Analogy in Morphology
Summary and Keywords
Analogy is traditionally regarded as one of the three main factors responsible for language change, along with sound change and borrowing. Whereas sound change is understood to be phonetically motivated and blind to structural patterns and semantic and functional relationships, analogy is licensed precisely by those patterns and relationships. In the Neogrammarian tradition, analogical change is regarded, at least largely, as a by-product of the normal operation (acquisition, representation, and use) of the mental grammar. Historical linguists commonly use proportional equations of the form A : B = C : X to represent analogical innovations, where A, B, and C are (sets of) word forms known to the innovator, who solves for X by discerning a formal relationship between A and B and then deductively arriving at a form that is related to C in the same way that B is related to A.
Along with the core type of analogical change captured by proportional equations, most historical linguists include a number of other phenomena under the analogy umbrella. Some of these, such as paradigm leveling—the reduction or elimination of stem alternations in paradigms—are arguably largely proportional, but others such as contamination and folk etymology seem to have less to do with the normal operation of the mental grammar and instead involve some kind of interference among the mental representations of phonetically or semantically similar forms.
The Neogrammarian approach to analogical change has been criticized and challenged on a variety of grounds, and a number of important scholars use the term “analogy” in a rather different sense, to refer to the role that phonological and/or semantic similarity play in the influence that forms exert on each other.
1. Analogy in Historical Linguistics
Linguists use the term “analogy” in a wide variety of senses. This article focuses on its use in the Neogrammarian tradition that dates back to the last quarter of the 19th century (see especially Paul, 1886) and continues to provide an important part of the theoretical foundation for historical linguistics in the early 21st century. Here, analogy refers to various types of innovations and changes in linguistic form that are licensed by, or at least sensitive to, structural patterns or semantic/functional relations among words or expressions. The key opposition is between analogical developments and “sound change,” which is understood to be purely phonetically motivated and thus “blind” to the patterns and relations responsible for analogical change. Many historical linguists also maintain a more or less clear opposition between analogy and reanalysis, which they associate roughly with the overt and covert sides of grammatical change, respectively (Andersen, 1980, 2001; Hopper & Traugott, 2003, pp. 39–70; Paul, 1886; cf. Timberlake, 1977). Typical examples of analogical change in morphology often involve lexical shifts from one inflectional pattern to another, as when the English verb glide–glode–glid(en) was regularized to glide–glided starting in the 17th century.
Analogical developments can involve patterns at any level of linguistic structure (Bloomfield, 1933, pp. 407–408, 440–442; Harris & Campbell, 1995; Kiparsky, 1974; Kroesch, 1926; Paul, 1886, pp. 85–98; Vennemann, 1972). This article focuses on the role of analogy in morphological change.
2. Innovation, Change, and Variation
At least since Paul (1886), historical linguists have recognized that the models and mechanisms of analogy that they propose deal with innovations in the speech and mental representations of individuals rather than directly with what we normally think of as ‘language change’, which involves the usage and norms of speech communities. The relationship between innovations and change, in this sense, is complex, and all change undoubtedly involves processes and forces that are not straightforwardly analogical(cf. ; Andersen, 1980, 2001; Bloomfield, 1933, p. 405).
Linguists realized that analogical change necessarily entails a period of variation between old and new forms long before they recognized that the same is true of sound change (Paul, 1886, pp. 94, 163–164), and for centuries grammarians have described widespread variation associated with ongoing analogical change (e.g., Wallis, 1688). Theoretical work on analogical change has always been strongly dominated, however, by a focus on individual innovations. This is partly because linguists who do serious work on analogical change are often motivated at least as much by an interest in the grammatical and cognitive processes and pressures behind analogy—and the light that historical data can shed on these matters—as they are by an interest in language change per se. Some varied examples of important work that illustrate this point include: Albright (2010); Albright and Fuß (2012, pp. 273–287); Bybee (1980); Enger (2013); Fertig (2016); Garrett (2008); Hare and Elman (1995); Hyman (2003); Kiparsky (2000, 2012); Kuryłowicz (1966); Maiden (2008, 2018); Paul (1877, 1886); and Wurzel (1984).
3. Types of Morphological Change Commonly Classified as ‘Analogical’
Many historical linguists distinguish a number of types of developments that they include under the “analogy” umbrella (Fertig, 2013, pp. 42–84; Hock, 1991, pp. 167–209, 2003; Hock & Joseph, 2009, pp. 150–171).
3.1 Proportional Analogy
Overviews often start with a core type of change usually labeled “proportional” or “four-part analogy.” The glide–glode→glide–glided development mentioned in section 1 can serve as an example (see (1)). The labels “proportional” and “four-part” are based on the traditional use of proportional equations with (minimally) four terms—three known and one unknown—as a model of the mechanism at work. Generally, the left-hand side of the equation contains two morphologically related forms (A and B). If A and B are two inflected forms of the same lexical item, then the third term (C) is typically a known inflected form of a target lexical item that corresponds in its morphosyntactic and morpho(phono)logical properties to A. Solving the equation involves first discerning the relevant formal relationship between A and B and then coming up with a form (X) that is related to C in the same way that B is related to A. (See Stroppa & Yvon, 2005, for a more formal account.)
Crucially, the proportional model entails an understanding of analogical innovation as replacement rather than modification of existing forms. The old form—which would be glode in example (1)—does not appear in the equation at all and is considered irrelevant to the process. In fact, it is widely believed that analogical innovation is most likely to occur when the innovator has no mental representation of the old form, or fails to access that representation.
Note also that phonological similarity of the model item(s) (represented by treat in (1)) to the target item(s) (glide in (1))—what Albright and Hayes (2002) call the “shared context”—plays no role in this Neogrammarian conception of analogy (see section 5).
