In my talk I will propose a lexical account of complex denominal verbs, i.e., denominal particle verbs and denominal prefix verbs, within the framework of Lexical Decomposition Grammar (Wunderlich 1994). If one requires that morphological derivation and semantic composition be isomorphic, the German complex denominal verbs give rise to apparent mismatches between morphology and semantics. I assume a morpheme-based account and then explore the minimum further assumptions necessary to account for these mismatches.

I assume that complex denominal verbs obey the same structural, semantic and conceptual constraints as simple denominal verbs. Denominal verbs are derived by the instantiation of abstract templates into which the base noun is integrated. These abstract templates account both for the distributional patterns of possible and impossible denominal verbs observed by Clark & Clark (1979) and Hale & Keyser (1992) and for the required semantic and referential shifts in the noun-verb conversion. Denominal verbs are structurally constrained by the Lowest Argument Restriction (LAR), which requires the base noun to saturate the lowest ranked argument of the abstract template.

I will show that complex denominal verbs exhibit the same degree of heterogenity as other complex verbs. This heterogenity consists of differences in the preverbs' argument structure effects on the bases and the mechanisms required for semantic composition. Preverbs can be classified as lexical arguments, lexical adjuncts and pure aspectual markers, according to these differences.

In particular, I will consider the following types of complex denominal verbs:

(1) a. er-schreinern `get sth. by doing carpentry', an-fiedeln `fiddle at'
    b. unter-kellern [under-cellar], auf-satteln `saddle up' 
       (complex locatum verbs)
    c. auf-bahren `lay on the bier', ein-rahmen `frame' 
       (complex location verbs)
    d. ver-stauben `get dusty', ver-silbern `silver plate'
    e. ver-slumen `become a slum', ver-sklaven `enslave'

I will argue that (1a) has to be considered as the default case: in the first step, the denominal base is derived, in the second step the preverb is added as a lexical adjunct. The most problematic cases are the complex denominal verbs in (1d/e): in (1d) the base noun is semantically integrated into the relation contributed by the prefix, in (1e) the prefixes appear to be pleonastic prefixes without a specific semantic contribution.

I want to show that (i) once the lexical entries of the preverbs are specified, the required results of the formation of complex denominal verbs are rendered automatically, (ii) no specific assumptions have to be made for the preverbs with respect to the denominal bases, and (iii) the lexical approach I advocate is simpler and more flexible than Hale & Keyser's syntactic approach.