(1) a. I don’t know if it is true. (Question) b. If it is true, I’ll eat my hat. (Protasis) (2) a. Had you known, you would have done otherwise. (Protasis) b. Had you known? (Question) The subordinate clauses in (1) can both be paraphrased as ‘either it is not true or it is’; those in (2) as ‘either you had not known or you had’. According to Langacker (1987: 57; 1991), the mechanisms at work in the do- main of lexical items are also applied to grammatical categories (e.g., nouns) and relations (e.g., subjects)—a view that condemns autonomous syntax since each category must then have a sense. This can be achieved by adopting the principle of schematicity, 4 whereby isomorphism is exploited to its maximum. 5 At the same time, however, motivation may play a role. Motivation is defined by Haiman (1985: 11) as follows: By motivation, I have in mind the property whereby diagrams exhibit the same relationship among their parts as their referents do among their parts. Most cases of diagrammatic iconicity are of this motivational type. For in- stance, the ‘‘schema,’’ so often used in Cognitive Linguistics, is in fact an example of motivational diagrammatic iconicity in linguistics (on schematicity, see Tuggy, this volume, chapter 4). Ideally, diagrams should show both isomorphism and motivation. However, the reason to use a diagram is precisely to simplify. As life is short and memory finite, so it is often only the essential attributes of the objects diagrams denote that are reproduced. Hence, a diagram often shows more isomorphism than motiva- tion, or conversely, more motivation than isomorphism, even if both are always present to some degree. 6 1.5. Iconicity and Markedness: Semantic Markedness and the Prototypical Speaker In linguistics, motivational iconicity has mostly to do with markedness. This concept is often captured in the slogan ‘‘The more form, the more meaning.’’ 7 The more (marked) complex the form of a sign (or of a constellation of signs), the more (marked) complex the meaning will be. This markedness principle can take the shape of a binary relation or of a cline from unmarked to most marked, a gradation gamut (Jakobson 1966). However, the question remains how to define complexity. With regard to form, complexity can be held to be a function of the number of morphemes (not of syllables). For instance, the plural book-s is more complex or marked than the singular book. Note that such oppositions do not hold in every case or in every language. Both the marked and the unmarked value may be ex- pressed in the same way, as in sheep. Still, a case such as sheep is not considered a counterexample; it is only when the reverse holds, that is, when the unmarked form 400 willy van langendonck is heavier than the marked one that we have an exception, provided that seman- tic markedness is preserved. More difficult to define is semantic markedness. Although in the case of num- ber we could argue that ‘more than one’ (book) is more complex than just ‘one’, this is not possible with the overwhelming majority of categories, such as definiteness, semantic roles, and so on. What we need here is not a logical or mathematical view of meaning but rather an experiential view (Lakoff 1982). 8 More specifically, it is helpful to adduce at this point the notion of ‘‘prototypical speaker’’ (Mayerthaler 1980;Van Langendonck 1999). As a human being, the speaker has certain biological, psycho- logical (perceptual), and cultural properties that can be called prototypical. 9 These prototypical properties are reflected in language as unmarked semantic categories and tend to be acquired first by children. A case in point is the spatial prepositions in, on,andat. Essentially, in refers to three dimensions (a container), on to two (a sur- face), and at may refer to any single dimension. Although in is logically the most complex, experientially it is the most ‘‘normal’’ case: the prototypical speaker is a container and lives in a three-dimensional space; in turn, surfaces are more impor- tant in daily life than points: for instance, as a rule we walk on a surface. Therefore, in is the unmarked and most frequent preposition and is acquired first; then comes on, and finally at. Logically, we would expect the reverse order, as did Clark (1973)atfirst when he suggested that the two-word sentence baby highchair meant baby at high- chair. However, it was found later that in several languages the initial zero preposition is replaced by in or its equivalents: baby in highchair. As a consequence, the putative order of acquisition at < on < in was wrong and had to be reversed. Other examples include dichotomies such as positive-negative, Agent-Patient, topic-comment, and subject-object. Since the prototypical speaker is obviously as- sumed to exist, it is natural that ‘positive’ will be semantically unmarked