Universität Duisburg-Essen

Margaret E. Winters Cognitive Grammar and Kurylowicz`s Laws of Analogy Series A: General & Theoretical Papers ISSN 1435-6473 Essen: LAUD 1988 (2nd ed. with divergent page numbering 2012) Paper No. 226 Universität Duisburg-Essen

Margaret E. Winters Southern Illinois University (USA) Cognitive Grammar and Kurylowicz`s Law of Analogy Copyright by the author Reproduced by LAUD 1988 (2nd ed. with divergent page numbering 2012) Linguistic Agency Series A University of Duisburg-Essen General and Theoretical FB Geisteswissenschaften Paper No. 226 Universitätsstr. 12 D- 45117 Essen Order LAUD-papers online: http://www.linse.uni-due.de/linse/laud/index.html Or contact: laud@uni-due.de ii

Margaret E. Winters Cognitive Grammar and Kurylowicz`s Laws of Analogy 1. Introduction 1.1. It is safe to say that Kurylowicz`s "La nature des process dits analogiques" (1966 1 ) has incited more controversial questions than it has settled on the nature and functioning of analogy. 2 Often cited and analyzed in introductions to historical linguistics and elsewhere, it begins with remarks about the relationship between basic and derived forms (158 161), followed by the famous (not to say notorious) six "laws" of analogy (162 174), each discussed and illustrated by a wide range of examples from various Indo-European languages. Of the six "laws", five are grammar-internal, having to do, in very general terms, with the direction of analogical extension, while the sixth is a statement of the relationship between language and society, between, roughly, langue and parole in the creation of new forms. 1.2. Several problems arise from this article, even if one accepts, as I do, the validity of this study of the nature of analogic processes and is not questioning the data used to illustrate the principles. The first problem is the exact nature of Kurylowcz`s use of loi "law", which seems at first to give a rather nineteenth-century feeling to his analysis. It is interesting, however, to note that while Kurylowicz, indeed, does use the word loi, he prefers formule or simply designates each of the six items by a Roman numeral without providing any other designation for them. In his conclusion he specifies that he is proposing directions of possible analogical change, but that the social factor decides if and to what extent analogy takes place. His well-known image of run-off mechanisms for rain-water (sewers, drains, etc) and their use when and only when it rains (174) seems to clarify further the nature of his "laws". 1 All citations will be from the article as reprinted in Hamp et al. 1966. It was originally published 1945 1949 Acta Linguistica 5: 121 138. 2 Geoffrey Nathan has discussed and criticized many versions of this paper, for which I am grateful. At the conference at Lille where this paper was presented orally, I benefited from discussion with Joan Bybee, Paulo de Carvalho, Peter van Reenen and Lene Schøsler. I appreciate their taking the time to give me their reactions and suggestions. They are not responsible, of course, for weaknesses and errors here. 1

A second problem is that of the interpretation of these laws 3 and of the article as a whole. Arlotto (1974) and Hock (1986), among others, devote pages to explicating them, analyzing both what is meant by a "basic" and "derived" form and how these interact according to the directionality principles that constitute the laws. I will return below to a discussion of some of the interpretation. Lastly, once the laws are clear, at least to the satisfaction of any analyst at a given moment, there is the question of the validity of Kurylowicz`s principles. Are they true? All of the time? Some of the time? How do they compare to other attempts to provide a typology for analogy, especially Manczak`s (1958)? Implied in all of this discussion are wider questions about the subject: what is the nature of analogy? Is there truly a division between the purely structural aspects of language and this aspect, always recognized as psychological? In the present paper I will consider in some detail two of Kurylowcz`s laws and discuss some of their implications for the study of language evolution. I shall do so within the framework of Cognitive Grammar, which, because of its stance on the close relationship between language function and human cognition in general, is particular well suited to such a study. The next section of the paper, therefore, is a brief overview of Cognitive Grammar and its applicability to the study of analogy and analogical change. The following section will return specifically to Kurylowcz`s second and fourth laws and consider them in light of Cognitive Grammar. I shall end with some further questions which arise from this discussion. 2. Cognitive Grammar Cognitive Grammar is a theory of language processing and production, based on the notion that in order to understand linguistic function, it is essential to integrate it with general cognitive functioning. It is a meaning-based theory, positing the fundamental nature of semantics (taken in the widest sense to include much of pragmatics too) as underlying all if linguistic function. Involved in processing language (and hence meaning) are the same activities that we use as human beings to interact in a wide number of circumstances with the world: comparison, assignment of saliency, entrenchment of certain routines (Langacker 1987) and categorization of the environment, both mental and physical (Lakoff 1987). 2.1. In ways similar to those in which we categorize colors, shapes, faces, music, we also categorize linguistically significant sounds, morphemes and words, phrases, constructions, sentences. They form what Lakoff (1987: 91ff) calls radial categories, arranged around 3 Although it is reasonably clear by Kurylowicz`s use of the word loi that he did not mean it in the Neogrammarian sense, he did use it and I will therefore continue what has been a convention among commentators to use the word "law" without further explanation. 2

