Event-related potential (ERP) investigations of auditory-visual (AV) interactions in humans have documented non-linear effects beginning ~50ms post-stimulus onset in response to rudimentary stimuli and during discrimination (Giard and Perronet, 1999 J Cogn Neurosci) or detection tasks (Molholm et al., 2002 Brain Res Cogn Brain Res). These findings and more generally the use of an additive model to test for multisensory effects with ERPs have been criticized, in part due to so-called ‘common’ activity related to task performance (Teder-Salejarvi et al., 2002 Brain Res Cogn Brain Res; Gondan et al., 2005 Percept Psychophys). An independent shortcoming of these prior works is that no analyses were performed that would discern whether non-linearities stemmed from modulations in response strength and/or response topography; the latter of which would indicate the recruitment of distinct configurations of brain generators during multisensory processing. Moreover, in one of these studies were source estimations performed, limiting the information these studies could provide on the likely underlying generators. Here, we performed electrical neuroimaging analyses of ERPs from 12 subjects in response to task-irrelevant auditory, visual, and AV multisensory stimuli. Importantly, because subjects were performing a motion detection task, attention was maintained on both sensory modalities (detailed in Cappe et al., 2009 Neuropsychologia), circumventing the criticisms of uncontrolled attention and the absence of a need to integrate multisensory information associated with passive paradigms (though see Vidal et al. 2008 Clin Neurophys for evidence of early non-linear effects during passive conditions). On the one hand, we replicate previous findings by showing there to be non-linear neural response interactions of individual ERP waveforms starting at ~55ms. Thus, comparably timed AV interactions can be reliably obtained when stimuli are task-irrelevant, but still attended. We then extend these findings to show that this initial non-linearity is the consequence of topographic modulations and not simply the result of pure amplitude enhancement. These effects indicate that multisensory interactions engage distinct configurations of brain networks, rather than simply modulating the strength of unisensory activity. Finally, we performed source estimations to identify the likely regions mediating these effects. We discuss our results both in terms of methodological/analytical considerations for multisensory research using ERPs and in terms of extant models of AV multisensory interactions.