Measure words, plurality, and cross-linguistic variation

In the context of Borer's (2005) theory of nominal classification, the aim of this paper is to explain why measure words in some languages (English, French, Hebrew) necessarily take an - s (two bottles of milk versus *two bottle of milk) while in other languages (Azeri, Persian, Ojibwe) measure words can surface without plural marking (the equivalent of two bottle of milk is grammatical). If we assume - s in English-type languages is responsible for division in measure constructions (as in Borer 2005), we face the following puzzle: What is responsible for division in Azeri-type languages in the absence of the plural? We argue that, for a number of reasons, it cannot be the numeral (two) and propose that division is performed, in the absence of a plural, by measure words themselves (as in Chierchia 1998; Stavrou 2003; Acquaviva 2008, among others). We argue that whether or not plural marking appears on the measure word depends on a higher projection that expresses the counting function (distinct from the classifying/measuring function, Rothstein 2010b). Measure constructions thus provide evidence for the idea that, in addition to the dividing plural, we need a higher, counting plural, bolstering the hypothesis that the plural comes in many flavours (Acquaviva 2008; Harbour 2008; Wiltschko 2008, 2012; Butler 2012; Mathieu 2012, 2013, 2014).