Paul (1886) regarded proportional equations like (1)—with whole surface word forms as the only allowable terms—as a rough model of the mental grammar, reflecting his “implicative” (Ackerman & Malouf, 2017; Stump, 2016, p. 257), word-and-paradigm, exemplar-based theoretical views, according to which subword morphological units such as roots, stems, and affixes are not represented as such in the grammar, and productive (implicit) morphological operations take fully inflected surface word forms as input and produce other forms in the paradigm as output (Blevins, 2016). Paul’s proportional equations have been adopted as a convenient model of analogical innovation by generations of historical linguists, many of whom do not share his theoretical views. If these equations are understood to capture the acquisition/construction of the mental grammar as well as its operation—with the grammar treated as a black box between relevant fragments of the input to the learner and the output of the language user—then they are largely compatible with a wide variety of morphological theories, including those that operate with abstract stems and affixes and explicit symbolic rules for combining subword elements or applying processes to stems. Proportional equations actually capture a stage in the acquisition process that is crucial for virtually any theory of morphology—the stage where learners have identified the morphosyntactic properties of many of the forms they have encountered and figured out which forms belong to which lexical items but have not (yet) abstracted out any explicit rules or subword elements (Goldsmith, 2006). Like the conception of “morphology as lexical organization” (Bybee, 1988), Paul’s proportional model treats this stage as essentially the end point of morphological acquisition. (As explained later in this section and in section 5, however, Bybee’s theory is fundamentally different from Paul’s in other respects.)
An implicative word-and-paradigm approach does entail one interesting hard constraint on (proportional) analogical change (cf. section 4). As Paul puts it: “one word can be subject to analogical influence from another in its inflection only if it [already] corresponds to the other word in the formation of one or more forms” (Paul, 1886, p. 95; translation from Fertig, 2015a, p. 94). To illustrate the kind of development that this constraint is intended to rule out, consider the (partial) early Old High German verb paradigms in (2):
During the Old High German period, the 2sg prs ind ending for the vast majority of verbs, represented in (2) by suochen, changed from -s to ‑st. Many scholars have seen preterite-present verbs like wiʒʒan, with 2sg weist, as the analogical source of this ‑st (e.g., Ringe & Taylor, 2014, pp. 353–355). When we try to write a proportional equation based on forms in (2) to capture this development, however, we find that none of the possibilities yields any obvious solution. In the proposed model lexeme wiʒʒan, both the presence of root-final ʒ(ʒ) in the rest of the paradigm but not in the 2sg and the distinctive agreement endings of preterite-present verbs mean that 2sg weist is not related to any other form in its paradigm in a way that can be paralleled in a normal verb like suochen. Two examples of the resulting unsolvable proportional equations are given in (3).
Because there are other historical processes—including but not limited to the “non-proportional” phenomena discussed in section 3.3—that can yield innovations resembling those attributed to proportional analogy without requiring any formal matches elsewhere in the paradigm, Paul’s proportional condition, by itself, does not make straightforward predictions about what changes can and cannot occur. Nevertheless, it can be an important consideration in choosing between alternative accounts of attested changes (Fertig, 2019).
Fundamentally incompatible with the proportional model of analogical change are theories that attribute morphological productivity largely to “product-oriented schemas” (Bybee & Moder, 1983; Bybee & Slobin, 1982; Kapatsinski, 2013). Product-oriented schemas are first-order generalizations about the phonological shapes of morphosyntactically defined sets of forms, such as past-tense verb forms or plural noun forms. These theories ascribe less significance to the second-order generalizations—involving paradigmatic relationships among forms—that analogical proportions are designed to capture (cf. ; Booij, 2018; Booij & Masini, 2015). Thus, for example, rather than characterizing the productive classes of German nouns with plurals ending in ‑e (/ə/), -er (/ɐ/), and -en (/ən/ or /n̩/) by positing operations that add these suffixes to the singular form, a product-oriented approach, in its purest form, would instead posit generalizations that capture the prosodic and segmental characteristics that are typical of German plurals, without any reference to the singular. This approach accounts for the high incidence of singular–plural identity among nouns whose singulars already satisfy one of the plural schemas, such as Wunder ‘miracle(s)’ or Becken ‘basin(s)’ (Köpcke, 1998; cf. Wiese, 2009).
Although some morphologists have used the term “analogy” to refer specifically and exclusively to marginal cases of semi-productivity based on minor, lexicalized patterns for which they do not want to posit abstract grammatical rules (Kiparsky, 1992, p. 56; cf. Clahsen, 1999; Pinker & Prince, 1994), most historical linguists have always understood analogy as encompassing all patterns that show any degree of productivity (cf. Blevins, 2016, pp. 97–116; Blevins & Blevins, 2009). Familiar examples of proportional analogical innovations often correspond to what many morphologists would regard as the application of highly productive morphological rules. The innovations consist of applying the existing rules in new contexts (Anderson, 1992, pp. 366–367). In many cases, the rule had previously been blocked in the relevant context by an irregular form or a more specific rule, in accordance with an “elsewhere” principle (Kiparsky, 1973), as in the glode→glided example. In other cases, the rule was previously subject to tighter positive conditions on its application, as when the German pattern for forming noun plurals with an ‑er suffix, which had originally been restricted to neuters, was extended to a small number of masculine nouns starting in the later Middle Ages: Mann–Männer ‘man–men’, Gott–Götter ‘god(s)’, etc. Such analogical extensions involving the relaxation of positive conditions on a morphological operation appear to be especially common in derivational processes that originally applied to a single lexical category—nouns, verbs, or adjectives—but are subsequently extended to apply to other input categories, sometimes ultimately shifting to apply primarily or exclusively to the new category. The German suffix ‑bar, for example, originally had a meaning similar to that of ‑ful and was used to create adjectives from nouns, as in fruchtbar ‘fertile, productive’ from Frucht ‘fruit’. In modern German ‑bar is a highly productive suffix for forming adjectives from verbs, with a meaning very close to that of English ‑able.
Prior existence of a fully productive rule is certainly not a necessary condition for analogical innovation, however. There are many attested changes based on very minor patterns. The English verb wear, for example, was originally a regular weak verb: wear–weared. The modern standard forms wore–worn, which first became common in the 16th century, are clearly attributable to an analogical change based on a pattern that only occurred in a handful of other verbs: bear, swear, tear. In some cases, the relevant pattern occurred in only a single lexical item or a single set of derivationally related words before it was extended analogically (Becker, 1990, p. 18; Cowgill, 1959). Of particular interest are cases where the emergence of a productive rule seems to go hand in hand with the creation of innovative analogical forms, as in the spawning of productive affixes based on blends: e.g., alcoholic→workaholic→‑(o/a)holic; marathon→walkathon→‑(a)thon (Fertig, 2013, pp. 68–69).
