The Dancing Mouse: A Study in Animal Behavior

by Robert Mearns Yerkes

CHAPTER XIII

HABIT FORMATION: THE LABYRINTH HABIT

The problem method, of which the ladder and door-opening tests of the preceding chapter are examples, has yielded interesting results concerning the individual initiative, ingenuity, motor ability, and ways of learning of the dancer; but it has not furnished us with accurate measurements of the rapidity of learning or of the permanency of the effects of training. In this chapter I shall therefore present the results of labyrinth experiments which were planned as means of measuring the intelligence of the dancer.

The four labyrinths which have been used in the investigation may be designated as A, B, C, and D. They differ from one another in the character of their errors, as well as in the number of wrong choices of a path which the animal might make on its way from entrance to exit. In the use of the labyrinth method, as in the case of the discrimination method of earlier chapters, the steps by which a satisfactory form of labyrinth for testing the dancer was discovered are quite as interesting and important for those who have an intelligent appreciation of the problems and methods of animal psychology as are the particular results which were obtained. For this reason, I shall describe the various forms of labyrinth in the order in which they were used, whether they proved satisfactory or not. At the outset of this part of my investigation, it was my purpose to compare directly the capacity for habit formation in the dancer with that of the common mouse. This proved impracticable because the same labyrinth is not suited to the motor tendencies of both kinds of mice.

[Illustration: FIGURE 25.—Labyrinth A. I. entrance; O, exit; 1, 2, 3, 4, blind alleys.]

The first of the four labyrinths, A, appears in ground plan in Figure 25. It was constructed of wood, as were the other labyrinths also, and measured 60 cm. in length and width, and 10 cm. in depth. The outside alleys were 5 cm. wide. In the figure, I marks the starting point or entrance to the maze, and O the exit through which the mouse was permitted to pass into its nest-box. Any turn in the wrong direction which the animal made in its progress from entrance to exit was recorded as an error. The four errors, exclusive of the mistake of turning back, which were possible in this labyrinth, are indicated in the figure by the numerals 1, 2, 3, and 4. By retracing its steps a mouse might repeat any one or all of these errors, and add to them the error of turning back.

In the experiments a mouse was permitted to enter the maze from a small box which had been placed by the experimenter at I, and an accurate record was kept of the number of errors which it made in finding its way from entrance to exit, and of the time occupied. Each of five dancers was given 31 tests in this labyrinth. The number of tests per day varied, as is indicated in Table 36, from 1 to 4. The results of the tests, so far as errors and times are in question, appear in the table. T at the head of a column is an abbreviation for time, E for errors.

The dancers did not learn to escape from this labyrinth easily and quickly. In fact, the average time of the thirty-first test (198") is considerably longer than that of the first (130"). The number of errors decreased, it is true, but even for the last test it was 6.6 as compared with only a little more than twice that number for the first test. The last column of the table furnishes convincing proof of the truth of the statement that the animals did not acquire a perfect labyrinth-A habit. Was this due to inability to learn so complex a path, or to the fact that the method is not adapted to their nature? Observation of the behavior of the mice in the experiments enables me to say with certainty that there was no motive for escape sufficiently strong to establish a habit of following the direct path. Often, especially after a few experiences in the maze, a dancer would wander back and forth in the alleys and central courts, dancing much of the time and apparently exploring its surroundings instead of persistently trying to escape. This behavior, and the time and error results of the accompanying table, lead me to conclude that the labyrinth method, as it has been employed in the study of the intelligence of several other mammals, is not a satisfactory test of the ability of the dancer to profit by experience. That the fault is not in the labyrinth itself is proved by the results which I obtained with common mice.

