“…Although the human vocal apparatus can produce many different vocal sounds, languages generally use only a small subset of these sounds, and these are not uniformly distributed between and within languages (De Boer, 2000). Many psychophysical procedures, such as adaptive staircase methods but also reverse correlation (Jack & Schyns, 2017), require presenting participants with random distributions of stimuli in feature space, which is often impractical using naturalistic vocalizations (Belin, Boehme, & McAleer, 2017). With voice transformations, experimenters can uniformly or adaptively sample a large space of prosodic variations, for example, all vibrato frequencies between 1 and 10 Hz, regardless of how common these may be in actual behavior, and are thus able to draw better inferences about how these features are processed.…”