Romance languages
Knud Togeby (1918-1974)
Sound law and Tendency. Togeby
Is dissimilation a sound law ? Suffix distribution. Suffix change. Tendency towards non-repetition.
Knud Togeby was one of the most important linguists in the second half of the 20th century and, especially, a leading scholar of the Romance languages. He has treated a large number of the most important issues of Romance philology, normally with implications for general linguistics. One of these subjects was Dissimilation, at that time discussed mainly in a historical perspective, as a kind of language change. Togeby's article gives us occasion for a discussion of some of the principles and of some of the differencies in point of view.
Togeby's article "Qu'est ce que la dissimilation ?", 1964, is a masterpiece. He reviews all important literature about dissimilation until 1964, and comes to the conclusion that almost all alleged cases of dissimilation can largely be explained in another way. T. mentions that there are a few cases of dissimilation that behave like "regular" sound laws, e.g. Latin kw - kw > k - kw (p. 649-50). In the present text it will be attempted to demonstrate that this is not a reasonable conclusion on the available facts.
During his article Togeby comments on a great amount of material from the classic discussion of dissimilation.
At p. 648-9, under the heading "Distribution de suffixes", Togeby declares: "Under the apparently regular dissimilations you will find even a group of cases, not real dissimilation in the right sense, but rather, as it were, a preventive dissimilation. Without that a phonetic change has actually taken place, certain suffixes are distributed in such a way that an accumulation of like consonants is avoided."
Togeby then mentions 5 classic, well-known examples, where two or three suffixes are distributed depending on the phonic shape of the last part of the base word: latin -alis / -aris; -lum / -rum, French -euil / -eul; -erer/-eler; -oter / -onner / -iner. Togeby has no problem of recognizing that in these cases the distribution of the suffixes is made according to the phonic shape of the base word (it is not clear why these five are mentioned and only them, probably they are meant as examples), but they are not examples of dissimilation, according to him, because the concept of "dissimilation" presupposes "sound change": there must be a shift from an (original) form of the word to another one.
At p. 649 he mentions though that at some moment we have got chevreuil, écureuil instead of chevreul, écureul (the circumstances of the last word are somewhat more complicated), but the passage from -euil [œj] to -eul [œl] is blocked in words containing ï, gn, (i)ll (palatal sounds). Togeby repeats: 'Not a real dissimilation, since there has been no change' ("pas dissimilation proprement dite, puisque aucun changement n'a lieu"). "Phonetic change" is something like e.g. Latin on that becomes French õ, while "distribution of suffixes" means something like touss-oter versus chant-onner, where we don't say that -oter has "become" -onner. Togeby is quite right, that these two situations are different, however, in our view, the point is not if a dissimilation (non-repetition) is phonetic or not; with Togeby's own words: 'Certain suffixes are distributed in such a way that an accumulation of like consonants is avoided.' ("certains suffixes sont distribués de façon à éviter la rencontre de consonnes identiques"). That is what we call dissimilation (or non-repetition): the avoidance of repetition.
While the previous paragraphs treated groups with large numbers of words, Togeby later (p. 659-660) deals with suffix change, which he says should be kept separate from suffix distribution, without giving an explanation of the difference. The terminology seems to imply the following distinction: distribution of suffixes occurs "in the beginning" when the words were formed, while suffix change may occur at a later stage of the history of the single word. This division allows on the one hand to recognize the purely phonetic distribution of suffixes which is then declared as not dissimilation, "because no change has taken place". On the other side some processes as Latin femin-inu> femin-ile are transferred from the phonetic area of dissimilation to the area of morphology, under the heading suffix change. That words can change suffix of many different reasons is obvious, but why should phonetic and morphological aspects be unable to work together in such a process ? It seems somewhat barren, what Togeby here does, to look for a word or group of words that have the same ending as the result of the change (-ile), and explain the suffix change by that alone (actually it gives no sense). Togeby says about dissimilation as the cause of linguistic changes, that it is not any cause (p. 651). He refers Grammont, that in his thesis (La dissimilation consonantique dans les langues indo-européennes et dans les langues romanes. Dijon, 1895. p. 128) explains that femin-inu becomes femin-ile (> italian femminile, which has formed a model for maschile, signorile and others) "non par la dissimilation, comme Meyer-Lübke [Grammaire des Langues Romanes, II, § 436], mais par l'intervention du suffixe -ile, cf. seruile, febrile, uirile" (Togeby p. 659). Is this a causal explanation? Many words in -inu, e.g. equinus, divinus, have not got the suffix -ile, so why exactly femininu, if not because of the urge to avoid inadvertent repetition ?
