Listeners' discrimination of read and spontaneous speech is primed by performance of a prior speech production task

Author(s): Rosanna Morris-Haynes, Laurence White and Sven Mattys


Distinguishing read and spontaneous speech seems intuitively to be a straightforward task, but listener performance in experimental studies is highly variable. Indeed, two recent studies showed chance-level discrimination performance, suggesting that – even with relevant prosodic cues available – listeners’ judgements may be heavily mediated by their contextual interpretation. Using lexically-identical map-task and read utterances previously found to be poorly discriminated despite available cues, we asked whether speech style identification could be primed by active familiarisation with the context of the speech production task. A between-subjects design with two conditions (priming vs no priming) was used. In both conditions, listeners completed a forced-choice speech style discrimination task on lexically-identical paired utterances. In the priming condition, prior to the discrimination task, listeners completed a communicative map task in pairs, equivalent to that used to generate the spontaneous speech stimuli. Although cues to speech style were available in the stimuli, performance in the no-priming condition was at chance. Discrimination performance was significantly better for subjects in the priming condition, suggesting that recent exposure to the production context of spontaneous speech promotes engagement of appropriate discrimination strategies. Indeed, subjective judgement data indicated that the priming condition increased listener awareness of relevant speech-style cues.