Perceptual learning of time-compressed speech across different training tasks

Even brief training can elicit perceptual learning of time-compressed speech, but whether the type of task performed during training influences learning remains unclear. This study investigated the effects of training tasks with different levels of participant engagement during training on time-compressed speech learning and its transfer to novel stimuli. Learning and cross-talker transfer were assessed by comparing post-training transcription accuracy among five training groups and a no-training control group. At test, all trained groups recognized time-compressed speech produced by the trained talker (learning) and a new talker (cross-talker transfer) more accurately than the no-training group. Outcomes did not depend on the training task. A surprising finding was that baseline recognition of time-compressed speech was more accurate in participants tested in 2023 than among participants tested in 2018. The findings underscore the robustness of learning of time-compressed speech. To conclude, consistent with findings from other forms of degraded speech, learning can result from different experiences that require different levels of engagement with speech stimuli. Learning and transfer both depend on the acoustic features of the training stimuli and their similarity to the transfer stimuli, as suggested by ideal observer models. Lexical context seems sufficient to drive learning.