one more yard and scores. Blending these two spaces gives us a blended space in which there is now an element that is ‘absence of one more yard of progress’. We might express this blend by making the negative element explicit: The Rams won by stopping the Titans from advancing one more yard. But the front-page headline was in fact: Rams win by a yard. It then becomes possible to refer to the Rams’ ‘one-yard win’. One-yard win is then identical in its integration patterns with caffeine head- ache. In fact, this example reveals something else. The more conventional pattern for expressions like winning by a yard and winning by a nose is, of course, that of a race where the winner crosses the finish line a yard ahead of the runner-up. This does not feel intuitively like a counterfactual expression—it seems as if we can ‘‘see’’ that fateful yard right there on the photograph of the finish. But if you think twice, you can see that this more standard notion of a ‘one-yard win’ is really also counterfactual. The crucial yard is the one that the loser failed to cover, just as the crucial yard in the Super Bowl win was the yard that separated the ball carrier from the goal line. The central problem of language, the one that must be solved if human language is to emerge, is that relatively few linguistic patterns—such as words, syntactic patterns, and suprasegmental patterns—must be app licable to vast ranges of conceptual structure. Language must be available to be used in any and every situation. Human language has this property of being ‘‘equipotential’’: for any situation, real or imaginary, there is always a way to use language to express thoughts about that situation. A word like food or there, for example, must apply very widely if it is to do its job. The same is true of grammatical patterns inde- pendent of the words we put in them. Take the resultative construction in En- glish, which has the form A—Verb—B—Adjective, where the Adjective de- notes a property C (see Goldberg 1995). It means ‘A do something to B with the result that B have property C’, as in Kathy painte d the wall white.Wewantitto prompt for conceptions of actions and results over vast ranges of human life: She kissed him unconscious, Last night’s meal made me sick, He hammered it flat, I boiled the pan dry, The earthquake shook the building apart, Roman imperialism made Latin universal,andsoon.Wefinditobviousthatthemeaningofthe resultative construction could apply to all these different domains, but applying it thus requires complex c ognitive operations. The events described here are in completely different domains (Roman politics versus blacksmithing) and have strikingly different time spans (the era in which a language rises versus a few seconds of earthquake), different spatial environments (most of Europe versus the stovetop), different degrees of intentionality (Roman imperialism versus a forgetful cook versus an earthquake), and very different kinds of connection between cause and effect (the hammer blow causes the immediate flatness of the object, but eating the meal one day causes sickness later through a long chain of biological events). This very simple grammatical construction allows us to perform a complex conceptual integration which in effect compresses over Vital Relations of Time, 390 mark turner Space, Change, Cause-Effect, Intentionality, and Identity. The grammatical con- struction provides a compressed input space with a corresponding language form. It is then blended in a network with another input that typically contains an unintegrated and relatively diffuse chain of events. So, if it is our job to turn off the burner under the pan that has zucchini in boiling water, and we forget about it and all the water evaporates, we can say, confessionally, ‘‘No zucchini tonight. I boiled the pan dry. Sorry.’’ In the diffuse input, the causal chain runs from forgetting to the invariant position of the burner knob, to the flow of gas, to the flame, to the temperature of the pan, to the temperature of the water, to the level of the water, to the dryness of the pan. The agent performs no direct or indirect action on the pan at all. But in the blend, the compressed structure associated with the grammatical construction is projected together with some selected participants from the diffuse chain of events in the diffuse input. In the blend, the agent acts directly on the pan. Moreover, although the boiling of the water is an event and its cause was some- thing the agent did or did not do, there is cause-effect compression in the blend so that in the blend, although not in the input spaces, boiling is an action the agent performed on the pan. As this example of the resultative construction in English shows, the simplest grammatical constructions require high abstraction over domains and advanced conceptual integration. The most advanced form of conceptual integration is called ‘‘double-scope’’ integration. Human beings are uniquely extraordinarily adept at double-scope integration. For them, it is a routine operation, and the one that makes human language possible. A double-scope conceptual integration network has in- puts with different (and often clashing) organizing frames and an organizing frame for the blend that includes parts of each of those frames and has emergent structure of its own. In