Predicted concurrent vowel identification scores by the computational model as a function of onset asynchrony.

The range of concurrent vowel identification scores (mean ± one standard deviation) obtained from 14 human subjects in Hedrick and Madix (2009) is shown as a blue-colored area. Model predictions in the current study are shown for the internal noise at the level of neurogram generation (A) and comparison (B). Note that a different range of the level of internal noise was used for the left and right panels (see legends). Error bars for the model predictions are not plotted for clarity, but standard errors across 14 model runs were generally below 3%.