Although the proportions do not hold, there is a direct correspondence between the two series of responses; hence it would seem that increased respiratory activity is an expression of agreeable feeling in pigeons, and this especially since the breathing, when varying at all in amplitude, usually became shallower, and also showed certain minor irregularities of contour, as often occurs in human respiration during moderate stimulation of a pleasant character.[196]

In the second series of respiratory tests four animals were used for fifteen days. Average results showed nothing as to the relation between intensity of stimulus and amount of quickening, since the three intensities used, 1, 2, and 4, produced reactions, respectively, as follows: 6.6%, 4.3%, and 6.4%. This may have been because the three intensities were employed each on different days. When the reactions are averaged according to daily succession, without regard to the intensities of the stimuli, we get the following results: first reaction, 8.0% rise in rate; second, 3.7%; third, 4.1%; fourth, 5.7%; and fifth, 6.9%. We should have expected the second daily response to be less vigorous than the first, since the animals were perhaps better prepared for the second stimulation. That the reactions increased thereafter was probably due, partially to summation, and partially to the fact that the short illuminations occasioned mental action (perception of interior of box, increased desire to escape, etc.) which involved heightened, rather than depressed, breathing activity, and thus worked directly against the dulling tendency of repetition.[197]

IV. INVESTIGATION OF ASSOCIATION

1. Labyrinth experiments. Four labyrinths in all were used (L, M, H, O). Each was constructed by attaching moveable wire partitions in a wooden box, covered with chicken wire. The pigeon was admitted through a small entrance compartment which was fastened at one end of the box, and which communicated with it by means of a lifting door, operated by pulling a cord from behind the observation curtain. Food was placed within the maze, and usually at the opposite end. Before beginning the tests the bird was allowed to become thoroughly familiar with the box without the partitions. After a few trials it learned to go to the food immediately upon entering the box. The partitions were then put into position, and the bird was tested as to the time required (except in the case of labyrinth O) and as to the method employed in reaching the food. The time was measured by means of a stop-watch, and the bird's horizontal movements were recorded on a small plot of the labyrinth; other general observations were added.

A. Habits in Labyrinth L

In this labyrinth (Fig. 1) six animals were tested once daily for thirty days, and five of these again after two and six weeks, respectively. On entering the labyrinth with the partitions in place the first time, a bird started on its usual direct course toward the food-box; running against the first partition it made vigorous efforts to push through, flying at the wire and often clinging to it for a short time; some of these random movements eventually brought it to the left of the compartment, and thence, through the opening, into the second compartment, and so on through the others, until finally it reached the food by a series of fortunate accidents.[198] The same general reaction was shown in case of the next few tests, except that fewer and fewer useless movements were made, and that the right ones were carried out with greater and greater precision. Later the animal had no difficulty in reaching the food; it did not run against the partitions, enter the blind alley, nor display such general signs of uncertainty as pausing and looking about. The process of learning in this case was obviously one of "trial and error," or the selection of useful movements. From the mass of random movements constituting the reaction to the unfamiliar environment, only those which enabled the bird to reach the food were retained and improved; the others gradually disappeared until finally the path taken became the shortest one possible, and was entered upon and pursued without hesitation by each animal as soon as it was allowed to enter the labyrinth. The time required for the tests is given in Table I.

It will be seen at a glance that the absolute time required for reaching the food varied for the individuals (see especially the results given by different birds in the case of test 1), but that the several periods for any one bird were relatively similar to those for another; and also that the time was long at first, but rapidly shortened from test to test, thus showing a steady advance in the learning process. Various lapses occurred (for example, A, 8; C, 13; E, 10) after the habit had been fairly well fixed.

In tests 18-22 the time-shortening was due principally to quickening of movements which had already become well defined. The great importance of visual data is brought out by the abrupt lengthening of the periods in the case of tests 23-25, and 26-30, where the light intensities were decreased. The lengthening was roughly proportional to the change of illumination. In the relative darkness the birds had to re-acquire the habits. The same mistakes were made as at first (running against partitions, and into the blind alley), yet here, as before, there was a ready adjustment. That the food was out of sight, or at least very much less visible, probably made no difference, since it was found that the birds would readily go to the old place after both food and food-box had been removed. In order to exclude the light entirely without making their movements invisible to me, I blindfolded the birds by means of a thin black hood, comfortably adjusted over their eyes and top of head; as a result, none was able to make the course in twenty minutes. The first turn, however, was usually made naturally, perhaps because associated with certain non-visual sense-data (sound of the lifting door, and perhaps tactual impressions of the close entrance compartment, etc.). Rats[199] seem far less dependent upon visual data than do pigeons. The great permanence of the pigeons' habits is shown by comparing the periods for tests 31-3, given after two and six weeks of rest, respectively, with those for tests 18-22.