such networks, both organizing frames make central contributions to the blend, and their sharp differences offer the possibility of rich clashes. Far from blocking the construction of the network, such clashes offer challenges to the imagination, and the resulting blends can be highly creative. Human beings are able to blend very different conceptual inputs in a ‘‘double-scope’’ way and to use language attached to the inputs in order to prompt for the new blend. Language is a consequence of our ability for double-scope integration. There is every reason to think that some species are able to operate efficiently in separate domains of, say, tool use, mating, and eating without being able to perform these abstractions and integrations. If that is so, then grammar would be of no use to them, because they cannot perform the conceptual integrations that grammar serves to prompt. But could they just have a simpler grammar? The only way they could have a simpler grammar and yet have descriptions in lan- guage for what happens would be by having separate forms and words for ev- erything that happens in all the different domains. But the world is infinitely too rich for that to be of any use. Trying to carry around ‘‘language’’ of that size would be crippling. The evidence does not suggest that primates have compensated for lack of language by developing, for example, one million special-purpose words, conceptual integration 391 each conveying a special scenario. On the contrary, while primate species have some specific ‘‘vocalizations’’ (e.g., in response to a potential predator), the best efforts to teach words to chimpanzees cannot get them past a vocabulary of about two hundred items. Having a handful of vocalizations is clearly a help, but evolution has found no use in trying to extend that strategy very far. The ex- traordinary evolutionary advantage of language lies in its amazing ability to be put to use in any situation. Double-scope conceptual integration is the key to the amazing power of the equipotentiality of language, which we take for granted and use effortlessly in all circumstances. To date, blending analyses of specific grammatical patterns include studies of single words like safe; fixed expressions like two-edged sword; clausal constructions like the resultative construction, the ditransitive construction, and the caused-motion construction in English; Y-of networks like Ann is the boss of the daughter of Max and Causation is the cement of the universe; nominal compounds like boat house, house boat, and jail bait; adjective-noun compounds like guilty pleasures, likely candidate, and red ball; morphological combinations in a single word like Chunnel; grammatical forms in American Sign Language that rely on blends involv- ing grounded mental spaces; forms that call for blends involving counterfactual spaces; forms with syntax that results from blending, as in double-verb causitives in French that use the verb faire; morphological causatives in Hebrew; and the mechanisms of polysemy. The studies in this list are presented, summarized, and cited in Fauconnier and Turner (2002). Others are presented online at Turner (1995–2006). REFERENCES Fauconnier, Gilles, and Mark Turner. 2002. The way we think: Conceptual blending and the mind’s hidden complexities. New York: Basic Books. Goldberg, Adele E. 1995. Constructions: A construction grammar approach to argument structure. Chicago: University of Chicago Press. Langacker, Ronald W. 1987. Foundations of cognitive grammar. Vol. 1, Theoretical prereq- uisites. Stanford, CA: Stanford University Press. Langacker, Ronald W. 1990. Concept, image, and symbol: The cognitive basis of grammar. Berlin: Mouton de Gruyter. 392 mark turner Langacker, Ronald W. 1991. Foundations of cognitive grammar. Vol. 2, Descriptive appli- cation. Stanford, CA: Stanford University Press. Travis, Charles, 1981. The true and the false: The domain of the pragmatic. Amsterdam: John Benjamins. Turner, Mark. 1995–2006. Blending and Conceptual Integration (Web site). http:// blending.stanford.edu. conceptual integration 393 chapter 16 ICONICITY willy van langendonck 1. Introduction Iconicity can be contrasted with ‘‘arbitrariness,’’ or in Peirce’s terms, iconic is the opposite of symbolic. What we now call ‘‘iconicity’’ was until fairly recently re- stricted to mainly onomatopoeia. As a consequence, iconicity was largely neglected in linguistic theory, as it constituted a marginal phenomenon in the lexicon of a language. At best, the notion was expanded to encompass ‘‘sound symbolism,’’ but at least in the Indo-European family, languages seemed to make little use of such a device. It was something to be found in music rather than in human natural language, where ‘‘arbitrariness’’ was thought of as being an essential feature. In its crudest formulation, Saussure’s principle of the ‘‘arbitrariness of the linguistic sign’’ stated that there was nothing ‘‘X-like’’ about a word ‘‘X’’ in any given language. The form and meaning of a word were regarded as independent of each other (see Haiman 1985: 2). This is, of course, especially the credo of autonomous linguistics. The principle of iconicity challenges the monopoly of arbitrariness. To be sure, this does not mean that there is more onomatopoeia or sound symbolism in the world’s languages than has been