Second thoughts: multiple P300s elicited by a single stimulus.

In a previous report (Johnson & Donchin, 1978), we manipulated the discriminability of tone pairs that delivered feedback information in a time-estimation paradigm. As in other experiments using feedback stimuli, the event-related potentials elicited by these stimuli did not return to baseline in the 800-ms poststimulus interval. Since we were interested in this "Slow Wave" activity, the poststimulus interval was lengthened to 1500 ms. Averages revealed that a second positive peak was present for some of the individual subjects. To investigate this activity further, the filtered singletrial waveforms were inspected visually. These data were characterized hy the presence of one, and occasionally two, positive peaks, with highly variable latencies, following the P300. These peaks were indistinguishable in frequency and general appearance from the P300s elicited by the feedback stimuli. After latency adjusting the waveforms on the peak of the second positive wave, amplitude and latency were quantified. Whereas P300 amplitude was directly related to stimulus discriminability and positive feedback elicited larger P300s than negative feedback, the amplitude of the second positive wave was constant across levels of discriminability and the same for both types of feedback. In contrast, the latencies of both waves were inversely related to stimulus discriminability and shorter following positive feedback than negative feedback. Evidence is presented to support our contention that these additional positive peaks represent P300 activity. The data are discussed in terms of what these multiple P300s reveal about human information processing. DESCRIPTORS: Event-related potentials, P300, Eeedback, Decision making. In a previous report (Johnson & Donchin, 1978), second after the occurrence of a flash of light. If the we examined the effect of the utility of a stimulus subject's estimate fell within a window of permison the amplitude of the P300 it elicits. Specifically, sible error, it was considered correct, otherwise it we instructed subjects to press a button exactly one was incorrect. .After the subject pressed the button, ,,, , , r 1 c re \ii u one of two tones was presented indicating to the We are particularly grateful to Samuel Suuon. Walter . , , j , Ritter. and Chris Wood for their thorough and thoughtful ^"^ject whether or not the interval had been estireviews. We also wish to thank Connie C. Duncan-Johnmated correctly. The intensity difference (i.e., disson for her numerous helpful suggestions and criticisms. cnminability) between the two tones was varied in Monica Fabiani and Demetrios Karis also commented on different series. The ERP data indicated that the an earlier draft of this paper. Preliminary reports of these amplitude of the P3OOs elicited by both tones was data were presented at the Twenty-Second Annual Meetdirectly related to stimulus discriminability. Since ing of the Society for Psychophysiological Research, 1982, the tones served to provide the subjects with knowland the Second International Conference on Cognitive ^ ĵgg ^f ^^e results of their actions, we refer to them Neuroscience, 1982. The initial data collection was sup^^ "feedback stimuli " ported by the Advanced Research Projects Agency's Cyj ^ ^ a%pccx% of the data are consistent with the bernet.c Technology Office (Program Manager. Craig , ^ j ^ ^ ^^at there is a relationship between the Fields . under ONR Contract No. N-000-14-76-C-0002 to T^ r , ^ ., , j . ,• j r th snd uth r utility of the feedback tones and the amplitude of " AdTess''requ°e'sts for reprints to: Ray Johnson, Jr., The ^^e P300. First, P300 amplitude was directly proNational Institutes of Health, NINCDS/Medical Neurolportional to the accuracy of the subjects' time esogy Branch, Building 10. Room 4N246, Bethesda, Marytimations and inversely proportional to the subland 20205. jects' a posteriori uncertainty (i.e., equivocation)