Institut de Lingüística Aplicada

Third Mediterranean Meeting on Morphology (MMM3)


Bernard Fradin (CNRS) and Françoise Kerleroux (Université de Paris 10)
How abstract are lexemes?

1. The lexeme is widely assumed to be the proper domain of morphology (a consequence of the rejection of morpheme-based morphology). A lexeme is identified by its category, its phonological form and its semantics. By definition, it is an abstract entity lacking inflection. As a consequence (i) WFRs only apply to underspecified entities, in which feature values relevant to syntax (inflection) or normally actualised in discourse (particular speech act) are left unspecified. Nevertheless, is the lexeme as abstract an entity regarding semantic representation (SR)? We would like to show that the way derivational rules semantically work implies that the basic entities to which they apply are fully specified in particular in what concerns the dimension of argument structure (Arg-str).

2. Clipping of deverbal nouns. French deverbal nouns in -(t)ion, -age, -ment are systematically ambiguous. They are interpreted as nouns denoting either events or results. These interpretations are correlated to the construction in which the noun appears (Grimshaw 1990). The clipping rule may apply to introduction as far as it denotes a resulting object, but may not when it denotes an event:

  1. a. Il s'oppose à l'(introduction+*intro) du loup à Paris.
    b. Il a apprécié l'(introduction+intro) de ton livre.

3. The -eur derivation. The lexical entry of French verb tomber 'fall' must specify that this lexeme shows up in two constructions: an inaccusative construction, C1: Arg-str , Sem (go-down'ox), Valency:SUJ:NP1; and a causative one C2: Arg-Str <NP1, NP2>, Sem. (causeoxo(go-down'oy)), Valency:SUJ:NP1, COMP:NP2. The fact that we can build a deverbal N in -eur only out of the causative version of the lexeme shows that WFRs are sensitive to argument-structure variations.

  1. a. Pierre tombe tout le temps. Pierre est un *tombeur.
    b. Pierre tombe toutes les filles. Pierre est un tombeur.

The data of §§2-3 indicate that the lexical object selected as base by the WFRs is not an abstract lexeme, in the sense that there exists no unit supposed to subsume the particular interpretations attached to introduction and tomber, because there is no way to subsume two different Arg-Str.

4. Delocutive derivation. Several nouns (diable, la vache, foutre) can be used as interjection in exclamative structures INTJ, S[EXCL]. E.g. for the noun diable 'evil' we have:

  1. Diable! Ça sent mauvais! 'The hell! It stinks'
  2. Diable! J'ai perdu mes lunettes. 'Damn! I have lost my glasses'.

(4) parallels other exclamative constructions where the interjection marks the fact that the speaker is surprised in an unpleasant way. In (3) the interjection indicates that what the predicate denotes in the sentence that follows is true at a very high degree. The so-called delocutive derivation (Benveniste 1966; Cornulier 1976) can build manner adverbs only out of interjection as far as it expresses the high degree:

  1. Ça sent diablement mauvais.
  2. *J'ai diablement perdu mes lunettes.

Sentence (6) roughly says that it stinks to such a degree that the speaker could say "diable!" to express how high this degree is (Anscombre 1979). These facts lead to the same conclusion: the base is not an abstract lexeme not only because diablement is built upon the interjective use of diable (which contradicts (i)), but also because there exists another derivation, namely (8) diable>diabolique>diaboliquement.

5. Conclusion.

  1. The border between morphological and syntactic units (lexemes and words (Aronoff 1994; Matthews 1974; Koenig 1999; Sag & Wasow 1999)) does not rest upon underspecification (lexemes) vs maximal specification (words). The lexeme is underspecified only for inflection but not for semantic alternations which show up in derivation as well as in syntactic constructions.
  2. To say that the units required by WF Morphology must exhibit the same information content (Arg-Str, SR) as the units used in syntax does not amount to saying that WFRs apply to syntactic units but only that this content must be encoded in the lexicon and is available as such by morphological rules and projectable as such by syntactic mechanisms.
  3. These conclusions tentatively disclaim the idea according to which the relationship between morphology and syntax results from a sequential organisation of components (WFRs precede syntax). We would say instead that both unit building components of grammar (morphology and syntax) exploit semantic properties of lexical units, each according to its program.