Suppletion, natural morphology, and diagrammaticity *.
This paper challenges the widespread view of suppletion as nothing but morphologically nonfunctional diachronic residue. Within natural morphology, suppletion has generally been treated as an "unnatural" phenomenon that must be accounted for in extragrammatical terms. A richer notion of diagrammatic iconicity, however, inspired largely by Bybee's work, leads to an analysis in which suppletion in semantically generic, grammaticalized or grammaticalizing items fits quite well into the overall diagrammatic structure of inflecting languages in a way that is consistent with the main principles of natural morphology. I consider Bybee's view that all iconicity/structure should be regarded primarily as an epiphenomenal artifact of use and change but argue that some aspects of morphological change can most easily be understood if we grant some degree of psychological reality to diagrammaticity.
Most contemporary morphologists have come to use the term suppletion to refer generally to any morphological relationship that is "maximally irregular" (Mel'cuk 1994: 339), in other words to any unique formal relationship between two word forms whose semantic relationship is expressed by a regular alternation elsewhere in the same language. This definition is purely synchronic and does not include the criterion that suppletive stems be etymologically unrelated, which is often regarded as the defining characteristic of suppletion in traditional historical linguistics (cf. Rudes 1980; Wurzel 1985: 117-118). A second kind of deviation from the traditional definition of suppletion is less obvious but equally important to the present investigation. Many studies that have tried to make some kind of sense of suppletion focus less on the irregularity per se and more on the extreme stem allomorphy of many suppletive paradigms. Although the argument is never made explicit, these researchers seem to feel that extreme stem allomorphy is really the essence of suppletion, and the irregularity of the alternations is merely an inevitable consequence of this degree of allomorphy. Furthermore, these studies all focus (often exclusively) on (1) inflectional as opposed to derivational suppletion; (2) stem as opposed to affixal suppletion. I follow these studies in all of these respects.
The view of suppletion as pure historical artifact or residue is extremely widespread among linguists today (see section 2.1 below). Both the proponents and the critics of natural morphology (NM) have, for the most part, adopted this view, treating suppletion as a highly "unnatural" phenomenon. The argument that suppletion is easily reconcilable with a naturalness theory that is broadly construed as concerned with all nonarbitrary aspects of the relation between form and function, including communicative function (Wurzel 1990; Harnisch 1990), does not alter the general consensus that from the narrower grammatical perspective of NM suppletion is indeed more or less unnatural (Dressier 1985b: 330-335, 1985c; Wurzel 1985: 121, 1990: 87-88; Bittner 1988: 416, 422; Mayerthaler 1981: 38-39; Carstairs-McCarthy 1992: 240; Mel'cuk 1994: 392; Werner 1989: 38-39; Ronneberger-Sibold 1988: 454 455; Markey 1985: 52). (2)
1. Principles of natural morphology and their relevance to suppletion
Mayerthaler (1981: 23-35, 1987: 48-49) proposes three universal principles of naturalness.
1. "Principle of constructional iconicity/diagrammaticity" is the name he gives to the familiar notion that the unmarked-marked (or basic-derived) semantic relationship tends to be reflected formally in some kind of "less" vs. "more" relationship. (3) In Mayerthaler's formulation: "If (if and only if) a semantically more marked category [...] is encoded as 'more' featured than a less marked category [...], the encoding of [the semantically more marked category] is said to be iconic. Otherwise it is non-iconic or even countericonic." His definition of what it means for one form to be "more featured" (4) than another (1981: 23-25) can be summarized in the following scale:
(1) [ILLUSTRATION OMITTED]
Thus, for example, if a plural form has an affix that the singular lacks, the marking of number is iconic (diagrammatic), since the plural is semantically marked with respect to the singular. If the plural stem undergoes some kind of modification, the marking is still iconic, although less so than with (only) an affix (cf. Dressler 1985b: 328). If the plural and the singular are identical the marking is noniconic, and if the singular has an affix that the plural lacks, the marking is countericonic.
2. "Uniform encoding" (or simply "uniformity") is maximal when every form always has the same meaning and every meaning is always represented by the same form. As Mayerthaler (1987: 49) points out, this is also just a new name for a very familiar notion. (5) Polysemy, homonymy, and syncretism, on the one hand, and synonymy and allomorphy, on the other, all count as violations of this principle.
3. "Transparency" includes both semantic transparency--the notion that the meaning of a word should be predictable from the meaning of its constituent morphemes ("the Fregean principle of compositionality")--and morphotactic transparency (segmentability). Fusion or overlapping of morphs is a violation of morphotactic transparency (cf. Dressler et al. 1987: 21, note 8). (6)
For Mayerthaler's system-independent principles, suppletion is indeed quite "unnatural" It always violates uniform encoding, at least as far as the stem is concerned, since the meaning of the lexical item is represented by a different form in each of the suppletive stems (Bittner 1988: 421), and--at least under one interpretation violation of morphotactic transparency is the other side of the same coin, since each stem gives us information about grammatical categories as well as lexical identity (cf. Mel'cuk 1994: 392; Bybee 1996: 252). On the constructional-iconicity scale, suppletive paradigms are, at best, less than maximally iconic (good-better) and are often minimally iconic (much-more, bad-worse) and sometimes even countericonic (little-less) (cf. Dressler 1985c: 43, 1987: 103-104; Wurzel 1990: 88). (7) The only one of Mayerthaler's naturalness (sub-)principles that suppletive inflectional paradigms do not generally violate is semantic transparency.
Wurzel's basic insight is that "naturalness" in a complex system amounts to system-internal consistency (1984, 1987, 1989). Furthermore, Wurzel argues that when inconsistencies do arise, which is inevitable due to the effects of extramorphological forces, one of the competing alternatives will normally be dominant, and it is these dominant phenomena that then determine what counts as natural within a less-than-perfectly-uniform system or subsystem. Dominance, according to Wurzel, is simply a matter of the relative frequency of competing phenomena within the system. Wurzel sees this kind of naturalness operating on two levels. He coins the terms "inflectional-class stability" (along with "marker stability") and "system congruity" to refer to system-dependent naturalness on the level of concrete, specific inflectional markers and on the level of more general patterns respectively.
Wurzel defines system congruity in terms of six parameters based on which the main "system-defining structural properties" (SDSPs) of an inflectional system or subsystem can be determined (1984: 82, 205, 1987: 63). These parameters include, for example the types of markers associated with each category (e.g. prefixes, suffixes, stem alternations) and whether a language has separate or combined expression of particular grammatical categories. Diachronically, morphological phenomena that are not system-congruous, that is, that are not in line with a language's SDSPs, tend to be targeted for elimination.
The principle of inflectional-class stability (8) (Wurzel 1984: 116-153, 207-210, 1987: 76-87) states that if a system or subsystem has inflectional classes, these classes will tend to correlate with extramorphological (phonological, semantic, syntactic) properties. When two or more classes correlate with the same extramorphological properties, they are in competition. (9) The stability of each competing class is a function of its size, that is, the number of lexical items that belong to it. Diachronically, the larger classes will tend to attract new members from the smaller classes with which they are in competition, and neologisms and loan words will also tend to gravitate toward the larger classes. A small class can be highly stable, however, if it is well defined by extramorphological properties. The modern German modal verbs are a classic example of a small, relatively stable inflectional class (Wurzel 1984: 149; Birkmann 1987).
Suppletion would appear to be at least as problematic for Wurzel's theory as for Mayerthaler's. Since there are seldom more than a handful of suppletive lexical items in a language, it is hard to see in what sense suppletion could ever be considered system-congruous, and to the extent that suppletive paradigms are in competition with other inflectional classes, we would have to say that the minimal size of the suppletive "classes" (a single lexical item) makes them maximally unstable (cf. Wurzel 1984: 67). Both of Wurzel's key principles thus lead us to regard suppletion as highly anomalous from the point of view of system-dependent naturalness. As Wurzel (1990: 88) puts it, "Bezogen auf das Flexionssystem und daruber hinaus das gesamte Sprachsystem ist Suppletion demzufolge nichtfunktional" (10) (cf. Wurzel 1985: 121; Bittner 1988: 416, 422).
2. Approaches to suppletion
2.1. Traditional views
The most widespread and obvious kind of explanation for the occurrence of suppletion in particular lexical items involves the effect of high token frequency. Suppletion is seen as an extreme manifestation of a general tendency for irregularity to occur in high-frequency items.
