The explanation of the Italic imperfects in *-fām Lat. -bam which is commonly accepted in the hand books is not without its difficulties, although it has many points in its favor. It is true that the derivation of e.g. legēbam from a participial or infinitival form of lego plus *bhu̯ā-m, the old injunctive of IE *bheu-, is unobjectionable from the phonetic point of view, and that both the second syllable of Osc. fu-fans and Lat. -bam -bās, etc., may without the slightest hesitation be derived from *bhu̯ā-m. It is also true that the derivation of the formation from one of the most common of IE roots and the analogy of periphrastic imperfects of the type of Engl, was reading are additional favoring considerations. Nevertheless there are two very serious difficulties which make it hard to agree with this hypothesis in its received form.

Syncretism, where a single form corresponds to multiple morphosyntactic functions, is petvasive in languages with inflectional morphology. Its interpretation highlights the contrast between different views of the status of morphology. For some, morphology lacks independent structure, and syncretism reflects the internal structure of morphosyntactic features. For others, morphological structure is autonomous, and syncretism provides direct evidence of this. In this article, I discuss two phenomena that argue for the second view. Directional effects and unnatural classes of values resist attempts to reduce them to epiphenomena of more general rule types and require purely morphological devices for their expression.

In a paper entitled Assimilation and Dissimilation, read at the Linguistic Institute in the summer of 1936, Roland G. Kent explained these two types of linguistic change as the products of psychological conditions. They are due to the fact that ‘the thinking of the speaker is ahead of his utterance’ and they are hence for the most part regressive. Regressive assimilation is caused by an attempt to catch up with the thought, resulting in the replacement of one phoneme by another more similar to or identical with a subsequent phoneme. Regressive dissimilation is caused by an endeavor to avoid identity with what is coming by preventing the recurrence of the same phoneme within the same form. Because of the psychological factors involved, assimilation and dissimilation prove to be semiregular processes, that is to say, they are linguistic changes very different from the gradual and unnoticeable alterations of ‘phonetic change’. They consist in a redistribution of phonemes which is unpredictable, even though under certain conditions they occur with considerable regularity.

Frequent words tend to shorten (see e.g. Schuchardt 1885, Hooper 1976), as do words that have a high probability of occurrence given a neighboring word (Jurafsky et al. 2001). This tendency has been cited in support of the claim that probabilities are an inherent part of grammar, and of syntax in particular. There is widespread consensus, however, that the syntax of natural languages cannot be captured in terms of item-to-item transitions (Chomsky 1957). Therefore, unless one considers probabilities of syntactic structures, rather than particular combinations of neighboring words, pronunciation variation cannot be said to reflect probabilistic effects in syntax. In this article, we report a case of pronunciation variation that reflects contextual probabilities of syntactic structures. The relevant probabilities are based on the probability of a given syntactic-structure, given a particular verb. We show that these probabilities affect American English /t,d/-deletion, as well as the durations of words and phrases. Our results are consistent with the notion that knowledge of grammar includes knowledge of probabilities of syntactic structures, and that this knowledge affects language production.*