In earlier reports we described contingent negative variation audiometry (CNV-A) and explored the accuracy of the method used in clinical conditions. Amplitude was found to be the factor which determines the accuracy of CNV-A. With this study we aimed to increase the amplitude of the CNV so as to improve the accuracy of CNV-A. The amplitude of the CNV is dependent on the content of the information of the S2 stimulus in the CNV paradigm. For this purpose we used a light stimulus (red or green) for the S2. The patient had to recognize the stimulus and act accordingly. The tests were performed on 12 volunteers in whom we simultaneously registered the CNV and the slow vertex response. The amplitude of the CNV was on average higher than with the previous CNV-A methods. The mean value of the absolute difference of the results between CNV-A and subjective tonal audiometry (ADG) was 6.9 +/- 5.5 dB. The mean difference between the results of both methods, however, was only 4.3 +/- 7.8 dB. The new CNV-A method is more accurate than the previous one. Nevertheless, we believe that not all possibilities for the improvement of the method have been explored. By taking into account the psychological factors which influence the amplitude of the CNV, and with a better computer analysis of the recordings, it will probably be possible to obtain even more accurate objective data on the threshold of perception of auditory stimuli with the aid of CNV-A.