It is sometimes considered characteristic of analogical change in morphology that it affects one word at a time rather than applying across the board to all candidates at once (Hock, 1991, p. 241; Kiparsky, 1974, p. 262). There are many morphological changes that do apply across the board, however. This is generally true of changes in the structure of paradigms or subparadigms, as opposed to shifts of lexical items from one class to another (Fertig, 2013, p. 56). In late Middle High German, for example, the 3pl present indicative verbal ending ‑ent was replaced by ‑en, which matched the corresponding ending in the subjunctive and the past tense (Paul, 1886, p. 95). This is traditionally treated as a proportional change, usually with the (identical) 1pl and 3pl subjunctive or past-tense forms ending in -en serving as the A and B terms of the equation, and the 1pl present indicative, also in -en, as the C form. There is no evidence that this change, or many others like it, applied one word at a time.
3.2 Back-Formation and Paradigm Leveling
There are two well-known types of analogical change that are often treated as distinct from proportional analogy, even though they can in most instances be modeled with proportional equations. In back-formation, one might say that a productive rule is applied in reverse: a previously non-existent input to a rule is inferred from what appears to be an output (cf. Becker, 1994). Identifying an innovation as a back-formation requires that there be a clear-cut asymmetrical basic→derived relationship between two forms, which is more common in derivational than in inflectional morphology. An often-cited case in English is the verb edit, created by back-formation from editor, on the model of many existing verb–noun pairs: confess–confessor; create–creator, etc. Examples of back-formation often entail a covert reanalysis that ascribes an innovative morphological structure to a word. Several English singular nouns were created through back-formation from forms ending in a /z/ that was reanalyzed as a plural ending, as in pea, created by back-formation from pease, originally a singular mass noun. A derivational example is lase from laser, originally an acronym in which the ‑er had no morphological status. As the edit example shows, however, back-formation does not necessarily entail such reanalysis—the ‑or in editor had presumably always been analyzed by speakers as an agentive suffix (cf. the related noun edition); the underlying verb edit just happened not to exist.
Another type of development that is arguably proportional but is often treated as distinct from proportional analogy is paradigm leveling, which is usually defined as the reduction or elimination of stem alternations within paradigms (without any other concomitant changes, such as the appearance of the ‑(e)d suffix in glided). The example in (4) involves an alternation between (originally) diphthongal and monophthongal vowels—orthographically <ai> and <a>, respectively—that arose in a number of verbs in Old French and was subsequently leveled in favor of the reflexes of the diphthongs.
Most cases of paradigm leveling can be modeled with proportional equations because there are usually existing non-alternating paradigms that correspond in their affixal inflectional patterns to the items undergoing leveling. A simple example is English old–elder–eldest → old–older–oldest. Any regular adjective could serve as a proportional model for this change, as in (5):
Cases of partial leveling, where some aspects of a stem alternation are eliminated while others are retained are sometimes claimed to be more problematic for proportional accounts since the model paradigm cannot be non-alternating but must instead match the innovative alternation of the partially leveled forms (Kiparsky, 1992, p. 58). Here too, however, the necessary models exist for many attested cases of partial leveling. An example is the leveling in English of root-final consonant alternations that had resulted from the operation of Verner’s Law in the paradigms of Germanic strong verbs. The s–r alternation that survives in was–were is an instance of this; it has been leveled in all other relevant verbs since Old English times. The retention of ablaut alternations in the root vowels makes this a partial leveling, as shown in (6).
There are some cases of leveling that do appear to defy a proportional account, however. Some of these may turn out to involve poorly understood regular sound changes, but this is not a possibility in all cases. Latin iecur (nom/acc sg)–iecinoris (gen sg) ‘liver’, for example, developed from earlier *yekor–*yekwinis, through the leveling of the -Vr of the nom/acc to all other forms in the paradigm. The retention of -in- in the oblique forms results in a unique paradigm for which no other item could have served as a proportional model (Anttila, 1989, p. 95; for further examples see Fertig, 2016, pp. 437–442).
Many linguists believe leveling to be fundamentally different from proportional analogy in that it is motivated not (merely) by a tendency to extend the dominant morphophonological patterns of a particular language but rather by a universal preference for invariant stems (or invariant morphemes) (Burzio, 2005; Curtius, 1860, p. 331; McCarthy, 2005; Steriade, 2000). For arguments against this position, see Fertig (2016); Garrett (2008); Hill (2007); and Paul (1886, p. 161) Paradigm leveling is undeniably an extremely common diachronic phenomenon in many languages (Mańczak, 1958), but a number of scholars have offered explanations for the prevalance of paradigm leveling that have nothing to do with any universal preference for uniformity (Albright, 2005, pp. 39–41; Bybee & Newman, 1995; Fertig, 2016, p. 451; Wurzel, 1984, pp. 169–172).
3.3 Contamination and Folk Etymology
Several influential scholars of the 19th and early 20th centuries restricted the label “analogy” to the (arguably) proportional phenomena discussed in sections 3.1–3.2 (Bloomfield, 1933; Kruszewski, 1884–1890; Paul, 1886; Saussure, 1995). This was consistent with their understanding of analogy as the fundamental principle underlying the normal, productive operation of the mental grammar (Blevins, 2016, pp. 97–116). For these scholars, studying analogical change meant studying the role that the acquisition, representation, and use of the mental grammar play in the historical development of languages. Since the second half of the 20th century, it has become more common to include other phenomena such as contamination and folk etymology under the analogy heading, sometimes grouping them together as “non-proportional analogy” (Anttila, 1989; Kiparsky, 1992, 2012). Definitions of contamination and folk etymology vary from scholar to scholar and are often vague. Both phenomena are generally understood to involve some kind of influence of one word or expression on the phonological shape of another and thus to constitute modification of an existing form, as opposed to the replacement entailed by proportional analogy. Typically, the expressions involved are not morphologically related but are connected by phonetic or semantic similarity or by frequent proximal co-occurrence in speech.
Many cases of contamination involve words belonging to a series, as when the Latin word for ‘nine’ novem acquired a final -m in place of earlier ‑n under the influence of decem ‘ten’. Similar cases involving adjacent numerals are legion (Osthoff, 1878). Contamination is often invoked to account for influence among forms within paradigms as well, however, especially in cases of suppletion. The paradigm of the verb ‘to be’, for example, includes forms from two or more originally distinct lexical items in many Indo-European languages. In prehistoric and medieval German, there are a couple of instances where an initial b- (compare English be, been) has spread to forms that originally did not begin with b-. Old High German 1pl present indicative birum and 2p birut must have developed from earlier *irum, *irut, for example, and the innovative Middle High German imperative sg form bis apparently developed from the older form wis (from the root wes-, reflected in English was, were) (Paul, 1920, p. 161). Here, the line between contamination and (partial) paradigm leveling is not always clear (cf. Gaeta, 2007), and some scholars have rejected the distinction between contamination and leveling entirely (Wundt, 1900, pp. 451–454; cf. Fertig, 2015b, p. 217, 2016, p. 443; Paul, 1920, p. 160).