TABLE 36

RESULTS OF LABYRINTH A TESTS WITH DANCERS

AVERAGE
TEST DATE No. 1000 No. 2 No. 6 No. 4 No. 5 FOR ALL
1905
T E T E T E T E T E T E

1 Nov 23 130" 14 100" 8 170" 13 60" 6 190" 26 130" 13.4
2 24 140 19 78 7 60 8 149 6 211 25 128 13.0
3 25 392 31 87 1 98 5 185 13 120 9 176 11.8
4 26 448 38 38 3 47 2 50 3 121 12 141 11.3
5 27 142 8 21 2 27 3 27 2 17 1 47 3.2
6 28 45 2 61 7 63 5 102 8 33 4 61 5.2
7 29 303 17 64 7 36 3 42 2 57 4 100 6.6
8 30 222 15 26 2 37 5 42 3 7 0 67 5.0
9 Dec 1 185 9 36 5 48 3 63 3 94 8 85 5.6
10 2 52 2 71 4 19 0 196 5 95 11 87 4.4
11 3 180 8 32 2 107 4 52 3 38 4 82 4.2
12 4 310 10 133 11 65 3 242 6 125 6 175 7.2
13 4 153 9 335 55 130 10 195 15 154 18 193 21.4
14 5 330 7 69 2 42 2 201 6 130 10 154 5.4
15 5 287 7 34 4 61 4 136 7 25 2 109 4.8
16 5 455 15 65 4 25 0 110 8 160 15 183 8.4
17 6 120 15 280 9 33 0 168 4 39 2 128 6.0
18 6 120 4 164 10 81 4 101 5 85 4 110 5.4
19 6 132 12 78 7 110 6 40 2 151 12 102 7.8
20 7 258 10 223 16 33 1 92 5 37 1 129 6.6
21 7 110 7 23 3 44 4 20 4 305 23 100 8.2
22 7 100 4 60 8 167 15 44 7 58 4 86 7.6
23 8 43 1 179 7 356 6 34 3 65 3 135 4.0
24 8 92 5 56 5 42 3 17 1 23 1 46 3.0
25 9 85 5 114 3 62 3 129 8 31 0 84 3.8
26 9 30 2 36 4 109 15 12 1 34 2 44 4.8
27 9 69 5 40 4 85 6 36 3 16 1 49 3.8
28 10 169 7 80 3 28 0 142 5 35 2 89 3.4
29 10 155 5 266 8 91 5 27 0 37 2 115 4.0
30 10 29 1 25 2 124 14 83 6 111 12 74 7.0
31 10 465 6 208 8 95 3 65 3 159 13 198 6.6

On the basis of two tests per day, two common mice, a white one and a gray one, quickly learned to escape from labyrinth A by the shortest path. The time of escape for the gray individual (Table 37) decreased from 180" in the first test to 21" in the tenth, and the number of errors from 6 to 1. Similarly in the case of the white individual, the time decreased from 122" to 8", and the errors from 5 to 1. A fraction of the number of tests to which the dancer had been subjected sufficed to establish a habit of escape in the common mouse. It is evident, therefore, that the dancer differs radically from the common mouse in its behavior in a maze, and it is also clear that the labyrinth method, if it is to be used to advantage, must be adapted to the motor tendencies of the animal which is to be tested.

TABLE 37

RESULTS OF LABYRINTH A TESTS WITH COMMON MICE

GREY MOUSE WHITE MOUSE TEST T E T E

1 180" 6 122" 5
2 26 2 80 6
3 37 1 56 4
4 18 0 27 1
5 68 2 33 2
6 10 1 19 1
7 11 1 17 1
8 13 1 17 1
9 10 0 8 1
10 21 1 8 1

The behavior of the dancer made obvious two defects in labyrinth A. Its passages are so large that the mouse is constantly tempted to dance, and it lacks the basis for a strong and constant motive of escape by the direct path. To obviate these shortcomings labyrinth B was constructed, as is shown in Figures 23 and 24, with very narrow passages, and a floor which was covered with the wires of an interrupted electric circuit so that errors might be punished. The length of this labyrinth was 52 cm. and the passages were 2.5 cm. wide and 10 cm. deep. Dancing in these narrow alleys was practically impossible, for the mice could barely turn around in them. In the case of all except the common mice and two dancers, a depth of 10 cm. was sufficient to keep the animals in the maze without the use of a cover.

As an account of the behavior of the dancer in labyrinth B has already been given in Chapter XI, I may now state the general results of the experiments. In all, thirty individuals were trained in this labyrinth. Each individual was given tests at the rate of one per minute until it had succeeded in following the correct path five times in succession. The weak electric shock, which was given as a punishment for mistakes, provided an activity-impelling motive for escape to the nest-box.

An idea of the extreme individual difference in the rapidity with which the labyrinth-B path was learned by these dancers may be obtained by an examination of Table 38, from which it appears that the smallest number of training tests necessary for a successful or errorless trip through the maze was one and the largest number fourteen. It is to be remembered that each mouse was given an opportunity to pass through the labyrinth once without punishment for errors, and thus to discover, before the training tests were begun, that a way of escape existed. This first test we may designate as the preliminary trial. Table 38 further indicates that the females acquired the labyrinth habit more quickly than did the males.

