Italic
PHILOLOGY AND DISSIMILATION
A typical example of the use of the idea of non-repetition or dissimilation in "traditional" lingustic studies is the following consideration of Alfred Ernout, in his important book on Latin morphology in a historical perspective (Ernout does not use the word 'dissimilation').
The Italic languages, Latin, Umbrian and Oscan have a suffix for the genitive plural of ā-stems, Latin *-ōm > -om > -um, Oscan -ūm, Umbrian -u, with later developments.
In Latin literature, before the classical period (Cicero), the suffix was -um, e.g. agricolum, where the -um is the result of the contraction and shortening of -ā-ōm. Later (around 200 BC) the ending (genitive plural) of the demonstrative pronouns -sōm penetrated into the nouns. When added to the -ā of the stem we got -āsum, and the intervocalic s changed to r (rhotacism), giving the well known form -ārum. By analogy also the second declension (dominus) got -ōrum instead of –um: dominōrum.
Ernout discusses the distribution of the new enlarged suffix -ōrum and the older version ‑um. In Cicero's time the suffix was almost exclusively ‑ōrum. That is, in most cases, but not always. Cicero himself, in his De Oratore, written in the year 46 BC, explains his own use of these forms, defending the partial use of -um basing his judgment on tradition (what the old poets said) and usage (what people use to say or write in Cicero's own time).
According to Ernout the exceptions (use of the old suffix -um in Cicero's time) are of three types: 1) In order to avoid the presence of too many r's near one another. 2) Because of the "excessive" length of some words. 3) In some set phrases.
That a word as horriferōrum cold be found awkward or clumsy, with 3-4 r's is not surprising; neither that somebody could prefer the older form horriferum. Can we be sure that the accumulation of r's is the problem ? Ernout gives 3 examples of words with an r in the root plus the final r of the stem: hoRRifeRum, baRbaRum, tRiumviRum; the other words given as examples of piling up of r's have only the final r of the stem: sēviRum (sēviRōRum), decemviRum (decemviRōRum) etc. So, apart from the three first words, the problem is the sequence -rōrum in posterōrum and the like. The other criterion is the length of words: montivagōrum is longer than montivagum and for that sake some people have preferred the shorter form. Maybe; the difference between the two possible forms consists of two sounds: ō + r. Many of the words with r in the stem are as long as montivagum or longer, so, maybe, the length and the piling up of r's have worked together. The third group is rather obvious: set phrases are known for being used long after the grammatical rules have been changed, like "to and fro". Thus people continued to say prō deum fidem instead of adopt the modern form prō deōrum fidem, simply by tradition, like we do also nowadays with old religious or legal phrases.
Now, why should Latin speakers avoid accumulation of r's in this specific case and only here ? Later in Ernout's book, treating the verbs, we learn about the forms with passive aspect. Here, corresponding to amās, 'you love' (active), we have amāreris, 'you are loved' (passive), and so on: monēreris, audiēreris, capereris; also legebāreris (imperfect indicative passive) 'you were being bound'. Or future tenses legēreris, amābereris, 'you will be loved', monēbereris. The climax is maybe legerēreris (imperfect subjunctive passive) 'that you were being bound'. Or, same tense, (non mentioned by Ernout) of the verb sero 'sow': sererēreris; the parts of this form are: ser-e-rē-re-ris: root (ser) + stem vowel (e) + suffix (re < se) + tense suffix (rē) + personal suffix (ris).
Also in the future imperative you meet repetitions: amātō '(you/he/she) do love !', amātōte '(you pl.) do love !', petitōte '(you pl.) do demand !', from petō (non mentioned by Ernout). These word forms full of repetitions receive no comment by Ernout. The reason why he discusses the problem of the assumed too many r's in horriferōrum and the like seems to be owed to the fact that the -ōrum-question is known from Cicero's book De Oratore, where he (Cicero) recommends the use of common sense. He does not - always - prefer one and the same of the two forms -ōrum or -um, he says procum instead of procōrum, but he will "absolutely not in any case" say triumuirōrum, only triumuirum. He wants to endorse the tradition or, maybe, good taste. So, Ernout's magnificent book does not treat the many accumulations of same consonant or vowel in Latin words or in the single forms of the words, except for the -ōrum-question of Cicero's.
Max Niedermann, in his Historische Lautlehre des Lateinischen, 2te Auflage, Heidelberg 1911, §46 Liquidae (in that edition p. 61-62), says "When two r's or two l's follow each other within a short distance, then they will attempt to differenciate, that is the articulation of one of the two sounds will, under the influence of the other one, be displaced or even altogether removed. This phenomenon, which is known under the name Dissimilation, ...". The choice of words is emblematic for Niedermann's period: "the two r's ... will attempt..", but it is more a question of style than of understanding of the linguistic facts. Niedermann cites cases like navalis but lunaris (not *lunalis), poclum but fulcrum (not *fulclum) and the world famous pelegrinus (not *peregrinus), originally per‑aegrinus from per agros =’ through fields’, that is from or to abroad. Niedermann gives a hint at a more generalizing view of the matter. In §98 he treats "Syllabic loss through haplology", cases like semodius instead of *semimodius, fastidium instead of *fastitidium, nutrix for *nutritrix. N. maintains that "when two syllables begin with the same consonant (or group of consonants), there is a tendency, in all languages, to leap over the first of them." "Leap over", I presume, is more or less the same as "being removed" cited before. N. gives no explanation of the alleged limited presence of haplology or dissimilation (only found in the –orum-words, with r and l in some words and the mentioned haplology cases. These cases are simply there, but why they are not more widespread remains without explanation or discussion.
Here is Ernout's text:
(Génitif pluriel de la deuxième déclinaison)
La désinence ‑ōrum est la seule désinence employée dans la langue littéraire classique. A l'époque de Cicéron le génitif en ‑um passait pour un archaïsme et ne s'employait plus que dans des circonstances particulières (Cicéron Orator 46, 155, 156). Néanmoins il s'est maintenu dans un certain nombre de cas:
1° pour éviter l'accumulation des r: par ex. dans nostrum līberum (Térence), sguāmigerum, horriferum (Lucrèce), posterum (Pervigilium Veneris), barbarum, triumvirum, sēvirum, decemvirum, quīndecimvirum.
2° dans des mots longs: montivagum, cōnsanguineum, magnanimum, omnigenum.
3° dans une série d'expressions spéciales aux langues techniques: praefectus fabrum, socium; deum dans prō deum fidem etc.; noms de monnaies et de mesures:
nummum, dēnārium, sēstertium, modium, jūgerum; Plaute a de même aureum, Pilipum, numerātum; enfin dans des noms de nombre fixés par l'usage: bīnum, trīnum, nūndinum,
ducentum.
Ernout, A., Morphologie historique du latin, nouv. éd., 1927, pp. 53-54.