time (Piéron, 12, Tarchanoff, 14), sometimes by absence or reduced extent of movement (Rosenbach, 13, Lombard, II, Bowditch and Warren, 1, Lee and Kleitmann, 10), and sometimes by a rise in the threshold of excitation (Tarchanoff, Piéron). In contrast, however, to the diminution of excitability which appears with sleep through the greater part of its course, Rosenbach found a temporary stage in which irritability was increased; this might appear just before the subject went to sleep and persist for a few minutes while his sleep was still light. Piéron reported a similar temporary increase just before waking. The similarity between the results found with sleep and those of the present experiment grows even closer when one recalls that in both cases muscular relaxation and lowered reflex irritability are accompanied by reduced sensitivity to sensory stimuli.

A current theory in regard to sleep is that this raising of the threshold for both reflexes and sensation is due to the shutting off of incoming sensory impulses; some of the excitations shut off are those from external stimuli, some from the muscles. These last are reduced by the process of relaxation. This theory applies equally well to the present experiment. In this case external stimuli, as well as afferent impulses from the muscles, were reduced with the condition of relaxation; for the subject relaxed with eyes closed, and the room was quiet except for the monotonous sound of the motor. An interesting question for further experiment is how much of this reduction of muscle contraction is necessary to obtain the decrease of excitability. With sufficiently trained subjects it may be possible to determine the result of relaxing the arm alone, or the rest of the body without the arm.

The most puzzling problem which remains is that presented by the exceptional results obtained with Subject G. The fact that this subject was less systematic than the others in his method of relaxing suggests that some difference in the kind or degree of his relaxation may have been responsible for the peculiarities of his reaction. It is further suggested by the observations of Rosenbach and Piéron that there may be an early stage of relaxation where excitability is increased

rather than diminished. This suggestion is borne out by the common clinical observation that a subject must relax to some extent before the knee-jerk can be obtained. It is possible that Subject G never passed such an early stage of relaxation. On the other hand, it is strange that, if such a stage exists, exaggerated reactions were not found at some time with other subjects; for their success in relaxing certainly varied from time to time, and in spite of all precautions stimuli were sometimes given when they were not fully relaxed. Further experiment is necessary to determine the validity of this conception.

Another suggestion in regard to Subject G came from the fact that he had had long experience with electric shocks in a conditioned reflex experiment, and had acquired a peculiar dread of them. It is possible to conceive that, if the neural mechanism for this response had thus been raised to an unusual degree of excitability, it might retain its readiness to respond after other interfering or inhibitory systems had yielded to the effect of relaxation.

REFERENCES

I. BOWDITCH, H. P., & WARREN, H. C., J. Physiol., 1890, 11, 25.

2. CORIAT, I. H., J. Abn. Psychol., 1912, 6, 329.

3. JACOBSON, E., Amer. J. Psychol., 1912, 23, 345.

4. JACOBSON, E., N. Y. Med. J., 1920, 111, 397.

5. JACOBSON, E., Ill. Med. J., 1921, 39, 185.

6. JACOBSON, E., J. Nerv. & Ment. Dis., 1921, 53, 283.

7. JACOBSON, E., J. Nerv. & Ment. Dis., 1924, 60, 568.

8. JACOBSON, E., Amer. J. Psychol., 1925, 36, 73.

9. Kleitmann, N., Amer. J. Physiol., 1923, 66, 67.

10. LEE, M. A. M., & KLEITMANN, N., Amer. J. Physiol., 1923, 67, 141.

II. LOMBARD, W. P., Amer. J. Psychol., 1887, 1, 358.

12. PIERON, H., Le problème physiologique du sommeil, Paris, 1913.

13. Rosenbach, O., Zsch. f. klin. Med., 1879, 1, 358.

14. TARCHANOFF, J., Arch. ital. de biol., 1894, 21, 450.

15. WILLIAMS, R. D., Psychol. Rev. Monog. Suppl., 1914, 17 (no. 75), 55-155.

LIMITS OF LEARNING BY TRIAL AND ERROR

BY JOSEPH PETERSON

George Peabody College for Teachers

The terms and the conception of 'trial and error learning' seem to have come from Lloyd Morgan.1 Previously to the development of this conception it was not understood that the individual is a mechanism whose 'will' is dependent upon the training of an innate organization or structure. Will was usually regarded as a force or faculty by which we could order or direct our acts in advance of learning those acts. Lloyd Morgan's conception was at first applied only to animals, and even today many writers on psychological topics to say nothing of writers on popular topics—seem to regard rational, voluntary behavior as simply the regulation of activity by certain ideas somehow mysteriously picked up, as it were, from being told how to do a thing or from seeing others do it. This erroneous view is found particularly in literature bearing on ethics; but it also occurs rather extensively in educational and even in psychological literature. Learning is frequently divided into learning 'by ideas' and learning by trial and error. In many instances the former kind of learning is regarded as being free from the random excess movements characterizing the latter kind. That is to say, the early trial-and-error stage of acts, which later are well controlled, is often overlooked.

