In calling out the numbers the experimenter followed chance-order schedules varied so that no one sequence of the stimuli would become fixed. Thus no associations were made except those between the letters and their respective numbers. The time of each response was measured by means of a record made on the revolving kymographic drum just above a scale of sec. units registered simultaneously with the responses by means of a Jacquet chronometer. The experimenter touched a telegraph key when the syllable revealing the stimulus-number in any case was called; as twelve', elev'en, four'teen, thirty-three', etc. He released it as soon as the subject gave the correct response. This method was preferred to the use of a standard chronoscope because the experimenter was obliged to determine whether the response was right or wrong and to disregard the latter, and it was desirable to have records of rapid serial responses. Throughout the experimental series each number-stimulus was called out immediately after the subject gave the right response to the preceding number. The records were shellacked, and, when they had become dry, the number of responses taking a longer period than two seconds were counted as errors. In a second count, with the more rigid criterion, all responses taking longer than one and one-half seconds were listed as errors. Two parallel records of the same reaction-series were thus obtained. In the case of one subject the experiment was continued till all the reactions of two successive days-a total of 312-gave no errors at all even by the 1-seconds' criterion. Table II. summarizes the data and Fig. I gives the learning TABLE II ERRORS MADE (Subj. K) in the first (a) and IN THE SECOND (b) half of the pRACTICE 1.5-sec. Criterion. 2-sec. Criterion.. abababa│bababababababab 73 5661|47|55|30|30|29|26|24|17 16 10 8 17 8737300 |63|52|45|38|28|23|24|14|15 14 7 8 44 5 I O O O OO O curves for a thirteen-year-old high-school boy. The errors made in the first and in the last half of each day's practice are given separately, to display possible effects of warming up and of fatigue. Each day's practice began just after the subject was dismissed from school in the early afternoon. A study of the curves (Fig. 1) reveals a greater irregularity FIG. 1. Learning curves of Subject K, plotted from the data of Table II. Curve 2: errors by the one-and-one-half-seconds' criterion. in Curve 2, obtained by the more rigid criterion of errors, than in Curve 1, showing that the finer the analysis the greater the irregularity in the responses. That is to say, in those cases where errors are not made evident by the coarser methods of analysis, finer analysis shows that the learning still has the characteristics of the trial-and-error process of adjustment. In both curves, however, the irregularities are largely concealed by the summing of errors in groups of 78 responses. The individual kymographic records must be examined for details of fluctuation. These were unfortunately not marked in a manner permitting us to identify the individual responses to the several letters, so that, as a consequence, we cannot determine the variabilities in response to any given letter in successive presentations, as would be desirable in a thorough study. As the experimenter was seated just before the subject, carefully watching his behavior so as to record as accurately as possible the moment when the correct letter was given, an unusual opportunity was afforded to observe the nature of those responses which were more or less delayed. The subject at first kept his eyes on the sheet. When a number within the clear range of his vision was called the response was direct and quick; but for other numbers the eyes made random, exploratory movements here and there, often recurring several times to a given part of the sheet. As the localization of certain numbers became better established by the subject, these ocular movements showed less variability and they were more predictable; and they finally disappeared and became unnecessary as the association of the numbers with their respective letters reached an approximately automatic stage. Hereupon the inhibitions and checks, which the instructions to get the 'right letter' had given to the responses, practically disappeared, and the stimulus directly and automatically brought about the proper reaction of the vocal mechanism. The subject remarked on several occasions that he experienced a feeling of released tension and responsibility at this final stage of the learning. The release of these inhibitions, which were precautionary against the giving of wrong responses, was, however, very gradual, and complete automaticity could be acquired only by long practice through successive days. This phenomenon of gradual release of inhibition is just what we see also in such trial-and-error processes as the learning of ball-tossing, in which the subject is at first, and for a long period, in a marked tenseness betraying serious conflicts among the several reaction-systems, a tenseness which makes it impossible for him to carry on a conversation or easy mental calculations while tossing the balls, but which gradually disappears until, at a final stage, he can toss the balls while answering difficult questions or calculating. In the present experiment conversations after each practiceperiod yielded evidence confirmatory of the observations by the experimenter. The subject reported conflicting tendencies to look hastily about in different directions on the sheet, with impulses to go back again and again to certain parts when the required number did not readily come into view, and on several occasions he mentioned the persistence of tendencies to certain eye-movements after the association in question had become so well established that seeing the number and letter was unnecessary. At this stage, however, the movements had settled down to fairly definite responses each with its pull in a given direction. Gradually these gave way to immediate and direct vocal reactions which were stripped of all such tendencies. In connection with experiments on rational learning, similar conflicting impulses and inhibitions of one actionsystem by others have been noted. In this case A, B, C, D, E, F, G, H, I and J are to be associated, respectively, with the numbers 9, 6, 2, 10, 8, 1, 5, 4, 7 and 3. The subject has nothing before him and does not know which of the ten numbers belongs to any one letter. All this he is to determine by his own trial guesses, until he remembers the several associations. In the early stages of this test the subject is likely to be so taken up with some detailed aspect of the complex situation presented to him that he overlooks the most obvious and well-known relations. He may, for instance, remember perfectly well that 9 is the number for A, the first letter called out, and therefore avoid guessing 9 for 1 Peterson, Jos., Experiments in rational learning, Psychol. Rev., 1918, 25, 443 other letters; and yet when he comes to determine the number for, say, F, which he has had difficulty in establishing, he seems to get into so tense a state as completely to release the inhibition on 9 for other letters than A, and so gives 9, among other random guesses, for F. When asked afterwards why he did this irrational thing, when his record shows that for two or more repetitions he had rightly given 9 for A each time, he replies: "I knew A was 9, but didn't think of it at the time." When in such a state of tension and inner conflict the subject proceeds very slowly, sometimes beginning to give a letter, as 'sev' for seven, and then retracting it; sometimes speaking it out completely, then immediately saying "no!" and guessing another number. Not infrequently after guessing one or two wrong numbers in a hesitating manner, he suddenly speaks out the correct one with certainty. Occasionally the confusion becomes so great and the range of orientation so limited that the same erroneous number is repeated several times among other numbers, as 5, 3, 6, 5, 4, 5, I (the correct number). This repetition of a number found to be incorrect is what I have called perseverative response, after Hamilton, who found it characteristic of certain animals and individuals of lower intelligence. It occurs in human subjects when they are in great confusion and inner tension. As the Rational Learning test nears completion there is a very noticeable release of tension, the speed of reaction becomes accelerated, and the hesitating, retracting responses change to reactions in which the numbers flow out, as it were, evenly and uniformly, and a look of certainty comes over the subject's face as the inhibitions disappear. The experimenter readily perceives whether a correct repetition of the numbers through an entire series from A to J is accidental or due to a real mastery of the problem. If there is still hesitancy and tension the subject usually makes some error in the next repetition. Two repetitions without error are required. One subject had trouble with the letter E, in the Rational Learning test. Its number is 8. He finally overcame the difficulty sufficiently to avoid error at this point, but in his anxiety about retaining |