prototypical members of the given semantic category or set. Extended from the prototypical or central member are others which share some features of the prototype, but differ in regard to others. These less prototypical members of the category or set can be understood as belonging to the set via these lines or extension back to the center, but may or may not be directly related to each other. A cup of butter (which often refers to two 4-ounce sticks and has no container involved at all) is only related to a trophy (also called a cup) by a consideration of the ways in which the central notion of "cup" has been extended quite far from the prototype in several different directions not necessarily related at all to each other (see Winters forthcoming for further discussion). It is important to note that no two items in a radial category share all the same features: some are added, eliminated, substituted for in each instance in relationship to the center. Important for categorization is the assignment of saliency to certain features (Langacker 1987: 39 40). Not all features of any given item (linguistic or non-linguistic) are of equal importance, and those which are considered more important are those which will guide the assignment of any item to a category. Not only is the choice of category to which anything is assigned determined by the choice of features used to categorize it, but also the place in the category (central, extended somewhat from the center, peripheral) comes from how the human mind views the various features. Obviously each human being does have the fresh task of assigning saliency anew to every item he or she encounters at every minute. Much of the categorization of the universe we consciously or unconsciously know is conventional, based on history and social norms. It is, therefore, to a large extent language or even dialect specific, on a continuum of degrees of what Langacker (1987: 59 60) calls entrenchment or conventionalization. 2.2. Diachronic change is often, consequently, change in categorization. As I have said elsewhere (Winters 1987), there are two ways in which set membership can change: either an item (again, sound, morpheme, construction, etc.) can change from one set to another (as happens when an indicative trigger becomes the trigger of the subjunctive) or its place within the set in relationship to other items can change (as when Latin passum becomes the unmarked negation particle in modern French [Winters forthcoming]). A refinement to this notion of diachronic must be added here: this kind of movement across categories or within categories is at least part of the time a result of changes in assignment of saliency to a given item or to a given feature of some item. With changes in saliency come changes in how subsequent items are scanned, compared, and assigned to categories. To use a trivial, non-linguistic example, our way of looking at people and grouping them changes if we are concerned with height (and thus assign saliency to that feature) instead of being concerned with eye color. 3

2.3. Analogical change can be characterized more precisely, therefore, as a subset of linguistic change, different in the kind of item being changed, perhaps, but not in the fundamental mechanism of change. When analogical change takes place, as is well known, morphemes and words become more like each other, or more like some basis of comparison, following lines of development which are morphological and not phonological; that is, these changes do not, most of the time, involve the normal course of sound change. What is happening, instead, is that they are being perceived, through one or more features, as being like other, better entrenches items, and undergo modification of some other feature or features to make them even more like the basis of comparison. To use an often-cited example, the German plural form Baüme did not always have the fronted diphthong. Speakers of German, comparing the morpheme of "plural" in this word to the morpheme of "plural" in others assigned over time a more salient position within the category of PLURAL to the feature of fronting on well-entrenched and frequently used forms. That feature was then perceived as part of the nature of pluralization in words where it was not organically present. In this particular case I would say that the form Baüme did not change category when it added the umlaut, but moved closer within the morphological set of PLURAL to what had become the more prototypical form for German nouns. In other cases the analogy does cause a full set change; to use an example I have also used elsewhere, the comparatively recent use of the subjunctive with French après que "after", which is generally recognized as being a consequence of its close (polarized) semantic relationship with avant que "before", constitutes a full category change for après que, a former trigger of the indicative, and places it near the periphery of its new semantic set of subjunctive triggers (Winters 1987: 612). 3. Kurylowicz`s Laws of Analogy 3.1. Kurylowicz`s second law states that: Les actions dites analogiques suivent la direction: forms de foundation forms fondées, dont le rapport découle de leurs sphères d`emploi. As Hock points out (1986: 213), this law should be divided into two section, one on the direction of change (from the fondation or basic form to the fondée or derived form) and one on the meaning of sphère d`emploi or sphere of usage. The direction of change, of course, is simply a restatement of the basic proportion of analogy, from a form which serves as base of comparison to the form which changes to become more like it. Within Cognitive Grammar it can be seen that the base form has the property of being better entrenched than the form which changes, and also has features which are perceived as more salient. The change, therefore, is a change in feature or features 4

