Institut de Lingüística Aplicada

Third Mediterranean Meeting on Morphology (MMM3)


Andrew Spencer and Ana Luis (University of Essex)
A Paradigm Function account of 'mesoclisis' in European Portuguese

Romance clitic systems exemplify many of the problems of studying the morphology/syntax interface. Portuguese future/conditional tense forms are unusual compared to those of other Romance languages in two principal ways. Whereas the tense/agreement marker is adjacent to the verb stem in (1a), as in other Romance languages, it must occur in the order given in (1b) if an object pronominal is present ('mesoclisis'):

  1. a. levaremos 'we will carry'
    b. leva-los-emos 'we will carry them' (NOT *levar-os-emos)

Portuguese verb forms generally bear only one stress. However, uniquely, the 'mesoclitic' forms are stressed both on the suffix AND on the verb stem:

  1. a. levarEmos vs. levA-los-Emos
    b. [levA-los]PW [Emos] PW

Mesoclisis has suggested to many that grammaticalization of the periphrastic future/conditional has not been completed in EP. Under this assumption, previous syntactic analyses (Duarte & Matos 2000, Raposo 2000) have derived the future/conditional marker as an inflectional affix in (1a), but as a separate prosodic word in (1b), as seen in 2(b). We show that this type of approach incurs a considerable cost. First, the future markers in (1a, b) are indistinguishable morphologically: i) they are formally identical, ii) always stressed and iii) always found at the right-periphery of the verb. The syntactic/phonological approach treats this identity as purely coincidental, and also completely fails to explain why the Fut./Cond. marker shows NONE of the properties of genuine auxiliary verbs in Portuguese. Second, previous treatments of 'mesoclisis' are based on the crucial assumption that the intervening object pronoun in (1b) constitutes a syntactically autonomous element (Duarte & Matos 2000, Martins 1994) that attaches postlexically to the verb (Vigario 1999). However, as can be seen in (1b, 2), some pronominals trigger idiosyncratic stem allomorphy (and also undergo idiosyncratic allomorphy), showing that they must be affixes (Crysmann 1997). Thus, pronominal enclitics must be realised as part of the morphology of the verb. This then permits us to analyse the Fut./Cond. markers as suffixes in both (1a) and (1b), automatically accounting for points (i - iii) above. We implement such a morphological analysis within the framework of Stump's (2001) realisation model.

We account for the double stressing in (2) by two simple assumptions: (i) stems and certain TAM suffixes (including the Future/Conditional) bear inherent accent, but the rightmost of two contiguous accents will erase that on the left, so that only the rightmost accent of a continuous sequence surfaces; (ii) pronominal clitics are unaccented. Therefore, the mesoclitic in (1b) breaks up the sequence of accents, allowing both to surface.