Folk etymology often involves a perception or assumption of morphological relatedness between items that are etymologically unrelated, as when English bridegroom acquired an r in its second component (cf. Old English brȳdguma) under the influence of groom. Folk etymology often affects compounds that have become opaque, often because one or both of their components has fallen out of use. This is the case with bridegroom; guma ‘man’ was quite common in Old English but subsequently disappeared from the language. Folk etymology often creates morphological structure where previously there had been none, especially in words that based on their length and prosodic structure sound like compounds. This is particularly common in loanwords, as in German Hängematte ‘hammock’ (literally hang-mat) < hamaca (borrowed by Christopher Columbus from Carib into Spanish). Affixes can also be affected by folk etymology, as in the case of the English superlative suffix ‑most (foremost, utmost, outermost, etc.), which owes its ‑o‑ vowel to folk-etymological association with the unrelated word most. Historically, the suffix ‑mest in Old English combines two inherited superlative suffixes: ‑m‑ and the familiar ‑est. Examples from Booij (2018) of Dutch agent nouns that have been altered to conform to the productive modern ‑er pattern (see (8)) can also be regarded as affixal folk etymology.
In practice, it is not always easy to determine whether an innovation constitutes contamination or folk etymology, or perhaps a combination of both. The development of the English word female, for example, from the earlier French borrowing femelle under the influence of the etymologically unrelated word male is sometimes cited as an example of folk etymology (e.g., Rundblad & Kronenfeld, 2003, p. 126) and sometimes of contamination (e.g., Hock, 1991, p. 197). Some scholars define folk etymology as crucially involving phonetic similarity between the affecting and affected words, whereas contamination requires a semantic connection, but many attested changes involve both semantics and phonetics. If we adopt the idea by Oertel (1901) that these phenomena both involve some kind of extra-grammatical “interference” among the mental representations of items, we can distinguish in principle between contamination and folk etymology as involving interference in production and perception respectively (cf. Fertig, 2016, pp. 442–446).
Folk etymology and contamination are problematic for Neogrammarian theory because they disrupt the tidy complementarity between regular sound change—which involves the modification of existing forms and is supposedly purely phonetically motivated and blind to grammatical patterns—and analogy (narrowly defined as including only the phenomena discussed in sections 3.1–3.2)—which involves the replacement rather than the modification of existing forms and is supposedly licensed entirely by grammatical patterns and blind to phonetic factors. Consequently, many historical linguists have marginalized the perception and production-based mechanisms behind contamination and folk etymology, believing them to be implicated only in changes affecting individual words and rarely considering the possibility that they might play an important role in more systematic processes, perhaps even in many changes that are usually regarded as strictly proportional (Fertig, 2015b, pp. 221–225). This marginalization has meant that relatively little progress has been made in understanding the mechanisms behind these types of change (cf. Coates, 1987; Maiden, 2008) since Paul (1886, p. 183) attributed (unintentional) folk etymology to mishearing and other early scholars noted the obvious connection between contamination and speech errors (e.g., Hermann, 1931, p. 16).
4. Constraining Analogical Change
The largely haphazard way in which historical linguists invoke analogy to account for attested changes that are at odds with more general patterns of sound change and the lack of any clear principles constraining such appeals to analogy have been major concerns throughout the history of modern historical linguistics, dating back at least to Scherer (1868). Only a few hard constraints—intended to rule out some imaginable analogical changes as impossible—have ever been proposed. Besides Paul’s proportional condition, discussed in section 3.1, the most important of these is probably the proposal by Kiparsky (1968, 2000, 2012) that every analogical change must entail a simplification or optimization of the grammar. The usefulness of this idea has been somewhat limited in practice due to the difficulty and theory-dependence of determining what counts as simplification/optimization.
Proposals for soft constraints that address the relative likelihood of changes are much more common. These constraints generally apply to the directionality of analogical change, whereby it is necessary to distinguish two main dimensions of directionality. On the interparadigmatic dimension, proposed constraints address the question of which derivational or inflectional patterns are most likely to spread at the expense of others. For example, given the existence of both ‑ed suffixation and ablaut to mark the present–past distinction in English verbs, one can ask which of the two should be expected to have more analogical success among verbs that would be potential candidates for either pattern. There are attested changes in both directions: glide–glode→glide–glided vs. dive–dived→dive–dove. Historical linguists have suggested a number of factors that might play a role in making some directions of change more likely than others, with relative type frequency being the most commonly cited factor (Bybee, 1995; Paul, 1877, p. 329; Wurzel, 1984). Thus, for example, inflectional classes to which large numbers of lexical items belong, such as the class of English verbs that forms its past tense and participle with the ‑ed suffix, are considered more likely to gain new members that formerly belonged to smaller classes than they are to lose members to those smaller classes. High token frequency, by contrast, is widely cited as a factor that generally promotes resistance to analogical change (Beckner & Wedel, 2009; Bybee, 1988; Wheeler, 1887, p. 39), but it is less clear what effect token frequency has on the likelihood that a pattern will expand at the expense of other patterns (Bybee & Moder, 1983). Section 5 discusses the role that phonological similarity plays in the relative likelihood of different analogical developments.
One important general question about constraints is the extent to which they reflect system-independent versus system-dependent preferences (Wurzel, 1984). The idea mentioned in section 3.2 that paradigm leveling is motivated at least partly by a universal preference for non-alternating stems is a prominent example of a proposed system-independent preference. By contrast, the purported tendency for larger inflectional classes to expand at the expense of smaller classes is clearly a system-dependent matter.