TABLE 38

RESULTS OF LABYRINTH-B EXPERIMENTS, WITH TWENTY DANCERS

MALES FEMALES

NO. OF NO. OF FIRST NO. OF LAST OF NO. OF NO. OF FIRST NO. OF LAST OF MOUSE CORRECT FIVE CORRECT MOUSE CORRECT FIVE CORRECT TEST TESTS TEST TESTS

76 8 14 75 4 15
78 5 20 77 7 11
86 13 22 87 12 22
58 2 14 49 1 5
50 6 23 57 3 20
60 13 37 59 14 28
410 6 20 415 4 13
220 4 8 225 6 18
212 3 7 211 6 10
214 10 28 213 5 14

AV. 7.0 19.3 AV. 6.2 15.6

A graphic representation of certain of the important features of the process of formation of the labyrinth-B habit is furnished by Figure 26 in which the solid line is the curve of learning for the ten males of Table 38, and the broken line for the ten females. These two curves were plotted from the number of errors made in the preliminary trial (P in the figure) and in each of the subsequent tests up to the sixteenth. In the case of both the males and the females, for example, the average number of errors in the preliminary trial was 11.3, as is indicated by the fact that the curves start at a point whose value is given in the left margin as 11.3. In the second training test the number of errors fell to 3.3 for the males and 2.7 for the females. The number of the test is to be found on the base line; the number of errors in the left margin. If these two curves of learning were carried to their completion, that for the males would end with the thirty-seventh test, and that for the females with the twenty- eighth.

[Illustration: FIGURE 26.—Curves of habit formation, plotted from the data of labyrinth-B tests with ten males and ten females. The figures in the left margin indicate the number of errors; those below the base line the number of the test. P designates the preliminary test. Males ____[solid line]; Females ——[broken line].]

Time records are not reported for these and subsequent labyrinth tests because they proved to be almost valueless as measures of the rapidity of habit formation. At any point in its progress through a labyrinth, the dancer may suddenly stop to wash its face, look about or otherwise examine its surroundings; if a shock be given to hurry it along it may be surprised into an error. It is my experience, and this is true of other animals as well as of the dancing mouse, that a long trip, as measured in time units, does not necessarily indicate the lack of ability to follow the labyrinth path correctly and rapidly. Hence, whenever it is possible (and the experimenter can always plan his tests so that it shall be possible), the number of errors should be given first importance and the time of the test second place. I have presented in Table 38 the number of the first correct test, and the number of the last of five successive correct tests. Space cannot be spared for records of the errors made in the several tests by each individual.

In general, labyrinth B proved very satisfactory as a means of testing the ability of the dancer to learn a simple path. The narrow passages effectively prevented dancing, and the introduction of the electric shock as a punishment for mistakes developed a motive for escape which was uniform, constant, and so strong that the animals clearly did their best to escape from the labyrinth quickly and without errors. This maze was so simple that it did not tend to discourage them as did the one which is next to be described. It must be admitted, however, that, though labyrinth B is perfectly satisfactory as a test of the dancer's ability to learn to follow a simple path, it is not an ideal means of measuring the rapidity of habit formation. This is due to the fact that the preliminary trial and the first training test play extremely different roles in the case of different individuals. A dancer which happens to follow the correct path from entrance to exit in the preliminary trial may continue to do so, with only an occasional error, during several of the early training tests, and it may therefore fail for a considerable time to discover that there are errors which should be avoided. The learning process is delayed by its accidental success. On the other hand, an individual which happens to make many mistakes to begin with immediately attempts to avoid the points in the maze at which it receives the electric shock. I was led to conclude, as a result of the labyrinth-B experiments, that the path was too easy, and that a more complex labyrinth would, in all probability, furnish a more satisfactory means of measuring the rapidity of habit formation.

[Illustration: FIGURE 27—A record sheet, showing the plan of labyrinth C (as made on the sheet by means of a rubber stamp) on which the experimenter recorded the path followed by the mouse. This sample sheet presents the path records for the first, fifth, tenth, and eleventh tests of No. 2 in labyrinth C. 1, 2, 3, 4, 5 designate the several errors of the labyrinth.]

On the basis of the supposition that a maze whose path was so complex that the animal would not be likely to follow it correctly in the early trials would be more to the purpose than either A or B, labyrinth C was devised. As is shown in the plan of this maze, Figure 27, five mistakes in choice of path were possible on the forward trip. These errors, as a rule, were more difficult for the dancers to avoid than those of labyrinths A and B. Those which are designated by the numerals 2, 3, and 4 were especially difficult. Error 4 was much more troublesome for left whirlers than for right whirlers because, after turning around abruptly at the entrance to the blind alley, the former type of dancer almost always followed the side wall of the maze so far that it missed the correct path. Undoubtedly the various errors are not of the same value for different individuals; but it would be extremely difficult, if not impossible, to devise a maze which should be equally difficult for several normal individuals.