to those which share the saliency. The notion of entrenchment is itself a radial category and can involve various ideas frequency. Most obvious for morphology is, probably, type frequency: there are, for example, simply more regular verbs (weak in Germanic, those belonging to the ARE class in Romance) in comparison to which others regularize. But there is also token frequency. The French first person plural verb ending, -ons is generally believed to have spread from the present tense form somes of the highly irregular verb être "to be" in early French. Here, then, a single instance, but one of exceptionally high token frequency, was the cause of an analogical change throughout the rest of the Old French verbal system. 4 Sphere of usage can be reinterpreted as the assignment of category, and place within category, to any form. Based on work considering a variety of languages (see, for example, Manczak 1958, Bybee 1985) a prototypical verb form, for example, is usually cited as present, indicative, active, third person and singular. A noun is singular, masculine in gender systems, and, in case languages, in the nominative. 5 This form of the noun serves as citation form in dictionaries and, often, when the word is used in isolation. It is also the subject of an active verb, a position of salience in a prototypical sentence (see, for example, van Oosten 1986), and, in many languages, the form of direct address. Diachronically it serves as the base for analogical change involving a model outside the paradigm (Latin fourth declension nouns usually merge with the second declension in large part because of the identity of nominative singular us forms), but not, interestingly, for change within a single paradigm. Here the sphere of usage factor has to be balanced against sheer frequency of forms: in Latin nouns whose nominative singular form had one fewer syllable than any other form, when the number of syllables was made equal within the paradigm, the majority of nominative singular (the significantly different form) changed in the direction of the rest of the paradigm. Examples include: 1. Classical Latin Late Latin Nom. sg. Full New. Nom. sg. mor- mort- mort- "death" leo leon- leon- "lion" flo- flor- flor- "flower" aesta- aestat- aestat- "summer" 4 Hock (1986: 215) suggests productivity as a measure of the basicness of a given form, but I believe he has fallen into a circular trap: is the form productive because it is basic or basic because it is productive? 5 These grammatical categories are, of course, based on Indo-European. Both Kurylowicz`s work and mine are within this family, and the analysis in this paper reflects this bias. The work should be extended eventually to a much wider number of diverse languages. 5

There are, naturally, some counterexamples: 2. sagui- sanguine- sangui- "blood" here- hered- here- "heir" In verbs, there is some question about the basicness for analogical change of the third person among the various categories which are cited. It is true, as Hock (1986: 220) points out, that there are reasons based on frequency to perceive that forms as basic. But citation forms of verbs tend to be the bare root (as often in English) or the infinitive (as in the Romance languages) or even more or less random (as in Latin as evidenced by dialogue in plays). In other cases clusters of forms seem to be involved rather than any one single base. Old French, as a result of the Latin stress system, had many verbs which, simply within the present tense system, had a diphthong on the singular forms and the third person plural, and a simple vowel for the first and second plural as well as the infinitive: 3. Infinitive: amer "to love" aim amons aimes amez aime aiment The modern verb has generalized the diphthongized form, which includes as part of the base the third person singular. There are, however, many verb which also became regular, but modeled on the infinitive, and the firt and second plural forms: 4. Infinitive: lever "to raise" lief levons lieves levez lieve lievent The assignment of saliency and therefore the condition of being basic is hierarchical, I believe, with certain features being salient, not automatically in an absolute sense, but if other features are not present. To return to example 1 and 2, nominative singular is the central member of the set of forms of a given paradigm, but only when some other features isn`t valued more highly, in this case the number of different forms having the same syllabic and stress pattern. In other cases as exemplified by 3 and 4, it is difficult to talk of saliency of one person in contrast to all the others, but only of tendancies toward saliency and basicness of sets of persons. 6

