Giulio Severijnen defends thesis 14 February

Link to live-stream

Talkers vary in how they speak, which leads to acoustic variability in segments (consonants and vowels) and in prosody (e.g., speech rate, lexical stress). In turn, the task for listeners is to correctly perceive speech from different talkers despite this variability. This thesis examined how listeners achieve this, particularly focusing on variability in prosody. The thesis first presents data that illustrates how exactly talkers vary prosodically, specifically regarding lexical stress production. Then, the thesis presents experiments that examined how listeners deal with that variability through perceptual learning, showing that listeners can adapt to how individual talkers produce lexical stress. Finally, the thesis asks how listeners deal with differences in speech rate and showed that speech is perceived relative to the rate in the surrounding sentence. In sum, the thesis shows that despite there being a large amount of prosodic variability in speech, listeners deal with that variability in two ways. First, listeners normalize incoming speech to the speech rate in the surrounding sentence. Second, listeners can adapt to how individual talkers produce lexical stress through perceptual learning. Prosodic variability thus really seems to be a blessing in disguise; listeners use variability in prosody to optimize perception of each individual talker.