L1 influence on the perception of consonantal and vocalic length contrasts

Phonemic length contrasts, which are signaled mostly by durational cues, are quite common in the segmental inventories of the world's languages. In phonological theory, they are encoded by means of one or two timing slots. The current study aimed at testing (a) whether listeners can redeploy an existing phonemic length distinction in their L1 to a new class of sounds in an L2 (e.g. from vowel length contrast in L1 to consonantal length contrast in L2 and vice versa) and (b) whether there is a correlation with level of proficiency in the L2 (Hayes-Harb & Masuda 2008). Previous research has shown that vocalic length contrasts in the L1 aid perception of this contrast in L2 (Hisagi et al. 2010). In cases without vocalic length contrasts in the L1, however, the literature is inconclusive: On the one hand, such contrasts have been found to be problematic (e.g., McAllister et al. 2002, Nenonen et al. 2003); on the other hand, vowel length distinctions in L2 perception are argued to be accessible independently of their existence in the L1 (Bohn 1995; Cebrian 2006). For consonantal length contrasts, there have also been mixed findings. While the absence of phonemic consonant length in the L1 yields poorer discriminability of this kind of contrast (Pohl & Grijzenhout 2010), Kingston et al. (2009) indicate little influence of L1 durational properties on L2 consonant length discrimination. In our study, three listener groups performed a speeded same-different task (inter-stimulus interval 1600 ms) involving vocalic and consonantal length contrasts as well as segmental contrasts: 14 native Italians, 11 native Germans with no knowledge of Italian, and 10 advanced German learners of Italian. Italian, but not German, distinguishes between singleton and geminate consonants (Esposito & Di Benedetto 1999) while German, but not Italian, distinguishes between phonemically short and long vowels (Hall 2000). CVC and VCV nonce words, differing in vowel quality as in the respective duration of the medial element, were recorded by a German-Italian bilingual (average durational ratio was 1.8). In VC(:)V nonce words, the initial vowel was of constant, ambiguous length so that listeners could only rely on consonantal cues. Analysis of responses showed no effect of language group or trial type (same/different) in the segmental contrast condition or in the vocalic length condition (both over 85% correct). The consonantal length condition yielded an interaction between language and trial type: in 'different' trials, the hit rate improved significantly (p<0.05) …