The Alternation Problem in Animal Psychology: Experiments With Fowl

In describing the experimental procedure I have attempted to give an accurate account of the bird’s behaviour. On the other hand, I have not hesitated to describe the reactions of the birds occasionally in a ‘subjective’ way—that they appeared to be ‘disappointed’ or ‘discouraged’, etc. I wish to state that I am fully aware of the fact that neither disappointment nor discouragement can be proved, and these expressions are used only because they give the shortest possible description of the facts observed and are not intended to indicate more than that.

The method employed in the present paper was as follows: the bird entered a shifting cage of 60 ×30 cm. (40 cm. high) consisting of wooden struts and a wooden bottom. Four sides were made of wire netting. One small side consisted of a sliding door (plywood, sliding sideways). The actual experiments were carried out in a ‘run’, 180 × 40 cm. (55 cm. high) also constructed of wooden laths supporting wire netting, which was open on both ends and at the bottom. In this run teak boards were placed to which the grains were glued with seccotine for the training experiments. These boards measured 150 ×15 cm. (2·2cm. thickness) weighing about 61b. each. They were frequently changed as a safeguard against ‘secondary cues’. The lid covering the grains was moved by a string coiled round a vertical metal rod at the far end of the run and actuated by an electric motor. A switch in front of the teak board, but outside the run, controlled the movement of the lid. The whole apparatus was placed on a large concrete bench which formed the bottom of the run. The bench stood on a heavily reinforced concrete floor so that any movement of the bench was practically excluded, especially as there was no traffic nearby. Light entered by vertical windows and from skylights.

During the first experiments, shifting cages were used at both ends of the run to prevent the birds from running back after they had pecked grains from the teak board in one direction, but later the far end (to the right of the observer) of the run was closed with a sheet of plywood and only the left or near entrance used.

The food used was barley grains prepared for human consumption (‘pot-barley’). These grains were chosen as I knew from previous experience that in the case of rice and shelled maize the grain is slightly transparent, and therefore a glued grain often looks somewhat different from a normal (loose) one. These pot-barley grains have an average size of 5 × 4 mm. with a shallow groove.

The ten subjects used were eight Gallus domesticus (mainly Rhode Island Reds) aged 4–6 months at the beginning of the experiments, and a pair of jungle fowls (G. bankiva) consisting of a middle-aged hen and a cock, the long spurs of which showed his advanced age. The birds were marked with coloured celluloid rings round their legs. Altogether, more than 4000 experiments were carried out with them.

After some preliminary training in order to accustom the birds to the unusual conditions of eating in the run and returning to one of the shifting cages, the experiments were begun with one cock and three hens.

During the first experiments in order to prevent the grains being moved by scratching, boards with small grooves for each grain were used (Honigmann, 1921), but this was found to be unnecessary and ordinary smooth (brown-stained) teak boards were used. The standard training board had a row of seven fixed grains, 6 cm. apart, alternating with seven loose grains. The normal row consisted therefore of fourteen grains of which the odd ones were fixed and the even ones were loose, with a distance of 3 cm. between every grain and its neighbour.

In detail the procedure was as follows: The bird was enticed to enter one of the shifting cages by means of food and then in the same way transferred to cage 1. The door of cage 1 was now closed and the run tilted about 50°—a wooden bar preventing it from tilting completely (90°). Then the teak board with the fixed grains became accessible to the experimenter as the run had no bottom. The eaten grains were now replaced by new ones and the run returned to its normal stable position—thus covering the board again. Now about 30–40 grains of rice were thrown into cage 1 through the wire netting, and while the bird was eating these grains the sliding door of cage 1 was slowly opened, thus giving access to the run. This feeding with rice took place between all experiments and prevented the bird from rushing into the run as soon as the sliding door was opened, thus allowing the experimenter to go back to a place in front of the run before the bird entered it and started eating barley from the board. The fact that the birds received about 30–40 grains before every experiment did not influence their willingness to eat a maximum of seven grains during the real tests. Rice was chosen as most of the birds like it and as it prevented the birds from getting tired of the kind of grain (barley) used for the actual tests.

