The ventriloquist effect results from near-optimal bimodal integration. This study investigates how visual and auditory stimuli are localized when presented separately or together. When visual localization is good, vision dominates and captures sound. However, for severely blurred visual stimuli, sound captures vision. For less blurred stimuli, perception follows the average position. The precision of bimodal localization is usually better than either unimodal presentation. These results are explained by a model of optimal combination of visual and auditory information, not by one sense capturing the other. Observers were asked to localize brief light "blobs" or sound "clicks" presented separately or together. In a trial, two sets of stimuli were presented successively, and observers indicated which appeared more to the left. Visual stimuli were low-contrast Gaussian blobs, and auditory stimuli were brief clicks. Results showed that for small blobs, vision dominated, while for large blobs, sound dominated. For mid-sized blobs, perception followed the average. The study found that the variance of the combined estimate is always less than either individual estimate, provided the underlying noise distributions are independent. This suggests that bimodal integration improves localization accuracy. The results support a model where visual and auditory information is combined by minimizing variance, leading to improved discrimination of bimodal spatial location. The study also found that for certain blob sizes, bimodal localization was better than either unimodal localization and near the predicted threshold. However, individual results showed some deviations from the predicted value. The results are consistent with other reports of sensory integration but differ slightly from some previous studies. The study concludes that the ventriloquist effect is a specific example of near-optimal combination of visual and auditory space cues, where each cue is weighted by an inverse estimate of noisiness, rather than one modality capturing the other.The ventriloquist effect results from near-optimal bimodal integration. This study investigates how visual and auditory stimuli are localized when presented separately or together. When visual localization is good, vision dominates and captures sound. However, for severely blurred visual stimuli, sound captures vision. For less blurred stimuli, perception follows the average position. The precision of bimodal localization is usually better than either unimodal presentation. These results are explained by a model of optimal combination of visual and auditory information, not by one sense capturing the other. Observers were asked to localize brief light "blobs" or sound "clicks" presented separately or together. In a trial, two sets of stimuli were presented successively, and observers indicated which appeared more to the left. Visual stimuli were low-contrast Gaussian blobs, and auditory stimuli were brief clicks. Results showed that for small blobs, vision dominated, while for large blobs, sound dominated. For mid-sized blobs, perception followed the average. The study found that the variance of the combined estimate is always less than either individual estimate, provided the underlying noise distributions are independent. This suggests that bimodal integration improves localization accuracy. The results support a model where visual and auditory information is combined by minimizing variance, leading to improved discrimination of bimodal spatial location. The study also found that for certain blob sizes, bimodal localization was better than either unimodal localization and near the predicted threshold. However, individual results showed some deviations from the predicted value. The results are consistent with other reports of sensory integration but differ slightly from some previous studies. The study concludes that the ventriloquist effect is a specific example of near-optimal combination of visual and auditory space cues, where each cue is weighted by an inverse estimate of noisiness, rather than one modality capturing the other.