There are at least two important variations on this view. In its best-known version, it regards suppletion as a pure historical artifact. Morphological irregularity arises due to extramorphological forces such as sound change. Its survival then depends mainly on the frequency of occurrence of the forms in question. Highly frequent forms are learned by rote and are therefore highly resistant to the analogical influence of related forms and parallel patterns (e.g. Nida 1963: 265). According to this widespread view, there is no synchronic reason for suppletion at all. It is not integrated into the grammatical system in any way, and it serves no function of any kind.
A recent statement of this position can be found in Singh (1996: 22). After arguing that natural morphology can and should account for almost all of what are normally regarded as exceptions in morphology, Singh adds,
All that is left is suppletion but suppletion cannot possibly be a matter of morphology. That requires both formal and semantic similarity. The right meta-language for morphology, in other words, is bound to expose exceptionality in morphology for what it really is: an artifact.
Elsewhere (1996: 8), it is clear that Singh has suppletion in mind when he speaks of "synchronically unexplainable diachronic residues." Similarly Lass (1990: 82) cites suppletion as an example of "marginal garbage or nonfunctional/nonexpressive residue."
Bybee, whose work will be discussed at some length below, is fundamentally in agreement with the traditional view that sees suppletion as a result of high token frequency, and she develops this view in some very important ways (1985a: 117-123, 1988: 131-134, 139). As we will see, treating suppletion as an artifact of extragrammatical forces does not set it apart from the rest of grammar within Bybee's framework, since she views all grammatical structure as a byproduct of language use and change.
2.2. Economy-based theories
An important variation on the token-frequency approach argues that suppletion in high-frequency items is in fact functional. This view is most developed in the work of Werner and that of Ronneberger-Sibold. They see suppletion as being "economical" from a communicative-functional perspective in a number of ways: (1) it is economical for high-frequency items to be short, and shortness is supposedly a characteristic of suppletive forms (Werner 1987a; 1989: 40-43; cf. Ronneberger-Sibold 1980: 145-147; Harnisch 1990: 58); (2) it is more efficient to access highly frequent forms directly than to produce them by rule (Ronneberger-Sibold 1980: 147-150, 1988: 456-457; Harnisch 1990: 59; Dressler 1985c: 49); and (3) the maximizing of formal differentiation in suppletive paradigms facilitates perception (Werner 1987a: 302, 306-307, 1989: 40; Ronneberger-Sibold 1987a: 249, 1988: 457-458; cf. Harnisch 1990: 59). Thus, according to this "economy" theory, suppletion is a consequence of token frequency, but it is not purely a historical artifact. It is retained in a language at least partly because of its functional usefulness for communication.
All those who regard suppletion as a consequence of token frequency would agree, however, that suppletion does not make much sense in terms of the synchronic grammatical system itself.
2.3. The natural morphologists
Wurzel (1990) accepts the general correctness of some aspects of Werner's economy-based analysis but points out that this analysis is fully compatible with naturalness theory, understood broadly as dealing with all nonarbitrary aspects of the relationship between form and function, including communicative function. Wurzel emphasizes, however, that the "communicative-pragmatic naturalness" that allows us to account for suppletion is a very different kind of principle from those of NM, which operate within the grammatical system and involve relationships among forms or between form and grammatical function.
Dressler, on the other hand, points out a variety of grammatical and extragrammatical factors that play some role in suppletion (1985b: 333-334, 1985c: 50-54), including semantic/pragmatic basicness, order of acquisition, the indexical properties of allomorphy, and the correlation between suppletion and membership in small, closed word classes (e.g. pronouns, auxiliary verbs, articles). (11) Dressler also offers some insightful analyses of specific instances of suppletion, (12) but for the most part he regards suppletion as morphologically peripheral and uses it as one of his three "devil's cases" of "highly unnatural phenomena." He accepts the key role of token frequency in the survival of suppletive forms: "Under which conditions are suppletions most resistant to analogical leveling? The more token frequency a form has the better and earlier it is stored in the brain and therefore it is retained best" (1985b: 333, cf. 1985c: 49).
Among the proponents of natural morphology, only Bittner (1988, 1996: 27-36) is determined to account for suppletion entirely in terms that are consistent with the spirit of the main principles of NM, although even he believes that the motivation for suppletion must be extragrammatical (1996: 35-36). Bittner's basic hypothesis, which he adopts from Osthoff (1899) and Wurzel (1985), is that suppletion is natural within a semantically or psychologically definable "Nahbereich"--a domain that is in some sense close to the speaker because we differentiate more among things that are close to us than among things that are farther away (cf. Harnisch 1990: 57-58; Mel'cuk 1994: 392; Mayerthaler 1981: 11-21; Ronneberger-Sibold 1988: 454-455).
Although this idea makes a great deal of sense, there are serious problems with existing attempts to define the Nahbereich and to bring all instances of suppletion under this umbrella. The most common strategy has been to define the notion of "closeness to the speaker" so vaguely and broadly that most of the lexicon would have to be considered part of this domain (e.g. Harnisch 1990: 57-58).
Both Bittner (1988: 420-421, 424, 1996: 31-32) and Wurzel (1985: 121-122) readily acknowledge that the notion of the Nahbereich is extremely problematic. (13) More importantly, Bittner recognizes that regardless of how this domain is defined many suppletive lexical items have nothing to do with closeness to the speaker in any sense and can instead best be characterized as grammatical function words, including copulas, auxiliary verbs, and pronouns (cf. Werner 1987a: 312-313). He does not seem to realize to what extent this admission undermines his account of suppletion, which is based on the intuitively plausible notion of greater differentiation among things that are close to us. Bittner has no clear explanation for why grammatical function words should be suppletive beyond an ad hoc appeal to markedness reversal (Mayerthaler 1981: 48-58), a concept that numerous scholars have rightly criticized as a deus ex machina (Wurzel 1984: 25-26, note 23; Carstairs-McCarthy 1992: 219-223; Meineke 1989: 328-329; Birkmann 1987: 20). (14)
Another problem with Bittner's account is that, unlike the other natural morphologists (Wurzel 1985: 124-125; 1990; Hentschel 1990: 67), he tries to argue that token frequency plays no role in the survival of suppletive forms (1988: 419-420, 424, 1996: 30-35). His point that the token frequency of a lexical item depends on its function/meaning is obviously true, but this is irrelevant to the question of whether token frequency might be (partially or entirely) directly responsible for the survival of suppletive forms (Harnisch 1990: 60; Greenberg 1966: 65-70; Croft 1990: 160).
The weakest point in Bittner's argument comes when he tries to offer empirical evidence for his claim that suppletion is independent of token frequency (1988: 421-422, 1996: 32-33). This evidence consists of five pairs of verbs, supposedly matched for token frequency, all of which were originally irregular or at least strong. Each pair consists of one verb that belongs to the semantic suppletion domain and has retained its irregularity and one that does not and has undergone at least partial transfer to the regular weak class. In Bittner (1988), the five pairs are (with the verb from the suppletion domain listed first in each case): (1) helfen 'help'-bauen 'build'; (2) tragen 'carry'-laden 'load', 'invite'; (3) sterben 'die'-horen 'hear'; (4) essen 'eat-hauen 'hew', 'strike'; (5) sehlafen 'sleep'-melken 'milk'. (15) Since one recent textbook reports this evidence uncritically (Carstairs-McCarthy 1992: 242-243), it is worth pointing out several serious flaws in it:
1. To begin with, pairings of this type never represent valid evidence for anything. One would undoubtedly be able to construct such pairs even if there were no correlation whatsoever between membership in the semantic "suppletion domain" and the presence of suppletion. At the very least, Bittner needs to show that it is not as easy to find pairs where the relationship is reversed, but he in fact acknowledges that words in the suppletion domain can also be subject to regularization (1988: 422), and examples of nonregularization among words outside the domain are plentiful.
2. At least four of the five verbs from the semantic suppletion domain are in fact not quite suppletive (cf. Mel'cuk 1994: 378). They belong to strong-verb classes with several members showing identical or virtually identical conjugation. Two of Bittner's verbs--helfen and sterben--in fact belong to the same class, which also includes bergen, verderben, werben, werfen, gelten, and schelten. The conjugation of tragen is identical to that of graben, schlagen, and fahren and that of schlafen to blasen and raten. Even the fifth verb essen only displays a very weak form of suppletion, the doubled prefix in the participle gegessen. Otherwise, its forms are exactly parallel to those of fressen, vergessen, and messen.
3. Bittner (1988) admits that his data on token frequency may not be totally reliable. This is a very serious problem, since the claim of approximate equality of frequency for the members of each pair is absolutely crucial to his argument. Would a large and representative sample of modern German truly show that schlafen and melken occur with roughly equal frequency?
