to the semantic features (semes) of “Fred [a human being],” contradictory to its (positive) value, contrary to the semantic features (semes) of “self-pity,” and adds to it a (negative) value. In the end, “wallow” may only retain the suggestion of a certain movement pattern and project a negative value to its subject and its prepositional phrase. As a pictorial equivalent, one may envisage some kind of double exposure depicting at the same time two full events comprising agent, object, and the action connecting them, the agent being in one case a pig and in another a human being. The problem, however, is not only how to determine the directionality of such a metaphor, but, more importantly, it is not easy to imagine how the residue pattern of wallowing should be visualized in the rendering of the second act involving the human being. Even a verbal label may not be sufficient to do the trick if it is not precisely the reproduction of the whole verbal metaphor.

Figure 6:

The mechanism of the metaphor “Fred wallows in self-pity”.

7 Third dilemma: Peirce on metaphors

The present discussion of the classical notion of metaphor leads us to the conclusion that most of what Lakoff and Johnson call metaphors are better understood as diagrams in the Peircean sense. Am I, therefore, arguing that the Peircean notion of metaphor corresponds to the classical notion — either to the Tension model, the Overlap model, or a combination of both? This leads to the third dilemma: the status of Peirce’s theory of metaphor. In one of his many discussion of Peirce’s notion of metaphor, Jappy (1996: 219) quotes the following evaluation of Peirce’s metaphor definition by Pharies (1985: 36): “This is one of the more obscure passages I have had occasion to quote, and I do not profess to understand it completely.” This definition, it will be remembered, runs as follows: “those [hypo-icons] which represent the representative character of a representamen by representing a parallelism in something else, are metaphors” (CP 2.277, EP 2: 274). Unperturbed, Jappy goes on to draw some schematic pictures meant to illustrate this relationship — but, as much as I have appreciated Jappy’s other contributions to the interpretation of Peirce, I find them unenlightening (see also Jappy 2010: 141ff). It is in the course of this discussion, however, that Jappy suggests the example-metaphor quoted above: “Fred wallows in self-pity.”

Elsewhere, generalizing from the numerous definitions given by Peirce, I have tried to assign an intuitive sense to his categories of Firstness, Secondness, and Thirdness (Sonesson 2009, Sonesson 2013a): everything said about Firstness, I suggested, boils down to a meaning roughly paraphrased as “something there;” those phrases describing Secondness are equivalent to “reaction to the appearance of something;” and Thirdness can be reduced to “observing the appearance as well as the reaction to the appearance.” I will not here repeat again all the considerations which led me to this very tentative result, because they are only marginally relevant to the present discussion. More importantly in this context, I have suggested in my work from very early on (Sonesson 1989) that what Peirce characterizes as a hierarchy of signs is really a hierarchy of meanings. That is, iconicity as such carries an experiential meaning, but two iconicities are needed to form an iconic ground and, for there to be a sign, there must also be a sign function. Indexicality, since it is a relationship, is already a ground, but the sign function is needed to make this into a sign. Peirce clearly thinks that symbolic signs are created ad hoc, and that is often the case — but, in the human Lifeworld, there are many symbolic grounds that are not primarily signs (traffic rules, rules of particular games, and so on, but also more generally rules of behaviour, whether they are universal or specific to certain cultures) but that may be turned into signs by espousing the sign function.

This whole argument depends on having a clear definition of the sign and, as far as I know, nobody has offered such a definition, except the one I have given in many of my papers (certainly very much inspired by the notion of Anzeichen proposed by Husserl and the idea of the semiotic function formulated by Piaget).^[17] It runs as follows:

1. It contains (a least) two parts (expression and content) and is as a whole relatively independent of that for which it stands (the referent);

2. These parts are differentiated from the point of view of the subjects involved in the semiotic process, even though they may not be so objectively, i.e., in the common sense Lifeworld (except as signs forming part of that Lifeworld);

3. There is a double asymmetry between the two parts because one part, expression, is more directly experienced than the other;

4. And because the other part, content, is more in focus than the other;

5. The sign itself is subjectively differentiated from the referent and the referent is more indirectly known than any part of the sign.

In this sense, there are a lot of iconicities and iconic grounds that are not signs, just as there are many indexicalities that are not signs; and, in spite of the symmetry suggested by Peirce, there are a lot of symbolicities which are not signs (see Table 2).

