Calibration and the comprehension of synthetic and natural speech
Bonnie L. Zeigler, George J. Boggs, Leah S. Kaufman · The Journal of the Acoustical Society of America · 1987
Investigations of “calibration” (the relationship between knowledge and the self-assessment of knowledge) have found that humans reading expository text cannot accurately judge their ability to draw inferences from the material read. Whether low calibration was related to the inference process or to the act of reading by presenting passages aurally, as either natural or phoneme-synthesized speech is examined. Sixty-four adults, divided into natural and synthetic-speech groups, heard 15 passages, with half the subjects permitted to listen twice. After each passage, subjects rated their comprehension, then judged the veracity of an inference drawn from the text. Calibration quotients, derived from the ratings and performance scores, were significant for both synthetic and natural groups. Performance, ratings, and calibration were slightly higher for natural versus synthetic subjects, but the difference was reliable only for confidence ratings. Thus previous low calibration may be more related to presentation mode than to inference processes. Better calibration for listeners may be due to a closer relationship between inferences and listening, which encourages extraction of themes as opposed to detail, than between inference and reading.