4. The historical details of the regularization of several of the verbs in the non-suppletion-domain group further undermines Bittner's case. The variation between strong and weak forms of laden is not the result of regularization, as Bittner claims, but rather of the confusion between the forms of what were originally two unrelated verbs: a strong one meaning 'load' (OHG [h] ladan) and a weak one meaning 'invite' (OHG ladon) (Kluge 1975; Paul 1989: 250). The leveling of Ruckumlaut in horen has nothing to do with the semantics of this verb, since this leveling has occurred without exception in all verbs with Ruckumlaut alternations other than e-a (Paul 1989: 257-258). Osthoff (1899: 12) would include hauen in the semantic suppletion domain, and Bittner gives no reason for his difference of opinion on this point. Finally, the primary meaning of bauen at the time when it underwent transfer to the weak class (pre-OHG) was not 'build' but rather 'dwell', 'inhabit'. It is from the same Indo-European root as English be/been, German bin/bist, Latin fui/futurus, Greek phyo/physis (Kluge 1975). Thus, far from supporting
Bittner's case, this item may well be a problematic example of regularization in a word from the heart of the semantic suppletion domain. Bittner (1988: 424, note 9) recognizes that, rather than trying to explain suppletion by means of an ad hoe principle, it would be preferable to search for ways in which suppletion within a particular domain and morphological regularity outside that domain both follow naturally from a single underlying principle of morphological organization (cf. Harnisch 1990: 60, note 8). More recently (1996: 36), he has concluded that Mayerthaler's principle of uniformity and transparency (i.e. biuniqueness, cf. notes 5, 6) is in fact all we need. This principle says that semantic sameness should correspond to formal sameness and semantic difference to formal difference. Since the forms in an inflectional paradigm have an element of semantic sameness as well as one of difference, we should expect to find suppletion wherever the latter is regarded as outweighing the former. According to Bittner, this will be the case when the perceived "closeness" of a concept leads one to make finer semantic distinctions.
Bittner is clearly on the right track here, but he still has no explanation for suppletion in grammatical function words. A question that Bittner implicitly raises but for which he has no answer is, do grammatical categories and concepts that are, in some sense, close to us have anything in common? There are occasional references in the suppletion literature to a vaguely defined notion of semantic, pragmatic, or cognitive basicness, which might be relevant here (e.g. Dressier 1985b: 333, 1985c: 50). (16)
Bybee et al. (1994:9 10) argue that lexical items that "encode major orientation points in human experience" should be grouped together with those that have the most generic meanings, since both groups belong to "the reference plane of basic, irreducible notions [...] which serves as the basis for grammatical meaning in human languages."
3. Diagrammatic iconicity
Mayerthaler's (1987: 48) use of the terms "diagrammaticity" and "constructional iconicity" to refer to just one simple type of diagram in morphology (cf. section 1.1 above) can create the misleading impression that this is the only way in which inflection is diagrammatic (cf. Dressler 1995: 35-36, note 6). In fact, this is just one of many types of morphological diagrammaticity. (17)
Under names such as "biuniqueness" and "isomorphism" (cf. note 5), Mayerthaler's principle of uniform encoding (sometimes in combination with his principle of transparency) is widely regarded as one of the most basic types of diagrammaticity in grammar: sameness of meaning is diagrammed by sameness of form, difference of meaning by difference of form (Haiman 1985; Waugh and Newfield 1995; Wurzel 1992: 231-232; McMahon 1994: 86). (18) Other types of morphological diagrams are discussed by Jakobson (1990), Andersen (1980:36-44), Shapiro (1983: 89-92, 146-190, 1991: 12-16), and Wiese (1996). Many scholars see diagrammaticity as the fundamental organizing principle of grammar: "a linguistic system, one may maintain, is a structure constituted chiefly by diagrammatic relations" (Andersen 1980: 36; cf. Anttila 1989: 16).
For the purposes of the present analysis, the most important extension of the notion of diagrammatic iconicity in inflectional morphology is that explored by Bybee. She shows that the degree of fusion between a stem and an inflectional affix and consequently the extent to which the shape of a stem changes across the various dimensions of a paradigm diagrams the degree of "semantic relevance" of each grammatical category to the meaning of the lexical item (1985a: 36-37, 1985b: 26-28, 1988: 129-131). (19)
Bybee's (1985a: 24) semantic-relevance hierarchy for verbal categories is shown in (2).
(2) valence > voice > aspect > tense > mood > number agreement > person agreement > gender agreement
In Bybee's theory, diagrammatic iconicity thus plays a central role in the organization of paradigms. It allows us to say where stem alternations are most likely to occur in a paradigm. Bybee's concept of iconicity would correctly predict, for example, that among the German strong verbs stem-vowel alternations associated with tense will be much more widespread and much more stable than those associated with person and number agreement. Bybee has little to say about the role of diagrammatic iconicity in the lexical distribution of stem alternations, however, so this part of her theory would not predict anything about which verbs are most likely to be strong and which weak, or which items are most likely to show suppletion.
Bybee actually discovered two kinds of diagrammaticity in the same phenomenon. The first has a syntagmatic dimension; the second is purely paradigmatic. Syntagmatic diagrammaticity can be seen in the fusion between stems and those affixes that represent a grammatical category that is highly relevant to lexical meaning. As Bybee (1985a: 29) puts it, "the psychological restructuring of two words into one depends on the relatedness of the semantic elements being joined, and their ability to form a coherent semantic whole." This first kind of diagrammaticity necessarily entails the second (purely paradigmatic) type that only becomes apparent when we consider at least two dimensions of a paradigm. If affixes expressing more "relevant" grammatical meanings have more of an effect on stem shape than those expressing less relevant meanings, it follows that the degree of change in stem shape across the various dimensions of a paradigm will diagram the degree of semantic distance between forms. In Bybee's words, "Degree of relatedness is diagrammed by morphophonemic alternations: The more closely related two forms are semantically, the more likely they are to be similar morphophonemically" (1988: 130).
In (3) below, we see parts of the paradigms of a strong and a weak German verb. Which of these paradigms is "more iconic"? Mayerthaler believed that this question had a straightforward answer. Since affixation is the optimal way of making a semantically marked form (such as the past tense) more "featured" (see section 1.1), the weak-verb paradigm, where past tense is marked by the dental suffix, is the more iconic one. It also displays a uniform (biunique) encoding of the stem and is therefore more iconic than the strong paradigm on two counts. According to Bybee, however, the fact that, in the strong verb, there is a stem-vowel alternation associated with tense while the shape of the stem remains constant across the person number forms within each tense diagrams the facts that (1) tense has more of an effect on the meaning of a verb than does person or number agreement; and consequently (2) all the forms of a given tense are more closely related semantically than are any forms across tenses. The paradigms of the regular weak verbs do not diagram these semantic relations in this way and are therefore, in this sense, less iconic than the strong verbs. Thus, while biuniqueness, transparency and "constructional iconicity" are certainly iconic, violations of these principles can themselves be iconic in more complex ways. (20)
(3) strong weak Present: singe singen lerne lernen singt singst lernst lernt singt singen lernt lernen Preterite: sang sangen lernte lernten sangst sangt lerntest lerntet sang sangen lernte lernten
4. The diagrammaticity of suppletion
It should be fairly obvious how this kind of complex syntagmatic and paradigmatic diagrammaticity can be extended to explain not just the distribution of stem alternations in paradigms, but also their distribution in the lexicon (cf. Osthoff 1899:44 45). Bybee herself points out that a given grammatical category may be more semantically relevant to some lexical items than to others (1985a: 93). Perfectivity, for example, affects the meaning of some verbs much more than it does others.
Accounting for suppletion involves a somewhat different kind of extension of Bybee's diagrammaticity. In comparing the paradigm of the highly suppletive German verb sein 'to be' (see  below) with that of a regular weak verb such as lernen 'learn, study' (see  above), for example, we note that the high degree of change in stem shape across the forms of the paradigm of sein contrasts with the stable shape of the stem of lernen. This is the same kind of contrast that Bybee found among the dimensions of a single paradigm. Can we see the same kinds of diagrammaticity in the contrast between the paradigm of sein and that of lernen? Is there any sense, first of all, in which "the relatedness of the semantic elements being joined" (i.e. stem and inflectional affixes) is greater in sein than in lernen? Or to put it slightly differently, is there any difference between the two verbs in the extent to which stem and inflectional affixes "form a coherent semantic whole"? I would argue that there is. The "lexical" meaning of sein is itself essentially grammatical. Suppletive paradigms for such words could thus be seen as merely one manifestation of the general tendency toward combined rather than separate expression of grammatical categories (cumulative exponence) in inflectional languages (Harnisch 1990: 60, note 8).
