FIG. 2.

Flow chart summary of the MPI model used in the present study. The internal noise model generates a “percept” by adding JND noise to the location of a vowel stimulus in the F1 vs. F2 space. The decision model categorizes the percept by selecting the closest response center (i.e. the subject’s expected location of each vowel) after adding noise proportional to the subject’s response center uncertainty (RCU). The six model parameters optimized in the present study are highlighted in black with white lettering.