with respect to ‘negative’. This is formally reflected in zero marking for positive statements or properties and overt marking for negative ones: It is raining versus It is not raining; happy versus un-happy. The prototypical speaker sees himself as an agent rather than a patient, and thus the notion of agent is a prominent or unmarked relation. Since the prototypical speaker is assumed to exist and is therefore topical, the topic of an utterance will be unmarked as well. From this, we can derive the idea that since the subject is the grammaticalization of the intersection of agent and topic, it will be the unmarked grammatical relation, as was, in fact, already stated by Jakobson (1966). Below (see especially section 2.2), I will discuss the iconic aspects of some mor- phosyntactic categories that exhibit a marked versus unmarked relationship. 1.6. Iconicity and Economy Haiman ( 1985: 18, part I) argues that in languages, as in diagrams, there is an inverse correlation between iconicity and economy. The tendencies to maximize iconicity and to maximize economy are two of the most important competing motivations for linguistic forms in general. An obvious example of this is the iconicity 401 treatment of number in Indonesian. The plural in Indonesian is iconically formed by reduplicating the word, thus while orang means ‘human being’, orang-orang means ‘human beings’. Significantly, the latter form applies not only to ‘two people’, but also to more than two. For reasons of economy, the morpheme is not repeated more than once according to the real quantity, although this would in fact be more iconic if one spoke about more than two people. By contrast, iconicity and economy can go together as well. To take number again, in many languages the singular is expressed by zero, the plural by an overt form, as with English book versus book-s. This is both economic and iconic and is in keeping with the slogan ‘‘The more form, the more meaning’’ (i.e., ‘one’ vs. ‘more than one’). In fact, we are confronted here with a special case of Zipf’s ([1949] 1965) principle of least effort. 2. Iconicity on the Phonic, the Morphological, and the Syntactic Level In the following sections, I will deal with some instances of iconicity and especially of diagrammatic iconicity in the various components of grammar (in a wide sense). 2.1. Iconicity on the Phonic Level On the phonic level, we primarily find imagic iconicity, which is well known as onomatopoeia—in fact, we should speak here of acoustic iconicity. Compared to normal vocabulary, onomatopoeia remains a marginal phenomenon in natural language, though there may be differences in the degree to which it is implemented from one language to another, as was observed above. The more onomatopoeic words get integrated in the linguistic system, the more they become symbolic and the more they lose their iconic value. For example, in its capacity of a noun, cuckoo can be pluralized (cuckoo-s); words that primarily function(ed) as interjections can become verbs like crack, squeak, hiss, hush, and so on. It is to be expected, then, that there will be certain differences between languages in rendering the same kind of interjection. Compare, in this respect, the rooster’s chant as it is conventionally rendered in different languages, and notice the inevitable degree of arbitrariness: cock-a-doodle-doo (English), kukeleku (Dutch), kikeriki (German), cocorico (French), kukareku (Russian), and so on. As for diagrammatic iconicity on the phonic level, it is difficult to distinguish between isomorphism and motivation. Moreover, diagrammatic iconicity is rather limited on the phonic level because it is hard to assign meanings to sounds as such (see, e.g., Birdsong 1995; Landsberg 1995). It is sometimes argued, for instance, that 402 willy van langendonck the sound [i] expresses smallness while [A], [o], and [u] are related to big sizes. In certain expressions, this appears to be justified; note English a wee li ttle bit . However, according to Hage ` ge (1982: 25), this correlation is statistically not significant. At best, we can maintain there is something to say for this ‘‘sense’’ at the level of performance. 