After the bird had finished pecking (or eating) it went back to cage 1. The sliding door was now closed and the result of the experiment recorded. Then the next experiment was prepared by replacing the eaten grains, etc.

Factors which disturbed the fowls and prevented them from working satisfactorily included sparrows intruding into the room through open skylights. They were not influenced by pigeons, parrakeets and monkeys in cages in the same room—only about 10 yards away from them. The first snow much upset one hen. Another source of trouble was the sight of food outside the run, so the grain used for refilling the row after the experiments had to be kept in covered boxes.

The first tests carried out with domestic fowls (cock 4, hens 1–3) revealed the expected individual differences in learning. The majority of the birds quickly realized that some grains were not available, but one hen (no. 3) continued for weeks to peck at every grain in the row—loose and fixed. This bird, during preliminary training on rows where all the grains were loose, had proved to be a very reliable worker by eating all grains untiringly, but this reliability was now shown to be a lack of adaptability.

Cock 4 and hen 2 made the quickest progress. As early as the second or third day in several tests, they pecked all the even grains (2, 4, 6, 8, etc.) without touching the odd ones. I had the impression that they made use of secondary cues instead of a true alternation. The glue underneath the fixed grains was hardly visible, but the fixed grains themselves were slightly dusty and therefore darker than the loose ones. Possibly these grains also had some marks visible to the bird. To eliminate both possibilities I darkened the loose grains artificially with dust and repeated the tests. The result was that the birds became confused and made many mistakes. This proved that there were no ‘positive’ marks on the fixed grains and that the glue did not play any part in the discrimination, but that the birds had learned to eat the grains of a lighter shade and to avoid the darker ones. Numerous tests carried out alternately ‘with normal and darkened loose grains gave the same result.

The fact that in the experiments described by Katz & Révész (1909) and Révész (1922) the birds learnt so very quickly—after only 15–30 trials^*—to peck every second grain, or even every third one from a series of grains, was so striking that it seemed important to find out the reason for this marked discrepancy between our results. Accordingly, I repeated the experiments using the technique described in their papers, and used as subject a bird (hen 3) which even after 170 tests had hardly showed any distinct progress. The grains were now glued to pieces of cardboard (15 ×28 cm.) instead of to the heavy teak boards. It was interesting to note that even this hen was able to solve the various ‘problems’ very quickly. I repeated most of the different tests mentioned in these papers with the same result. But as soon as I made a crucial test in which all the grains on the cardboard were loose the bird ate them all. Pecking on cardboard causes minute movements of the loose grains, and these movements are perceived by the birds. In addition, the appearance of rice grains is somewhat altered if they are glued on to a board or cardboard. But even when barley grains glued on cardboard were used, the effect was similar.

These results make it probable that the apparently successful training to alternate reported in earlier papers was due to secondary cues. On the other hand, there was no proof that such an achievement is impossible. I therefore continued the training, and actually several birds learned to perform a true alternation, i.e. to select every second grain out of a line in which all grains were loose and without any difference in shade, etc. To achieve this, however, not 15–30, but 500–600 experiments performed in 17–20 training days were necessary. It is significant that these figures differed very little in the case of the single birds. We have to consider now each bird separately in order to follow adequately the progress made in each case.

This bird as we have seen ‘trained himself’ to eat light grains and to avoid darker ones. This secondary cue was eliminated by darkening the loose grains, but for the following experiments boards were used which were always freshly stocked with barley grains. The last grain in a row proved to have a very strong attraction for the bird and was often pecked afterwards, e.g. 2, 4, 6, 8, 10, 12, 14, 13.

