Miller (1956) has suggested that the number of stimuli which can be identified in an "absolute" discrimination task is 7-1 2, about 2.8 bits of transmitted information. Considering the abiliry of the human operator to handle very large amounts of stimulus information, under optimum conditions one might intuitively expect absolute discrimination to exceed Miller's magical 7 by at least an order of magnitude. A preliminary investigation suggested that "absolute" pitch discrimination improves as a result of experience without informative feedback or knowledge-of-results ( K R ) and that K R training results in additional improvement. Rather marked differences across individuals were also observed. The experimental hypotheses suggested by the preliminary investigation are: (1) "absolute" pitch discrimination improves as a function of experience without K R , ( 2 ) once the improvement stated in (1) has taken place, K R training will result in further improvement, and ( 3 ) performance differs across Ss. Ss were 22 introductory psychology students. The signal and response arrays were designed in a manner such that they each contained an ordinal structure and could be encoded easily by S. The stimuli were 18 simple sine-wave tones arranged in three scales with frequencies corresponding to the musical notes A, B, C, E, P , and G. The responses were coded alpha-numerical symbols. Thus A1 indicated a tone with a frequency of 220 Hertz, A 2 indicated 880 Hertz, and A3 indicated 3520 Hertz. K R was not available during the first four sessions composed of a total of 756 test tones plus 72 warm-up tones. An a priori performance stability criterion was reached by the end of the fourth session calling for implementation of K R training for the fifth session. During the K R condition, E correctly identified each tone after S had responded to it. K R was not available for Session 6; thus performance on Sessions 4 and 6 could be compared in a pre-treaunent vs post-treatment manner to test the effects of K R training. Sessions 5 and 6 were each composed of 216 test tones plus 1 8 warm-up tones. Bits of information transmitted was used (mere ly ) as the criterion measure; and appropriate non-parametric statistical tests were conducted to test the experimental hypotheses. The results supported all of the experimental hypotheses. Improvement was from 2.17 bits in Session 1 to 2.38 bits in Session 6. The results did not indicate that K R training had been continued long enough to produce a maximum effect; thus more training might have resulted in further improvement. Further research into the effects of more extensive and intensive K R training, and possibly other training techniques, is indicated.