TABLE I. TIME REQUIRED TO REACH FOOD IN LABYRINTH L

1: With 18-candle-power illumination of the room.

2: Same illumination; tests given after the animals had heard four other pigeons pecking in the labyrinth.

3: With 2-candle-power illumination, other conditions the same.

4: With a slight illumination through single curtain, other conditions the same.

5: After two weeks' rest, conditions as in 2.

6: After six weeks' rest, conditions as before.

Let us now notice the gradual progress of learning in three important parts of the maze, as shown in Fig. 2. It will be seen that in the beginning the animals started upon their usual course and pressed against the first partition (stage 1), but that later they touched it less and less (stages 2 and 3), and that finally they avoided it entirely (stage 4). The adjustment here was fairly simple: the sound made by the opening of the entrance door, and the glimpse thus given of the labyrinth, gradually came to be conditions of the movements of turning to the left, on emerging from the entrance, and passing along the compartment toward the opening, where impressions, mostly visual, in the same manner determined the movements of turning to the right and entering compartment 2.

The blind alley was naturally a decided obstacle. The pigeons learned to avoid this compartment by going around it only after many unsuccessful attempts to go through it. During the first test the animals entered it many times (stage 1); on emerging they returned to the second or the first compartment, only to encounter the pen again when they re-advanced toward the food; finally, on reëmerging from the annoying enclosure, perhaps for the eighth or tenth time, they might happen to turn to the right instead of going forward as usual toward the entrance of the box, and thus make their way along the new passage and reach the food. For the next few tests they usually entered the blind alley, but less frequently, and they remained for shorter periods (stages 2-4). Later they merely entered (stage 5); and still later they passed very near the opening without entering, or only paused a moment before it (stage 6); and finally they passed it without the slightest hesitation, walking briskly, but with well-directed movements, midway between the partitions (stages 7-8). The act of turning seemed to be an especially important factor in this habit. We notice that it was a turn to the right (most probably accidental) which first enabled the animals to get beyond the opening of the blind alley; that this same act was repeated in each successive trial until, by the gradual shortening of the loop forming the path taken by the animals in passing into, and from, the labyrinth it was finally reduced to a mere pause (stage 6); and that this later disappeared entirely, leaving only the left turn, which instead was now conditioned by the visual data at that part of the labyrinth and carried the animal past the entrance of the blind alley.

The animals did not come in contact with the second partition until they had almost learned to pass the opening of the blind alley (see stage 6). This was probably because the turn to the left which was made on approaching x was associated with visual data derived from points farther along the course (y), and when the animals reached z, compartment 2, these same data were received and were sufficient to occasion the turn to the left there also, thus bringing the birds against the partition. The adjustment was made principally on the basis of new sense-data arising from running against the wire, looking at it more closely, etc. For a few trials the birds made the turn at x too quickly, and thus failed to reach the third compartment. One of my most intelligent subjects made this mistake in the third test, and again in the sixth and seventh, and retained the act almost unchanged through the tenth, eleventh, twelfth, thirteenth, fourteenth, sixteenth, seventeenth, twentieth, and twenty-first tests, so strong was the tendency to continue a movement once begun, though it was really disadvantageous.[200]

After reaching the food and satisfying their hunger, the animals often returned to the maze passages, seeking an exit; but they never "explored" passages or showed other evidence of "free curiosity" and "desire to know all their new surroundings," as Small reports concerning rats.[201]

B. Habits in Labyrinth M

Five of the animals previously used were next tested twice daily, forenoon and afternoon, for five days, in a larger, more complicated maze. It had two blind alleys, and the food-box was near the centre (see Fig. 3). The animals' general behavior was similar to that before observed. The periods are given in Table II.

