Taking Scope: The Natural Semantics of Quantifiers

Chapter 3 The Natural History of Scope

Chapter 3 The Natural History of Scope It was a good-natured crowd, and a good time was had by all. —Lothian and Borders Constabulary, Edinburgh, Beltane 2004 Linking semantic scope directly to syntactic combinatorics makes it possible to explain a number of asymmetries between and among the scope-taking possibilities for universal and existential nominals in interaction with natural language syntax. These asymmetries present a challenge to all frameworks that attempt to capture scope phenomena in terms of uniform operations over generalized quantifiers, such as “quantifying in,” “quantifier raising,” or type shifting rules. The observations of Partee, Geach, and others concerning the surfacecompositional behavior of quantifiers with respect to distribution over natural language coordination have already been noted. A number of further asymmetries between and among universal and existential nominals of various kinds are set out below. The literature in this area is extensive and ramified, and the critical data are frequently in dispute. It is sometimes important to avoid getting distracted by details that in the end are not as important as the broad generalization noted by Farkas (1981), Fodor and Sag (1982), and Abusch (1994), among many others, which is that universal quantifers like each and every behave quite differently in languages like English from the other existential or individualdenoting quantifers.1 1. Since the primary focus of this book is on the varieties of scope-taking behavior exhibited by quantifiers, rather than on the quantifiers themselves, it examines only a representative sample of quantifier determiners in English and other languages. For a more extensive account of the detailed semantics of compound nonuniversal quantifiers, such as the partitives like nearly half (of) the students, in terms that appear to be in principle compatible with the approach followed here, the reader is directed to Carpenter 1997. 30 Chapter 3 3.1 Asymmetries in Scope Taking The summary below roughly follows Winter 2001, 166–7, Beghelli and Stowell 1997, 73–4 (among other papers in Szabolcsi 1997c), and Szabolcsi 2010, except where noted. First, all nonsingular so-called quantifiers distribute over existentials that they command. Thus all of the following have a reading in which there is a different pizza for each boy: (1) a. Every boy ate a pizza. b. The boys ate a pizza. c. Three boys ate a pizza. d. At least three boys ate a pizza. Second, the “distributive universal” quantifiers every and each can, in addition , distribute over quantifiers that command them, as in (2a): (2) a. At least one referee reviewed every paper. (∀≥1/≥1∀) b. At least one referee recommended that we should accept every paper . (∀≥1/≥1∀) c. At least one referee recommended that every paper should be accepted . (#∀≥1/≥1∀) More controversially, the present work assumes that such scope inversion of universals resembles wh-dependency in being both unbounded, as in (2b), and sensitive to some (though not all) “island constraints,” as in (2c), where scope alternation over the matrix subject is inhibited by the analog of the Fixed Subject Condition, parallel to the extractions in (3).2 (3) a. The papers that some referee recommended were terrible. b. The papers that some referee recommended that we should accept were terrible. c. #The papers that some referee said that should be accepted were terrible . Lakoff (1970d, 407–9) seems to have been the first to propose that scope inversion was both unbounded and limited by islands (independently supported by Rodman 1976). Both claims were contested by May (1977, 1985), Farkas (1981,1997b,2001), and Farkas and Giannakidou 1996,) although their examples against unboundedly inverting universals appear to be confounded with 2. Unlike extraction of subjects, scope inversion of embedded universal subjects is disallowed even from bare complements. This divergence is explained in section 8.5. The Natural History of Scope 31 subject islands like that in (2d), and only consider inversion over a/an indefinites (cf. Szabolcsi 2010, 91). (Farkas herself notes that determiners like some support bound readings under inversion more readily—see 1981n2— and that on occasion even indefinites do so—see 1997b, 212.) The literature has remained conflicted ever since, with Cooper (1983), Williams (1986), and Reinhart (2006) among those taking Lakoff’s and Rodman’s position, and Cecchetto (2004), Johnson (2000), and Szabolcsi 2010 among those taking May’s and Farkas’s. Experimental work by Syrett and Lidz (2005, 2006) suggests...