Rice University
Department of Linguistics Colloquium

It is widely assumed in the linguistic literature that arguments and adjuncts differ in that arguments are licensed by a predicate but adjuncts are not. Licensing means that the nature and number of arguments for a predicate is strictly specified, while that of adjuncts is simply unspecified. For example, in (1) the verb only allows for one argument. In (2) the verb requires two arguments. (3) shows the lack of any such restrictions on adjuncts.

(1) I fell (*a giraffe).

(2) I saw a giraffe.

(3) I saw a giraffe (in the parking lot) (yesterday).

Some suggested criteria for distinguishing arguments from adjuncts include: necessity for semantic completeness, limited vs. unlimited number, and whether selectional restrictions apply (e.g. Kroeger 2004, Haegeman 1994, Bresnan 2001).

This widely-accepted sharp demarcation between arguments and adjuncts does not hold when more cross-linguistic data are considered. Skou and One, two languages spoken in North-Central New Guinea (Donohue 2006, Laycock 1975, Crowther 2001, Sikale et al 2002), and Tsou, an Austronesian language spoken in Taiwan, are languages in which the nature and number of adjuncts is strictly specified, contrary to the widely-held assumption above. For instance, in Skou, the number of adjuncts allowable per clause is restricted to one. Furthermore, the syntactic manifestation of arguments licensed by a particular predicate differs depending on the nature and the position of a co-occurring adjunct. In Tsou, the frame-semantic knowledge of verbs places selectional restrictions on co-occurring adjuncts, again contrary to the assumption above. These facts suggest that adjuncthood may be a gradient notion and the boundary between adjuncts and arguments is not discretely demarcated. The need therefore arises for re-evaluating the status and categorization of adjuncts in the representation of argument structure.

© 2007 Gujing Lin
Last updated 26 Nov 07