Noun-verb symmetries in nahuatl nominalizations

In the literature on the argument structure of nominalizations, one can find mainly two approaches: the first approach (e.g., Grimshaw 1990) assumes that argument inheritance is restricted to event nominals or event-related nominals, whose inherited arguments are obligatorily realized. This view contrasts with the assumption in the second approach that all arguments of the base verb are inherited by the nominal unless they are explicitly bound, but that internal arguments of the noun/nominal are optional (e.g., Bierwisch 1989). The first approach predicts noun-verb asymmetries in argument linking for the class of non-event nominals, while the second approach predicts that the argument linking of nouns/nominals differs from that of verbs in general. Classical Nahuatl, an Uto-Aztecan language, provides evidence against both approaches: in all classes of nominals (event, agent, instrument, locative, and internal argument nominals), unspecific affixes saturate arguments in the base verbs and the derived nominals, independent of whether an event reading is possible or not. Only under the assumption of obligatory argument inheritance can the distribution of active and inactive stems and the distribution of reflexives be explained for the nominals. I will show that none of the nominals fall under the notion of 'mixed categories' (such as English gerunds), which are expected to display a verb-like argument linking.