consumed in repeating the F-lists than in repeating the S-lists or the B-lists, but the same time was spent in repeating the S-lists and the B-lists.

Results of individual subjects.-Although the summaries given in Tables V, VI, VII, and VIII seem to support the theory of backward association, these results are 4 sets of average results for 4 groups of subjects, and they tell little about the performance of the individuals. Comparisons have therefore been made for each of the 31 subjects. Table X shows a sample result.

TABLE X

INDIVIDUAL RESULTS FOR SUBJECT O.F.O., WHO USED 27 LISTS OF 9 FAMILIAR WORDS

The F-lists are superior to the S-lists or the B-lists for all subjects. Table XI shows the number of subjects who gave results in favor of backward associations, and the number of subjects who gave the opposite results. If we consider Table XI as a whole, it appears that, in the procedure which we used, backward associations are formed in about 2 cases out of 3. There is a possibility of course that "chance" factors may account for the one individual out of 3 who gives results which are opposed to the theory of backward association,—and the theory may still be true as a general principle. On the other hand, backward associations may be formed in one individual

who uses one method of learning, but not in another individual who uses a different method of learning. The latter alternative seems to be more strongly supported.

TABLE XI

NUMBER OF SUBJECTS WHO GAVE Results in Favor of, OR OPPOSED TO, BACKWARD

While the subjects were learning the practice lists, a definite effort was made to limit the learning process to the formation of associations in the forward direction, so that the conditions of the experiment would be suitable for testing the theory of backward association. When nonsense syllables and disconnected words are used as material, however, it seems to be practically impossible to limit the learning process in this strict manner. The following irregularities and difficulties were observed while the subjects were engaged in the rote learning of the practice lists.5

We attempted to observe the learning more carefully by requiring the subject to repeat the items aloud; but the subject's thoughts do not coincide with the oral pronunciation of the items. We also tried to follow the process more closely by observing eye-movements; but when the subject is looking at one item and pronouncing it he is frequently apprehending some other. The fluctuations and irregularities which occur are apparently beyond control." In spite of the most rigid

"We have pointed out some of these irregularities in a previous article, The concept of backward association, Amer. J. Psychol., 1924, 35, 217–221.

instructions, the subject also makes mistakes in pronouncing the syllables or words. When these mistakes occur, there is a tendency for the subject's eyes and thought processes to shift backwards in the list. It is important to note that these fluctuations constitute practice in the backward direction.

Further irregularities occur in the recitation. We observed a marked tendency for the subject to look away from the printed material from time to time and to attempt a private recitation. The items actually reproduced in these recitations are of course not always arranged in the proper order. The subject's own incorrect order frequently contains pairs and groups of items which are arranged in the backward direction. Further opportunity for practice in the backward direction is afforded when the subject makes his final recitation for the experimenter; and in such cases it is quite impossible to keep the subject from thinking rather vividly about his own incorrect arrangement of the items.

When the unavoidable irregularities occur, the associations which are established in the order 5-4-3-2-1 are not the result of the formation of associations in the order 1-2-3-4-5, but simply the result of forward learning in the order 5-4-3-2-1. The irregular associations formed while learning the practice lists also have the definite advantage of recency when the subject passes on to the test lists a few minutes later.

The irregularities just referred to affect the B-lists considerably, since the items in each B-list are adjacent to each other and are printed on the same card, when compared with the previous arrangement of the same items in the practice lists. These irregularities, however, do not affect the S-lists, because each of the items in a single S-list was previously a part of a different practice list, and the subject held in his hand and learned only one practice list at a time. If we consider the arrangement of all of the items in the practice lists, then the items in each S-list were previously separated from each other by about 14, 17, 20, or 26 intervening items, and they were also arranged in the backward direction. Therefore, since the learning of the B-lists begins above zero, on account of the faulty control, and since the learning of the

S-list begins at zero, the results given above are prejudiced in favor of the theory of backward association, and will have to be discounted to some extent.8

This interpretation seems to harmonize with some general tendencies which may be noted in Tables V, VI, VII, and VIII. It would be expected from the preceding discussion that an increase in the overlearning of the practice lists would be accompanied by greater reinforcing effect in the B-lists. It can be observed that as one passes from Table VI to Table VIII, there is an increase in the superiority of the F-lists and the B-lists over the S-lists. It would also follow from our discussion that a relatively larger number of irregularities and fluctuations should occur in learning long lists than in learning short lists; and it may be noted that the reliability of the results in favor of backward association increases as one passes from the short lists of Table V to the long lists of Table VIII. It is more difficult to control the learning process with long lists, and they generally require relatively more learning for one correct recitation.

We have shown that the learning of rote material under ordinary laboratory conditions results in the formation of backward associations, and the theory is therefore true as a practical proposition under these conditions. However, the irregularities and fluctuations which occur make it very doubtful whether the theory will apply to all learning situations and to all types of material. The question remains as to whether it is possible to control the conditions of learning in such a way that no backward associations will be formed. From a more practical point of view, it is also very desirable to know whether learning the more common meaningful material of everyday life results in the formation of backward associations. An attempt is made to answer these questions in Part II of the present paper.

It should be observed that the interpretation of the results which we have given also applies in larger measure to the results obtained by Ebbinghaus, op. cit., pp. 90-117.

PART II. TESTS WITH LOGICAL MATERIAL

INTRODUCTION

The experiment here to be described was a natural outgrowth of the difficulties and irregularities which were noted in Part I. Although both experiments are concerned with the same problem, the type of material and the procedure used in the two cases are different. The practice lists in the first experiment were learned to the point of one correct recitation on two different occasions; the tests with the practice lists were made a few minutes later, and the remote associations had the definite advantage of recency. The logical material in the present experiment was greatly overlearned, and suitable tests were given to determine whether this additional practice had resulted in the formation of a greater number of backward associations. The logical material was also retained over a period of several days, and tests were given to determine whether the remote associations had been forgotten, and whether the effect produced was a facilitation or possibly even an inhibition of backward associations. In contrast with the rote material and the laboratory situation of the first experiment, the subjects in the present experiment were simply given a paragraph of logical prose material and were instructed to take it home and memorize it at their leisure. When the subject is left to his own devices in this way, the conditions of learning are fairly natural and are similar to the common learning situations of everyday life, and some of the criticisms of the laboratory experiments in learning nonsense syllables are avoided.

MATERIAL

9

The prose passage selected is an adaptation from James' chapter on habit (Principles of Psychology, Vol. I, 127). It is quoted in full as follows.

'We were partly guided in the selection of the logical prose material by the factors of natural interest and literary style. The general appropriateness for the occasion was also considered, because at the time the experiment was carried out the subjects were taking the writer's summer course in elementary psychology, in which the topic of instinct had just been discussed.