Turning to the intraparadigmatic dimension of directionality, the question here is which form within an inflectional paradigm—or among a set of derivationally related lexemes—will serve as the unchanging “anchor” around which other forms undergo analogical change (the C form rather than the unknown X in a proportional equation, see section 3.1). This question is most often applied to paradigm leveling. Can we predict, for example, whether knife–knives would be more likely to level to knife–knifes or to knive–knives? Many linguists have maintained that analogical innovations (tend to) follow a basic → derived direction (Andersen 1980; Kuryłowicz, 1966), but the usefulness of this claim depends on having objective criteria for determining which form is basic. This may be relatively clear in many cases of derivational morphology, where the basic form is the one that does not have a derivational affix, but it can be a much less straightforward matter in inflectional paradigms. Inflectional basicness is often defined in morphosyntactic terms; for example, singular is (universally?) the morphosyntactically unmarked, and thus basic, number (Andersen, 1980; Lahiri, 2000). Others argue for a form-based notion of inflectional basicness: the basic form is the one from which the morphophonological make-up of other forms in the paradigm can best be predicted—and it is up to the learner to figure out which form this is (Albright, 2010; Albright & Hayes, 2003; Vennemann, 1972). Finally, some argue that what really matters is relative token frequency: the anchors in analogical change are simply the forms that language users/learners hear most often (Bybee, 2007; Wheeler, 1887, p. 21).
5. Alternative Conceptions of Analogical Change
Section 3 describes what might be regarded as the mainstream approach to the role of analogy in morphological development, following a tradition that leads from Paul (1886) through Saussure (1995), Bloomfield (1933), Kuryłowicz (1966), Wurzel (1984), and many others, to scholars such as Kiparsky (1965, etc.) and Hock (1991). Within this tradition, there are important differences on many key questions. Paul and Jespersen (1922), for example, see language use as the main driver of analogical change, with the mental representations of new forms being gradually strengthened while those of old forms are gradually weakened as the former are heard, uttered, and brought into consciousness more often than the latter. Kiparsky and many others, especially since the 1960s (e.g., Andersen, 1980; King, 1969) focus instead on the role of transmission to new learners, seeing analogical change as largely attributable to imperfect learning (cf. Beckner & Wedel, 2009).
Nevertheless, scholars in the mainstream camp generally agree on certain fundamental points: (i) that the core types of analogical change are essentially by-products of the normal functioning (acquisition, representation, use) of the mental grammar; (ii) that these same core types involve replacement rather than modification of existing forms (words or morphemes); (iii) that the fundamental opposition between phonetically motivated sound change and structurally motivated analogy is valid and important, even if not completely airtight; and (iv) that the role of analogy in morphological innovation corresponds to the conception of it in cognitive science, according to which analogy involves purely relational parallels, so that “literal similarity” (Gentner & Smith, 2013, p. 672) or “object commonality” (Ambridge, Theakston, Lieven, & Tomasello, 2006, p. 175) are factored out (cf. Stroppa & Yvon, 2005).
This fourth point means that the only thing that matters for the potential of one set of items to affect another analogically is the part of the word that differs across a set of paradigmatically related forms. In the proportional equation in (1) this would be the ‑(e)d suffix that is present in treated and glided but not in treat and glide. Paul refers to this part of the word—which can consist of affixes, alternating segments, zero, or even a pattern with no stable phonological substance, such as reduplication (cf. Wurzel, 1989)—as the “formal element” and the rest of the word as the “material element”, and he maintains that when mature speakers use their mental grammars to generate forms “only the formal element [. . .] enters consciousness while the various material elements mutually inhibit each other” (Paul, 1886, p. 91, translation from Fertig, 2015a). Paul thus considers the degree of phonological similarity between model items and target items in the part of the word that does not differ across the paradigm, what Albright and Hayes (2002) call the “shared context”—to be irrelevant to the essence of analogy. For the change that has occurred in (especially American) English dive–dived→dive–dove, for example, all verbs that already had the same ablaut alternation would be considered equally valid analogical models. The phonological similarity of drive–drove to dive–dove would not make it more relevant as a model than, say, write–wrote, and conversely any verb with /aɪ/ as root vowel in the present tense, such as whine, should, in principle, be a candidate for the change that has occurred in dive, regardless of its degree of similarity to verbs that already follow the relevant ablaut pattern.
There are alternative traditions of thinking about analogical change that deviate from mainstream views on all four of the points enumerated above. In some of these traditions, one finds a conception of core analogy that resembles the mainstream notion of contamination, treating it as a matter of forms influencing related forms, rather than of existing forms being replaced by analogical creations (Curtius, 1860; Joseph, 1998). The innovative forms are thus considered to be altered continuations of—rather than independent of—the older forms (Hermann, 1931; Wundt, 1900; cf. Fertig, 2015b).
Some proponents of alternative conceptions of analogy—most famously Schuchardt (1972)—also question the strict Neogrammarian dichotomy between sound change and analogy, along with the mainstream view of sound change as constantly disrupting grammatical systems—because it is motivated entirely by forces that lie outside those systems—and analogy constantly reacting to these disruptions and restoring some degree of “harmony” (Brugmann, 1876, pp. 317–318 n.33; Paul, 1877, p. 328).
Considering the usual meaning of analogy in cognitive science, it is perhaps surprising that many linguists have come to use the term to refer specifically to the role that phonological (and sometimes semantic) similarity plays in the influence that forms exert on each other (Bybee & Beckner, 2014, pp. 506–507; Chapman & Skousen, 2005; Hermann, 1931), thereby turning away from Paul’s focus on abstract relational parallels. In areas of linguistics outside morphology, one still finds views very close to Paul’s: “When an analogy is made, the objects involved are effaced; the only identity they retain is their role in the relational structure" (Tomasello, 2003, p. 164). In late 20th and early 21st century work on morphological productivity and change, however, one often finds “analogy” used in a way that highlights the role of object commonality: “To evaluate the [. . .] hypothesis that all morphology is analogical, we implemented a purely analogical model, which evaluates novel [English past-tense forms] based solely on their similarity to existing verbs” (Albright & Hayes, 2003, p. 119).
Paul and many of his followers were clearly wrong to have ignored the relevance of phonological similarity among words to morphological productivity (cf. Blevins, Milin, & Ramscar, 2017), but analogy was a good name for Paul’s imperfect theory. It is less clear that it is the best name for approaches that focus on the role of literal similarity.
