Background: Several studies have demonstrated that perceptual decision making can be remarkably consistent with statistically optimal models. For example, human auditory and visual localization judgments closely resemble that of a Bayesian causal inference model (Körding et al., 2007), where the observer infers the underlying causal structure of the environment based on the available evidence and prior knowledge. While most of these studies characterize perceptual inference within a static environment, there is little known as to how this inference process changes when exposed to a dynamic environment. One of the mechanisms of the nervous system for coping with exogenous or endogenous changes is the continued maintenance of sensory processing through crossmodal recalibration. A common example is the Ventriloquist Aftereffect: the shift in perceived location of sounds (in isolation) that occurs after repeated exposure to consistent spatial discrepancy between auditory and visual stimuli. From a computational point of view, this change could reflect a shift in the auditory sensory representations (i.e., shift in auditory likelihood distribution), a decrease in the precision of the auditory estimates (i.e., increase in spread of likelihood distribution), a shift in the auditory bias (i.e., shift in prior distribution), or an increase/decrease in certainty of the auditory bias (i.e., the spread of prior distribution), or a combination of these. Purpose: We aimed to computationally characterize this perceptual recalibration effect. Because the aforementioned causal inference model allows estimation of likelihoods and priors, we can empirically test which one(s) of these quantities undergoes change after recalibration. Methods: During an exposure phase, observers performed a visual contrast change detection task, while passively exposed to flashes and noise bursts at varying locations along azimuth but with a consistent spatial discrepancy. In pre- and post-exposure sessions, observers were presented with auditory and/or visual stimuli and asked to localize each of the stimuli. Using the Bayesian causal inference model, for each observer, we empirically and quantitatively tested for any parametric changes in prior or likelihood distributions after exposure. We used normal distributions to represent the likelihoods and priors over space, which resulted in 7 free parameters: the standard deviation of the auditory and visual likelihoods (2) and spatial prior (1), the mean of the spatial prior (1), the prior probability of a common cause (1), and the shift in the mean of visual and auditory likelihood functions (2). These parameters were fitted using 900 data points obtained from each subject in each of the pre-test and post-test sessions separately. Results: We found that after exposure, the auditory likelihood functions were shifted to the right (paired ttest, p<0.001) when the visual stimulus was presented to the right of the auditory stimulus during exposure, and vice versa (p<0.001) when the visual stimulus was presented to the left during exposure. No statistically significant shift was found in the mean of the prior over space, in the prior probability of a common cause, in the mean of the visual likelihood, or in the visual and auditory noise. All tests were corrected for multiple comparisons. Conclusion: This suggests that sensory recalibration observed in the ventriloquist aftereffect is achieved by updating the mean of the auditory likelihood using recent perceptual estimates. Recent studies have provided evidence that human responses follow the posterior distribution of combined cues; our results suggest that this internal estimator can also drive the recalibration process.