Event-related brain potentials (ERPs) were recorded to fragmented pictures of objects that were named correctly or were not to investigate the time course of visual object identification. The first ERP difference distinguishing identified from unidentified pictures estimates the upper limit of the time by which human brain regions have begun to activate long-term memory (LTM) representations specifying the identity of a visual object. Data from 15 young adults indicate that this time varies with the extent to which object parts are recoverable from the visual input, being ~200ms earlier with recoverable than unrecoverable parts. Successful identification is evident by ~300ms when object parts and overall structural configuration are readily recoverable but not until ~550ms when object parts are difficult or impossible to recover (i.e. too poorly specified by the available contours to be recovered). In both cases, successful identification is associated with greater relative positivity. However, unidentified recoverable pictures are associated with an enhanced frontal negativity (N350), linked to object matching operations, not seen for non-recoverable pictures. Taken together, these results implicate two distinct processing sequences in the successful identification of visual objects.