Second, recalling Bybee's formulation that "[t]he more closely related two forms are semantically, the more likely they are to be similar morphophonemically" (1988: 130), is there any sense in which the forms of lernen are more closely related to each other semantically than are the forms of sein? Again, there clearly is. Lernen has a much more specific lexical meaning than sein. We could say that the element of semantic sameness is stronger in the paradigm of lernen than in that of sein, and this is diagrammed by the relative degree of formal sameness of stem shape in the two paradigms. Indeed, since many linguists have argued that words such as sein "do not have a meaning of their own" (Heine 1993: 23), what could be more diagrammatic than for such words not to have a form of their own either? (21)
This analysis predicts that suppletion should be most common among grammaticalized words and among those lexical items whose semantic generality makes them likely candidates for grammaticalization (Bybee et al. 1994:5 6). (22) As Giacalone Ramat points out (1995:121), grammaticalization typically involves a weakening or breakdown of the direct iconic (biunique) relation between meaning and form but can at the same time give rise to "new configurations of iconicity." There is frequent discussion in the grammaticalization literature of a type of iconicity whereby grammatical morphemes tend to be formally smaller than lexical stems, just as they are semantically "smaller" (Givon 1985: 198), and diachronically, weakening of semantic content tends to go hand in hand with phonological "erosion" (Heine 1993: 95, 109-112; Bybee et al. 1994: 19-21, 106-124). Suppletion can be regarded as one manifestation of this widely recognized type of iconicity. A strong-suppletive stem is as "small" as it could possibly be in the sense that it has no fixed phonetic substance whatsoever. Similar reasoning leads Bybee et al. (1994: 110-112) to regard stem allomorphy, and especially allomorphy that shows morphological conditioning, as one indication of the extent of grammaticalization of an item.
Strong support for this analysis can even be found in some of the studies that have purported to show a connection between suppletion and some notion of closeness to the speaker. Wurzel (1985), for example, sets out to demonstrate that suppletion is characteristic of the comparatives and superlatives of the dimensional adjectives as a semantic class, but his findings in fact show that suppletion is very common only among the most semantically generic members of this class, those corresponding to big and small, which are of course susceptible to semantic weakening to 'much'/'many' and 'little'/'few' and further grammaticalization to comparative/superlative markers. Among the other dimensional adjectives, with more specific meanings, such as 'long', 'short', 'high', 'wide', 'thick', 'far', Wurzel's own table (1985: 132) shows that suppletion is quite rare in the languages examined.
The iconicity of lesser degrees of stem allomorphy in less highly grammaticalized words can perhaps be seen in the Funktionsverben in languages such as English and German. The verbs that participate in constructions such as Abschied nehmen 'take leave' and zum Abschluss bringen 'bring to a conclusion', where the normal semantics of the verb is suspended, tend to be strong or irregular, with some degree of stem allomorphy. An anonymous referee for Linguistics has pointed out that the English phrasal verbs, such as bring up and write off, and the German separable-prefix verbs, such as anfangen and einsehen, also tend to be strong or irregular. This is a more problematic kind of data, however, since these constructions range from full retention to near full suspension of the semantics of the basic verb, and even a single such construction can have several meanings, covering the full range of degrees of idiomaticity.
There are, admittedly, some instances of inflectional suppletion that do not appear to lend themselves especially well to the analysis presented here. Verbs for 'eat' (Osthoff 1899: 7), for example, have a relatively specific lexical meaning and, at least in many languages, show no tendency toward grammaticalization. (23) It is no coincidence, however, that suppletion in such verbs is always associated with tense and/or aspect. Markey (1985: 59) recognizes suppletion "that is promoted by the lexical semantic requirements and restrictions of aspectual strategies" as a special case that can be accounted for in straightforward semantic terms:
Radicals with a lexically durative value (e.g. *bher- 'carry') were ill-suited to express the notion of accomplished process as conveyed by, for example the aorist, and, hence, radicals with lexically non-durative values (e.g., *[??]en-k'achieve, attain') were spliced into a paradigm, thereby rendering it suppletive.
In the case of 'eat', another factor may also play a role. Anyone who has experienced both hunger and a full belly knows how relevant tense/aspect is to eating. It may be necessary to distinguish between such "category-specific" suppletions, which result from the high relevance of one grammatical category to the meaning of a particular lexical item, and "general" suppletion, which primarily affects grammaticalized and grammaticalizing words and can potentially be associated with any grammatical category, although it should still obey Bybee's semantic relevance hierarchy.
5. Some problems with pure token-frequency-based accounts
Even if one accepts the plausibility of the analysis presented here, a case still has to be made to show that it is needed. If suppletion can be explained adequately in terms of token frequency, why should we even consider alternative accounts? Semantic generality will invariably and necessarily correlate strongly with token frequency, so no simple empirical test can decide between the two (cf. Heine 1993:110-111). Several kinds of arguments can be offered, however, in support of the position that token frequency alone cannot be the whole story of suppletion.
5.1. Problems with specific theories
Werner's theory relies crucially on the argument that suppletion can be accounted for by the fact that it is economical for highly frequent forms to be short or compressed (cf. section 2.2 above). He offers little evidence, however, that the forms in suppletive paradigms are consistently shorter than the nonsuppletive alternatives would be. Ronneberger-Sibold (1980: 146) claims that suppletive forms are in fact shorter "[i]n den meisten Fallen" and gives a few examples (cf. Harnisch 1990: 58; Pike 1965: 205), but Wurzel (1985: 121, 140, note 9, 1990: 91) and Bittner (1988: 421, 1996: 34) both explicitly contradict this claim and point out a couple of cases that are especially problematic for Werner's position: (1) the comparative of German gut was originally monosyllabic baz; the partial regularization that took place when the comparative suffix -er was added made the form just as long as regular *guter but left the stem suppletion intact; (2) the suppletive preterite wurde of the ultra-high-frequency verb werden is longer than the regular strong preterite ward that it is replacing. Wcrner's attempts to defend himself against such criticism are not convincing. He explains away cases like baz > besser by making an ad hoc appeal to formal differentiation (cf. section 2.2) rather than shortness as the determining factor, but more generally he resorts to the curious argument that the debate over suppletion should focus exclusively on the "extremen Fallen" where we do in fact find formal compression (1987a: 306, 310-311). Werner also never makes it clear how length is to be measured, whether one should count segments, morae, syllables, milliseconds, or something else. Even Ronneberger-Sibold (1980: 146-147) questions the central role that Werner attributes to formal shortness in accounting for suppletion.
Even if Werner could produce convincing evidence of a strong tendency for suppletive forms to be shorter than the alternatives, he would still need to present a plausible mechanism that would allow for the desirability of compactness to play a major role in the development and/or survival of suppletive paradigms. Werner appears to be rather unconcerned with this issue, but when he does address it explicitly (1987a: 289-304, 1989: 41), he suggests a simplistic analogy with well-known lexical shortening processes, such as contraction, abbreviation, and clipping, in which speakers deliberately shorten words that they use often (cf. Zipf 1935:20-39). In general, Werner appears to see the connection between language use and grammatical structure in terms of "teleology of purpose" (Anttila 1989: 403; cf. Shapiro 1991; Itkonen 1982). This also explains how Harnisch (1990: 60) can regard "haufig Gebrauchtes sollte [...] m6glischst kurz sein," (24) as a statement about causation. But even if one believes that "final causes seek out efficient causes" (Anttila 1985: 7), a description of how the efficient causes work is still an essential component of any complete account of a type of change. (25)
Ronneberger-Sibold (1980:147-150, 1987b: 529, 1988: 456-457, 459-461) emphasizes another way in which suppletion in high-frequency forms is supposedly economical, claiming that lexical access of irregular forms requires less processing than the production of regular forms. Many other linguists have made this same point (e.g. Dressler 1985a: 325; Bybee 1996: 249), and there appears to be substantial experimental evidence supporting it (Harnisch 1990: 59). (26) With this as her main argument, however, it is not clear whether Ronneberger-Sibold really belongs in Werner's camp. Unlike the arguments based on shortness and maximal differentiation, the notion of the greater efficiency of direct lexical access does not favor suppletion per se, since, as Ronneberger-Sibold herself emphasizes, there is nothing to prevent a speaker from accessing a regular form directly rather than producing it by rule. Suppletion merely arises as a historical byproduct of direct lexical access. This is not so much a variation on Werner's position as it is a refinement along Bybeean lines--of the suppletion-as-artifact view (cf. Ronneberger-Sibold 1987a: 249, 259).
