Institut de Lingüística Aplicada

Third Mediterranean Meeting on Morphology (MMM3)


Marc Plénat & M. Roché (University of Toulouse le Mirail)
Prosodic Constraints on Suffixation in French

Various specialists have demonstrated that a number of processes of word-formation in French are governed by a rigid prosodic constraint imposing a disyllabic template on the output (cf. Plénat (1984), Weeda (1992), Scullen (1997) on reduplicating hypocoristics and Plénat (1991) on Javanese). It seems also possible that apocopated forms follow the same constraint too, although a number of exceptions still await a proper explanation (cf. Weeda (1992), Ronneberger (1996), Scullen (1997)). Another interesting case is that of the oralisation of acronyms in which the choice between reading the form and spelling it out largely depends on the possibility of obtaining a disyllabic output, but where other factors do play a role in determining a monosyllabic or a polysyllabic pronunciation (cf. Plénat 1993, 1998). This somewhat less rigid version of the constraint imposing a disyllabic output also seems to be in operation in a certain number of slang suffixations, in which a monosyllabic base yields a disyllabic output (cf. e.g. dame > damoche, Bastille > Bastoche). There again, faithfulness to the base requires compromises to be made when the latter is made up of more than two syllables (cf. Plénat 1997, 1999). When we speak of a constraint (in the singular form), this is merely shorthand. The analysis of the facts demonstrates that there are several requirements: the output must include at least two syllables and no more than two (and this of course can easily be characterised in terms of the following OT constraints: FootBin, AllFeetLeft, and ParseSyll; cf. McCarthy & Prince, 1993)).

In this paper, we would like to show that the set of constraints which define the optimal prosodic word instead of affecting the whole lexeme can actually apply to the stem alone (the stem being defined here as the manifestation of the surface form of the base lexeme). A survey by Roché (1996) has already demonstrated that names in -esse (referring to female entities) become dramatically less acceptable as soon as the base lexeme is over two syllables long (cf . ? ?crocodilesse ou ? ?hippopotamesse). We intend to demonstrate here that the requirement that the stem should be a prosodic word triggers phenomena of internal sandhi (which can only be observed if we take into account the many neologisms given by a computerised data base).

To support this idea, we will examine two types of phenomena which are merely exemplified here through lack of space:

  1. Many truncation processes are blocked as soon as the result leads to the formation of a stem including more than two syllables, as in:
    bourgeois > bourgeoisissime bruxellois > bruxellissime
    lyonnais > lyonnaisissime milanais > milanissime
    nerveux > nervosissime rigoureux > rigourissime
    The rimes of the form vowel + sibilant are dropped in front of -issime (dissimilatory truncation) provided the base is at least made up of three syllables.
  2. Interfixation (i.e. the insertion of morphological material without semantic relevance between the base and the suffix) is virtually only attested in the case of monosyllabic bases. Let's consider feminine diminutives in -ette. Only disyllabic bases ending in a dental obstruent can give rise to an infix (e.g. culotte > culottinette): in this case, the cause is dissimilation. In all other cases, infixation is only attested when the base is monosyllabic; but the frequency of the phenomenon is a function of the nature of the final consonant (goutte > gouttelette, where the infix is the result of dissimilation but snobe > snobinette or snobette, where dissimilation is not at work).

It seems noteworthy that the application of processes of deletion or expansion depends on the possibility of obtaining a disyllabic stem. This seems to provide evidence in favour of the thesis that constraints are output-driven. We are not able to state whether these constraints can block certain derivations once and for all. What can be demonstrated, however, is that they have a definite effect on the probability of certain virtual derivatives occurring and that the choice between competing bases or allomorphs can depend on their ability to obey length constraints.