It is generally agreed that I-level predicates cannot delimit quantificational domains that consist of a multiplicity of situations. I-level predicates are taken to have no situation argument in their argument structure that could supply the necessary variable to the logical structure of quantificational sentences (see Kratzer 1988, 1995), or the situation variable introduced by I-level predicates is made unavailable for binding by higher operators in complex structures by independently motivated mechanisms (see de Hoop & de Swart 1989, 1990; de Swart 1991, 1993; Chierchia 1995, for example). Hence, if a generic quantifier, an adverb of quantification or some other similar operator has an I-level predicate as a restrictor, quantification tends to involve individuals: cp. Oranges are (often) sweet.

I will examine this standard view of I-level predicates in the light of the use of nominal I-level predicates as restrictors in generic sentences that express generalizations over a multiplicity of situations. I will focus on Russian examples like Uçcitelem on çcasto bolel - 'He was often ill, if/when(ever) he was a teacher', 'As a teacher, he was often ill' (one possible reading). I will argue that the instrumental case suffix in Russian is here best not analyzed as a shifting device that takes the meaning of I-level predicates and returns their S-level correlates. A similar argument can be made for the restrictive as in English, contrary to Katz (1993) and Jäger (2000), for example. I will also argue that predicative Russian nominals in the instrumental case generally retain their inherent I-level status, and the semantics of the instrumental case is crucial in licensing the binding of their situation variable by generic (and other) quantifiers as well as the use of I-level predicates in non-quantificational contexts in which S-level predicates freely occur.

My analysis has the following consequences for the nature of I-level predicates: First, the uniqueness presupposition associated with I-level predicates (see de Hoop & de Swart 1989, 1990; de Swart 1991, 1993; Chierchia 1995) is too strong. Certainly, not all I-level predicates describe properties that cover at most one significant portion of an individual's life-time, and consequently I-level predicates do not constitute a homogeneous class, at least not in this respect. Second, the existence of Russian I-level nominal predicates in the instrumental case does not invalidate Carlson's (1977) original claim that all nominal predicates are I-level predicates. Third, the use of nominal I-level predicates as restrictors in generic sentences that express generalizations over a multiplicity of situations can be viewed as providing further support for the assumption that all predicates have a situation argument in their argument structure (see Parsons 1985, 1990; Higginbotham 1985, 1988).