The historical-artifact view is undoubtedly part of the story of suppletion, but there are many unanswered questions here too. Most obviously, the claim that suppletion can be attributed to resistance to analogical change is seldom subjected to any empirical scrutiny. When we do examine attested diachronic developments in suppletive paradigms, we often find a remarkably large amount of analogical change. The development of the verb sein in German provides an excellent illustration of this. A greatly simplified overview of the many changes that some of the irregular forms of this verb have undergone is shown in (4). (27) The development of this verb is similarly complex in the other Germanic languages, and many other suppletive paradigms also show a great deal of analogical change (cf. Mel'cuk 1994: 394-396).
(4) Present indicative:
1s *im > bim (bin)
2s *is > bis > bist
1p *ezum > birum (birn) > sin (sein)/seind/sint
2p *ezup > birut (birt)/bint > sit (seid)/sint
3p sint > sindun (sintun) > sin (sein)/seind/sint
Imperative singular: wis > bis > sei
Infinitive: wesan > sin (sein)
Preterite indicative singular: was > war
Preterite participle: gewesen/gewest/gesin (gesein)
Thus, "resistance to analogical change" does not appear to be an accurate characterization of the historical behavior of suppletive paradigms. Rather, in at least some cases, these paradigms manage to maintain their suppletive character in the face of pervasive analogical change. Markey (1985: 60) points out that even "[i]f suppletion is leveled out, it will be reintroduced later."
It is also not clear how the resistance-to-change account would deal with the extremely common phenomenon of "relexification" (to borrow a term from pidgin and creole studies) in suppletive paradigms, as in OE ga:n-e:ode > ModE go-went; OE yfel-wiersa-wierrest > ModE bad--worse-worst (cf. Osthoff 1899: 5-6; Mel'cuk 1994:394 395; Wurzel 1985: 125; Ronneberger-Sibold 1987a: 249).
5.2. General problems
Any approach that attributes suppletion entirely to the effects of high token frequency must account for the tremendous differences in the amount of suppletion that we find in different languages and in particular for the strong typological tendency for suppletion to be relatively common in inflectional languages and much rarer in agglutinating languages (Mel'ruk 1994: 391; Dressier 1985b: 334, 1985c: 43-48). All of the forces that have been proposed to account for the direct connection between frequency and suppletion--whether they involve communicative efficiency or resistance to analogical change--would presumably be universal. The fact that some languages fail to develop suppletive paradigms in spite of these universal forces and in particular the correlation between the prevalence of suppletion and other morphological characteristics of languages indicates that suppletion--like other aspects of grammatical structure (Bybee et al. 1994: 115-119)--can only arise and survive where it is compatible with the overall character of the morphological system. (28) Dressler (1985b: 334, 1985c: 43) recognizes this when he attributes the greater incidence of suppletion in inflectional languages to the fact that these languages do not value transparency and constructional iconicity as highly as do agglutinating languages. The parallel mentioned in section 4 between suppletive paradigms for grammatical function words and combined expression (cumulative exponence) of grammatical categories in affixes brings suppletion even more in line with the structural properties of inflectional languages. (29)
6. Summary and conclusions
The analysis presented here is intended as a challenge to the widespread view that suppletion must be accounted for in purely extragrammatical terms since it fails to make the kind of sense that the rest of grammar does. We have seen that some of the natural morphologists, as well as many of their critics, have held this view. But just how "unnatural" is suppletion in semantically generic, grammaticalized, or grammaticalizing items? It violates the simple kinds of diagrammaticity described in Mayerthaler's principles of uniformity, transparency, and constructional iconicity, but many linguists have shown that there are more complex kinds of diagrams in morphology that necessarily entail violations of the simpler ones. Wurzel believed that suppletion could not be system-congruous, but we have seen in sections 4 and 5.2 that a strong case can be made that suppletion is in fact system-congruous in many inflectional languages, where cumulative exponence of grammatical categories is the norm. Last, it is not clear that suppletion should be seen as violating Wurzel's principle of inflectional-class stability at all. One could argue that an idiosyncratic suppletive paradigm is motivated by the idiosyncratic extramorphological properties of the item in question and is thus not in competition with any inflectional class (cf. Wurzel 1984: 67). (30)
If one accepts the conclusion that suppletion is consistent with the kinds of diagrammatic patterns found in the rest of grammar, one is still left with the crucial question of whether the apparent diagrammatic structure of grammar in general is ultimately anything more than an artifact of use and change. While Bybee describes some of the most interesting diagrammatic patterns yet discovered in morphology, she emphatically rejects the view of diagrammaticity as "the essence of language" (Jakobson 1990), arguing instead that all grammatical structure is in fact epiphenomenal and that "the real insights into human language come not from examining the synchronic iconicity, but more from understanding the dynamic processes that create that iconicity" (Bybee et al. 1994: 22, 106-107).
There is undoubtedly a great deal of truth in this view, but a complete theory must be able to account for the fact that morphological change is constrained by the general typological characteristics and by the dominant structural properties of a language (Bybee et al. 1994: 115-119; Wurzel 1984: 81-115, 1987: 62-76). As mentioned in section 5.2 above, the forces exerted by language use are presumably largely universal, but the diachronic effects that those forces have on a system depend to a high degree on the general nature and specific properties of that system.
One way to account for these facts is to grant some degree of psychological reality to diagrammaticity. Andersen sees speakers making diagrammatic sense of their language in the abductive acquisition process, "the process of grammar formation, in which every speaker of the language grasps its overall design and formulates extensive, detailed rules" (1980: 18, cf. 1973). A theory such as this does not necessarily contradict Bybee's claim that "language use is the primary determinant of structure" (1996: 247). The byproducts of language use may well serve as the main raw material for morphological structure. The extent to which learners embrace or resist these byproducts, however, will depend on how easy it is for them to make morphological (diagrammatic) sense of them (cf. Wurzel 1984: 189-194; 1989:189-195), and this in turn will depend on how compatible they are with the global and local diagrammatic characteristics of their grammar (cf. Shapiro 1990: 27). Thus, one language might accept a phenomenon such as suppletion while another resists or eliminates it in spite of the strong extramorphological forces working in its favor in high-frequency items (cf. Ronneberger-Sibold 1987b: 527-529).
State University of New York at Buffalo
Received 5 March 1998
Revised version received
21 September 1998
Andersen, Henning (1973). Abductive and deductive change. Language 49, 765-793.
--(1980). Morphological change: towards a typology. In Historical Morphology, Jacek Fisiak (ed.), 1-50. Trends in Linguistics, Studies and Monographs 17. New York: Mouton.
--(1991). On the projection of equivalence relations into syntagms. In New Vistas in Grammar: Invariance and Variation, Linda Waugh and Stephen Rudy (eds.), 287-311. Current Issues in Linguistic Theory 49. Philadelphia: Benjamins.
Anttila, Raimo (1985). Dynamics in morphology. Acta Linguistica Academiae Scientiarum Hungaricae 25, 3-30.
--(1989). Historical and Comparative Linguistics, 2nd ed. Current Issues in Linguistic Theory 6. Philadelphia: Benjamins.
Aronoff, Mark (1976). Word Formation in Generative Grammar. Linguistic Inquiry Monographs 1. Cambridge, MA: MIT Press.
--(1994). Morphology by Itself. Linguistic Inquiry Monographs 22. Cambridge, MA: MIT Press.
Battistella, Edwin L. (1996). The Logic of Markedness. New York: Oxford University Press.
Beard, Robert (1995). Lexeme-Morpheme Base Morphology. Albany: State University of New York Press.
Birkmann, Thomas (1987). Prateritoprasentia. Morphologische Entwicklungen einer Sonderklasse in den altgermanischen Sprachen. Linguistische Arbeiten 188. Tubingen: Niemeyer.
Bittner, Andreas (1988). Regulare Irregularitaten. Zeitschrift fur Phonetik, Sprachwissenschaft und Kommunikationsforschung 41, 416-425.
--(1996). Starke "schwache" Verben--schwache "starke" Verben. Deutsche Verbflexion und Naturlichkeit. Studien zur deutschen Grammatik 51. Tubingen: Stauffenburg.
Braunmuller, Kurt (1984). Morphologische Undurchsichtigkeit--ein Charakteristikum kleiner Sprachen. Kopenhagener Beitrtige zur germanistischen Linguistik 22, 48-68.
Bybee, Joan L. (1985a). Morphology. A Study of the Relation between Meaning and Form. Philadelphia: Benjamins.
--(1985b). Diagrammatic iconicity in stem-inflection relations. In Iconicity in Syntax, John Haiman (ed.), 11-47. Typological Studies in Language 6. Philadelphia: Benjamins.
