Finally, in Table 45 I have arranged the results of what may fairly be called the continuous training method. In connection with several of the labyrinth experiments of Chapter XIII continuous training proved very satisfactory. It therefore seemed worth while to ascertain whether the same method would not be more efficient than any other for the establishment of a white-black discrimination habit. That this method was not applied to ten individuals as were the two-five-test, the ten-test, and the twenty-test methods is due to the fact that it proved practically inadvisable to continue the tests long enough to complete the experiment.

I have usually designated the method as one hundred or more tests daily. I applied this training method first to individuals Nos. 51 and 60. At the end of one hundred and twenty tests with each of these individuals I was forced to discontinue the experiment for the day because of the approach of darkness. In the table the end of a series for the day is indicated by a heavy line. The following day Nos. 51 and 60 succeeded in acquiring a perfect habit after a few more tests.

TABLE 45

EFFICIENCY OF TRAINING. WHITE-BLACK TESTS AT THE RATE OF 100 OR MORE PER DAY

SETS 51[1] 60 87 Av.

OF 10

A 5 5 6 5.3 B 5 3 7 5.0

1 6 6 5 5.7 2 3 2 5 3.3 3 5 4 7 5.3 4 7 4 5 5.3 5 6 2 3 3.7 6 1 1 3 1.7 7 4 2 3 3.0 8 3 3 0 2.0 9 2 2 3 2.3 10 5 0 2 2.3 11 1 2 2 1.7 12 2 1 1 1.3

13 4 1 2 2.3 14 1 2 1 1.3 15 3 1 5 3.0 16 3 3 2 2.7 17 1 0 1 0.7 18 2 0 1 1.0 19 0 0 2 0.7 20 0 0 0 21 0 1 0.3 22 - 23 - 24 -

[Footnote 1: Age of No. 51, 22 weeks. Age of No. 60, 17 weeks. Age of No.

87, 8 weeks.]

The results of the continuous training method for these two mice were so strikingly different from those yielded by the other methods that I at once suspected the influence of some factor other than that of the number of tests per day. The ages of Nos. 51 and 60 at the time of their tests were twenty-two and seventeen weeks, respectively, whereas all the individuals used in connection with the other efficiency tests were four weeks of age. It seemed possible that the slow habit formation exhibited in the continuous training experiments might be due to the greater age of the mice. I therefore selected a healthy active female which was only eight weeks old, and tried to train her by the continuous training method.

With this individual, No. 87, the results were even more discouraging than those previously obtained, for she was still imperfect in her discrimination at the end of two hundred and ten tests. At that point the experiment was interrupted, and it seemed scarcely worth while to continue it further at a later date. The evidence of the extremely low efficiency of the continuous method in comparison with the other methods which we have been considering is so conclusive that further comment seems superfluous.

We are now in a position to compare the results of the several methods of training which have been applied to the dancer, and to attempt to get satisfactory quant.i.tative expressions of the efficiency of each method. I have arranged in Table 46 the general averages yielded by the four methods. Although these general results hide certain important facts which will be exhibited later, they clearly indicate that an increase in the number of tests per day does not necessarily result in an increase in the rapidity of habit formation. Should we attempt, on superficial examination, to interpret the figures of this table, we would doubtless say that in efficiency the two-five-test method stands first, the continuous-test method last, while the ten-test and twenty-test methods occupy intermediate positions.