3.2. Kurylowcz`s fourth law states that: Quand à la suite d`une transformation morphologique une forme subit la différenciation, la forme nouvelle correspond à sa fonction primaire (de fondation), la forme ancienne est reserve pour la fonction secondaire (fondée). It is a statement of what happens when doublets arise through analogical change: the nonchanged form is used for secondary functions, while the analogically created form takes on the primary meaning of the word. The primary function of the form is again the central or prototypical meaning within a category, and the law can be restated to reflect what occurs when there is a shift in category organization so that a new prototype replaces an older one. The older form does not disappear, but becomes part of the radial set extending out from the new prototype. In this case, the fact that any form is made up of a series of features helps explain how the new prototype arises. In these cases we find two different features coinciding: one of the central meaning of the morpheme and one of the high frequency (and therefore high saliency) morphological marking. To use one of the most often cited examples, brethren becomes specialized to church-related use alongside brothers because the s (regular plural) morpheme reinforces the centrality of the more usual meaning of brother. Kiparsky (1974) cites numerous counterexamples to this law (of the type louses "unpleasant people", Maple Leafs "members of the hockey team of "hat name", badder "tougher"), and argues that Kurylowicz`s statement should be reversed, that in the majority of cases, all other things being equal, the analogically derived form will have a secondary meaning. Hock (2986: 226 227) defends Kurylowicz, on the grounds that the semantic differentiation (of lice and louses, for example) predates the analogical morphological change, and that these examples are therefore irrelevant to the interpretation of the law. The two meanings coexisted within the radial category of meaning, therefore, before morphological differentiation became part of the language. It still leaves us with the question, however, of why in the time-honored examples, regularity of morphological marking coincides with basic meaning, while it is the derived meaning in the examples proposed by Kiparsky which exhibits morphological regularity. This may be a genuine example of polarity in language, with competition between saliency (here in the case of non-prototypicality) and reinforcement of two kinds of prototypicality a suggested above. 7

4. Summary and Conclusions 4.1. What I have attempted to illustrate in this paper is that the theory of Cognitive Grammar can shed some light on the nature of analogical change and on the interpretation of Kurylowicz`s laws of analogical processes. It is necessary to start with the notion of the radial semantic set, as used not only for lexical items, but for morphemes such as case, number and person markers. Analogical change depends on the saliency of some features of the prototypical member of the set, against which other members are compared and then, in many cases, changed. Saliency in itself is not a monolithic construct, but is arranged in a hierarchy of features which are language and time specific and therefore not predictable. 4.2. There is no place in the scope of this paper to test the other four of Kurylowicz`s laws. I believe, however, that the first, third, and fifth are also understandable within the framework outlined above, and that such an analysis will shed further light on the nature of analogy. In addition, there is much work to do on the actual assignment of saliency to given features, linguistic or non-linguistic. The question goes beyond the scope of linguistic per se; it is a matter of cognitive psychology, I should think, and will be answered as we answer questions about how much of the world we are born ready to understand and how much we must learn about. Child cognitive development will not be sufficient on its own, moreover, since adults are capable of learning new ways of seeing the world and, both linguistically and extralinguistically, of making analogical leaps leading to new organization and new insight. 8

References Arlotto, Anthony. 1972. Introduction to Historical Linguistics. Boston: Houghton-Mifflin. Bybee, Joan L. 1985. Morphology: A Study of the Relation between Meaning and Form. Amsterdam & Philadelphia: John Benjamins. Hock, Hans Heinrich. 1986. Principles of Historical Linguistics. Berlin, New York & Amsterdam: Mounton-de Gruyter. Kiparsky, Paul. 1974. "Remarks on Analogical Change." Historical Linguistics, Vol. II ed. By J. M. Anderson & C. Jones, 257 275. Amsterdam & Oxford: North Holland. Kurylowicz, Jerzy. 1945 49. "La nature des procès dits analogiques ". Acta Linguistica 5, 121 138. (rpt. 1966 Readings in Linguistics 2 ed. by E. Hamp et al., 158 174. Chicago: University of Chicago Press). Lakoff, George. 1987. Women, Fire, and Dangerous Things. What Categories Reveal about the Mind. Chicago: University of Chicago Press. Langacker, Ronald W. 1987. Foundations of Cognitive Grammar. Vol. 1 Theoretical Prerequisites. Stanford, CA: Stanford University Press. Manczak, Witold. 1958. "Tendances generals des changements analogiques. " Lingua 7, 298 325 & 387 420. van Oosten, Jeanne. 1986. The Nature of Subjects, Topics and Agents: A Cognitive Explanation. Bloomington: Indiana University Linguistics Club. Winters, Margaret E. 1987. "Syntactic and Semantic Space: The Development of the French Subjunctive. " Papers from the VIIth International Conference on Historical Linguistics ed. by A. Ramat et al., 607 618. Amsterdam & Philadelphia: John Benjamins. Winters, Margaret E. Forthcoming. "Innovation in French Negation: A Cognitive Grammar Account. " To appear in Diachronica. 9