Maximum Likelihood Integration of rapid flashes and beeps.

Maximum likelihood models of multisensory integration are theoretically attractive because the goals and assumptions of sensory information processing are explicitly stated in such optimal models. When subjects perceive stimuli categorically, as opposed to on a continuous scale, Maximum Likelihood Integration (MLI) can occur before or after categorization-early or late. We introduce early MLI and apply it to the audiovisual perception of rapid beeps and flashes. We compare it to late MLI and show that early MLI is a better fitting and more parsimonious model. We also show that early MLI is better able to account for the effects of information reliability, modality appropriateness and intermodal attention which affect multisensory perception.