Is the syllabification of Irish a typological exception? An experimental study

We examined whether Irish speakers syllabify intervocalic consonants as codas (e.g., poca 'pocket' /po:k.@?/ CVC.V), as claimed by many authors, but contrary to claims in phonological theory of a universal preference for syllables with onsets. We conducted a perception experiment using a part-repetition task and presented auditory stimuli consisting of VCV items with a single medial consonant (Cm), varying in the length of V1 and the manner of articulation of Cm (e.g., poca /po:k@?/ 'pocket', lofa /lof@?/ 'rotten'), as well as VCCV items varying in the length of V1 and consonant sequence type (e.g., masla /masl@?/ 'insult', canta /ka:nt@?/ 'hunk'). Response patterns were in line with many, though not all, of the findings in the literature for other languages: Listeners preferred syllables with onsets, often treated Cm as ambisyllabic, syllabified Cm as a coda more often when V1 was short, and dispreferred stops as codas. The results, however, did not completely support the Syllable Onset Segmentation Hypothesis (SOSH), which proposes differing roles for syllable onsets and offsets in word segmentation. For VCCV items, listeners show a great deal of variability in decisions not only about where the first syllable ends, but also about where the second syllable begins, a variability that could not be explained by the number of legal onsets possible for a given consonant sequence. We examined the hypotheses that variability in perception can be accounted for by (1) variability in production (in the signal) and (2) phoneme pattern frequency. We searched pattern frequencies in an electronic dictionary of Irish, but found no support for an account in which language-specific syllabification patterns reflect patterns of word-initial phoneme sequences. Our investigation of potential acoustic cues to syllable boundaries showed a gradient effect of vowel length on syllabification judgments: the longer the V1 duration, the less likely a closed syllable, in line with results for other languages. For Irish, though, this pattern interestingly holds for all consonant manners except stops. The phonetic analyses point to other language-specific differences in phonetic patterns that cue syllable boundaries. For Irish, unlike English, consonant duration was not a more important cue to syllable boundaries than vowel duration, and there was no evidence that relative duration between the two consonants of a medial sequence signals syllable boundaries. The findings have implications not only for the syllable structure of Irish and theories of syllabification more generally. They are relevant to all theoretical and applied work on Irish that makes reference to the syllable.