PhD Defence Eva Reinisch on June 29

June 17, 2010

Utterances can be spoken at various speaking rates, depending on the speaker and the context in which they occur. When processing speech, listeners take this durational variation into account. The duration of sounds not only varies due to changes in speaking rate, but also due to structural properties of language, such as lexical stress or word boundaries.

Lexical stress

'Spoken language unfolds over time in a continuous stream of sounds', Eva Reinisch states in the introduction to her thesis. 'In contrast to written words, spoken words do not become available all at once, but gradually unfold with the speech signal.' In her thesis Reinisch shows that listeners use not only information about the sounds of a word as soon as they come available, but also lexical stress. That is, listeners immediately distinguish between words like 'alibi' (word initial stress) and 'alinea' (stress on the second syllable) on the basis of their stress patterns.

Speaking rate

Duration is the most reliable cue to lexical stress in Dutch. Reinisch therefore asked whether the perception of lexical stress depends on the speaking rate of the preceding utterance. Her research shows that word fragments such as 'ali' are indeed more often judged to be the beginning of 'alibi' (an initially stressed word) when they followed a fast sentence than a slow one. Relative to the fast sentence the initial syllable of 'ali' sounds longer and thus more stressed than when it follows a slow one.

When during the word recognition process is speaking rate information taken into account? Reinisch used eye-tracking to tap into the earliest moments of word recognition. She found that speaking rate is used to interpret durational information very early during word processing. As the speech signal unfolds, listeners not only update the current acoustic information to rapidly recognise upcoming words, but also update their current estimates of speaking rate.