TABLE II. TIME REQUIRED TO REACH FOOD IN LABYRINTH M

Although this maze was much more difficult, it will be seen that the animals learned the route to the food far more readily than before. The first period in this series was only about one fourth as long as the first period in the other, and the course was mastered sooner (by the fifth trial instead of by the ninth). There was less pressing against the wire than before, and unsuccessful movements were sooner discontinued. This improvement was probably due entirely to experience gained in dealing with the first maze. Thorndike speaks of the gradually increasing ability of animals to deal with successive contrivances.[202] The average results given in Tables I and II are plotted in Fig. 4, next page.

C. Habits in Labyrinth H

Since hearing is an important sense in pigeons, we should expect them to be capable of useful acoustical associations. Several things occurred in the course of the two preceding experiments which seemed to indicate that this is true; for example, although I could move about, rather noisily, in the darkened room, without apparently disturbing any of the birds, some few showed signs of fright (moving about restlessly) on hearing the low, grating noises made by lifting the hanging door of the cage, sounds which had always preceded the handling of the subjects before experimentation, and which had probably become signs to them of being taken.

To investigate this kind of association I constructed a labyrinth (see Fig. 5) in connection with which sounds could be utilized as one form of sense-data. The passages were so arranged that along the route leading to the food there were three blind alleys, which the animals would surely enter before mastering the course. In another part of the room were placed, very close to each other, two electric gongs of the same size, but of different material. One was of metal and gave a clear ringing sound; the other was of wood and gave a low rattling noise.

When an animal was learning the route (as in the other two mazes) I sounded the gongs, the metallic, as the bird approached and entered the blind alleys, by openings M, O, and R, and the wooden, as it emerged from them and proceeded along the proper course, and occasionally after it had reached the food. The ringing sound was also given after the animal passed P and was approaching Q. When the new route was fairly well learned, I changed the order of the sound stimuli, ringing one gong at the places where the other had previously been sounded, and compared the records thus obtained with those obtained when the sounds were given in the original order. As an animal is liable to become confused by the sounds, or else quickly accustomed to them, I thought it best to give only a few trials, one trial with the usual order of sound stimulations, the next immediately following with the reversed order, and so on till four pairs of records had been secured, the series of trials being completed in a single day. Four animals were thus tested. The periods of the various trials are shown in Table III.

TABLE III. SOUND ASSOCIATION, LABYRINTH H

Time required to reach food under different sound conditions

In the case of thirteen of the sixteen tests given with reversed acoustical conditions (see column II) the periods were longer than the corresponding ones given alternately with them for comparison, and there was an average time-lengthening of 5.2 seconds per trial, or 34.2%. The following is a short description of the animals' reactions to the changed conditions. It corresponds to the time-values expressed in column II of the table.

Bird A: test 1, animal undisturbed; test 2, drew back from S, turned to the left and went toward R, but later returned and passed S without pausing; test 3, paused at O for a short interval, but did not enter the blind alley; test 4, paused at O again, later drew back from S, turned to the left and entered blind alley 3; it soon escaped, and this time passed S without being disturbed, although it paused at T and U.

Bird B: tests 1 and 2, animal apparently undisturbed; test 3, a few slight pauses at openings; test 4, drew back from S, entered blind alley 3, but soon escaped and passed S without hesitation.

Bird E: test 1, undisturbed; tests 2 and 4, paused at openings; test 3, turned back from S, entered blind alley 3, and paused at several places later when passing toward F.

Bird G: test 1, many pauses; test 2, turned from S and entered blind alley 3; test 3, undisturbed; test 4, drew back from S, went toward R, but did not enter the blind alley.

As the animals gave little attention to the wooden gong, but were always sensitive to the metallic one, their observed movements probably must be accounted for chiefly on the basis of certain visual and organic sense-data now with, and now without, the ringing sound. The data governing the start (as already noticed) were probably sufficient for the avoidance of the first blind alley when the gongs were reversed. In case of the other two, however, the birds had come to depend upon acoustical data, and when these were lacking as they approached the openings O and Q, the left turn could not readily be initiated, hence certain hesitations and misdirections of movement frequently occurred. Experience with the blind alley in the first experiment assisted the animals in dealing with the second blind alley here, but mistakes were made. Visual data usually were sufficient to produce the proper turn at Q, but when the ringing sound was given just afterwards, it sometimes occasioned the left turn, thus bringing the animals toward the opening of the blind alley. While the tests given were not such as would indicate how far pigeons can discriminate sounds, they certainly show that these birds are capable of useful sound associations, although visual ones are evidently of greater importance to them.

