Figure 6

Two models of the interaction between the central tendency and the multisensory integration. (a) The first model assumes that the unisensory information is first biased by the prior integration, and then the biased unisensory representations are integrated by the maximum likelihood estimation. (b) The second model assumes that the unbiased unisensory representations are first integrated by the maximum likelihood estimation, and then the prior integration biases this multisensory representation of duration. Capitals in the legends represent the type of distribution: Likelihood (L), prior (P), posterior (R), and distributions in the multisensory integration (D). These graphs show examples for a stimulus duration of 0.45 sec, thus all the likelihood functions center at 0.45 sec. (c) Measured and modeled audiovisual central tendencies from the first model. (d) Measured and modeled audiovisual central tendencies from the second model. Symbols in (c) and (d) indicate the combination of auditory and visual noise levels: high visual and high auditory noise (purple), low visual and high auditory noise (blue), and high visual and low auditory noise (magenta). The solid line indicates the linear fit of the measured central tendency to the predicted central tendency. The shaded region indicates the 95% confidence interval of the linear fit.