--(1988). Morphology as lexical organization. In Theoretical Morphology, Michael Hammond and Michael Noonan (eds.), 119-141. New York: Academic Press.
--(1996). Productivity, regularity and fusion: how language use affects the lexicon. In Trubetzkoy's Orphan (Proceedings of the Montreal Roundtable "Morphonology: Contemporary Responses," Montreal, September 30--October 2, 1994), Rajendra Singh (ed.), 247-269. Current Issues in Linguistic Theory 144. Philadelphia: Benjamins.
--; Perkins, Revere; and Pagliuca, William (1994). The Evolution of Grammar. Chicago: University of Chicago Press.
Carstairs, Andrew (1988). Nonconcatenative inflection and paradigm economy. In Theoretical Morphology, Michael Hammond and Michael Noonan (eds.), 71 77. New York: Academic Press.
Carstairs-McCarthy, Andrew (1992). Current Morphology. New York: Routledge.
--(1994). Inflection classes, gender, and the principle of contrast. Language 70, 737-788.
Croft, William (1990). Typology and Universals. New York: Cambridge University Press.
Dressier, Wolfgang U. (1985a). Morphonology: The Dynamics of Derivation, Kenneth C. Hill (ed.). Linguistica Extranea, Studia 12. Ann Arbor, MI: Karoma.
--(1985b). On the predictiveness of natural morphology. Journal of Linguistics 21, 321-337.
--(1985c). Sur le statut de la supplrance dans la morphologie naturelle. Langages 78, 41-56.
--(1987). Word formation (WF) as part of natural morphology. In Leitmotifs in Natural Morphology, Wolfgang U. Dressier (ed.), 99-126. Studies in Language Companion Series 10. Philadelphia: Benjamins.
--(1995). Interactions between iconicity and other semiotic parameters in language. In Iconicity in Language, Raffaele Simone (ed.), 21-37. Current Issues in Linguistic Theory 110. Philadelphia: Benjamins.
--; Mayerthaler, Willi; Panagl, Oswald; and Wurzel, Wolfgang U. (1987). Introduction. In Leitmotifs in Natural Morphology, Wolfgang U. Dressler (ed.), 3-22. Studies in Language Companion Series 10. Philadelphia: Benjamins.
--; and Barbaresi, Lavinia Merlini (1994). Morphopragmatics. Diminutives and Intensifiers in Italian, German, and Other Languages. Trends in Linguistics, Studies and Monographs 76. New York: Mouton de Gruyter.
Fertig, David (1998). Grammaticalization and the development of suppletion. Paper presented at the German Department Roundtable, University of Wisconsin, Madison.
Giacalone Ramat, Anna (1995). Iconicity in grammaticalization processes. In Iconicity in Language, Raffaele Simone (ed.), 119-139. Current Issues in Linguistic Theory 110. Philadelphia: Benjamins.
Givon, Talmy (1985). Iconicity, isomorphism and non-arbitrary coding in syntax. In Iconicity in Syntax, John Haiman (ed.), 187-219. Typological Studies in Language 6. Philadelphia: Benjamins.
--(1995). Isomorphism in the grammatical code. In Iconicity in Language, Raffaele Simone (ed.), 47-76. Current Issues in Linguistic Theory 110. Philadelphia: Benjamins.
Greenberg, Joseph H. (1966). Language Universals. Janua Linguarum, Series Minor 59. The Hague: Mouton.
Grimm, Jacob (1965 ). Kleinere Schriften, III. Hildesheim: Olms.
Haiman, John (1985). Natural Syntax. New York: Cambridge University Press.
Harnisch, Rudiger (1990). Morphologische Irregularitat-Gebrauchshaufigkeit-psychische Nahe. Ein Zusammenhang im empirischen Befund und in seiner theoretischen Tragweite. In Naturalists at Krems (Papers from the Workshop on Natural Phonology and Natural Morphology, Krems, 1-7 July 1988), Julian Mendez Dosuna and Carmen Pensado (eds.), 53-64. Acta Salmanticensia, Estudios Filologicos 227. Salamanca: Ediciones Universidad.
Haspelmath, Martin (1990). The grammaticization of passive morphology. Studies in Language 14, 25-72.
Heine, Bernd (1993). Auxiliaries. New York: Oxford University Press.
Hentschel, Gerd (1990). Naturlichkeit vs. Frequenz Uberlegungen zu betonungsabhangiger Allomorphie in der Deklination russischer Substantive. In Naturalists at Krems (Papers from the Workshop on Natural Phonology and Natural Morphology, Krems, 1-7 July 1988), Julian Mendez Dosuna and Carmen Pensado (eds.), 65-76. Acta Salmanticensia, Estudios Filologicos 227. Salamanca: Ediciones Universidad.
Itkonen, Esa (1982). Short-term and long-term teleology in linguistic change. In Papers from the 3rd International Conference on Historical Linguistics, J. Peter Maher, Allan R. Bomhard, and E. F. Konrad Koerner (eds.), 85-118. Current Issues in Linguistic Theory 13. Amsterdam: Benjamins.
Jaeger, Jeri J.; Lockwood, Alan H.; Kemmerer, David L.; Van Valin, Robert D.; Murphy, Brian W.; and Khalak, Janif G. (1996). A positron emission tomographic study of regular and irregular verb morphology in English. Language 72, 451-497.
Jakobson, Roman (1985). Mark and feature. In Selected Writings, vol. 7, Stephen Rudy (ed.), 122-124. New York: Mouton. (Reprinted from World Papers in Phonetics: Festschrift for Dr. Onishi's Kiju. Tokyo: Phonetic Society of Japan, 1974.)
--(1990). Quest for the essence of language. In On Language, Linda R. Waugh and Monique Monville-Burston (eds.), 407-421. Cambridge, MA: Harvard University Press. (Reprinted from Diogenes 51, 1966.)
Kiparsky, Paul (1992). Analogy. In International Encyclopedia of Linguistics, vol. 1, William Bright (ed.), 56-61. New York: Oxford University Press.
Kluge, Friedrich (1975). Etymologisches Worterbuch der deutschen Sprache, 21st ed. New York: de Gruyter.
Lakoff, George; and Johnson, Mark (1980). Metaphors We Live By. Chicago: University of Chicago Press.
Lass, Roger (1990). How to do things with junk: exaptation in language evolution. Journal of Linguistics 26, 79-102.
Liszka, James Jakob (1996). A General Introduction to the Semiotic of Charles Sanders Peirce. Bloomington: Indiana University Press.
Markey, T. L. (1985). On suppletion. Diachronica 2, 51-66.
Matthews, P. H. (1991). Morphology, 2nd ed. New York: Cambridge University Press.
Mayerthaler, Willi (1981). Morphologische Naturlichkeit. Linguistische Forschungen 28. Wiesbaden: Athenaion.
--(1987). System-independent morphological naturalness. In Leitmotifs in Natural Morphology, Wolfgang U. Dressier (ed.), 25-58. Studies in Language Companion Series 10. Philadelphia: Benjamins.
McMahon, April M. S. (1994). Understanding Language Change. New York: Cambridge University Press.
Meineke, Eckhard (1989). "Naturlichkeit" und "Okonomie" Neuere Auffassungen des Sprachwandels. Sprachwissenschaft 14, 318-356.
Mel'cuk, Igor (1994). Suppletion: toward a logical analysis of the concept. Studies in Language 18, 339-410.
Newmeyer, Frederick J. (1992). Iconicity and generative grammar. Language 68, 756-796.
Nida, Eugene A. (1963). The identification of morphemes. In Readings in Linguistics, 3rd ed., Martin Joos (ed.), 255-271. New York: American Council of Learned Societies. (Reprinted from Language 24 , 414-41.)
O'Neil, Wayne (1978). The evolution of the Germanic inflectional systems: a study in the causes of language change. Orbis 27, 248-286.
Osthoff, Hermann (1899). Vom Suppletivwesen der indogermanischen Sprachen. Akademische Rede zur Feier des Geburtsfestes des hochstseligen Grossherzogs Karl Friedrich. Heidelberg: Horning.
Paul, Hermann (1989). Mittelhochdeutsche Grammatik, 23rd ed., Peter Wiehl and Siegfried Grosse (eds.). Sammlung kurzer Grammatiken germanischer Dialekte A.2. Tubingen: Niemeyer.
Pesetsky, David (1997). Optimality theory and syntax: movement and pronunciation. In Optimality Theory, Diana Archangeli and D. Terence Langendoen (eds.), 134-170. Malden, MA: Blackwell.
Pike, Kenneth L. (1965). Non-linear order and anti-redundancy in German morphological matrices. Zeitschrift fur Mundartforschung 31, 193-221.