commonly assumed. Accepting double articula- tion as an unchallengeable universal, Haiman (1985: 3) argues that there is no con- stant correlation between submorphemic sounds and meanings; put differently, words of similar sound will not necessarily be words of similar meaning: we should not expect, and do not find, semantic homogeneity among words like pod, pot, and pox (see also Bauer 1996). Admittedly, sound symbolism cannot be completely ignored, as certain languages clearly make more use of it than others. For example, sound symbolism is prominent in Yoruba and especially Japanese, where gitaigo is an important aspect of linguistic elements (Kakehi, Tamori, and Schourup 1996). Still, this chapter will follow Haiman in maintaining that iconicity is not primarily to be found in the lexicon, at least not in the inventory of its roots, which are ar- bitrary for the most part. Rather, it should be looked for in the system of gram- matical rules for combining these roots to express complex concepts. Thus, Hai- man’s concern—and ours—is with the grammars of languages. 1.1. Iconicity and Linguistics Iconicity can be found not only in language but also in other domains of the world of signs. In general, there is iconicity if something in the form of a sign reflects something in the world (normally through a mental operation). For language, this means that something in the form of a linguistic sign reflects (through its meaning) something in its referent (Mayerthaler 1980, 1988). In this respect, it is no coincidence that it was a philosopher who elaborated on different kinds of signs, among them ‘‘icons’’: Charles Sanders Peirce established the classic conception of iconicity. 1 In linguistics, however, ‘‘iconicity’’ is a relatively recent term, even though the phenomenon as such has been studied ever since the Ancient Greeks, more in particular since Plato’s Cratylus (see for a short overview Swiggers 1993). The Stoic philosophers went particularly far in this: in order to show original insights in nature and hence the ‘‘natural’’ basis of language, they argued that at least the first words in a language imitated nature by means of on- omatopoeia or ‘‘articulatory mimesis.’’ For example, the liquid sound [r] imitated flowing, as in Greek rhe ^ oo ‘to flow’. Certain etymologists in the Middle Ages and even up to the eighteenth century defended similar views (e.g., C. de Brosse and Antoine Court de Ge ´ belin). From the nineteenth century, this naive conception of iconicity has been rejected. Modern ideas of iconicity date back only to the twentieth century and even then did not find a place either in nonfunctional structuralist or in generative theories. Thus, Saussure spent only a few pages on onomatopoeia ([1916] 1967 101–2) and the notion ‘‘motive ´ ’’ (180–84). He introduced the concept of motivation to describe compound signs like the French word dix-neuf ‘nineteen’. Although the components of this word are themselves arbitrary, the compound is, in contrast, ‘‘relatively motivated.’’ Thus, Saussure’s notion of motivation appears to coincide with Peirce’s concept of ‘‘diagrammatic iconicity.’’ Saussure also seems to link ar- bitrariness to the lexicon, and motivation to the grammar. Finally, he pointed to the phenomenon of analogy or isomorphism in case paradigms (221; for analogy, see Tuggy, this volume, chapter 4). For Jakobson, as for Benveniste, Bolinger, Greenberg, and others, many lin- guistic universals reflect, in a rather obvious way, common perceptions about our world. For example, Jakobson (1965) pointed out that the relationship between the elements of a syntagm containing two sentences corresponds to the relationship between the events described by those sentences: after all, sentences, like events, occur in time, and the medium of language is structurally adapted to the iconic iconicity 395 display of temporal succession. On the morphological level, Jakobson linked ico- nicity to the phenomenon of markedness (see below). Benveniste (1946) consid- ered the three ‘‘persons’’ in verbal and pronominal paradigms and argued that the third person is in fact the nonperson since it is the person that is absent from the speech act situation. In a very large number of languages, this nonperson is icon- ically represented by a nondesinence in the verbal paradigm. Such is the case in Hungarian, where we find: la ´ t-ok [see-1sg] ‘I see’, la ´ t-sz [see-2sg] ‘you see’, la ´ t-Ø [see-3sg] ‘(s)he sees’. A notable exception is English, where we encounter the opposite situation, at least in the present tense: I see, you see, (s)he see-s. It should be remarked that iconicity is not always thought of as reflecting the external or mental world. Greenberg (1995) deals with a few cases of language- internal iconicity. These will not be gone into here. Whether iconicity is language-external or language-internal, the statistical nature of the phenomenon would be a sufficient reason for generativists to ignore it. However, the resistance of generativists against iconicity is even more funda- mental. In Noam Chomsky’s opinion, the structure of the grammatical system does not reflect the properties of the world but relatively independent properties of the human mind (see especially Chomsky 1972, 1980; cf. Newmeyer 1992). The fol- lowing quotation illustrates his view: Our interpretation