Two experiments investigated the time-course of object knowledge subsystems and how they might help explain novel object categorization. Experiment 1 compared ERP repetition effects in a 3 study (same or different exemplars of repeated categories, new category) x 2 typicality (high, low) design during category and recognition decisions. Repetition effects for same repeated objects but not different exemplars were found on the frontopolar N3, putative index of form representations. Effects on a posterior N400, thought to reflect amodal semantic representation, were found only when participants categorized. A centrofrontal N3 did show generalization, but only for high typicality items when participants were categorizing. Task differences were not found on the LPC /P600, even when analyses were limited to correct recognition trials. For Experiment 2, people categorized objects during an indirect memory test in a 2 repetition (new, old perceptually similar category exemplar) x 2 naming (same or different subordinate name from study to test) design. Results showed generalization to perceptually similar category exemplars on the frontopolar N3. Repetition and naming did not interact, inconsistent with semantic influence on object model selection. Taken together, these experiments show that distinct subsystems represent different aspects of object knowledge. From ∼ 150 - 350 ms, form-specific representations from a PRS were active and indexed by the frontopolar N3, as was a system that showed greater generalization to different exemplars which may reflect mapping onto conceptual representations. From 300 -500 ms evidence for two types of semantic representation, one modality specific (anterior N400) and one amodal (posterior N400) was found. Representations active after 500 ms may reflect conscious recollection or evaluation of memory match.