2.2. Diagrammatic Iconicity on the Morphological Level On the morphological level, we find iconicity to various degrees in the inflectional endings of nouns, verbs, and adjectives, and, further, in the markedness relation between ‘positive’ and ‘negative’. Nouns usually have such features as number, definiteness, humanness, and gender. With regard to number, isomorphism can take a strong form, as in the above example from Indonesian, insofar as the plural orang-orang refers to two people. A weaker form of isomorphism is encountered in the English plural where the plural -s is arbitrary in itself, but isomorphic insofar as it is used consistently to indicate plural meaning. In both cases there is also some motivation: the unmarked singular has zero form while the marked plural has one more morpheme. From an experiential perspective, the singular is semantically unmarked since the prototypical speaker is a single person, not a chorus. Also, it is easier to perceive or, more generally, to deal with one object at a time than with several (see also Langacker 1991: 74–81). With regard to definiteness, we may start from the fact that whereas the indef- inite article or value is used to introduce a referent into the universe of discourse (e.g., There is a girl in the garden), the definite article or value ideally occurs when the referent is presupposed to exist and to be unique in the universe of discourse for both speaker and hearer (e.g., The girl smiled at him). The feature ‘definite’ appears to be unmarked with respect to ‘indefinite’ (Van Langendonck 1979; Mayerthaler 1980, 1988). From an experiential view, this squares well with the fact that proto- typical speakers are ‘‘definite’’ in the sense that they are presupposed to exist and to be unique. This is borne out by some formal markedness phenomena. Although no argument can be drawn from the behavior of common nouns, it remains true that in inherently definite categories such as proper names and personal pronouns, this definiteness feature is mostly not expressed by an overt morpheme (e.g., Jo hn, Mary, Paris; I, you, he, she, it ). A further reflex of the unmarked status of definiteness is that in Dutch and French, definite noun phrases are pronominalized by one morpheme while indefinite noun phrases are pronominalized by two morphemes; compare: (3)a.Hij bezit een paleis.—Hij bezit er een. ‘He possesses a palace.’— ‘He possesses one.’ Hij bezit het paleis.—Hij bezit het. ‘He possesses the palace.’—‘He possesses it.’ b. Il posse ` de un palais.—Il en posse ` de un. Il posse ` de le palais.—Il le posse ` de. English does not provide further corroboration for this opposition but is not a counterexample either since both definite and indefinite pronominalization are iconicity 403 realized by one morpheme each. The situation is not reversed, that is, definiteness is not more marked than indefiniteness. With regard to humanness, we may expect that since the prototypical speaker is human by definition, the feature ‘human’ will be unmarked vis-a ` -vis ‘nonhu- man’. This is formally reflected by zero forms for ‘human’ in certain constructions. In many languages the imperative (meant for animate beings) has no ending (e.g., Swahili come! njoo!, Dutch kom!, Spanish ven!). Personal proper names usually have no article (e.g., English John, Mary, French Jean, Marie). Neither do names of places exhibiting some human organization, such as cities, villages, and countries. This is clearer in Germanic languages than in French, where only cities have no article (compare London—Londres versus France—la France). A final example is the impersonal passive in Dutch: in this specific construction type, the agent is always human, also if it is not expressed, as in Er wordt gelachen [there become.3sg laugh.ppart] ‘People are laughing’. Finally, with regard to gender, it is found that in most cultures the prototypical speaker is a man, not a woman. As will be clear, this is not a biological criterion but a purely cultural one, and hence subject to change. In English, there are still mor- phological traces of the markedness of feminine gender: compare the zero form of masculine titles with the ending -ess for feminine titles (e.g., count vs. count-ess, prince vs. princ-ess). Turning from nouns to verbs, categories such as tense, aspect, and mood often manifest iconic reflexes of marked versus unmarked values. For tense, we know that since the prototypical speaker speaks in the present by definition (the time of the speech act), the present tense is semantically unmarked with respect to the past. This is iconically reflected by the fact that most languages have zero for the present but an overt form for the past. Compare English I work with I work-ed and French je travaille with je travaill-ais. As for mood, it can safely be stated that the speech act time is also the most real time. Potential or unreal events are rendered by po- tentialis and irrealis, which show a more marked form than the present or even the past indicative. Consider French je travaill-er-ais ‘I would work’. Looking at adjectives, we typically find iconicity in the expression of the de- grees of comparison. In a number of languages, the comparative is more marked than the positive degree and often the superlative is more marked than the com- parative. Indeed, a comparative statement can be paraphrased with a negative statement (e.g., John is taller than Mary ¼ ‘Mary is not so tall as John’), while the superlative would imply that nobody is as tall as John. This is reflected morpho- syntactically: tall < tall-er < tall-est. In Latin, we find a similar gradation: longus < long-ior < long-issimus. In some other languages, the negation in the comparative is overtly present (see Van Langendonck, Swiggers, and Van de Velde, forthcoming), for example, in Abipon (Guaycuru): (4) Negetink chik naa ˆ , oagan nihirenak la naa ˆ .