At a certain time there were a series of words in -inu and some others in -ile, some few items in group one assume the same ending as group two, in what does the "explanation" consist ? Togeby considers phrases like "the change of r to l, because of the influence of another r in the word" by Rebecca R. Posner, Consonantal dissimilation in the Romance Languages, 1961, p. 1, as lacking sense ("La situation dissimilatrice ne suffit donc pas pour expliquer si une dissimilation a lieu ou non" (Qu'est-ce que .. "p. 651). Here T. is using (ibid. p. 659), following Grammont, the expression "par l'intervention du suffixe -ile"; it sounds just like "because of the influence of another r", and can much less than that claim to be a causal explanation. It is well known that repetition of the same consonant may cause difficulties of one kind or another and it is demonstrated that there is a tendency to avoid such repetition. That the presence of two suffixes, more or less synonymous, in itself brings about the replacement of one with another (in one or more words) appears rather doubtful. There must be something else to release a simply possible change, for instance, that one suffix is (highly) dominant in relation to the other one, in its use in speech or in running texts, in general, or, better yet, in a certain group of words which are linked together in a semantic field. In this case it should be, for instance, virile which would have attracted femininu, but hardly seruile, febrile. To render probable such a thing one needs detailed knowledge of the use of virile and femininu during the relevant period. However, Wilhelm Meyer-Lübke, Grammaire des Langues Romanes, II, § 436, describes the situation the other way round: [In Italy you have] "giovanile from lat. juvenilis and feminile, by dissimilation from femininus, which two have given rise to fantile, fratile, maschile, monachile, signorile;" In other words, feminile is the forerunner, the model for the other words in -ile. There is no basis for considering femininu attracted by a group of cognate words in -ile.
Creativity and derivation
There is a special side of the issue dissimilation and derivative suffixes where I think it is particularly clear that (Grammont and) Togeby are mistaken. When they claim that by distribution of suffixes no (sound) change takes place, it is interesting to study the formation of ethnic names, where there are so many variants, and where new forms are being created currently. The reason for this is that many ethnic names are words used relatively little, they belong to the margins of the language, and because you do not always know what the ethnic name of a certain town can be, what people usually say. This kind of variation can be easily studied in, e.g., H.-J. Wolf, Die Bildung der französischen Ethnica (Bewohnernamen), 1964, p. 113-129, E. Seyfried, Die Ethnika des alten Italiens, 1951, p. 164 8, M. Faust, Die antiken Einwohnernamen auf -itani, -etani, 1966, p. 128-9, 137-8, Cappello and Tagliavini, Dizionario degli etnici e dei toponimi italiani, 1981, O. Hjordt-Vetlesen, Romanske ethnica: derivation og dissimilation, Part 1, Italien, 1979.
Grammont and Togeby apparently agree that the phonetic changes are independent of the morphological circumstances. When to the city name Milano people use milanese as ethnic name, and not, for example *milanano, it is undeniable that this is due to the discomfort of repetition in *milanano that does not exist in milanese. It can hardly be proved for just Milano's case, especially because -ese is the dominant suffix for ethnic names in standard Italian, but it has been demonstrated statistically (in my little dissertation Romanske ethnica: derivation and dissimilation, Part 1, Italien 1979) that city names with a n as one of the last phonemes the suffix -ano occurs rarely, to a degree that is statistically significant. However, not all occurrences of the sequence "ano" in the Italian language are treated like that, some are, some are not. In the case of ethnic names, there is a tendency in that sense, which means, that the morphologic context is important, togehter with the phonetic context.
