Donnerstag, 9.30 Uhr
Grammaticalization has traditionally been seen as a process which involves "semantic bleaching" (Gabelentz) and a process whereby lexical items become grammatical morphemes (Meillet, Hopper and Traugott). The notion of semantic bleaching is understandable intuitively, but how and why this should affect the syntax of a language is not easily explicable. In particular, it is not easy to see how to model such a change within a derivational system such as Principles and Parameters. LFG, by contrast, a lexically based theory of syntax, which factors out semantic, functional and syntactic information into different but parallel levels of representation, would appear to be an ideal framework for examining such grammaticalization processes.
A particularly interesting topic in this connection concerns the historical origin of 'raising' verbs. Synchronically, these verbs have a grammatical function (subject, object) which has no associated theta-role, but diachronically they develop from verbs which did assign a theta-role to the relevant argument function: e.g. Eng. tend < Fr. tendre < Lat. tendere 'stretch out, offer'; Fr. sembler 'seem' < Lat. simulare 'to pretend'. We can thus hypothesize that 'semantic bleaching' in these instances corresponds to the historical dissociation of function and theta-role.
In the present paper I develop an account of how that dissociation takes place for the class of subject-to-subject raising verbs which have their origin in verbs of perception and which synchronically are markers of epistemic modality, e.g. seem, appear. Evidence for the connection between the active form of perception verbs and current epistemic 'raising' verbs comes from two independent sources: firstly, the fact that synchronically in many languages the word corresponding to 'seem' is directly related to the word for 'see' and secondly, diachronically we can trace the development of a 'raising' verb, e.g. Fr. sembler 'seem' from an active lexical verb e.g. Lat. simulare 'pretend' where perceptual reanalysis has taken place.
Synchronic relatedness of forms is seen in the application of various related syntactic processes: (i) via passivization of the perception predicate: Latin videre'to see' videri 'to be seen, to seem'; (ii) via reflexive morphology (the reflexive passive): Turkish görmek 'to see' görünmek 'to be visible, to be seen, to seem'; (iii) via other detransitivizing affixation: Zulu bona 'to see' bonakala 'to seem, to look at (copulative)'. Common to all is the suppression of the external (perceiver) argument and the presence of a morphological reflex of the process. In Latin, for example, passivisation is used as a strategy to suppress or demote the perceiver argument of the active verb 'see' and promote an argument which may then be interpreted either as the object of direct physical or of more abstract mental perception. Wheras the former does not require a clausal or adjectival complement the latter does, and hence the subject of the matrix verb can be interpreted as satisfying the theta requirements of the embedded predicate. Similar accounts can be developed when the syntactic process intervening between the active perception verb and the derived and grammaticalized raising verb involves reflexivization or detransitivization.
However, argument restructuring is not always morphologically signalled (cf. Dative Shift in English) or lexically marked (cf. English hear (active) and sound (copulative) vs. taste (active) and taste (copulative)), restructuring can then take place covertly, as I argue for the development of simulare to sembler. The development of simulare 'pretend' to sembler 'seem' can be seen as an anticausative process, whereby the 'agent' and 'cause' are removed from the semantic structure. Crucial to the process of grammaticalization to a 'raising' verb is therefore: 1) the presence of secondary predication; 2) passivization or change of perspective.