Ackerman, F., & Malouf, R. (2017). Implicative relations in word-based morphological systems. In A. Hippisley & G. Stump (Eds.), The Cambridge handbook of morphology (pp. 297–328). New York: Cambridge University Press.Find this resource:
Albright, A. (2005). The morphological basis of paradigm leveling. In L. J. Downing, T. A. Hall, & R. Raffelsiefen (Eds.), Paradigms in phonological theory (pp. 17–43). Oxford: Oxford University Press.Find this resource:
Albright, A. (2010). Base-driven leveling in Yiddish verb paradigms. Natural Language and Linguistic Theory, 28, 475–537. doi:10.1007/s11049-010-9107-zFind this resource:
Albright, A., & Fuß, E. (2012). Syncretism. In J. Trommer (Ed.), The morphology and phonology of exponence (pp. 236–288). Oxford: Oxford University Press.Find this resource:
Albright, A., & Hayes, B. (2002). Modeling English past tense intuitions with minimal generalization. Morphological and Phonological Learning: Proceedings of the 6th Workshop of the ACL Special Interest Group in Computational Phonology (SIGPHON) (pp. 58–69). Philadelphia: Association for Computational Linguistics.Find this resource:
Albright, A., & Hayes, B. (2003). Rules vs. analogy in English past tenses: A computational/experimental study. Cognition, 90, 119–161. doi:10.1016/S0010-0277(03)00146-XFind this resource:
Ambridge, B., Theakston, A. L., Lieven, E. V. M., & Tomasello, M. (2006). The distributed learning effect for children’s acquisition of an abstract syntactic construction. Cognitive Development, 21, 174–193.Find this resource:
Andersen, H. (1980). Morphological change: Towards a typology. In J. Fisiak (Ed.), Historical morphology (pp. 1–50). New York: Mouton de Gruyter.Find this resource:
Andersen, H. (2001). Actualization and the (uni)directionality of change. In H. Andersen (Ed.), Actualization linguistic change in progress (pp. 225–248). Philadelphia: John Benjamins.Find this resource:
Anderson, S. R. (1992). A-morphous morphology. Cambridge, U.K.: Cambridge University Press.Find this resource:
Anttila, R. (1977), Analogy. New York: Mouton de Gruyter.Find this resource:
Anttila, R. (1989). Historical and comparative linguistics (2nd ed.). Amsterdam: John Benjamins.Find this resource:
Becker, T. (1990). Analogie und morphologische Theorie. Munich: Fink.Find this resource:
Becker, T. (1994). Back-formation, cross-formation, and “bracketing paradoxes” in paradigmatic morphology. In G. Booij & J. van Marle (Eds.), Yearbook of morphology 1993 (pp. 1–25). Boston: Kluwer.Find this resource:
Beckner, C., & Wedel, A. (2009). The roles of acquisition and usage in morphological change. Annual Meeting of the Berkeley Linguistics Society, 35(1). doi:10.3765/bls.v35i1.3593Find this resource:
Blevins, J. (2016). Word and paradigm morphology. Oxford: Oxford University Press.Find this resource:
Blevins, J. P., & Blevins, J. (2009). Introduction: Analogy in grammar. In J. P. Blevins & J. Blevins (Eds.), Analogy in grammar: form and acquisition (pp. 1–12). Oxford: Oxford University Press.Find this resource:
Blevins, J. P., Milin, P., & Ramscar, M. 2017. The Zipfian paradigm cell filling problem. In F. Kiefer, J. P. Blevins, & H. Bartos (Eds.), Perspectives on morphological organization (pp.141–158). Leiden, The Netherlands: Brill.Find this resource:
Bloomfield, L. (1933). Language. New York: Holt.Find this resource:
Booij, G. (2018). Partial motivation, multiple motivation: The role of output schemas in morphology. In G. Booij (Ed.), The construction of words advances in construction morphology. Dordrecht, The Netherlands: Springer.Find this resource:
Booij, G., & Masini, F. (2015). The role of second order schemas in the construction of complex words. In L. Bauer, L. Körtvélyessy, & P. Štekauer (Eds.), Semantics of complex words (pp. 47–66). Dordrecht, The Netherlands: Springer.Find this resource:
Brugmann, K. (1876). Nasalis Sonans in der indogermanischen Grundsprache. In G. Curtius & K. Brugmann (Eds.), Studien zur griechischen und lateinischen Grammatik (Vol. 9, pp. 285–338). Leipzig: Hirzel.Find this resource:
Burzio, L. (2005). Sources of paradigm uniformity. In L. J. Downing, T. A. Hall, & R. Raffelsiefen (Eds.), Paradigms in phonological theory (pp. 65–106). Oxford: Oxford University Press.Find this resource:
Bybee, J. (1980). Morphophonemic change from inside and outside the paradigm. Lingua, 50, 45–59. doi:10.1016/0024-3841(80)90079-0Find this resource:
Bybee, J. (1988), Morphology as lexical organization. In M. Hammond & M. Noonan (Eds.), Theoretical morphology (pp. 119–141). New York: Academic Press.Find this resource:
Bybee, J. (1995). Regular morphology and the lexicon. Language and Cognitive Processes, 10, 425–455. doi:10.1080/01690969508407111Find this resource:
Bybee, J. (2007). Frequency of use and the organization of language. Oxford: Oxford University Press.Find this resource:
Bybee, J. L., & Beckner, C. (2014). Language use, cognitive processes and linguistic change. In C. Bowern & B. Evans (Eds.), The Routledge handbook of historical linguistics (pp. 503–518). Abingdon, U.K.: Routledge.Find this resource:
Bybee, J. L., & Moder, C. L. (1983) Morphological Classes as Natural Categories. Language, 59, 251–270. doi:10.2307/413574Find this resource:
Bybee, J. L., & Newman, J. E. (1995). Are stem changes as natural as affixes?. Linguistics, 33, 633–654. doi:10.1515/ling.19184.108.40.2063Find this resource:
Bybee, J. L., & Slobin, D. I. (1982). Rules and schemas in the development and use of the English past tense. Language, 58, 265–289. doi:10.2307/414099Find this resource:
Chapman, D., & Skousen, R. (2005). Analogical modeling and morphological change: the case of the adjectival negative prefix in English. English Language & Linguistics, 9(2), 333–357. doi:10.1017/S136067430500167XFind this resource:
