Conditions on /t/-insertion in Maltese numeral phrases : a reassessment

Abstract

There has, for a considerable period, been disagreement and confusion as to the conditions governing the appearance of the /t/ morpheme that sometimes intervenes between the numerals 2–10 and a following plural noun in Maltese, as in ħames skejjel / ħamest iskejjel ‘five schools’ (e.g. Aquilina 1965: 118; Borg 1974; Cremona 1938: 204–205). In recent work (Lucas & Spagnol 2016) we reported on a native-speaker production experiment designed to improve our understanding of this issue. The results of that experiment suggested that the key factor determining /t/-insertion was onset of the plural noun: CV-initial plurals virtually never permit /t/-insertion, whereas CC-initial and V-initial plurals at least sometimes do. Number of syllables also appeared to be a relevant factor, in that, e.g., monosyllabic CC-initial plurals were found to strongly favour /t/-insertion, disyllabic CC-initials less so, and polysyllabic CC-initials not at all. The present work builds on this earlier research, arguing that a more accurate and more general statement of the conditions on /t/-insertion is one that makes reference primarily to morphological pattern, rather than to onset and number of syllables. This conclusion stems from a new production experiment focusing specifically on /t/-insertion with CC-initial disyllabic plurals. The experiment tested combinations of numerals with a number of both ”sound” (suffixing, non-stem-altering) plurals and ”broken” (non-suffixing, stem-altering) plurals. The latter fell into one of three patterns: CCVVCV(C), CCVjjVC and CCVCVC. The basic prediction was that the broken plurals would, in general, be much more favourable to /t/-insertion than the sound plurals. This prediction was borne out (broken plural mean insertion rate: 32%; sound plural mean insertion rate: 5%). Additionally, we predicted that broken plurals of the CCVCVC pattern, such as gwerer ‘wars’, in which two consonants occupy the initial root-consonant slot in the basic, highly /t/-resistant, CVCVC pattern (cf. Mifsud 1994), would be less favourable to /t/-insertion than the other CC-initial broken plural patterns tested. This too was borne out (mean insertion rates: CCVCVC 23%; CCVjjVC 37%; CCVVCVC 55%). Taken together, these two findings show that morphological pattern should be taken as the key determinant of /t/-insertion, with onset and number of syllables contributing only secondarily.