To test the effects of perceived linguistic meaning on evoked potential (EP) waveform, an experiment was performed using the stimulus homophone "led" or "lead" in the single ambiguous phrase "it was /'led/". The phrase was presented aurally in sets of 60 repetitions, instructing the subject before each set as to the specific meaning of the stimulus word to be perceived. Averages of one hundred responses were obtained for each of the two meanings of the stimulus. Responses were recorded from scalp locations over Broca's and Wernicke's areas and homologous right hemisphere loci. Average waveforms evoked by the two meanings of the stimulus word were dissimilar for the left anterior locus and similar for the other three loci as indicated by correlations. The application of the Discrimination Index to these data disclosed that the EP waveform differences observed at the left anterior locus would not be attributed to random variability. Stepwise Discriminant Function Analysis on the same data revealed that responses from individual subjects could be classified according to experimental conditions at a greater-than-chance level. Discrimination was best for responses from left hemisphere loci. There was some commonality of discriminative EP components across subjects, specifically components late in the EP epoch. These data are interpreted as demonstrating an EP correlate of the processing of the contextual meaning of words.