Dataset of speech produced with delayed auditory feedback

Speakers use auditory feedback to monitor their speech output and detect any deviations from their expectations. It has long been known that when auditory feedback is artificially delayed by a fraction of a second, speech may be severely disrupted [[1], [2], [3]]. Despite the long history of using delayed auditory feedback (DAF) in experimental research on speech motor control, its effects remain relatively poorly understood. To our knowledge, there are currently no publicly available research datasets containing recordings of speech produced with DAF. Here we describe a large dataset of speech produced with DAF using modern experimental methods with systematic controls and varied speaking materials, including phonotactically legal, nonword syllable sequences and American English sentences. Auditory feedback latencies were tightly controlled and included a zero / minimal delay (∼12 ms), 150 ms, 200 ms, and 250 ms. The dataset includes simultaneous audio recordings from the microphone (production) and headphone (feedback) channels. It also includes recordings and annotations of reading passages and multiple other demographic and acoustic measures that serve as covariates of interest from each participant. The complete dataset, which is made available in two segments (one fully open access and one password restricted) includes speech audio recordings from 55 participants, 42 of whom completed a second session with similar testing materials. This dataset is valuable for researchers interested in theoretical aspects of speech sensory-motor control and for researchers interested in developing speech analysis tools.