Fig. 1.

(A) Distribution of VOT values for voiced and voiceless stops in English (from acoustic measurements in Allen & Miller, 1999). (B) Two-category cues can vary in how reliable they are (and, in turn, how highly they should be weighted) from reliable cues with distinct categories (top left) to unreliable cues with overlapping categories (bottom right). Cues of intermediate reliability (top right and bottom left) require us to take into account both the distance between categories and the variability within them.