D. Habits in Labyrinth O

I next made tests in which tactual and electrical sense-data could also be utilized. In one of the passages of a simple labyrinth was placed a board 8 in. square and 3/4 in. thick, over which were stretched copper wires which formed a series of interrupted electrical circuits. By closing a key a bird could be stimulated whenever it stepped upon the wire surface. A second key was connected with a metallic gong. When an animal on its way through the maze first stepped upon the wire surface, electrical and acoustical stimuli were given; later it was allowed to walk across the board without being thus stimulated; afterward acoustical stimuli were given it at various parts of the maze.

Eight animals were used. All were found quite sensitive to the electrical shocks, and when next tested they avoided the board, especially if the gong sounded as they approached. Some would show signs of uneasiness anywhere in the maze on hearing the gong. When the board was so placed that they had to pass over it in reaching the food, when once on it they moved very leisurely, often lingering; and if they stepped upon the wire surface in the darkened maze, they showed no evidence of being frightened. Evidently no association had been formed between the peculiar tactual stimulus of touching the wires and the electrical shocks which had at first been given. But the tactual stimulus may have been below the threshold. Yerkes[203] saw evidence of association of this kind in the frog; this animal, however, is probably much more sensitive to tactual stimulation received from surfaces over which it passes than is the pigeon.

The results of these four experiments indicate that the pigeon easily acquires complicated labyrinth habits; that these remain fixed for some weeks at least; that acoustical, visual, and certain organic data are the most important sensory factors; and that the process of learning is one of "trial and error," in which the animal comes to form such a close connection between the sense-data of the interior of the box and those other sense-data arising from movements involved in reaching the food, that when the box impressions are again encountered the other sense-data are revived and readily condition the proper movements. How much memory of eating was involved in these tests cannot be told; but it was certainly not an essential part of the mental act.[204] Proper guidance throughout the course was the main thing, and this was determined by definite sense-data. That recognition, discrimination, and perhaps choice were to some extent present seems likely from the animal's hesitating movements at certain critical points. Thus it is highly probable that when the bird approached a blind alley which it had always entered before (see Fig. 2, stage 6), two alternatives were recognized, to enter, as before, or not to do so, as was usual thereafter, and that the pause had for its mental correlate a state closely bordering upon what in us would be deliberation.

2. Release experiments. Under this heading are included certain cage experiments in which some act, such as touching a lever, pecking, or stepping upon a platform, resulted in the opening of the door, and thus enabled the animal to escape and secure the food lying in view without. The animal was admitted to the cage through an entrance compartment as in the case of the maze trials. Before being tested it was allowed to become familiar with the cage and to reach the food directly by passing out through the open door. When first in the cage the animals did not seem to notice the release apparatus, and hence they probably did not begin learning the method of escape until later when they entered the cage and found the door closed, and the ordinary exit thus obstructed.

A. Latch Tests

The cage here employed was an 18-inch cubical box. The top was of chicken wire and the bottom and three sides of heavy boards; the fourth side was formed by narrow vertical bars and a wire door which opened inwardly and was held by a latch working on the outer surface of the bars. At first a long wooden latch was used, which the animals raised when reaching out for food. As this seemed an unnatural act, downward pressure was substituted by attaching to the latch, now made smaller and of brass, a string which ran over a pulley above the door and down into the cage. As nooses did not seem adapted to the birds, the end of the string was attached to a wooden lever which worked on the inner surface of the bars, about three inches from the floor. Eight animals were tested four times daily (twice in the forenoon and twice in the afternoon) for ten days. The time required to escape and the animals' behavior were recorded as in the case of the labyrinth tests.

