multisensory integration and attention have been studied independently and a vast amount of literature exists for both phenomena. researchers have recently raised questions, however, as to how these two processes interact. for example, does multisensory integration occur automatically without the need for attention; or does integrative processing require attention for it to occur? if the latter, do attention and integration act in parallel throughout all information processing levels, or does one operation need to exist to advance the other? the present study sought to answer these questions through a series of within-subject tasks spanning multiple layers of the processing hierarchy. forty-five participants completed three tasks involving audiovisual, integrated stimuli in which they discriminated the location of a visual target stimulus from nontarget distractors while being simultaneously presented with congruent auditory tones. the first task involved the discrimination of shapes and was shown to be preattentive in nature, with no facilitatory effect being observed in response to simultaneously presented visual and auditory stimuli. the second task involved the discrimination of die-point stimuli, which required high attentional demand. a trend towards intersensory facilitation was observed in this task, but was not significant. the third and final task involved the discrimination of integrated shapes and die-points, which also required a high attentional load. findings from this conjunction search revealed a surprising reversal of intersensory facilitation. these results suggest that attention has a limited capacity in terms of multisensory processing, and that specific intersensory facilitation requires a unique amount of attentional involvement. results are discussed in terms of feature integration theory, the perceptual load hypothesis, and attentional inhibition. this study also highlights the need for multisensory research to pay close attention to the influence of methodology, task sensitivity, and cognitive hierarchy when interpreting results.