Figure 1.

Examples of gradient and categorical mapping of speech cues. The bottom level represents encoding of a continuous acoustic cue such as VOT. The top level represents speech categories; in classic views this may be phonemes or phonological features, but this could also be syllables or even lexical items. The important thing is that information is more or less discretely represented. A) Categorical encoding of cues leads to a sharp category boundary at both levels. B) Cue encoding is gradient, but speech categories are activated discretely, leading to an abrupt boundary only for the latter. C) Both cue encoding and activation of speech categories are gradient.