Clahsen, H. (1999). Lexical entries and rules of language: A multidisciplinary study of German inflection. Behavioral and Brain Sciences, 22(6), 991–1060. doi:10.1017/S0140525X99002228Find this resource:
Coates, R. (1987). Pragmatic sources of analogical reformation. Journal of Linguistics, 23, 319–340. doi:10.1017/S0022226700011294Find this resource:
Cowgill, W. (1959). The inflection of Germanic ō-presents. Language, 35(1), 1–15. doi:10.2307/411329Find this resource:
Curtius, G. (1860). Das dreisilbengesetz der griechischen und lateinischen betonung. Zeitschrift für vergleichende Sprachforschung, 9, 321–338.Find this resource:
Enger, H-O. (2013). Inflectional change, ‘sound laws’ and the autonomy of morphology: The case of Scandinavian case and gender reduction. Diachronica, 30, 1–26.Find this resource:
Fertig, D. (1999). Analogical “leveling” from outside the paradigm: stem-vowel changes in the German modals. Diachronica, 16(2), 233–260.Find this resource:
Fertig, D. (2000). Morphological change up close. Tübingen, Germany: Niemeyer.Find this resource:
Fertig, D. (2013). Analogy and morphological change. Edinburgh: Edinburgh University Press.Find this resource:
Fertig, D. (2015a). Analogy [translation of Paul (1920), Ch. 5]. In P. Auer & R. W. Murray (Eds.), Hermann Paul’s “Principles of language history” revisited (pp. 83–98). Berlin: Mouton de Gruyter.Find this resource:
Fertig, D. (2015b). Two conceptions of analogical innovation/change. In P. Auer & R. W. Murray (Eds.), Hermann Paul’s “Principles of language history” revisited (pp. 209–236). Berlin: Mouton de Gruyter.Find this resource:
Fertig, D. (2015c). Analogy and morphophonological change. In P. Honeybone & J. Salmons (Eds.), The Oxford handbook of historical phonology (pp. 205–218). Oxford: Oxford University Press.Find this resource:
Fertig, D. (2016). Mechanisms of paradigm leveling and the role of universal preferences in morphophonological change. Diachronica, 33(4), 423–460. doi:10.1075/dia.33.4.01ferFind this resource:
Fertig, D. (2019). Morphological change through phonological analogy: 2nd person singular -s→-st and related developments in Germanic. Journal of Germanic Linguistics, 31(1), 1–42. doi:10.1017/S147054271800003XFind this resource:
Gaeta, L. (2007). Is analogy economic? In F. Montermini, G. Boyé, & N. Hathout (Eds.), Selected proceedings of the 5th Décembrettes: Morphology in Toulouse (pp. 20–33). Somerville, MA: Cascadilla Proceedings Project.Find this resource:
Garrett, A. (2008). Paradigmatic uniformity and markedness. In J. Good (Ed.), Linguistic universals and language change (pp. 125–143). Oxford: Oxford University Press.Find this resource:
Gentner, D., & Smith, L. A. (2013). Analogical learning and reasoning. In D. Reisberg (Ed.), The Oxford handbook of cognitive psychology (pp. 668–681). Oxford: Oxford University Press.Find this resource:
Goldsmith, J. (2006). An algorithm for the supervised learning of morphology. Natural Language Engineering, 12, 353–371.Find this resource:
Hare, M., & Elman, J. L. (1995). Learning and morphological change. Cognition, 56, 61–98. doi:10.1016/0010-0277(94)00655-5Find this resource:
Harris, A. C., & Campbell, L. (1995). Historical syntax in cross-linguistic perspective. Cambridge, U.K.: Cambridge University Press. doi:10.1017/CBO9780511620553Find this resource:
Hermann, E. (1931). Lautgesetz und Analogie (Abhandlungen der Gesellschaft der Wissenschaften zu Göttingen, Philologisch-Historische Klasse, neue Folge 23, 3). Berlin: Weidmannsche Buchhandlung.Find this resource:
Hill, E. (2007). Proportionale Analogie, paradigmatischer Ausgleich und Formerweiterung: ein Beitrag zur Typologie des morphologischen Wandels. Diachronica, 24, 81–118.Find this resource:
Hock, H. H. (1991). Principles of historical linguistics (2nd ed.). Berlin: Mouton de Gruyter.Find this resource:
Hock, H. H. (2003). Analogical change. In B. D. Joseph & R. D. Janda (Eds.), Handbook of historical linguistics (pp. 441–460). Oxford: Blackwell.Find this resource:
Hock, H. H., & Joseph, B. D. (2009). Language history, language change, and language relationship (2nd ed.). Berlin: Mouton de Gruyter.Find this resource:
Hopper, P. J., & Traugott, E. C. (2003). Grammaticalization (2nd ed.). Cambridge, U.K.: Cambridge University Press.Find this resource:
Hyman, L. M. (2003). Sound change, misanalysis, and analogy in the Bantu causative. Journal of African Languages and Linguistics, 24, 55–90. doi:10.1515/jall.2003.004Find this resource:
Jespersen, O. (1922). Language: Its nature, development and origin. London: Allen & Unwin.Find this resource:
Joseph, B. D. (1998). Diachronic morphology. In A. Spencer & A. M. Zwicky (Eds.), The handbook of morphology (pp. 351–373). Malden, MA: Blackwell.Find this resource:
Kapatsinski, V. (2013). Conspiring to mean: Experimental and computational evidence for a usage-based harmonic approach to morphophonology. Language, 89, 110–148. doi:10.1353/lan.2013.0003Find this resource:
King, R. D. (1969). Historical linguistics and generative grammar. Englewood Cliffs, NJ: Prentice-Hall.Find this resource:
Kiparsky, P. (1965). Phonological change (Unpublished doctoral dissertation). MIT, Cambridge, MA.Find this resource:
Kiparsky, P. (1968), Linguistic universals and linguistic change. In E. Bach & R. T. Harms (Eds.), Universals in linguistic theory (pp. 170–202). New York: Holt, Rinehart and Winston.Find this resource:
Kiparsky, P. (1973). “Elsewhere” in phonology. In S. Anderson & P. Kiparsky (Eds.), A Festschrift for Morris Halle (pp. 93–106). New York: Holt, Rinehart and Winston.Find this resource:
Kiparsky, P. (1974). Remarks on analogical change. In J. M. Anderson & C. Jones (Eds.), Historical linguistics II (pp. 257–75). New York: American Elsevier.Find this resource:
Kiparsky, P. (1992), Analogy. In W. Bright (Ed.), International encyclopedia of linguistics (Vol. 1, pp. 56–61). Oxford: Oxford University Press.Find this resource:
