Oddities of numeral morphosyntax: from Arabic to Maltese

A well-known curiosity of Semitic morphosyntax is the phenomenon of reverse gender agreement in the cardinal numerals 3–10, e.g., Classical Arabic kitāb- ‘book ’ (masc.; pl. kutub-) triggers ostensibly feminine agreement (the morpheme-at-) on a modifying numeral (xams-at-u kutub-in ‘fiveNOM booksGEN’), while-at- is obligatorily absent on a numeral modifying a feminine noun (xams-u banāt-in ‘fiveNOM girlsGEN’). This peculiar system has been refunctionalized in many of the (caseless) present-day spoken Arabic dialects, such that numerals carry what looks like a feminine suffix essentially only in ‘absolute ’ position (i.e. in the absence of an overt head noun), as in Cairene Arabic xamsa ‘five’. (The original final /t / of the feminine suffix is in most varieties preserved only in the so-called ‘construct state’). Otherwise a reflex of the bare Classical (“masculine”) form appears, irrespective of the gender of the noun modified, as in Cairene xamas kutub ‘five books’, xamas banāt ‘three girls’. In a large subset of these dialects, however, spoken from Egypt and Sudan in the west to Iraq in the east (Mörth 1997), there is a relic of the original system in the form of a t- prefixed to vowel-initial plural patterns just in case they are preceded by a numeral between 3 and 10, but irrespective of the gender of the noun, as in Cairene xamas t-iʾlām ‘five pens (m.)’, xamas t-ifxida ‘five haunches (f.)’. If any morphology can be described as ‘junk ’ (Vincent 1995: 345–6), this ‘connecting t- ’ – wit