TABLE 46

EFFICIENCY OF TRAINING

Number of Errors in White-Black Series for Different Methods of Training

SETS OF 10 2 OR 5 TESTS 10 TESTS 20 TESTS 100 OR MORE PER DAY PER DAY PER DAY TESTS PER DAY

A 5.8 5.8 5.9 5.3 B 5.7 5.7 5.9 5.0

1 6.4 5.3 4.8 5.7 2 4.2 3.5 4.8 3.3 3 3.4 3.2 3.4 5.3 4 3.1 2.5 3.0 5.3 5 2.7 2.5 1.9 3.7 6 1.5 1.3 1.9 1.7 7 0.9 1.5 0.9 3.0 8 0.7 0.8 1.1 2.0 9 0.5 0.4 0.9 2.3 10 0 0.5 0.8 2.3 11 0 0.4 0.4 1.7 12 0 0.2 0.3 1.3 13 0 0 2.3 14 0 0 1.3 15 0 0 3.0 16 2.7 17 0.7 18 1.0 19 0.7 20 0

We may now apply to the results of our efficiency-of-training tables the method of measuring efficiency which was mentioned at the end of the preceding chapter as the _index of modifiability (that number of tests after which no errors occur for at least thirty tests)_. By taking the average number of tests for the several individuals in each of the Tables 42, 43, 44, and 45 we obtain the following expressions of efficiency:--

METHOD INDEX OF MODIFIABILITY (EFFICIENCY)

Two-five-test 81.7 2.7 Ten-test 88.0 4.1 Twenty-test 91.0 5.3 Continuous-test 170.0 4.8

Since the difference between the indices for the ten-test and the twenty- test methods lies within the limits of their probable errors (4.1 and 5.3) it is evident that it is not significant. Except for this, I think these indices may be accepted as indications of real differences in the value of the several methods of training.

A somewhat different interpretation of our results is suggested by the grouping of individuals according to s.e.x. In Table 47 appear the general averages for the males and the females which were tested by the several methods. The most striking fact exhibited by this table is that of the high efficiency of the twenty-test method for the females. Apparently they profited much more quickly by this method than by the ten-test method, whereas just the reverse is true of the males. I present the data of this table merely to show that general averages may hide important facts.

TABLE 47

EFFICIENCY OF TRAINING

CONDITION MALES FEMALES INDEX OF MODIFIABILITY INDEX OF MODIFIABILITY 2 or 5 tests per day 85.0 80.0 10 tests per day 72.0 104.0 20 tests per day 94.0 88.0 100 or more tests per day 160.0 180.0

From all considerations that have been mentioned thus far the reader would be justified in concluding that I made a mistake in selecting the ten-test method for my study of the modifiability of the behavior of the dancer.

That this conclusion is not correct is due to the time factor in the experiments. If the dancer could acquire a perfect habit as a result of twelve days' training, no matter whether two, five, ten, or twenty tests were given daily, it would, of course, be economical of time for the experimenter to employ the two-test method. But if, on the contrary, the two-test method required twice as many days' training as the five-test method, it would be economical for him to use the five-test method despite the fact that he would have to give a larger number of tests than the two- test method would have demanded. In a word, the time which the work requires depends upon the number of series which have to be given, as well as upon the number of tests in each series. As it happens, the ten-test method demands less of the experimenter's time than do methods with fewer tests per day. The twenty-test method is even more economical of time, but it has a fatal defect. It is at times too tiresome for both mouse and man.

These facts indicate that a balance should be struck between number of tests and number of series. The fewer the tests per day, within the limits of two and one hundred, the higher the efficiency of the method of training, as measured in terms of the total number of tests necessary for the establishment of a perfect habit, and the lower its efficiency as measured in terms of the number of series given. The greater the number of tests per day, on the other hand, the higher the efficiency of the method in terms of the number of series, and the lower its efficiency in terms of the total number of tests. By taking into account these facts, together with the fact of fatigue, we are led to the conclusion that ten tests per day is the most satisfactory number.

If my time and attention had not been fully occupied with other problems, I should have determined the efficiency of various methods of training in terms of the duration of habit, as well as in terms of the rapidity of its formation. As these two measures of efficiency might give contradictory results, it is obvious that a training method cannot be fairly evaluated without consideration of both the rapidity of habit formation and the permanency of the habit. A _priori_ it seems not improbable that slowness of learning should be directly correlated with a high degree of permanency. By the further application of the method which I have used in this study of the efficiency of training we may hope to get a definite answer to this and many other questions concerning the nature of the educative process and the conditions which influence it.

