Controversy exists about whether associative learning occurs without awareness. In an earlier study using subthreshold (subliminal) stimuli, we reported evidence that such learning could occur as measured by event-related brain potentials [ERP; Cons. Cognit. 6 (1997) 519]. In the present study, we extend these findings by changing several aspects of the methodology in order to provide a more stringent test of this effect and to examine its generality. We used two matched words (murder and cancer) as conditional stimuli (CS); a 100 dB white noise blast as unconditional stimulus (US); a CS-US interval of 3 s; and a full-factorial design with CSs counterbalanced. The conditioning-acquisition phase occurred when the CSs were perceptually unconscious, as confirmed by a subsequent behavioral task. The conditioning-acquisition and postconditioning-extinction phases were examined for ERP evidence of associative learning. The clearest and strongest evidence for associative learning without awareness was observed in the ERP component measures (up to 1 s, poststimulus) in the postconditioning-extinction phase. The CS+ was significantly more positive than the CS- in the P3b-LP component region, which is highly consistent with the results of our earlier study. Differences also were observed in the P1-P2 components. In an unexpected finding, we observed a significant positive slow potential shift for the CS+ in the region between 1 and 3 s poststimulus. We discuss these results and their implications for our understanding of associative learning and awareness.
Copyright 2004 Elsevier B.V.