EMB: Morphological embedding and phonetic reduction
Summary
Morphologically complex words are words that consist of two or more constituents, such as deepen, consisting of the base deep and the suffix -en, or careless, consisting of the base care and the suffix -less. While words like these seem structurally similar, the boundary between base and suffix has been found to differ. For example, the boundary in words containing the suffix -less is stronger than in those containing -en. Consequently, listeners are likely to decompose careless into its two constituents, and to process them separately, while deepen is more likely to be processed as a single unit without decomposing it first. The pronunciation of complex words is also affected by different boundary strengths. The two constituents at a weaker boundary show more phonological integration than those at a strong boundary, and are therefore more likely to be acoustically shortened or even deleted than in words with a strong boundary.
But the effect of different degrees of boundary strengths extends beyond complex words with only two constituents. It also constrains the order of attachment if the complex word contains two or more suffixes. In general, the boundary in the embedded form (the boundary between the base and the first suffix), is weaker than the outer boundary formed by the second suffix. For example, in the word owlishness, the boundary between owl and -ish is weaker than the boundary between the embedded form owlish and the outer suffix -ness. The reverse case, that is, a complex word in which the inner boundary is stronger than the outer boundary, is extremely rare in English.
While there is strong empirical evidence for these boundary strength differences in tri-constituent words like owlishness, it is as yet unknown whether these differences are also reflected in the speech signal. The fact that more phonetic reduction is found if the boundary is weaker in words with two constituents gives rise to the hypothesis that in tri-constituent words, more phonetic reduction is found at the inner boundary between the embedded constituents than at the outer boundary. In other words, the embedded constituents should have a relatively short acoustic duration, and should be more prone to segmental deletions, than the outer constituent.
The present project will test this hypothesis. It will first replicate the results from a pilot study that investigated tri-constituent compounds such as pizza home delivery, and which found that the embedded constituents (home delivery) were on average relatively shorter than the outer constituent (pizza). Critically, it will use data from speech production experiments as well as from corpora of spoken English to test the predictions of the hypothesis in an empirically rigorous way. In addition, a speech perception experiment will reveal whether listeners are sensitive to the different degrees of phonetic reduction due to differing boundary strengths, and whether they use it to facilitate the recognition of the internal structure of morphologically complex words.