In certain skilled acts in which errors are not very obvious, learning, or the acquirement of skill, is not infrequently regarded as merely the result of repetition. On this view practice simply involves a repetition of perfectly accurate acts for the result of greater speed in the performance. Here we think of such things as addition by adults who are careful and accurate, as typewriting by one who has good control of 1 See his Introduction to comparative psychology, 1894, Chap. 12.

finger-movements and who uses the visual method so carefully that no wrong keys are hit, or as card-distribution, in which great caution is taken against errors. In such activities, with the proper care, no 'errors' are made, yet speed increases regularly with practice; and the usual formula for improvement is very simple just repeat the performance many times. Such learning as this, in which improvement seems to be only a gain in speed, and in which all gross acts can be voluntarily controlled, with the proper precautions, is by some writers regarded as qualitatively different from learning by trial and error, which displays many excess movements. Trial-and-error has therefore seemed to present a difficulty to theory that has not appeared in the other kind of learning, which we may, for the present, call rational, this difficulty being how the errors are eliminated.

Experiments have made it obvious that learning by trial and error is not accomplished simply by repetition; for this process, even if it were possible before the act is learned, would fix wrong as well as right acts. Careful tabulations. of successive acts in their actual order, moreover, have shown that the laws of frequency and recency do not account for the selection that goes on in such learning. The view, held by certain writers, that each particular act is a discrete thing, and that in learning by trial and error new acts are not acquired but that erroneous or irrelevant acts are simply dropped off, leaving finally only those which from the first were appropriate, has of late also received some hard knocks. Experiments on reflexes, like those by Sherrington and by Dodge, as well as numerous studies on more far-reaching coördinations involving attitudes or sets, have shown that in any particular act practically the entire organism is involved in a degree, and that therefore in learning there is an overlapping of impulses, a conflict among different systems of coördination. Thus for any particular consummatory re

2

1 Peterson, Jos., Frequency and recency factors in maze learning by white rats, J. Animal Behavior, 1917, 7, 344-364; Learning when frequency and recency factors are negative, J. Exper. Psychol., 1922, 5, 270–300.

2 Sherrington, C. S., The integrative action of the nervous system, 1906.

Dodge, R., Adequacy of reflex compensatory eye-movements including the effects of neural rivalry and competition, J. Exper. Psychol., 1923, 6, 169–181.

sponse there results a balancing of the different factors involved, until finally the successful expression of one system results in the suppression of others. The question as to how learning goes on has therefore taken on of late so many interesting and new aspects that the whole process of learning is coming to be regarded as a central problem in the field of psychology.

The present experiment, begun about seven years ago, was devised to explore the alleged differences between the acquirement of skill in voluntarily controlled acts on the one hand and learning by trial and error on the other. The learning is of a kind that does not, on superficial consideration, seem to allow of any errors, in the usual sense of this term; all responses are correct, but with practice they constantly increase in rapidity. The subject is given a sheet of paper on which are typewritten, in order, the letters of the alphabet, these letters being numbered in a random order from 11 to 36, as shown in Table I. He is instructed as follows:

On this sheet you will find the several letters of the alphabet each associated with a certain number, written directly under it. The problem is to see how soon you can learn to give each letter immediately when its number is called. I shall call out these numbers in a chance order and for each number given you are to speak out as quickly as possible the corresponding letter. As soon as you give your response each time another number will be called, so that you will have to keep responding to successive stimuli until a certain series has been completed. The series gone through each day will consist of 156 numbers involving each of the letters of the alphabet six times. You may keep this sheet before your eyes during the experiment, and must not respond in any case until you get the right letter. In case you make a mistake you must correct it at once, before the next number is called. With the apparatus which you see here [the operation of which was shown to S] I shall keep an accurate record of the time of each of your responses, and shall count as errors all responses which are not made within a period of 1.5 or 2 seconds after the stimulus has been given. Do the very best you can and see how soon you can go through the entire list of 156 responses with no errors. You will be told each day before practice the number of errors made in the last practice.

TABLE I

B C D E F G H 33 24 22 36 18 31 20

A3 N3

9335

32 12 26

I J K

21 30

17

13

X Y Ꮓ

19 28 23 15