Ronneberger-Sibold, Elke (1980). Sprachverwendung--Sprachsystem Okonomie und Wandel. Linguistische Arbeiten 87. Tubingen: Niemeyer.
--(1987a). Verschiedene Wege zur Entstehung von suppletiven Flexionsparadigmen, Deutsch gern--lieber--am liebsten. In Beitrage zum 3. Essener Kolloquium uber Sprachwandel und seine bestimmenden Faktoren, Norbert Boretzky, Werner Enninger, and Thomas Stolz (eds.), 243-264. Bochum--Essener Beitrage zur Sprachwandelforschung IV. Bochum: Brockmeyer.
--(1987b). A performance model for a natural theory of linguistic change. In Papers from the 7th International Conference on Historical Linguistics, Anna Giacalone Ramat, Onofrio Carruba, and Giuliano Bernini (eds.), 517-533. Current Issues in Linguistic Theory 48. Philadelphia: Benjamins.
--(1988). Enstehung von Suppletion and Naturliche Morphologie. Zeitschrift fur Phonetik, Sprachwissenschaft und Kommunikationsforschung 41, 453-462.
Rudes, Blair A. (1980). On the nature of verbal suppletion. Linguistics 18, 655-676.
Shapiro, Michael (1983). The Sense of Grammar. Bloomington: Indiana University Press.
--(1990). On a universal criterion of rule coherence. In Contemporary Morphology, Wolfgang U. Dressier, Hans C. Luschutzky, Oskar E. Pfeiffer, and John R. Rennison (eds.), 25-34. Trends in Linguistics, Studies and Monographs 49. New York: Mouton de Gruyter.
--(1991). The Sense of Change. Bloomington: Indiana University Press.
Singh, Rajendra (1996). Natural phono(morpho)logy: a view from the outside. In Natural Phonology: The State of the Art. Bernhard Hurch and Richard A. Rhodes (eds.), 1-38. Trends in Linguistics, Studies and Monographs 92. New York: Mouton de Gruyter.
Trudgill, Peter (1983). On Dialect. New York: New York University Press.
--(1992). Dialect typology and social structure. In Language Contact. Ernst Hakon Jahr (ed.), 195-211. Trends in Linguistics, Studies and Monographs 60. New York: Mouton de Gruyter.
Waugh, Linda R.; and Newfield, Madeleine (1995). Iconicity in the lexicon and its relevance for a theory of morphology. In Syntactic Iconicity and Linguistic Freezes, Marge E. Landsberg (ed.), 181-221. Studies in Anthropological Linguistics 9. New York: Mouton de Gruyter.
Werner, Otmar (1987a). Naturlichkeit und Nutzen morphologischer Irregularitat. In Beitrgige zum 3. Essener Kolloquium uber Sprachwandel und seine bestimmenden Faktoren, Norbert Boretzky, Werner Enninger, and Thomas Stolz (eds.), 289-316. Bochum-Essener Beitrage zur Sprachwandelforschung IV. Bochum: Brockmeyer.
--(1987b). The aim of morphological change is a good mixture not a uniform language type. In Papers from the 7th International Conference on Historical Linguistics, Anna Giacalone Ramat, Onofrio Carruba, and Giuliano Bernini (eds.), 591-606. Current Issues in Linguistic Theory 48. Philadelphia: Benjamins.
--(1989). Sprachokonomie und Naturlichkeit im Bereich der Morphologie. Zeitschrift fur Phonetik, Sprachwissenschaft und Kommunikationsforschung 42, 34-47.
Wiese, Bernd (1996). Iconicity and syncretism. In Theoretical Linguistics and Grammatical Description. Papers in Honour of Hans-Heinrich Lieb, Robin Sackmann (ed.), 323-344. Current Issues in Linguistic Theory 138. Philadelphia: Benjamins.
Wurzel, Wolfgang Ulrich (1984). Flexionsmorphologie und Naturlichkeit. Studia Grammatica 21. Berlin: Akademie-Verlag.
--(1985). Zur Suppletion bei den Dimensionsadjektiven. In Studien zur Morphologie und Phonologie I, W. U. Wurzel (ed.), 114-143. Linguistische Studien, A 126. Berlin: Akademie der Wissenschaften der DDR, Zentralinstitut fur Sprachwissenschaft.
--(1987). System-dependent morphological naturalness in inflection. In Leitmotifs in Natural Morphology, Wolfgang U. Dressler (ed.), 59-96. Studies in Language Companion Series 10. Philadelphia: Benjamins.
--(1989). Inflectional Morphology and Naturalness (translation of Wurzel 1984). Boston: Kluwer.
--(1990). Gedanken zu Suppletion und Naturlichkeit. Zeitschrift fur Phonetik, Sprachwissenschaft und Kommunikationsforschung 43, 86-91.
--(1992). The structuralist heritage in natural morphology. In Prospects for a New Structuralism, Hans-Heinrich Lieb (ed.), 225-241. Current Issues in Linguistic Theory 96. Philadelphia: Benjamins.
Zipf, George Kingsley (1935). The Psycho-biology of Language. Boston: Houghton Mifflin.
Zwicky, Arnold M. (1990). Inflectional morphology as a (sub)component of grammar. In Contemporary Morphology, Wolfgang U. Dressier, Hans C. Luschutzky, Oskar E. Pfeiffer, and John R. Rennison (eds.), 217-236. Trends in Linguistics, Studies and Monographs 49. New York: Mouton de Gruyter.
* would like to thank the participants at the Third Germanic Linguistics Annual Conference at UCLA (April, 1997) for their questions and comments on an earlier version of this paper. Special thanks to Raimo Anttila, Joe Salmons, and the anonymous referees for Linguistics. Correspondence address: Department of Modern Languages and Literature, 910 Clemens Hall, State University of New York at Buffalo, Buffalo, NY 14260, USA. E-mail: firstname.lastname@example.org.
(1.) See Mel'cuk (1994) for a more formal definition.
(2.) See Trudgill (1983: 102-107), however, for a very different view of what should be regarded as "natural" in historical linguistics.
(3.) Cf. Givon's "quantity principle" (1995: 49-51) as well as his "meta-iconic markedness principle" (1995: 68). The same basic insight is discussed by many others, including Jakobson (1990) and Lakoff and Johnson (1980: 127-128).
(4.) Carstairs-McCarthy (1992: 218, 263, note 2) proposes "markerful" as an alternative to Mayerthaler's own "featured" to translate the German merkmalhaft. "Featured" is indeed an extremely poor choice in light of Jakobson's (1985) important distinction between "mark" and "feature" in a completely different sense. I will nevertheless use "featured" in this article for the sake of consistency with Mayerthaler's English publications. Even the original German terms are potentially confusing, since Jakobson and Trubetzkoy originally used merkmalhaft/merkmallos in precisely the sense of Mayerthaler's markiert/unmarkiert (cf. Shapiro 1983: 75).
(5.) Dressier and many others call it "biuniqueness." Haiman (1985), Itkonen (1982: 90), McMahon (1994: 86), and Waugh and Newfield (1995) prefer "isomorphism," although this term is potentially confusing, since its original use in mathematics is broader (Croft 1990: 164-165; Newmeyer 1992: 760, note 8), and many linguists also use it in a more general sense (Shapiro 1990: 25, 31-32, note 1, 1991: 11, 13; Dressler 1985a: 368, note 168; Givon 1995; Giacalone Ramat 1995: 122-123). Waugh and Newfield (1995: 211, note 16) give a long list of the many different names by which this principle has been known (cf. also Shapiro 1991: 10-11; Haiman 1985: 14). Aronoff (1994: 8) offers a brief overview of the long history of the concept.
(6.) The different senses in which Mayerthaler uses the term "transparency" have led to considerable confusion. When first laying out his principles of naturalness (1981: 35), he uses "transparency" to refer to the "no polysemy/homonymy/multifunctionality" part (as opposed to the "no synonymy/allomorphy" part) of the principle of uniformity (cf. Wurzel 1984: 23, 1989: 11-12). Later in the same work (1981: 131) and in Mayerthaler (1987), he uses it in the morphotactic-segmentability/semantic-compositionality sense outlined above (cf. Wurzel 1984: 29, 1989: 17). Carstairs-McCarthy (1992: 219) and McMahon (1994: 98-99) have picked up only on the first usage, while Hentschel (1990: 66) and Kiparsky (1992: 59) combine the two. Even in the latter sense, uniformity and transparency are often treated as two sides (paradigmatic and syntagmatic, respectively) of a single "one function--one form" principle (Dressler et al. 1987: 7; Wurzel 1992: 231 232; Bittner 1996: 10; Bybee 1996: 253).