In order that records of the path followed by a mouse in test after test might be kept with ease and accuracy by the experimenter, the plan of this labyrinth, and also that of labyrinth D, were cast in rubber. The outlines of labyrinths C and D which appear in Figures 27 and 28 respectively were made with the rubber stamps which were thus obtained. Figure 27 is the reproduction of a record sheet which presents the results of the first, the fifth, the tenth, and the eleventh tests of No. 2 in labyrinth C. The path followed by this individual in the first test was far too complex to be traced accurately on the record sheet. The record therefore represents merely the number of errors which was made in each region of the maze. For the fifth test, and again for the tenth and the eleventh, the path was recorded accurately. This simple device for making record blanks which can readily be filled in at the time of the experiment should recommend itself to all students of animal behavior.

In labyrinth C ten pairs of dancers were given continuous training tests at the rate of one test per minute until they were able to follow the direct path correctly. Because of the difficulty in learning this maze perfectly, it was not demanded of the mice that they should follow the path correctly several times in succession, but instead the training was terminated after the first successful trip.

TABLE 39

RESULTS OF LABYRINTH-C EXPERIMENTS, WITH TWENTY DANCERS

MALES FEMALES

NO. OF NO. OF FIRST NO. OF NO. OF FIRST MOUSE CORRECT TEST MOUSE CORRECT TEST

2 11 29 15 30 33 49 34 50 49 57 15 52 22 59 15 58 16 215 10 60 17 415 10 76 3 75 8 78 6 77 11 86 5 87 9 88 25 85 11

AV. 18.7 AV. 13.8

The results of the experiments with this labyrinth as they are presented in Table 39 indicate that its path is considerably more difficult for the dancer to learn than that of labyrinth B, that the females learn more quickly than the males, and finally, that individual differences are just as marked as they were in the case of the simpler forms of labyrinth. It therefore appears that increasing the complexity of a labyrinth does not, as I had supposed it might, diminish the variability of the results. Certain of the individual differences which appear in Table 39 are due, however, to the fact that in some cases training in labyrinth B had preceded training in labyrinth C, whereas in the other cases C was the first labyrinth in which the animals were tested. But even this does not serve to account for the wide divergence of the results given by No. 2 and No. 50, for the latter had been trained in B previous to his training in C, and the former had not been so trained. Yet, despite the advantage which previous labyrinth experience gave No. 50, he did not learn the path of C as well in fifty tests as No. 2 did in eleven. The facts concerning the value of training in one form of labyrinth for the learning of another, as they were revealed by these experiments, may more fittingly be discussed in a later chapter in connection with the facts of memory and re-learning.

[Illustration: FIGURE 28.—Plan of Labyrinth D, as reproduced from a print made with a rubber stamp. I, entrance; O, exit; numerals 1 to 13, errors.]

Labyrinth C is a type of maze which might properly be described as irregular, since the several possible errors are extremely different in nature. In view of the results which this labyrinth yielded, it seemed important that the dancer be tested in a perfectly regular maze of the labyrinth-D type. The plan which I designed as a regular labyrinth has been reproduced, from a rubber stamp print, in Figure 28. As is true also of the mazes previously described, it provides four kinds of possible mistakes: namely, by turning to the left (errors 1, 5, 9, and 13), by turning to the right (errors 3, 7, and 11), by moving straight ahead (errors 2, 4, 6, 8, 10, and 12), and by turning back and retracing the path just followed. The formula for the correct path of D is simple in the extreme, in spite of the large number of mistakes which are possible, for it is merely "a turn to the right at the entrance, to the left at the first doorway, and thereafter alternately to the right and to the left until the exit is reached." This concise description would enable a man to find his way out of such a maze with ease. Labyrinth D had been constructed with an exit at 10 so that it might be used as a nine-error maze if the experimenter saw fit, or as a thirteen-error maze by the closing of the opening at 10. In the experiments which are now to be described only the latter form was used.