Kiparsky, P. (2000). Analogy as optimization: “exceptions” to Sievers’ Law in Gothic. In A. Lahiri (Ed.), Analogy, levelling, markedness (pp. 15–46). New York, NY: Mouton de Gruyter.Find this resource:
Kiparsky, P. (2012). Grammaticalization as optimization. In D. Jonas, J. Whitman, & A. Garrett (Eds.), Grammatical change origins, nature, outcomes (pp. 15–51). Oxford: Oxford University Press.Find this resource:
Köpcke, K-M. (1998). The acquisition of plural marking in English and German revisited: Schemata versus rules. Journal of Child Language, 25, 293–319.Find this resource:
Kroesch, S. (1926). Analogy as a factor in semantic change. Language, 2, 35–45. doi:10.2307/408785Find this resource:
Kruszewski, M. (1884–1890). Prinzipien der Sprachentwickelung. Internationale Zeitschrift für allgemeine Sprachwissenschaft, 1(1884), 295–307; 2(1885), 258–268; 3(1887), 145–187; 5(1890), 133–144, 339–360.Find this resource:
Kuryłowicz, J. (1966). La Nature des procès dits “analogiques.” In E. P. Hamp, F. W. Householder, & R. Austerlitz (Eds.), Readings in linguistics II (pp. 158–174). Chicago: Chicago University Press.Find this resource:
Lahiri, A. (2000). Introduction. In A. Lahiri (Ed.), Analogy, levelling, markedness (pp. 1–14). New York: Mouton de Gruyter.Find this resource:
Maiden, M. (2008). Lexical nonsense and morphological sense: On the real importance of ‘folk etymology’ and related phenomena for historical linguists. In Þ. Eyþórsson (Ed.), Grammatical change and linguistic theory (pp. 307–328). Amsterdam: John Benjamins.Find this resource:
Maiden, M. (2018). The Romance verb morphomic structure and diachrony. Oxford: Oxford University Press.Find this resource:
Mańczak, W. (1958). Tendences générales des changements analogiques. Lingua, 7, 298–325, 387–420. doi:10.1016/0024-3841(57)90101-8Find this resource:
McCarthy, J. J. (2005). Optimal paradigms. In L. J. Downing, T. A. Hall, & R. Raffelsiefen (Eds.), Paradigms in phonological theory (pp. 170–210). Oxford: Oxford University Press.Find this resource:
Oertel, H. (1901). Lectures on the study of language. New York: Scribner’s.Find this resource:
Osthoff, H. (1878). Formassociation bei Zahlwörtern. In H. Osthoff & K. Brugmann, Morphologische Untersuchungen auf dem Gebiete der Indogermanischen Sprachen, vol. 1 (pp. 92–132). Leipzig: Hirzel.Find this resource:
Paul, H. (1877). Die Vocale der Flexions- und Ableitungs-Silben in den aeltesten germanischen Dialecten. Beiträge zur Geschichte der deutschen Sprache und Literatur, 4, 315–475. doi:10.1515/bgsl.1877.1877.4.315Find this resource:
Paul, H. (1886). Principien der Sprachgeschichte (2nd ed.). Halle: Niemeyer.Find this resource:
Paul, H. (1920). Prinzipien der Sprachgeschichte (5th ed.). Halle: Niemeyer.Find this resource:
Pinker, S., & Prince, A. (1994). Regular and irregular morphology and the psychological status of rules of grammar. In S. D. Lima, R. L. Corrigan, & G. K. Iverson (Eds.), The reality of linguistic rules (pp. 321–351). Philadelphia: John Benjamins.Find this resource:
Ringe, D., & Taylor, A. (2014). The development of Old English: A linguistic history of English, vol. 2. Oxford: Oxford University Press.Find this resource:
Rundblad, G., & Kronenfeld, D. B. (2003). The inevitability of folk etymology: A case of collective reality and invisible hands. Journal of Pragmatics, 35, 119–138. doi:10.1515/9783110940756.105Find this resource:
Saussure, F. de (1995). Cours de linguistique générale (C. Bally & A. Sechehaye, Eds.). Paris: Payot.Find this resource:
Scherer, W. (1868). Zur Geschichte der deutschen Sprache. Berlin: Duncker.Find this resource:
Schuchardt, H. (1972). Über die Lautgesetze: Gegen die Junggrammatiker. Berlin: Oppenheim.Find this resource:
Steriade, D. (2000). Paradigm uniformity and the phonetics–phonology boundary. In M. Broe & J. Pierrehumbert (Eds.), Papers in laboratory phonology 5 (pp. 313–334). Cambridge, U.K.: Cambridge University Press.Find this resource:
Stroppa, N., & Yvon, F. (2005). An analogical learner for morphological analysis. Proceedings of the Ninth Conference on Computational Natural Language Learning (pp. 120–127). Stroudsburg, PA: Association for Computational Linguistics.Find this resource:
Stump, G. T. (2016). Inflectional paradigms: Content and form at the syntax-morphology interface. Cambridge, U.K.: Cambridge University Press.Find this resource:
Timberlake, A. (1977). Reanalysis and actualization in syntactic change. In C. N. Li (Ed.), Mechanisms of syntactic change (pp. 141–180). Austin: University of Texas Press.Find this resource:
Tomasello, M. (2003). Constructing a language: A usage-based theory of language acquisition. Cambridge, MA: Harvard University Press.Find this resource:
Vennemann, T. (1972). Phonetic analogy and conceptual analogy. In T. Vennemann & T. H. Wilbur (Eds.), Schuchardt, the Neogrammarians, and the transformational theory of phonological change (pp. 181–204). Frankfurt: Athenäum.Find this resource:
Vennemann, T., & Wilbur, T. H. (1972). Schuchardt, the Neogrammarians, and the transformational theory of phonological change. Frankfurt: Athenäum.Find this resource:
Wallis, J. (1688). Grammatica linguae Anglicanae (3rd ed.). Hamburg: Gotfried Schultzen.Find this resource:
Wheeler, B. I. (1887). Analogy and the scope of its application in language. Ithaca, NY: John Wilson and Son University Press.Find this resource:
Wiese, R. (2009). The grammar and typology of plural noun inflection in varieties of German. Journal of Comparative Germanic Linguistics, 12(2), 137–173. doi:10.1007/s10828-009-9030-zFind this resource:
Wundt, W. (1900). Völkerpsychologie, vol. 1: Die Sprache (part 1). Leipzig: Wilhelm Engelmann.Find this resource:
Wurzel, W. U. (1984). Flexionsmorphologie und Natürlichkeit. Berlin: Akademie-Verlag.Find this resource:
Wurzel, W. U. (1989). Von der Inadäquatheit einer Affixmorphologie. Linguistische Studien, Reihe A, 194, 277–298.Find this resource: