In sentences like Every teacher laughed we think of every teacher as a unary (=type (1)) quantifier - it expresses a property of one place predicate denotations. In variable binding terms, unary quantifiers bind one variable. Two applications of unary quantifiers, as in the interpretation of No student likes every teacher, determine a binary (= type (2)) quantifier; they express properties of two place predicate denotations. In variable binding terms they bind two variables. We call a binary quantifier Fregean (or (...) reducible) if it can in principle be expressed by the iterated application of unary quantifiers. In this paper we present two mathematical properties which distinguish non-Fregean quantifiers from Fregean ones. Our results extend those of van Benthem (1989) and Keenan (1987a). We use them to show that English presents a large variety of non-Fregean quantifi ers. Some are new here, others are familiar (though the proofs that they are non-Fregean are not). The main point of our empirical work is to inform us regarding the types of quantification natural language presents - in particular (van Benthem, 1989) that it goes beyond the usual (Fregean) analysis which treats it as mere iterated application of unary quantifiers. Secondarily, our results challenge linguistic approaches to "Logical Form" which constrain variable binding operators to "locally" bind just one occurrence of a variable, e.g., the Bijection Principle (BP) of Koopman and Sportiche (1983). The BP (correctly) blocks analyses like For which x, x's mother kissed x? for Who did his mother kiss? since For which x would locally bind two occurrences of x. But some of our irreducible binary quantifiers are naturally represented by operators which do locally bind two variables. This paper is organized as follows: Section 1 provides an explicit formulation of our questions of concern. Section 2 classifies the English constructions which we show to be non-Fregean. Section 3 presents the mathematical properties which test for non-Fregean quantification and applies these tests to the constructions in Section 2. Proofs of the mathema tical properties are given in the Appendix. (shrink)

In this paper, I want to contribute to understanding and improving on Keenan'sintriguing equivalence result about reducible type quantifiers (Keenan, 1992).I give an alternative proof of his result which generalizes to type quantifiers, andI show how the reduction of a reducible type quantifier to (the composition of) ntype quantifiers can be effected.

avant propos This paper is basically Keenan (1992) augmented by some new types of properly polyadic quantification in natural language drawn from Moltmann (1992), Nam (1991) and Srivastav (1990). In addition I would draw the reader's attention to recent mathematical studies of polyadic quantiicationz Ben-Shalom (1992), Spaan (1992) and Westerstahl (1992). The first and third of these extend and generalize (in some cases considerably) the techniques and results in Keenan (1992). Finally I would like to acknowledge the stimulating and constructive (...) discussions ofthe earlier paper with many scholars, notably Dorit Ben-Shalom, Jaap van der Does, Hans Kamp, Uwe Mormich, Arnim von Stechow, Mats Rooth, and Ede Zimmermann. And I repeat here the acknowledgment in the earlier paper to Jim Lambek, Ed Stabler and two anonymous referees for Linguistics and Philosophy (the latter responsible for substantial improvements in the proofs - see footnote 10). (shrink)