Can the dancer learn a regular labyrinth path more quickly than an irregular one? Again, I may give only a brief statement of results. Each of the twenty dancers, of Table 40, which were trained in labyrinth D had previously been given opportunity to learn the path of C, and most of them had been trained also in labyrinth B. All of them learned this regular path with surprising rapidity. The numerical results of the tests with labyrinths B, C, and D, as well as the behavior of the mice in these several mazes, prove conclusively that the nature of the errors is far more important than their number. Labyrinth D with its thirteen chances of error on the forward trip was not nearly as difficult for the dancer to learn to escape from as labyrinth C with its five errors. That the facility with which the twenty individuals whose records are given in Table 40 learned the path of D was not due to their previous labyrinth experience rather than to the regularity of the maze is proved by the results which I obtained by testing in D individuals which were new to labyrinth experiments. Even in this case, the number of tests necessary for a successful trip was seldom greater than ten. If further evidence of the ease with which a regular labyrinth path may be followed by the dancer were desired, it might be obtained by observation of the behavior of an individual in labyrinths C and D. In the former, even after it has learned the path perfectly, the mouse hesitates at the doorways from time to time as if uncertain whether to turn to one side or go forward; in the latter there is seldom any hesitation at the turning points. The irregular labyrinth is followed carefully, as by choice of the path from point to point; the regular labyrinth is followed in machine fashion,—once started, the animal dashes through it.

TABLE 40

RESULTS OF LABYRINTH-D EXPERIMENTS, WITH TWENTY DANCERS

MALES FEMALES

NO. OF NO. OF FIRST NO. OF LAST OF NO. OF NO. OF FIRST NO. OF LAST OF MOUSE CORRECT TWO CORRECT MOUSE CORRECT TWO CORRECT TEST TESTS TEST TESTS

2 3 7 29 10 11
58 7 10 49 7 8
30 9 10 57 3 6
60 10 14 215 6 10
402 10 11 415 7 8
76 4 7 75 4 13
78 4 5 77 11 12
86 3 9 87 4 9
88 4 8 85 3 4
90 7 8 83 4 7

Av. 6.1 8.9 Av. 5.9 8.8

From the results of these labyrinth experiments with dancers I am led to conclude that a standard maze for testing the modifiability of behavior of different kinds of animals should be constructed in conformity with the following suggestions. Errors by turning to the right, to the left, and by moving forward should occur with equal frequency, and in such order that no particular kind of error occurs repeatedly in succession. If we should designate these three types of mistake by the letters r, l, and s respectively, the error series of labyrinth C would read l-l-r-s-l. It therefore violates the rule of construction which I have just formulated. In the case of labyrinth D the series would read l-s-r-s-l-s-r-s-l-s-r-s- l. This also fails to conform with the requirement, for there are three errors of the first type, four of the second, and six of the third. Again, in a standard maze, the blind alleys should all be of the same length, and care should be taken to provide a sufficiently strong and uniform motive for escape. In the case of one animal the desire to escape from confinement may prove a satisfactory motive; in the case of another, the desire for food may conveniently supplement the dislike of confinement; and in still other cases it may appear that some form of punishment for errors is the only satisfactory basis of a motive for escape. Readers of this account of the behavior of the dancing mouse must not infer from my experimental results that the electric shock as a means of forcing discrimination will prove satisfactory in work with other animals or even with all other mammals. As a matter of fact it has already been proved by Doctor G. van T. Hamilton that the use of an electric shock may so intimidate a dog that experimentation is rendered difficult and of little value. And finally, in connection with this discussion of a standard Labyrinth, I wish to emphasize the importance of so recording the results of experiments that they may be interpreted in terms of an animal's tendency to turn to the right or to the left. My work with the dancer has clearly shown that the avoidance of a particular error may be extremely difficult for left whirlers and very easy for right whirlers.

I hope I have succeeded in making clear by the foregoing account of my experiments that the labyrinth method is more satisfactory in general than the problem method as a means of measuring the rapidity of habit formation in the dancer, and I hope that I have made equally clear the fact that it is very valuable as a means of discovering the roles of the various senses in the acquirement of a habit (Chapter XI). From my own experience in the use of the labyrinth with the dancer and with other animals, I am forced to conclude that its chief value lies in the fact that it enables the experimenter so to control the factors of a complex situation that he may readily determine the importance of a given kind of sense data for the formation or the execution of a particular habit. As a means of measuring the intelligence of an animal, of determining the facility with which it is capable of adjusting itself to new environmental conditions, and of measuring the permanency of modifications which are wrought in its behavior by experimental conditions, I value the labyrinth method much less highly now than I did previous to my study of the dancer. It is necessarily too complex for the convenient and reasonably certain interpretation of results. Precisely what is meant by this statement will be evident in the light of the results of the application of the discrimination method to the dancer, which are to be presented in the next chapter. The labyrinth method is an admirable means of getting certain kinds of qualitative results; it is almost ideal as a revealer of the role of the senses, and it may be used to advantage in certain instances for the quantitative study of habit formation and memory. Nevertheless, I think it may safely be said that the problem method and the discrimination method are likely to do more to advance our knowledge of animal behavior than the labyrinth method.