On the 16th day of training the cock selected the grains in perfect alternation during Exps. 1,3,5, 12 and 15 out of 25 made during the day, and the following day in Exps. 1, 2, 5, 6, 7. In the next experiment the whole row consisted of loose grains and the result was Exp. 8/493^*—eaten 2, 4, 6, 8, 10, 12, 14, 13—the first real success. The next two experiments—again all grains loose—failed, the cock being rather excited, but in Exp. 11/496 he ate 1, 3, 5, 7, 9, 11, 13 (odd ones this time), and in Exp. 12/497 performed in exactly the same way as in Exp. 8/493. Corresponding results were obtained on the following day, e.g. Exp. 10/509 = 2, 4, 6, 8, 10, 12, 14 and Exp. 11/510 = 1, 3, 5, 7, 9, 11, 13, 14, 12 eaten, thus showing that there is no distinct tendency to start with the first grain of the row.

While the previous experiments were carried out with a row of grains having a standard distance of 3 cm. in the following series the first tests were made with unequally spaced grains (Exp. 13/511). The distance between grain 1 and 2 was still 3 cm., between 2–3, 3–4, 6–7 and 7–8, 4–5 cm. and between 4–5 and 5–6, 6 cm. Grains 8–14 had the usual distance of 3 cm. The result was that the cock ate the whole row except four grains (1, 6, 9, 14).

I should now have tried again to offer the same scheme to the cock—a row of fourteen loose grains with the usual distance of 3 cm. each—and perhaps the result would now have been better (see below, p. 146), but I had the impression that some fresh training was necessary and unfortunately used a new board (with fourteen grains—the odd ones fixed) painted black, instead of being stained brown. The bird was completely upset by this unexpected colour, gazed intently at the board and finally started pecking at it, but not at the grains. Even when one of the usual boards was now used he refused to eat, and after three more unsuccessful tests the experiments had to be abandoned for the day.

Next morning the unfortunate impression created by the black board had disappeared and the cock ate quite correctly from a normal board (with the odd grains fixed) all Even grains three times in succession; but afterwards he could not refrain from pecking The fixed no. 13. In the next experiment he pecked 2, 3, 4 but otherwise ate all even grains without pecking at odd ones. I mention these details in order to show how impossible it is to predict the behaviour of any bird in a single experiment. During the 6th experiment of the day he had the unusual misfortune to start with fixed grain no. 3 instead of 2, and since his alternating training was retained quite well, he pecked 3, 5, 7, 9, 11, 13, without any success as all these grains were fixed. He became very excited and received plenty of mixed food in order to ‘restore his confidence ‘. He had indeed deserved this ‘consolation’ as, at this stage of the training, the perfect eating of all odd grains was a positive achievement. In the following test he started correctly with 2, 4, 6 but missed no. 8 and went on to the fixed grains 9, 11—then he stopped again ‘disappointed’. The next two experiments were still very unsatisfactory but 10/538 gave a perfect result. I therefore took the risk of offering again a test row (all fourteen grains loose) and he ate correctly 2, 4, 6, 8, 10, 12. Then, after a slight pause, 13 and 14.

In order to eliminate the possible—though very improbable—influence of smell, the odd grains remaining from the above experiment were now moved to the positions of the even ones. In the next test he actually ate again the odd grains—leaving the grains he had not taken in the previous test. Thus twice the same grains were refused. But this was merely a coincidence. The remaining even grains were again moved to the positions of the odd ones and this time (13/541) he ate them nearly completely (5–13) thus eliminating the possibility of any secondary cue communicated by the sense of smell.

Similar tests were carried out on the following day. In six test experiments (all grains loose, distance 3 cm.) twice he ate the even ones and four times the odd ones with very few mistakes. Now for the first time since the unsuccessful experiment 511, the distances between the grains were again made unequal, but the transition was effected much more carefully than before. The alteration was here only 0·5 cm. (compared with 1·5 cm. in Exp. 511) and the position was as follows: the distance between grains 1–2, 2–3, 12–13 and 13–14 was the original 3 cm.; 3–4, 4–5, 5–6, 9–10, 10–11 and 11–12 was 2·5 cm. and 6–7, 7–8 and 8–9 was only 2 cm. The cock reacted to the experiment (11/552) this time with perfect alternation (eating all odd grains and no. 14 as well). A repetition yielded exactly the same result. The next time he ate the even ones plus grain 3, and then, in three successive trials, the odd grains almost correctly.