When they first entered the box (singly as in the other experiments) and found the usual exit closed they made various attempts to push through between the bars, springing and often flying about with great force and persistency. In course of their random movements they touched the bar and opened the door and thus escaped. Later the unnecessary movements were mostly dropped and the necessary ones became highly specialized. The first association was established between the box impressions received on entering and the movements involved in approaching the front of the box and depressing the lever; later a connection was formed between the sensations of touching the lever, of hearing the sound of the opening door, of feeling the jar, etc., and the movements of turning away from the lever and passing out. The sight of the opening door seemed to be of less service to the birds than the sound and jar. Each animal soon came to touch the bar at the point of least resistance, and usually with considerable precision. The time required by the several birds is shown in Table IV, next page. The daily average results are plotted in Fig. 6, above.

TABLE IV. TIME REQUIRED TO ESCAPE FROM CAGE BY USING LATCH

The periods here were similar to those given in the maze tests—the time was long at first, then it shortened very rapidly for a few trials, then more slowly but still constantly, until the act became thoroughly familiar. The process was one of "trial and error" throughout. As before, various lapses occurred, even although the animals were as persistent as usual in their efforts to escape. When the lever was moved to the side or back of the box, none of the animals could escape. In general, pigeons show less ingenuity in dealing with latches than do sparrows, according to Porter's[205] observations, although in some other tests they are equally apt.

B. Pecking Tests

The preceding series of trials proved the animals' ability to utilize certain touching or clawing movements, at first accidental, in making an escape, and showed that these could become highly specialized. Desiring to carry out similar tests in the case of pecking movements, which are quite as natural to pigeons, I arranged a contrivance by means of which the act of pecking at corn-grains fastened to a small piece of cardboard (placed just outside the cage, but within easy reach through an opening in the wire) would open the cage-door by making a delicate electrical contact. Four animals were tested.

On entering the cage they endeavored to escape as before; failing in this they began pecking about until they found the corn-grains and made the contact which opened the door and allowed them to escape to the food without. Three of the animals made their escape in this way several times; but the habit seemed to be one that could not be readily learned, as the successive periods showed little shortening.

Thinking that the pecking of things at a definite place perhaps complicated the matter, I removed the electrical apparatus and arranged to open the door myself by pulling a string whenever the pigeon pecked anywhere upon entering the box. Preliminary experiments with Bird J indicated some ability to profit by this kind of experience. As the act of pecking could be used to advantage in a series of imitation trials, this animal only was allowed to learn to escape by actually pecking; the others were reserved for the imitation tests next to be reported.

C. Imitation Tests

In the experiments already reported the animals were used individually and usually out of sight of the others, although in the same room and within hearing of them. When efforts were made in some of the experiments to test the animals in a separate room, signs of fear and discontent were often noticed, and it was necessary to return to the first room to continue the tests. Some instances were noticed in which a pigeon would do what it saw another doing. For example, one of my subjects would not eat one day, being ill apparently; but when I put two others into the compartment with it, and they began eating the food lying about, it also began pecking. Its act could not have been due to its only then happening to see the corn, for it had before looked toward the food when this was thrown to it.

Desiring to test, under definite conditions, the imitative ability of these animals, I arranged trials in which birds were allowed to see a useful, but simple act performed by another bird, and then were given an opportunity to execute the act themselves. Using the animal which I had trained for the purpose, I allowed its series of acts (entrance to box, pecking, release, and food-eating without) to be observed by another animal, confined in a small wire compartment (similar to the entrance compartment before used) attached to the side of the large cage. Care was taken to see that the confined bird was observing, or at least was looking toward the acting one; in case of doubt, the trial was repeated. Later the trained animal was replaced by the observing one, and the latter's reaction was noted. Five animals, in all, were tested, and each was given two opportunities to escape after having seen the trained animal perform the act ten successive times. None of them, however, showed any signs of trying to escape by repeating the movements so often performed by the bird familiar with the act, but each rushed against the sides of the cage and tried to push through at various places, just as the trained bird had done when first learning the habit.

As the act, or series of acts, was rather too complex to be easily observed and utilized by the other pigeons, I arranged two much simpler tests. In one case the leading bird was taught to open the cage-door by stepping upon a platform (the lowering of which made an electrical contact); in the other, to avoid a blind alley, enter a short passage, and ascend a wooden plane (inclined at an angle of thirty degrees) which led into another box containing food. In these tests it was more difficult for the series of acts to be viewed, but the animals, singly as before, were placed at a point of vantage and apparently saw the movements of the other animals.