CHAPTER XVI

THE DURATION OF HABITS: MEMORY AND RE-LEARNING

The effects of training gradually disappear. Habits wane with disuse. In the dancer, it is not possible to establish with certainty the existence of memory in the introspective psychological sense; but it is possible to measure the efficiency of the training to which the animal is subjected, and the degree of permanency of habits. The materials which const.i.tute this chapter concern the persistence of unused habits, and the influence of previous training on the re-acquisition of a habit which has been lost or on the acquisition of a new habit. For convenience of description, I shall refer to certain of the facts which are to be discussed as facts of memory, with the clear understanding that consciousness is not necessarily implied. By memory, wherever it occurs in this book, I mean the ability of the dancer to retain the power of adaptive action which it has acquired through training.

I first discovered memory in the dancer, although there was previously no reason for doubting its existence, in connection with the ladder-climbing tests of Chapter XII. In this experiment two individuals which had perfectly learned to escape from the experiment box to the nest-box by way of the wire ladder, when tested after an interval of two weeks, during which they had remained in the nest-box without opportunity to exercise their newly acquired habit, demonstrated their memory of the method of escape by returning to the nest-box by way of the ladder as soon as they were given opportunity to do so. As it did not lend itself readily to quant.i.tative study, no attempts were made to measure the duration of this particular habit. At best the climbing of a wire ladder is of very uncertain value as an indication of the influence of training.

Similarly, the persistence of habits has been forced upon my attention day after day in my various experiments with the mice. It is obvious, then, that the simple fact of memory is well established, and that we may turn at once to an examination of the facts revealed by special memory and re- learning experiments.

The visual discrimination method, which proved invaluable as a means of measuring the rapidity of habit formation, proved equally serviceable in the measurement of the permanency or duration of habits. Memory tests for discrimination habits were made as follows. After a dancer had been trained in the discrimination box so that it could choose the correct electric-box, white, red, blue, or green as it might be, in three successive daily series of ten tests each, it was permitted to remain for a certain length of time without training and without opportunity to exercise its habit of visual discrimination and choice. At the expiration of the rest interval, as we may designate the period during which the habit was not in use, the mouse was placed in the discrimination box under precisely the same conditions in which it had been trained and was given a series of ten memory tests with the box to be chosen alternately on the right and on the left. In order that the entire series of ten tests, and sometimes two such series given on consecutive days, might be available as indications of the duration of a habit, the mouse was permitted to enter and pa.s.s through either of the electric-boxes without receiving a shock.

Had the shock been given as punishment for a wrong choice, it is obvious that only the first test of the memory series would be of value as an indication of the existence of a previously acquired habit. Even under the conditions of no shock and no stop or hindrance the first test of each memory series is of preeminent importance, for the mouse tends to persist in choosing either the side or the visual condition (sometimes one, sometimes the other) which it chooses in the first test. If the wrong box is chosen to begin with, mistakes are likely to continue because of the lack of punishment; in this case the animal discriminates, but there is no evidence that it remembers the right box. Likewise, if the right electric- box is chosen in the first test, correct choices may continue simply because the animal has discovered that it can safely enter that particular box; again, the animal discriminates without depending necessarily upon its earlier experience. I have occasionally observed a series of ten correct choices, made on the basis of an accidental right start, followed by another series in which almost every choice was wrong, because the animal happened to start wrong.

As the results of my tests of memory are of such a nature that they cannot advantageously be averaged, I have arranged in Table 48 a number of typical measurements of the duration of visual discrimination habits. In this table I have indicated the number and age of the individual tested, the habit of discrimination which had been acquired, the length of the rest interval, the result of the first test (right or wrong), and the number of errors made in each series of ten memory tests.

TABLE 48

MEASUREMENTS OF THE DURATION OF A HABIT

Memory

ERRORS NO. AGE NAME OF TEST REST FIRST FIRST SECOND INTERVAL CHOICE SERIES SERIES