Togeby was very fond of debate, also the polemic ones. His article as a whole is a masterful exposition of the facts of the case. It has a clear trend, written with a sharp pen, humorous and ironic, often tending towards the sarcastic in front of other scholars, which could hurt the more sensitive ones (in fact, there is little doubt that Togeby in this article let himself carry away by a topic that traditionnally provides ample opportunity for demeaning polemics). Immediately in 1964, the article was commented by Henri Guiter, disciple of Grammont, who felt himself piqué over "une attaque acerbe de l'œuvre de Grammont, et ceci d'un ton dont la malveillance est souvent inacceptable" and he quotes 5-6 passages, such as "La force de la position syllabique, base du système de Grammont, ne joue en réalité aucun rôle". In his resentment Guiter fails to discover that Togeby actually agrees with Grammont at some points (Revue des langues romanes 76, 1964, p. 97-8).
Togeby lists a number of reasons why the concept of dissimilation practically is devoid of content. P. 665 it is affirmed that even those words (often discussed in the traditional debate about dissimilation) that cannot be explained by other means, cannot prove that dissimilation is a real cause of change, since there are other words, without a structure that gives rise to dissimilation, that show the same changes. In other words, when you besides xx > xy also have zx > zy, one can not assume the repetition of xx to have been the (contributory) cause of the change. This argument may sound alluring, but is obviously not sustainable; if it were, you could similarly reject the explanation of many other linguistic changes, by finding some words that have gone through the same development without the presence of the concerned phonetic situation. In fact, it is clear that one and the same sound change in different words can have different causes.
Togeby finds, p. 652-3, the fact that one can find assimilations and dissimilations in about the same (opposite) situations, e.g. mappa > nappe (dissimilation) matta > natte (assimilation), or that one can find dissimilation and assimilation which succeed each other over time in the same word (liliu> *ljilju > *jilju > it. giglio), shows that dissimilation is not the cause of the changes. After our point of view the two kinds of language change are not in such a contrast. (Paul Passy, in his Étude sur les changements phonétiques et leurs caractères généraux, Paris 1891, affirms that .. il peut fort bien y avoir assimilation à un moment, dissimilation à un autre; ou encore assimilation dans un sens, dissimilation dans un autre. P. 101). By the way, the two functions, assimilation and dissimilation, most often works in structurally different situations: Assimilation occurs mostly in direct contact (noctem > it. notte) while dissimilation predominantly is found where there is no such contact: *caeluleum > caeruleum (to be sure, there are examples of contact dissimilation: English fifth /fifþ/ not rarely is pronounced /fift/).
P. 649-50 Togeby discusses two groups of Latin words that practically everywhere are dissimilated: the quinque-group with kw + kw, and the peregrinus-group with r + plosive + r. In these two groups Togeby finds that dissimilation as sound law is working. Now the question is: Can one imagine such a sound change, whose essence is the speaker's (unconscious) avoiding of repetition, limited to these two groups of words - or to these two sound combinations, kw + kw and r + plosive + r, or is it more likely that the reason why it might look so, is an inadequately thorough investigation ?
Some years after Togeby changed somewhat his position. In an article on Spanish and Portuguese verbs in -ir he writes (p. 258): "Le seul facteur qui puisse rendre compte de l'apophonie propre à la conjugaison en -ir, est le facteur qui fonctionne encore synchroniquement: la dissimilation du i en e sous l'influence d'un i tonique d'une syllabe suivante, car c'est justement cet i tonique qui caractérise la conjugaison en -ir." ("L'apophonie des verbes espagnols et portugais en -ir", Romance Philology 26 (1972): 256-164.) In other words, dissimilation is not a phenomenon found in one or two cases alltogether, it is actually found here and there, and also as a factor that can change the grammatical structure