Figure 3. Group-averaged ventriloquist effects as a function of stimulus onset asynchrony (SOA) before and after training.

Perceptual training caused a reduction in the magnitude of the ventriloquist effect across all SOAs. These reductions were particularly large for intermediate SOAs indicative of a narrowing of the temporal binding window. Changes between the pre- and post-training integration windows were quantified by fitting each dataset with a Gaussian function (R² = 0.96 for both pre and post-training fits). The inset provides a summary of the changes to the standard deviation and amplitude of the Gaussian function following training. Data are expressed as learning ratios, calculated by dividing the post-training estimate of each parameter by the pre-training estimate. Both values are significantly less than one, indicating that these values are reduced in the post-training data. Error bars represent ± 1 standard error across participants.