of the world is based in part on representational systems that derive from the structure of the mind itself and do not mirror in any direct way the form of things in the external world. (Chomsky 1981: 3) Haiman (1985: 7) suspects that generativists are driven to such statements not by the data they consider but by the assumption that the only interesting universals are those which seem to be arbitrary or pointless from a formal or functional point of view. Only these arbitrary universals can provide unambiguous evidence for a specifically human linguistic faculty which Chomsky has come to describe as an organ. (Chomsky 1976: 57; see also Chomsky 1980) According to Haiman (1985: 8), the breakdown of iconicity into arbitrariness does not result from mysterious human genetic predispositions, but from relatively familiar principles such as economy, generalization, and association. In fact, these are the principles that are responsible for the ‘‘impairment’’ of iconicity in dia- grams outside as well as inside language. A theory that suits diagrammatic iconicity particularly well is Cognitive Lin- guistics (see Langacker 1990: 1–2). As is well known, the credo of this framework is diametrically opposed to the generativist view. Explanations in Cognitive Lin- guistics tend to be cognitive, functional, pragmatic, or experiential (see also Kleiber 1993; Givo ´ n 1994). The notion of iconicity fits perfectly in this view, since it assumes that a number of linguistic structures reflect the world’s structure and not the brain’s. Moreover, the semantic interpretation of the related concept of markedness is best interpreted from an experiential point of view, as will be argued below. 2 According to Kleiber (1993: 105), there seems to be a curious contradiction between Langacker’s promoting cognitivism in linguistics and the fact that in his 396 willy van langendonck main publications (1987, 1991) such terms as iconic and iconicity are absent in the glossaries and are used sparingly in the text, whereas its contrary, the word sym- bolic, is omnipresent. However, Kleiber goes on, this contradiction is only appar- ent. In fact, the iconicity of isomorphism is promoted to the rank of theoretical founding principle, and the iconicity of motivation shows up at regular intervals. Since both kinds of iconicity are diagrammatic in nature, let us first look at dia- grams, and more generally, at icons as signs. 1.2. Icons as Signs: The Trichotomy Symbol, Index, and Icon In Peirce’s (1974: 2.249) terminology, ‘‘A Symbol is a sign which refers to the Object that it denotes by virtue of a law, usually an association of general ideas, which operates to cause the Symbol to be interpreted as referring to that Object.’’ Most words of natural languages are symbols, or ‘‘conventional’’ signs, or in Saussure’s terminology: ‘‘arbitrary signs.’’ Indeed, in different, genetically unrelated languages the same sense is expressed by different forms; for example, lion in English cor- responds to simba in Swahili. The second sign distinguished by Peirce is the index: ‘‘An Index is a sign which refers to the Object that it denotes by virtue of being really affected by that Object’’ (2.248). There is a kind of causal relationship between the index and the object; for example, smoke is an index of fire, but if the fire is put out, the remaining smoke is no longer an index. In natural language, we have the indexicals this, that, here, now, and so on. However, insofar as they are conventional signs, these in- dexicals are also symbols. The third sign is the icon, the one that interests us here: ‘‘An Icon is a sign which refers to the Object that it denotes merely by virtue of characters of its own, and which it possesses, just the same, whether any such Object actually exists or not’’ (2.247). Everything can be an icon of everything provided it resembles a cer- tain object and it is used as a sign for that object. As I have already suggested, it should be emphasized that these three kinds of signs hardly ever occur in their pure form. Most actual signs are a mixture of the three types, as is illustrated by the indexicals above, or by instances of onomato- poeia. This, eventually, is why dogs say wafwaf in Dutch, but wauwau in German (see also below). 1.3. Three Kinds of Icons: Imagic Icon, Metaphor, Diagram Concentrating on icons, Peirce again distinguishes three subclasses: image, met- aphor, and diagram. The image is the prototype of an icon. It is a simple sign that resembles its referent by virtue of sensory characteristics. These may be visual, as in iconicity 397 a photograph, a statue, or a painting, but they may also be auditory, as in program music, that is, music that renders feelings or perceptions. In natural language, the obvious example of imagic iconicity is onomatopoeia, for example, in interjections like cuckoo and cock-a-doodle-doo (see below). A metaphor, in Peirce’s view, brings out the representative character of a sign by representing a parallelism in something else; for example, a lion may represent a (brave) man. It may suffice here to give metaphor its place in the classification of iconic phenomena and to refer the reader to the extensive literature on this topic, especially in Cognitive Linguistics (see Grady, this volume, chapter 8). Our main concern in this section is with the diagram, or diagrammatic icon. According to Peirce (1974: 2.277), diagrammatic icons are ‘‘those which represent the relations of the parts of one thing by analogous relations in their own parts.’’ In other words, a diagram is a systematic arrangement of signs that do not nec- essarily resemble their referents but whose mutual relations reflect the relations between their referents. More specifically, the constellation of the object and of its diagram is similar, but the individual referents and the individual signs themselves need not resemble each other. Again, we have to point to a continuum: there is a cline from an almost pure image, for instance, a photo (with a resemblance between individual referents and individual signs) and a ‘‘pure’’ diagram, where there is no such resemblance, for instance, a technical diagram, a scheme, or a Gestalt such as the one in figure 16.1. The individual signs, that is, the circle, the dash, and the dots, bear no resem- blance to reality, but the constellation does, so we can still identify this picture as a ‘face’. 1.4. Diagrammatic Iconicity: Isomorphism and Motivation The linguistic importance of iconicity lies in the recognition of diagrammatic iconicity. Haiman (1985: 11) distinguishes two aspects in a diagram: isomorphism and motivation. 3 Isomorphism is defined as follows: By isomorphism, a one-to-one correspondence alone is intended, without regard for the relative position, importance, mutual relevance, or any other property Figure 16.1. Representation of a face 398 willy van langendonck of points in a diagram. Violations of isomorphism are: many-to-one, one-to- many, one-to-zero, and zero-to-one relations between points in the diagram and points in ‘‘reality.’’ All of these violations occur in language, and are familiar as homonymy, synonymy, polysemy, empty morphs, and ‘‘signe ze ´ ro.’’ (11) Isomorphism in linguistics can be seen as a variant of the higher, more gen- eral principle of ‘‘one meaning, one form,’’ which is as old as European linguistics. It was also connected early with psychological factors that aim to eliminate pur- poseless variety. It is well known that young children reject such phenomena as homonymy and synonymy. Why should the word bank refer to a building (finan- cial institution) as well as to a riverside? As to synonymy, its very existence is often disputed: real synonymy is claimed not to occur (Bolinger 1977). As to homonymy, it has been ascertained that when homonymy starts disturbing communication, borrowing and innovation are used to undo it. Anttila (1972: 89) seems to have been the first to recognize isomorphism as a kind of iconicity: ‘‘Language has a general iconic tendency whereby semantic sameness is reflected also by formal sameness.’’ Notwithstanding certain difficulties pointed out by Kleiber (1993), I will follow Anttila (1972) and Haiman (1980: 516)in their claim. At the level of the lexicon, it was structuralism that used to posit isomorphism as an important principle. While recognizing polysemy, Goossens (1969: 98–106) provided evidence from dialect geographical data in favor of ‘‘polysemiophobia.’’ Geeraerts (1997: chapter 4) discusses J. Goossens’s arguments in the light of pro- totype theory. Isomorphism seems to conflict with polysemy and prototypicality, and one should at least admit that ‘‘the isomorphic principle cannot be maintained in its rigid form’’; instead, it should be considered a tendency. Geeraerts (1997: 124) eventually reconciles the two phenomena by concluding: Prototype theory, one could say, specifies what is to be understood by ‘one form, one meaning’: according to the prototypical conception of categorization, the iso- morphism between form and content applies to conceptual categories as a whole, that is, to prototypically organized bundles of nuances, and not to the nuances within these categories. In the realm of syntax, the Katz-Postal hypothesis (Katz and Postal 1964) implied a rejection of isomorphism. In arguing that transformations do not change meaning, the authors had to admit that both neutralization (many deep structures, one surface structure, i.e., homonymy) and diversification (many surface struc- tures, one deep structure, i.e., synonymy) must exist. It has become more and more apparent, however, that this thesis is untenable. What is more, in a number of cases syntactic ‘‘homonymy’’ may be motivated; that is, apparent homonomy turns out to be a case of isomorphism after all. This is borne out by the fact that the ‘‘homonymous’’ constructions occur in unrelated languages. For example, morphological and syntactic similarity between conditional protases and polar questions is the result of a meaning common to both construc- tions (Haiman 1974, 1980: 518): iconicity 399 . position of the burner knob, to the flow of gas, to the flame, to the temperature of the pan, to the temperature of the water, to the level of the water, to the dryness of the pan. The agent performs. represent the relations of the parts of one thing by analogous relations in their own parts.’’ In other words, a diagram is a systematic arrangement of signs that do not nec- essarily resemble their. like the resultative construction, the ditransitive construction, and the caused-motion construction in English; Y -of networks like Ann is the boss of the daughter of Max and Causation is the