An invisibly small thumb-contraction was conditioned under secondary positive reinforcement (money) in four adult human subjects without their observation of the response. Electromyo-graphic detection enabled the experimenter to reinforce the response by advancing on the subject's illuminated scoreboard the count of nickels earned. A light-beam galvanometer recorded on photosensitive paper not only those instances of the response which were of the size pre-selected for reinforcement but also those too small or too large to qualify. From the developed record cumulative response curves were constructed for each of the variously sized subclasses of the operant. Histograms, too, were plotted showing response-frequency by subclass for each 10-min interval of the experimental session. Before conditioning, response frequency was radically skewed toward the large-amplitude end of the distribution. The effect of conditioning was to normalize the distribution, with the middle-sized subclass (the one reinforced) becoming modal. This entailed reduced frequency of responses in subclasses smaller than the one reinforced. In extinction the original skew was strikingly restored in three of the four cases.