Syllabic dissimilation. Grammont.

Maurice Grammont published his dissertation La dissimilation consonantique dans les langues indo-européennes et dans les langues romanes, Dijon 1895, the most ambitious attempt of investigate what Dissimilation is, and how it works. Through almost half a century he defended his "laws" of dissimilation in a rather hot debate between him and his followers and the opposers, between who believe that dissimilation is a kind of Sound Law, like many other changes in different languages, and those that think that dissimilation is practically a Fata Morgana, a hotchpotch of different changes without any kind of ordered direction or common reason. See the survey of Knud Togeby, "Qu'est-ce que la dissimilation?" - Romance philology, 17 (1964): 642-667. Berkeley. (Reprinted in his Immanence et structure, Copenhagen, 1968, p. 96-121. = Revue romane, Numéro spécial 2.).

Grammont treats the issue of "syllabic dissimilation" more times. In his thesis, it is in the chapter "La superposition syllabique", p. 147-161, where he explains that "syllabic dissimilation" does not exist, supposedly because the undissimilated form of that kind of words never existed. His favourite example is the Greek word κελαινεφής 'covered with dark clouds', composed of κελαινo- and -νεφής (κελαινός, black, νέφoς, sky) and his main argument is that one cannot call this phenomenon dissimilation because the form *κελαινoνεφής never existed. In a chapter of the same name in Traité de phonétique, 1933, he repeats this claim, but I have not seen any place where he explains why the undissimilated forms must have existed (i.e. documented), before one can speak of dissimilation. And from where does Grammont know that *κελαινoνεφής has not been said or written? Considering all the data we have, it is actually extremely probable that people originally said *κελαινoνεφής, probably they "always" have used both forms. The fact that this single form is not recorded in literature is a fate that this word shares with countless others whose actual existence we mostly do not doubt. The numerous asterisk forms in our historical grammars are normally considered to have existed, of course, with more or less probability. In Kristoffer Nyrop, Grammaire historique de la langue française, Vol. 2, ch. 4, "Les terminaisons" (§§ 48-61, pp. 32-45), namely verbal suffixes, there is a number of such asterisk forms: *siam, *sias, *siat, *tragere, *sentivi, *senti(vi)sti, *cantumus, *cantamus, *debumus, *dormemus, *dormumus, *vidīmus, *perdēmus, *vendēmus, etc., a total of approximately 40. These forms are reconstructed on the basis of the actual documented forms, like one reconstructs *κελαινoνεφής. I wonder if Grammonts followers want to affirm that these word forms have not existed? It's granted that most scholars consider these forms to have existed, although they do not occur in the survived texts.

The whole idea of Grammont's appears quite hypothetical and thoughtful: he claims that the undissimilated form has not existed - although he does not support it with arguments or just make it probable; there are plenty of words of the same type where both forms are documented, that is they are documented in both a dissimilated and in an undissimilated version. This in itself is a crucial and sufficient refutation of the idea: an idea simply unfounded. Examples: τετράδραχμoν, the origin of τεδράχμoν, is, according to Grammont courante, but forme refaite(Traité .. p. 332), and the same applies to a number of other documented undissimilated Greek forms: αμφιφoρεύς, πιμτότες, ̔υψηλoλόφoς, ̔υδρoρoσάτoρ, ̔ημιμέδιμνoν. To demonstrate or render plausible that these forms are restored, in the sense that the dissimilated forms in a first period have reigned supreme, and that the undissimilated forms later have been introduced as a result of a restoration, could be theoretically possible, but Grammont does not suggest any form of documentation for that.

κελαινεφής 'covered with dark clouds'

More precisely Grammont's argumentation runs as follows: *κελαινoνεφής and many other similar examples are not documented, what shows that they have not existed, and that the dissimilated word is formed directly from the two parts (NB the reason why one of the two (partially) identical syllables are not found in the dissimilated word, is that the repetition is unfunctional. The repetition is only present when the two syllables are together in the (undissimilated) derived form; in other words, it is possible to reach the repetition-free form only through the form with repetition); when you actually find many word pairs consisting of the same word recorded in an undissimilated and in a dissimilated form, a fact that in itself disproves the hypothesis that has been put forward, you (Grammont) declare the undissimilated forms secondary restorations. That means, first the dissimilated form is formed by the two parts without that the undissimilated form ever existed, then, at a later date (the succession is not further defined), the undissimilated form is created, and, mind you, this could, according to Grammont himself, happen at any time. The whole line of thought is based on two false claims: that the undocumented undissimilated forms never existed, and that the documented undissimilated forms are secondary to the dissimilated. These ideas are in theory possible, but there is no reason to believe that things are that way, since there is another, simpler explanation.

