@PHDTHESIS\{IMM2005-05212, author = "T. S. Andersen", title = "Model-based Assessment of Factors Influencing Categorical Audiovisual Perception", year = "2005", month = "mar", school = "Helsinki University of Technology", address = "", type = "", url = "http://lib.tkk.fi/Diss/2005/isbn9512275481/isbn9512275481.pdf", abstract = "Information processing in the sensory modalities is not segregated but interacts strongly. The exact nature of this interaction is not known and might differ for different multisensory phenomena. Here, we investigate two cases of categorical audiovisual perception: speech perception and the perception of rapid flashes and beeps. It is known that multisensory interactions in general depend on physical factors, such as information reliability and modality appropriateness, but it is not known how the effects occur. Here we parameterize the effect of information reliability for both our model phenomena. We also describe the effect of modality appropriateness as that of a factor that interacts with the effect of information reliability for counting rapid flashes and beeps. Less explored is whether multisensory perception depends on cognitive factors such as attention. Here we show that visual spatial attention and attentional set influence audiovisual speech perception. Whereas visual spatial attention affected unimodal perception prior to audiovisual integration, attentional set influenced the audiovisual integration stage. We also show a strong effect of intermodal attention on counting rapid flashes and beeps. Finally, we introduce a quantitative model, early maximum likelihood integration (MLI), of the interaction between counted flashes and counted beeps. We compare early {MLI} to the Fuzzy Logical Model of Perception (FLMP) which is a {MLI} model based on categorical percepts, and show that early {MLI} fits the data better using fewer parameters. Early {MLI} is also able to incorporate the effects of information reliability and intermodal attention in a more efficient way than the {FLMP}.", isbn_issn = "{ISBN} 951-22-7548-1" }