Of the five birds tested in the platform experiment, four utterly failed to escape in the two trials given. The fifth, in its second test, went to the platform promptly and thus made its escape, but the success may have been accidental, or due to the animal's experience of seeing and approaching the platform in its first test. In the labyrinth experiment only one bird (second test) avoided the blind alley and went directly up the inclined plane to the food; this success was probably due to experience in the first test. There was certainly no evidence that the animals had grasped the nature of the problem; nothing to indicate that the performance of the trained animal had supplied data for the guidance of future conduct, that is, for the conditioning of the necessary movements, in this case those of pecking, stepping upon a platform, or avoiding a blind alley and ascending an inclined plane.[206]

These results are similar to those which other experimenters have secured in the case of chicks, cats, dogs, monkeys,[207] and also rats.[208] Although the method I employed is doubtless open to the criticism of being artificial,[209] some value at least should be attached to the results; if so it would seem probable that imitation in pigeons is not above the "instinctive" stage, and that learning depends entirely upon first hand experience, upon really doing the thing, and not upon merely seeing it done.

3. Position, Color and Form Tests. The apparatus used in these experiments consisted of small boxes, six inches in height, and open at the top. Sometimes they were exactly similar, and sometimes they differed in color or in form. They were moveably attached, six inches apart, to a board which was placed in a large wire-covered box, having an entrance compartment as in the case of the mazes and cages. Food was placed in one of the small boxes, and the pigeons were allowed to find it twice; later each bird was tested as to its ability to return to the food by depending upon the position of the box in the group, or upon its color or its form. Tests were given in series of six, and the box which was first approached was recorded as the animal's choice for that test. If it made a wrong selection, it was allowed to look about until it found the box containing the food, but in no case was it permitted to satisfy its hunger until the last test of the series. The animal apparently did not see the food until it approached the box, and subsequent tests demonstrated that it was not guided by odor.

A. Position Tests

In this series of trials I used at first six, later nine food-boxes, four inches square and covered with dark gray paper. The board to which the several receptacles were fastened was shifted at irregular intervals to various oblique angles; this was done to prevent the animal from being assisted by the position of the food-box in the larger box rather than in the group of similar boxes. After a bird had been tested sufficiently for one position it was then used, for a week or so, in some other experiment, and thus given an opportunity to forget, to some extent at least, the old experience before being taught to find the food in a box placed elsewhere in the group. For the positions 2, 3, and 4, in the group of six similar boxes, eight animals were each given thirty tests in series of six, as stated above. For the positions 5, 6, and 7, in the group of nine similar boxes, the experiments were shortened. Six animals were given twenty-four trials each, two animals for each of the three positions.

The animals quickly learned the position of the food-box and passed to it promptly when released from the entrance compartment. Changing the position of the board to which the food-boxes were attached did not affect the animals' ability to reach the food readily. They usually selected the proper box as before, although frequently they went around the end of the board and approached the food from the opposite side. The general distribution of choices in the case of positions 2, 3, and 4 is given in Table V; the rate of learning, in Table VI.

With a single exception (Bird B, box 3) the box containing the food was far more often chosen than any one of the empty boxes, and usually more often than all of them combined, the average right choices being: position 2, 62%; position 3, 57.7%; position 4, 53.3%. It will be seen that the animals were more successful in finding the food in the second position than in either the third or the fourth, that is, positions nearer the end were more easily located. But what is of greater interest to us is the rate of learning to go to the right box. This is indicated by the increasing number of right choices from series to series. (Table VI.) That the increase was small was due to the fact that the animals learned so quickly in the first series of six tests that little improvement could be made thereafter; what they could learn they acquired early in the experiment. There is some evidence of improvement after the first series in the case of box 4, a more difficult position, and the average for all three boxes shows a slight improvement from series 1 to series 5, although a falling-off is seen in the last series: 26, 27, 27, 31, 28. The same general features appear in the case of the incomplete tests (positions 5, 6, and 7). For each box there were, on the average, 26 right choices in the possible 48. There were more right choices in the case of box 7 than in the case of either box 6 or box 5, box 7 being nearer one end (box 9). There was little evidence of learning after the first series of trials.

TABLE V. ASSOCIATION OF POSITION: GENERAL DISTRIBUTION OF CHOICES