(7.) It is surely no coincidence, however, that it is precisely in the case of little-less that we find a comparative adjective than is shorter than the positive grade.
(8.) The alternative name "inflectional-class motivation," suggested by Wurzel (1992: 226, 237), perhaps captures the essence of the principle better.
(9.) I follow Carstairs-McCarthy (1994: 740, note 2) in avoiding Wurzel's potentially confusing term "complementary" to refer to competing classes.
(10.) "In terms of the inflectional system and beyond that 'the entire linguistic system, suppletion is consequently non-functional."
(11.) This last connection was explored in much greater detail by Pike (1965), who points out that the inherent efficiency of suppletion is only practicable in closed classes where there is no need for productive morphological rules. I would like to thank Raimo Anttila for calling my attention to Pike's work on suppletion.
(12.) The best examples involve singular--plural pairs among the 1st and 2nd person personal pronouns. The analysis is best developed in Dressler and Barbaresi (1994: 60-63). The basic insight that the formal difference between the singulars and plurals of the 1st and 2nd person pronouns is related to the fact that 'we' and 'you' are semantically not straightforward plurals of 'I' and 'thou' is discussed by Grimm (1965 : 239-240) and by the second-century (AD) Alexandrian grammarian Apollonius Dyscolus (Mel'cuk 1994: 387). Osthoff (1899: 45, 74, note 108) agrees with Dressier in regarding this as a case of suppletion. Mel'cuk, on the other hand, argues that these pairs should simply be regarded as unrelated lexemes.
(13.) See Ronneberger-Sibold (1987a: 249-250) for further criticism of the notion of the Nahbereich.
(14.) Bittner (1996: 35) argues for a "modified application of the concept of markedness reversal" (my translation), but it is not at all clear what he has in mind. Shapiro (1983: 93-94) defends the unconstrained use of markedness reversal.
(15.) Bittner 1996 replaces helfen with sprechen and changes some of the pairings, but the problems described here still apply.
(16.) Markey (1985: 62-63) argues that suppletion in the copula can, like many other instances of suppletion, be seen as "an iconic index of the fundamental contrast of ego vs. 'other'." He gives no indication of how this type of analysis could be extended to other grammatical function words, however.
(17.) Dressier (1985b: 329, 333, 1985c: 51-52) might consider many of the types of iconicity discussed in this article to involve metaphors rather than (or in addition to) diagrams. As Waugh and Newfield (1995: 191) point out, most linguists who make use of Peircean semiotics conflate diagram(maticity) and metaphor(icity) and use the former term indiscriminately to refer to both. Dressier regards the distinction as crucial, since he claims that diagrams are more natural than metaphors, but he does not make his criteria for distinguishing between the two explicit. From his examples it would appear that he only recognizes syntagmatic diagrams as diagrams, since all of his "metaphors" in fact participate in paradigmatic diagrams. Indeed, as far as I can see, a paradigm of conventionalized metaphors must necessarily constitute a diagram. For example, Dressler regards morphologically subtractive diminutives as bearing a metaphoric relationship to the small entities to which they refer and sees no diagrammaticity here, but the paradigmatic relationship between the formally shorter diminutive and the longer basic word obviously constitutes a diagram of the semantic relationship. For discussion of the distinction between syntagmatic and paradigmatic diagrammaticity, cf. Matthews (1991: 225-245), Andersen (1991), Battistella (1996: 62), and Haiman (1985: 4-5). For more on Peirce's diagram/metaphor distinction, cf. Shapiro (1983: 184-185) and Liszka (1996: 37-38, 131, note 32 with further references). Lakoff and Johnson (1980: 126-138) also see "metaphor" playing a role in many of the types of iconicity discussed here, but they are concerned with the conceptual metaphors that underlie form-meaning parallels rather than with the semiotic status of the parallels themselves.
(18.) Newmeyer (1992: 760, note 8) "doubt[s] that Peirce would have regarded isomorphism as a species of iconicity." Dressler (1995: 35, note 1) says that biuniqueness "is related to, but not identical with, the parameter of iconicity." Unfortunately, neither elaborates any further on these views.
(19.) Some neogrammarians also noticed this connection between the paradigmatic distribution of stem alternations and the semantic relevance of grammatical categories to lexical meaning (Osthoff 1899: 44-45, 55, note 7).
(20.) Dressier has also made this point on a number of occasions (e.g. 1995: 26-27, 1985a, 323-325, 1987: 104-105; Dressler et al. 1987: 21, note 8; Dressler and Barbaresi 1994: 47, 62-63). He discusses a way in which violations of morphotactic transparency can be diagrammatic, claiming that where the meaning of a derived word is the sum of the meanings of its parts, we tend to find that the parts are relatively distinct and segmentable, whereas we tend to find more fusion of morphs where the meaning of the whole is less compositional. In Dressler's words, there is "a diagrammatic relationship between morphotactic and morphosemantic transparency/opacity" such that "a relatively high degree of morphosemantic transparency tends to be reflected in a relatively high degree of morphotactic transparency" (1995: 27; cf. also Aronoff 1976: 26, 40, note 8).
(21.) Zwicky (1990: 230) characterizes grammatical function words as "pure combinations [...] of syntactic and grammatical category features, with no other semantics" (cf. also Pesetsky 1997: 164).
(22.) Rudes (1980: 655, 668, 672) speaks of "general meanings" and "neutrality of meaning" as a characteristic of suppletive lexical items and of "semantic weakening" in association with the creation of suppletive paradigms.
(23.) An anonymous referee for Linguistics has pointed out that verbs meaning 'eat' do occasionally become grammaticalized as passive morphemes (cf. Haspelmath 1990: 41, 64, note 9).
(24.) "Frequently used items should [...] be as short as possible."
(25.) Croft's (1990: 156) claim that "[t]he hidden assumption for the connection between frequency and zero expression is the principle that people will shorten the linguistic expressions that are used most commonly for economy" is based on misreading of Greenberg (1966: 65-70), who emphasizes (1966: 70) that "frequency is itself but a symptom." Greenberg's point is that speakers will naturally tend to give no formal marking to whatever is most common or normal in their experience. He sees high token frequency as another, independent consequence of this commonness or normalness. There is no "hidden assumption" of a causal connection between the two.
(26.) Jaeger et al. (1996), however, have recently found that, at least in one experimental setting, the production of irregular forms takes significantly longer than that of regular forms.
(27.) In many cases, the chronology here is only rough because the reality has always included a tremendous amount of synchronic variation, and we still see this very strikingly in the modern dialects.
(28.) Sociolinguistic factors often play an important role here as well (Trudgill 1992; Braunmuller 1984; O'Neil 1978). The ultimate account of suppletion will have to consider all the complex interactions among (1) universal extramorphological forces; (2) extragrammatical forces that are dependent on the sociolinguistic context; and (3) the morphological system. Werner (1987b: 603) suggests that the morphologicaltypological differences among languages might be entirely attributable to sociolinguistic factors, but he offers no arguments or evidence in support of this position.
(29.) All the key points made about suppletion in this article lend themselves well to an analysis within the framework of Beard's (1995) lexeme--morpheme base morphology, which complements the iconicity-based analysis presented here. Specifically, (1) the fact that the most common suppletive items are grammatical function words; (2) the view of suppletion as a special case of cumulative exponence; and (3) the fact that suppletion is common in inflecting languages but rare in agglutinating languages all follow from Beard's central thesis that grammatical morphemes (including free-standing function words) are not lexical items and therefore do not necessarily have any fixed underlying phonological representation associated with them. Rather, the stems of grammatical function words, like all other grammatical morphemes, acquire their phonological shape from a morphological-spelling mechanism that is capable of much more complex kinds of mapping between form and meaning/function than the isomorphism that is characteristic of the lexicon (cf. Fertig 1998).
(30.) There may be another sense in which some (though not all) suppletive paradigms belong to stable inflectional classes if Carstairs-McCarthy (1994: 739-740, 757-759; Carstairs 1988) is right that only affixal inflection should count for principles such as paradigm economy, no blur, and, presumably, inflectional-class stability.
|Printer friendly Cite/link Email Feedback|
|Publication:||Linguistics: an interdisciplinary journal of the language sciences|
|Date:||Dec 1, 1998|
|Previous Article:||Infinitival copular complement clauses in English: explaining the predominance of passive matrix verbs (1).|
|Next Article:||Indexed stems and Russian word formation: a network-morphology account of Russian personal nouns (1).|
|Collateral adjectives, Latinate vocabulary, and English morphology (1). (Linguistics).|