The nature and early time course of the initial processing differences between visually matched linguistic and nonlinguistic images were studied with event-related potentials (ERPs). The first effect began at 90 ms when ERPs to written words diverged from other objects, including faces. By 125 ms, ERPs to words and faces were more positive than those to other objects, effects identified with the P150. The amplitude and scalp distribution of P150s to words and faces were similar. The P150 seemed to be elicited selectively by images resembling any well-learned category of visual patterns. We propose that (a) visual perceptual categorization based on long-term experience begins by 125 ms, (b) P150 amplitude varies with the cumulative experience people have discriminating among instances of specific categories of visual objects (e.g., words, faces), and (c) the P150 is a scalp reflection of letterstring and face intracranial ERPs in posterior fusiform gyrus.