This poses the question how we should understand the Peircean metaphor. While still remaining an iconicity (Firstness), it is also a sign (Thirdness), but this is something it has in common with the image and the diagram. Since the image applies Firstness (iconicity) to itself, we can consider it to be unproblematic. As we have seen above, it is not difficult to think of the way a Secondness (a relation) can be applied to a Firstness (iconicity): indeed, we can think of it both as a diagrammatic ground (which may be case with what Lakoff and Johnson call “primary metaphors”), that is, combined with Secondness, as well as a diagrammatic sign, combined with Thirdness, which would be the case with Lakoff and Johnson’s other “metaphors” and “dead metaphors” (see Section 4 above). It is much more difficult to imagine how iconicity (Firstness) and sign character (Thirdness) can be combined once again with Thirdness. Is this simply Peirce’s way of saying that a sign has been substituted for another (more proper) sign, as in the classical definition of metaphor (referred to in Section 2 above)?

But if Thirdness should be understood as “observing the appearance as well as the reaction to the appearance,” according to the interpretation of Peirce suggested by Sonesson (2009, 2013a), we should rather conceive the metaphor as some kind of meta-statement. Lenninger (2019, this volume ) also takes the Peircean metaphor to be a “metarepresentation.” This suggests that metaphors, in the Peircean sense, possess, in some sense, a meaning which is even further removed from direct perception than what I have so far supposed. Perhaps this is why Jappy gives such a complex example of a metaphor as “Fred wallowed in self-pity.” It involves not only substituting Fred for the pig and self-pity for the mud, but also changing the nature of what happens between them. Whatever metaphoric wallowing is, it certainly must be something very different from what the pig habitually does. If so, metaphors (and other rhetorical figures) in the classical sense are only diagrams in Peirce’s sense, but with the added effect of tension between the domains. While I do think there is something to be gained in clarity by giving up the broad Lakovian notion of metaphor for some variety of the classical notion, I am not sure this is true of narrowing it down even further to the Peircean notion if it is taken to be some kind of meta-statement. In my conception, metaphors such as “Fred wallowed in self-pity” are simply particularly intriguing case of true metaphoricity.

8 Some conclusions

In this paper, I have argued that the classical notion of metaphor is much more fruitful than the notion introduced by Lakoff and Johnson: indeed, I have claimed that, while my notion of metaphor may not be identical to the notion favoured by Peirce (Section 7), Lakoff and Johnson’s metaphor is only a diagram in the Peircean sense. Suppose that Lakoff and Johnson (or their followers, and the followers of their followers when all the former are dead) would accept that what they call metaphors are really diagrams. This would certainly be much more than a change of terms because, although Peirce’s notion of metaphor remains obscure, his distinction between image, diagram, and metaphor supposes there to be a structure to iconicity. The difference between diagram and metaphor allows us to spell out the difference between similarity judgements, many of which are part of what is taken for granted in the (particular socio-cultural) Lifeworld in which we are living, and metaphors, which are in some way or other trying to go beyond this world taken for granted.

There are, however, at least two classical models of the metaphor, which I have termed the Overlap model (Section 3) and the Tension model (Section 2). At the end of the journey, it seems that only the Overlap model is valid for all kinds of metaphor: it certainly accounts for the difference between metaphors and similarity judgments. If there is also tension, then it comes in very different degrees: if we take the catachresis (the dead metaphor) to be (close to) the zero degree, then we can modify it or add to it in order to obtain a real metaphor. This is also true of many of Lakoff and Johnson’s so-called metaphors, such as “LIFE IS A JOURNEY.” Both catachresis and these kinds of “metaphors” are really the opposite of metaphors in the classical sense, that is, they are part of what is taken for granted in the (particular socio-cultural) Lifeworld which forms the background against which metaphors stand out.

This is even more true of what Lakoff and Johnson have come to term “primary metaphors:” they are basically literal projections of human embodiment to the environment. Thus, if the catachresis-type of Lakovian metaphors may still fulfil the Overlap model, if not the Tension model, none of the models are fitting for the so-called primary metaphors. In the second part of the paper, we discussed the possibility of pictorial metaphors, which I vindicated, arguing that the denial of their possibility is based on a confusion having to do with the surface serving as the expression for pictorial signs which obey other mechanisms than is the case with verbal metaphors. While admitting that directionality is not always intrinsic to pictorial metaphor, unlike the case of the verbal one, I have shown that not only verbal labels but also genre, the nature of the depicting surface, and other kinds of world knowledge often assign such a directionality also to the former. We also saw that some kinds of metaphor seem to be impossible, or at least difficult to bring about, in the verbal domain and others in the pictorial domain. This all serves to suggest a more complex and thus more productive framework for metaphor research.