Eugen Hill (Universität zu Köln)
Zur diachronen Typologie der morphologischen Null (Watkins' Gesetz und Watkins' Transformationen)
Vortrag im Rahmen des Linguistischen Kolloquiums
The talk deals with patterns of argument indexing on verbs. It addresses a conflict between synchronic typological preferences and diachronic trends in the development of argument indexing systems. The talk aims at establishing a framework accommodating both the synchronic data and the diachronic observations.
As is well-known, argument indexes on verbs commonly emerge either by
(a) univerbation of the predicate with personal pronouns, or by (b) univerbation of a nominal predicate with inflectional forms of the copula. These observations are in harmony with the fact that synchronically 3sg.-zero patterns are more common than patterns with a zero-index in the 1/2sg. (cf. recently Bickel et al. 2015, Siewierska 2004, 2010). A 3sg.-zero pattern must be expected when person indexes are grammaticalized from personal pronouns because (a) many languages do not possess a personal pronoun in the 3rd person (cf. Cristofaro 2013, Bybee 1988), (b) in the 3rd person the use of a personal pronoun is often redundant because the relevant argument is expressed by a noun phrase. A similar conclusion is valid for those languages where the person indexes on verbs succeed the copula. In pro-drop languages, the copula is obligatory in the 1st and 2nd persons (because otherwise the subject would not be encoded) but redundant in the 3rd person (where the subject is mostly expressed by a noun phrase).
However, recent investigations also reveal that synchronically 3sg.-zero patterns are rarer than patterns indexing all three persons (cf. again Bickel et al. 2015). This implies that in 3sg.-zero patterns the empty slot must be frequently filled by secondary processes (cf. Mithun 2013, Fuß 2005). The latter finding is clearly in conflict with the well-known cross-linguistic trend to reshape patterns indexing all three persons into secondary 3rd-zero patterns by generalizing the inherited form of the 3sg. as the bare stem (so-called ‘Watkins’s law’, cf. recently Miller 2010, Croft 2000, Koch 1995).
I will argue that this conflict between a diachronic trend implied by synchronic preferences (to fill the 3sg.) and a counter-trend directly observed in the diachrony (‘Watkins’s law’) can be resolved by assuming that evolution of argument indexing is typically conditioned by two different factors: (a) regular sound change, (b) system pressure from within or outside the relevant language system.
This assumption explains the tendency to secondarily fill the empty slot in the 3sg. as well as both the frequency of ‘Watkins’s law’-transformations and the absence of comparable generalizations of the 1sg. or 2sg. A generalization of the 1sg. or 2sg. as the bare stem requires a pre-existing 1/2sg.-zero model which can only emerge by sound change. By contrast, a generalization of the 3sg. may follow a 3sg.-zero model emerged either by sound change or by univerbation (with personal pronouns or with the copula). It follows that ‘Watkins’s law’-generalization of the 3sg. is fed by two distinct sources and must, therefore, be more frequent.