During the following days cock 4 did not work well. It was not because of the absence of hunger for he ate well in cage 1 but refused to do it in the run. The bird had a complete rest of 10 days. Then (27th training day, Exp. 1/616) I started with a test experiment— all fourteen grains loose in the standard distance of 3 cm.—and he ate almost correctly 4, 6, 8, 10, 12, 14, 13. This was very remarkable after a complete pause of 10 days. Unfortunately he again showed no ‘interest’ in succeeding experiments. Three days later (28th day of training) the first test was made in the 6th experiment of the day (6/624). He started with grain 2 but then ate the odd ones (2, 5, 7, 9, 11, 13, 14). The following tests, however, gave quite erratic results. Five more days of complete rest were therefore interposed and I started again with a test experiment (627) (all grains loose). The result was exactly the same as 9 days before in Exp. 1/616. During the next twenty-eight training experiments he ate correctly all even ones, the odd ones being fixed—in Exps. 5, 6, 10, 13, 16, 18, 21, 22, 24, 28 and 29 of the day. In Exp. 15/641 he started wrongly with 1, 3, 5, but corrected himself for the first time by going back and starting now 2, 4, 6, etc. This behaviour, however, must not be over-estimated, as the fixed grain were again distinctly darker on this day in order to make the first experiment with 3 ‘moving lid’ less difficult for the bird. As was to be expected, the presence of this new device (speed 5 · 25 cm./sec.) surprised the cock so much that he did not eat. He looked at the moving lid instead of at the row of grains. When six grains were visible he started eating the even (loose) ones, and when he had finished the last (third) grain the whole row was already visible and was correctly eaten. But this result was of course not produced under the desired conditions. The speed had to be reduced to 2 · 15 cm./sec., but then he pecked at every grain as soon as it became visible (Exp. 36/662). The grain, actually motionless, but apparently moving relative to the lid, had the same attractive power for the bird as a really moving grain. The experiments were repeated several times but always with similar results, although tests interposed between them (without lid) proved that the bird was then very well able to peck only the loose grains. Of course it was quite hopeless to try the moving lid over a row of ‘all-loose’ grains. (For more experiments with the moving lid see hen 1.)

The first test of the following day Exp. 1/667 grains loose) was a failure (1, 2, 4, 5, 6, 8, 10, 12, 13, 14). No fresh training was made with fixed odd grains, but the test simply repeated (Exp. 2/668), and this time surprisingly the cock ate the even grains perfectly, though he took grains 13 and 9 on his way back. So it was only ‘lack of interest’ or ‘attention’ that caused the failure in the first test. In the following test the distances between the grains were increased to 5 cm., with the result (Exp. 670) that the whole row, except grains 5 and 9, was eaten at once. More surprising still in the next test, when the distance was again reduced to 3 cm., all even grains were eaten in correct alternation and none of the odd ones touched. All grains were, of course, loose.

The last mentioned tests may seem to contradict the results obtained with hen 1. But in fact they do not, and we shall discuss the reason for this apparent discrepancy later (see Exps. 634 – 36 and 690 of hen 1).

This specimen was very quiet and from the beginning much tamer than all other birds. Her progress in learning was not very quick, rather it was steady. She, too, trained herself to eat darker grains. Small changes in the apparatus upset her as well as the other birds. The first test (all grains loose) was made in Exp. 15/489 of the 16th training day but there were still many mistakes. Even on the 19th day the first tests (Exps. 13/546–16/549) were not very successful, but Exp. 17/550 was a perfect solution : all the even grains were eaten while no odd grains were touched. All were, of course, loose.

Immediately afterwards—perhaps too early—I made a test with unequal distances between the grains (distances identical with Exp. 511 of cock 4). The result was the same as in the case of the cock (nearly all eaten).