A better description is probably that it is quite common that you have double forms of the same word in circulation for a period, and that over time there will be made a selection based on some, not always clear, criteria, what we call language use. This is a familiar situation generally for changes in language, and also for the here discussed changes. It turns out, not surprisingly, that there actually is a number of other, similar situations in other languages where Grammont's imaginative explaining away (neither) finds any evidence. P. 334 of the Traité de phonétique, he mentions the italian forms filogo, qualcosa, contractions of filologo, qualche cosa (besides sotterra, that very likely is not a contraction of sotto terra); it does not seem that filogo were ever dominant and that filologo then came back, and the same applies to qualcosa; we say in Italian both domattina and domani mattina, is there any way to prove that domani mattina is "restored"? No, there is nothing to suggest that domattina originally were formed without the existence of domani mattina and that domani mattina were later restored by the speakers. But even if it should be common that the undissimilated forms which "correspond" to the dissimilated, were missing (which is unproven and unlikely), it is pointless to argue that the language user has reached the dissimilated form without that the undissimilated one in any way had existed in the speaker's consciousness. It is also worth noting that Grammont does not suggest that other, "normal", dissimilations don't exist or are not "real" dissimilations, if the undissimilated forms are undocumented; e.g. if we did not have Latin peregrinu,French pèlerin would not be a dissimilated form ? Theoretically, one could argue that pèlerin had arisen in this form directly from per and ægrinus without that there had been any "real" sound change.

In Traité de phonétique Grammont has included a chapter on "superposition syllabique préventive" (p. 335-6 - which was not treated in the thesis), where the point is not syllabic superposition but rather how to avoid creating situations where this could occur. He cites a number of examples of choice of suffixes, including what we might call Wackernagel's distribution (Latin -nV- + -a/inus -> -ensis / -as, ecc., by city names), Spitzer's distribution (Spanish -a- + -ino vs -i- + -ano) - without mentioning their names - and also a Greek example: -ῐκός is a common ethnic suffix, but it is avoided where it would lead to -ῐκῐκός; in that case is used instead ́-ιoς; it applies in this case only to identical syllables: -ῑκῐκός (with long i) occurs, apparently without creating problems.

Brugmann

In his article "Das Wesen der lautlichen Dissimilationen", Abhandlungen der Phil.-hist. Klasse der Sächs. Gesell. der Wiss., 27, H. 5, Leipzig, 1909, p. 139-178, Karl Brugmann believes, against Grammont, that it is irrelevant whether the undissimilated form of the dissimilated word is documented or not; what matters is the cause of the phenomenon, and it cannot be denied that with κελαινεφής was meant *κελαινoνεφής (pp. 149-150). As we said, if the speaker had not somewhere in his linguistic life an experience of the undissimilated form, how would he then have reached the dissimilated ?

Niedermann

Max Niedermann in his article "Tendances euphoniques en Latin", Mélanges Bally. Genève, 1939, p. 423-38 (espec. p. 435-6) comments these ideas of Brugmann, in total disagreement with him. He explains the difference between the two types of changes *athenanus > atheniensis(choice of suffix) and *agrestris > agrestis (dissimilation in the traditional sense) in these terms: the first change is due to a 'deliberate action', while the other is 'unconscious, automatic'. Niedermann considers choice of suffixes as part of stylistics, while the 'real' dissimilations belong to the historical phonetics. Against this we must argue that our knowledge of the mental conditions of language activities does not allow us to distinguish with certainty between conscious and unconscious actions, let alone that today we are moving away from even sustain the idea of two such sharply divided spheres of our emotional and mental life.

Togeby

See the page Romance languages.