Probing the Syllabic Structure of Words using the Audio-Visual McGurk Effect

Empirically, the onset-rhyme structure of the syllable has been revealed in several experiments using word-games. The simplest games used monosyllabic words C(C)VC(C) presented as auditory stimuli (Trieman, 1983, 1985). In most of Treiman’s work, participants were asked to coin a new word from given pairs of words. The participants created new words using perceived inner boundaries, usually located between onset and rhyme. They took either the onset of the first word with the rhyme of the second, or the onset of the second and the rhyme of the first. These studies concluded that the participants made onset-rhyme partitioning more often than by chance – suggesting the empirical reality of the boundary in the cognitive word model of participants. Many studies have shown that a syllable has an internal hierarchical structure and is made up of two main constituents: the consonantal onset and the rhyme. Most experiments to test the cognitive reality of the syllable in the mental models of humans have involved word games using words with concealed parts. Here we outline an alternative way of testing for syllabic structure using McGurk fusion. The experiments develop a method for locating syllable boundaries in polysyllabic words, based on the coda-onset fusion rate differences found in our earlier work with monosyllabic words. For the sake of simplicity, the fusion rate measurements are made on polysyllabic words in which an internal consonantal site is given audio-visual incongruence. A word-internal site may be the coda of a first syllable or the onset of a second syllable. The aim of the research was to determine whether a fusion in the wordmedial consonant of polysyllabic words confirms the traditional or another syllabification scheme. The results from our first study show that the McGurk fusion rates can be used to locate syllable boundaries in polysyllabic words. The findings show that in 65% of the English words investigated the coda of the traditional morphological stem has the fusionrate behaviour of an empirical coda. In further polysyllabic games using auditory priming stimuli and visual targets, Mehler et al (1981) investigated a bimodal monitoring task put to French participants. The participants were presented with an auditory word stimulus for a short duration and the target was a short-duration visual presentation of the first fragment of a text word. The reaction times for identifying the fragment were found to be much shorter when the target corresponded to the whole first syllable of the priming stimulus than when the target was a part syllable. For example, French speakers identified the target syllable /pa/ much faster when primed by ‘pa.lace’ than by ‘palm.ier’. Mehler et al concluded that word-recognition is dependent upon good placement of masking boundaries at syllabic breaks in words. (These empirical boundaries follow traditional syllable boundaries, at least with francophone participants). In a separate study, Segui (1984), reported similar findings and added that the first syllable can be considered as important key to lexical access. However, in some experiments with English speakers, investigators found no difference between responses after priming by whole or part syllables (Cutler et al, 1986). For example, English participants primed by ‘balance’ and ‘balcony’ responded similarly to the targets with ‘ba-’ and ‘bal-’. A possible explanation is ambisyllabicity in English, both ‘bal.ance’ and ‘ba.lance’ being acceptable syllabifications of ‘balance’. Corresponding findings by Zwitserlood et al (1993) on English and Dutch participants in priming and masking experiments were similar. Dutch participants given ‘bu(k)en’ (= to stoop), ‘ro(k)en’ (= to smoke), ‘me(n)en’ (= men), showed marginal preference for the coda option in the