(Stassen1985: 184) dog not bad yet tiger already bad ‘A tiger is more ferocious than a dog.’ 404 willy van langendonck Iconicity on the morphological level can, finally, also be found in the relation between positive and negative. Since the prototypical speaker is presupposed to ‘exist’, existence will be unmarked with respect to nonexistence; that is, ‘positive’ is valued higher than ‘negative’ (see also Hamilton and Deese 1971 for a psychological experiment). Very often, then, negation will require a special morpheme, compare: (5) happy vs. un-happy, animate vs. in-animate, honest vs. dis-honest, entangle vs. dis-entangle, sense vs. non-sense, etc. It has also been argued that the first adjective in such pairs as big—small, tall— short, long—short, thick—thin, high—low can be explained in the same way; that is, the first element of the pair is positive, the second negative. The problem is that there is no iconic relation here nor is it even immediately obvious why the first elements should be regarded as positive. This may be right in a metaphorical sense, but then we need to know why. An explanation can be sought in relating the ‘positive’ meanings to properties of the prototypical speaker. For example, an expected and even desirable property of the prototypical speaker and of nature in general seems to be to grow and expand, not to shrink. 2.3. Diagrammatic Iconicity on the Syntactic Level On the syntactic level, diagrammatic iconicity manifests itself especially as iconicity of distance (or closeness) in syntactic constructions (patterns) and as word-order iconicity. 2.3.1. Iconicity of Distance (or Closeness) in Syntactic Patterns Elements that occur closely together and form a unity in experience will tend to be related to each other by the prototypical speaker on the content level as well. Conversely, what is separated will be seen as unrelated conceptually. Haiman (1983: 782) speaks about the reflection of conceptual distance or separation by physical (morphosyntactic) distance or separation. Put in a simple slogan, ‘‘The distance between expressions corresponds to the conceptual distance between the ideas they represent.’’ In this way, we are even able to formulate universal tendencies that we can regard as constraints or meta-constraints on grammatical structures. To illustrate this, I will discuss a number of different syntactic constructions. One relevant phenomenon is conjunction reduction. Originally, Generative Grammar attempted to derive a sentence like (6a) from the one in (6b): (6) a. We can do it quickly and well. b. We can do it quickly and we can do it well. In (6b), it is possible to do something well under one set of conditions and to do it quickly under another; but grouping quickly and well together, as in (6a), implies that these are realized under the same set of circumstances (Haiman 1983: 808). Note also that this is a case where iconicity and economy are in harmony. iconicity 405 Another example of the iconicity of distance is found with verbs of perception. Usually, a distinction is made between direct perception (as in 7a) and indirect perception (as in 7b): (7) a. I hear him sing(ing). b. I hear that he sings/is singing. Generative attempts to derive (7a) from (7b) by some kind of raising appear to have failed. The reason is that different things are meant by the two constructions, which leads to different syntactic possibilities. In the that-clause, we have two propositions that can differ in terms of tense and modality; 10 consider in this re- spect also (7c) and (7d): (7) c. I hear that he sang/was to sing. d. *I hear him being to sing. Sentence (7c) encodes an indirect perception: there clearly are two events, iconi- cally represented by two propositions. In sentence (7a), this is not the case: we have to do here with one event, one propositional Gestalt with only one tense and one modality (cf. 7d). 11 Further, distance iconicity also appears to be at work in causatives. In causative constructions, we can observe a cline from direct to indirect causation. Direct cau- sation implies a unity of action, volition, place, and time: the event is one Gestalt, and the responsibility of the causer will here be the greatest. If causation is less direct, the event seems to be spread more in time and space, and less responsibility can be invoked on the part of the causer. The continuum from direct to indirect causation is reflected in the grammar of natural languages through a continuum of lexical via morphological to syntactic causative constructions (Comrie 1981: 165; Haiman 1983: 782; Song 1992 on Korean). Partly, this is observable in one and the same language. Consider the Dutch expressions for the notion ‘killing’, which show a cline in formality: 12 (8) doden–doodmaken–doen sterven–maken dat x sterft ‘kill–make dead–make die–make that x dies’ These alternatives require appropriate sentences, as exemplified in (9): (9)a.De misdadiger doodde zonder scrupules. ‘The criminal killed unscrupulously.’ b. De boer maakte het dier dood. the farmer made the animal dead ‘The farmer killed the animal’ c. De zure regen doet de bomen sterven. the acid rain does the trees die ‘The acid rain causes the trees to die.’ d. Het slikken van een doosje pillen maakte dat de man stierf. the swallowing of a box pills made that the man died ‘The swallowing of a box of pills caused the man to die.’ 406 willy van langendonck We can see here that the responsibility and control in the causation of the event diminishes from left to right: for a crime only doden will be used, as in (9a). 13 The compound doodmaken in (9b) is used especially in the case of an animal patient killed by a human causer. Doen sterven does not require a human subject; the object is often a plant, as in (9c). In the periphrastic construction maken dat x sterft in (9d), the responsibility of the causer is lowest; subject and object may even be abstract propositions. Finally, distance iconicity affects the expression of certain grammatical rela- tions. By definition, prepositional objects show a greater physical distance with the verb than direct objects. This seems to correspond with greater conceptual dis- tance. Consider the famous pair of examples in (10): (10) a. He smeared the wall with paint. b. He smeared paint on the wall. In the so-called holistic reading of (10a), the whole wall is painted whereas this is normally not the case in (10b), where a ‘partial’ reading is most likely. The greater affectedness of the patient in (10a) is reflected by the fact that the wall in (10a), being a direct object, is closer to the verb than in (10b), where it is separated from the verb by a preposition. 2.3.2. Word-Order Iconicity Several phenomena resort under word-order iconicity (see, e.g., Van Langendonck 1995; compare for English: Brinton 1987;Givo ´ n 1994; Thompson 1995;Fischer1997). As a first example, word-order iconicity is found in the ordering of events in narrative sequences, which tends to reflect closeness in time. Genette (1976: 226) points out that already in the eighteenth century, it was considered a rule that the order in discourse was a true reflection of individual thought. Greenberg (1966b: 103) formulated it as follows: ‘‘The order of elements in language parallels that in physical experience or the order of knowledge.’’ Paraphrasing Jakobson (1965), we can state that in a narrative sequence utterance 2 follows utterance 1, just like event 2 follows event 1. There is an element of closeness here, in that the closer to each other events occur in time, the closer they are positioned in discourse. This is es- pecially evident in asyndetic and coordinate structures, where a series of coded events is least ‘‘disturbed’’ by intervening linguistic elements: (11) Veni, vidi, vici. (12) She married and got pregnant. Subordinating conjunctions with a particular meaning may undo the iconic order, as in (13a), whereas others are subject to a fixed iconic order, as in (13b) and (13c): (13) a. She got pregnant after she married. b. He ran too hard so (that) he fell down. c. *So (that) he fell down he ran too hard. A special case of closeness in events is represented by so-called serial verbs. In cer- tain languages, a series of closely related events that form a unity is expressed by a iconicity 407 corresponding juxtaposition of the verbs indicating the events. Consider the fol- lowing sentences in Akan (from Schachter 1974: 254): in (14), Kofi’s going precedes his coming back, while in (15), the action of pouring causes the flowing to happen. (14) Kofi k OO e baae. Kofi went came ‘Kofi went and came back.’ (15) Kofi de aburow gu nsum. Kofi takes corn flows water.in ‘Kofi pours corn into the water.’ At the same time, it must be pointed out that representations of events are not necessarily ordered iconically in discourse. Langacker (1991: 502 ) speaks of cases where the speaker scans backwards mentally through conceived time, as in the following passage: (16) Professor Muddle died last night at the age of 75. He suffered from an inflamed ego for several years prior to his death. He taught theoretical basket weaving for almost four decades at MIT. He received his Ph.D. from that institution in 1948. We might be inclined to think that all natural languages conform at least to the most general ordering principle as exemplified in Veni, vidi, vici. However, even this kind of motivation is not universal: a number of South-East Asian languages de- viate. A native speaker of Burmese will understand that a series of instructions are to be followed in the order given only if the sequence is marked by extra gram- matical signs. Otherwise, the instructions are to be carried out simultaneously (Haiman 1980: 533). A second kind of iconic ordering involves the concept of closeness to the speaker. What is nearest to the speaker in a literal (physical) or in a metaphorical sense is mentioned first, especially again in asyndetic or in fixed coordinate struc- tures. This principle is found to be at work in various ways. Literal closeness to the speaker is involved in spatiotemporal conjunctions like the English, Polish, and French freezes (fixed collocations, see Landsberg 1995): (17) a. English: here and there, this and that, now and then, sooner or later b. Polish: tu i tam ‘here and there’ 14 c. French: c¸a et la ` ‘here and there’, t ^ oot ou tard ‘sooner or later’ The initial motivation for putting here, now, and sooner in first position apparently derives from the speaker’s viewpoint. Next to literal closeness, we also find metaphorical closeness to the prototyp- ical speaker. The properties that are metaphorically ‘‘nearest’’ to the prototypical speaker tend to be mentioned first. Recall the unmarked features positive, human, masculine, definite, and so on mentioned above (see section 2.2). Usually, these come first in coordinate constructions, especially in freezes: 408 willy van langendonck (18) happy or unhappy, yes or no, big and small, tall and short, a man or a mouse, husband and wife, 15 this and something else Further, in the expression of grammatical relations, the reason for putting the indirect object before the direct object is obviously that the former mostly refers to a human referent but the latter to an inanimate one, as in: (19) John gave Mary a book. A third example of word-order iconicity involves several formal reflections of closeness in content. We will look at four principles where we see that elements that are close in content in some way or other tend to be placed together. The first three concern adjacency, 16 the last one concerns the placement of ‘‘relators’’ in between their ‘‘relata.’’ Simple Adjacency The first principle is that of simple adjacency of head and modifier (Hudson 1984: 98)—or, the principle of head proximity (Rijkhoff 1992: 229)—the principle whereby a modifier is put as closely as possible to its head. One effect of this prin- ciple is the tendency for objects to accompany the verb they depend on or for adjectives to accompany the noun by which they are governed. Of course, the modifier may be separated from its head by other modifiers of the same head, as in (20), where the modifiers of the noun dolls are as close as possible to each other. (20) She liked those three nice little wooden dolls. When all modifiers precede (as in 20) or follow their head, the order of the modifiers with respect to each other is determined by a second principle, namely, relative adjacency. Relative Adjacency This principle implies that, in neutral ordering, the modifiers that are closest to the head in content are generally placed closest to it as well. The effect of relative adjacency is found most clearly in the noun phrase, as illustrated in (21): (21) those three nice little white wooden dolls In (21), we see that the qualifying adjectives appear closer to the noun than the quantifying modifiers. Further, absolute objective qualities like wooden and white are closest to the noun, whereby those qualities involving substance (wooden) are expressed closest to the noun. Relative objective qualities (size, e.g., little) are ex- pressed farther away from the noun, and subjective qualities like nice are expressed still farther. Farthest away from the noun is the determiner, since it serves only to pick out the referent. This order appears to be widespread cross-linguistically (see Hetzron 1978; Seiler 1978; Posner 1986). Nevertheless, Heine (1980) observed that in some East-African languages the deictic and quantificational elements are nearer iconicity 409 . iconicity in the expression of the de- grees of comparison. In a number of languages, the comparative is more marked than the positive degree and often the superlative is more marked than the com- parative (marked) complex the form of a sign (or of a constellation of signs), the more (marked) complex the meaning will be. This markedness principle can take the shape of a binary relation or of a cline. exist and is therefore topical, the topic of an utterance will be unmarked as well. From this, we can derive the idea that since the subject is the grammaticalization of the intersection of agent