During succeeding days the same phenomenon as in cock 4 was observed, namely, that when the first test was unsuccessful (all grains loose) the alternation nevertheless returned without any training. Even a test with unequal distances of the grains varying from 4 to 2 cm. was successful (Exp. 11/615). The distances were: grains 1–2 and 2–3, 4 cm. ; 3–4 and 4–5, 3·5 cm.; 5·6 and 6·7, 3 cm. ; 7·8 and 8·9, 2·5 cm. ; and from 9 to 14 every grain had a distance of 2 cm. from its neighbour. The hen ate 1, 3, 5, 8, 10, 12, 14 and 13, 9, 6, on her way back—making only one mistake in the alternation. This is remarkable as the pecked distances varied here from 8 cm. (grain 1–3) to 4 cm. (grain 10–12 and 12 –14), but the gradual variation of the distances is apparently an indispensable help.

On the 29th training day I started the new task of training the bird to eat two grains and to leave the third one. This differs from the scheme used by previous workers where the birds are reported to have selected every third grain. I assumed it would be easier to train them to eat two grains than to leave them. The new training board had therefore fifteen grains of which nos. 1, 4, 7, 10 and 13 were fixed. The distances were again 3 cm.

The first experiment (Exp. 715) gave a very significant result. The following grains were eaten: 2, 5, 8, 11, 14, (15). To a casual observer this would have appeared a successful training to select every third grain, but actual observation of the pecking bird revealed an entirely different solution. The bird started with grain 2 and went on—according to the previously trained alternation—to grain 4. This, however, was now fixed, so the hen pecked 5 and ate it. The successful pecking resulted again in alternation and the hen now pecked grain 7 which was also fixed. Grain 8 was now pecked and eaten and 10 pecked. Grain 10 was fixed, therefore 11 eaten and 13 pecked unsuccessfully. Then nos. 14 and 15 were eaten—the latter, being the last grain, was irresistible. The following scheme shows the reaction of the bird quite clearly:

This was not a coincidence. Exactly the same response was observed in the following experiments (716–721) with the only difference that the hen twice started with the (fixed) grain 1 which of course made no difference to the final result. Later she occasionally ate some of the remaining loose grains on her way back. Even 2 days later (30th training day) the first five experiments gave exactly the same result. From the 6th experiment onwards the first progress in the new training scheme was observed but during the whole day (47 experiments) no perfect solution was obtained. On the following day for the first time she pecked the right grains but often not in the right order, e.g. she pecked 3, 2 instead of 2, 3 etc. (‘inversion’).

Exp. 8/783 of the 31st training day gave the first perfect result: 2, 3–5, 6–8, 9–11, 12–14, 15 (on a training board with fixed grains 1, 4, 7, 10, 13) and many more followed during the same and the following day. But the fact that the hen in these successful trials ate rather slowly and sometimes nearly pecked a wrong grain but resisted at the last moment, made me suspicious. When now in Exp. 21/819 a row of all-loose grains was offered the result was a complete failure. She started 2, 1, 3, 4, but then ate all grains till no. 12–leaving 13–15 as she apparently did not ‘feel entitled’ to eat more than twelve grains. On the other hand, when the training board was used again immediately afterwards, she did not touch the grains 1, 4, 7, 10, 13 in Exps. 820-822. So it was quite clear that the bird had again trained herself to discriminate between different shades. To start darkening the fixed grains again seemed futile. Therefore I now tried a ‘figurai’ help. The loose grains were placed in a line parallel but 1 cm. left of the fixed ones. It was surprising to see how little this new position helped the bird. Even when the distance of 1 cm. was increased to 3 cm. the results did not greatly improve. Exp. 35 of the day gave a positive result and the distance was again reduced to 1 cm. and after a perfect solution (Exp. 41) the grains were placed in one row. The training was continued for 10 days, but without any progress, i.e. as soon as a board was freshly stocked with barley, the results were negative.

As there was no prospect of any success the scheme was now altered to two fixed and one loose grain in a row of fifteen—the grains 3, 6, 9, 12, 15 being loose. This was the original scheme used by Katz & Révész (1909, p. 106), but here, too, I was not able to observe any progress although the experiments were continued until 26 February 1941 (317 experiments in 11 training days). Often a success seemed to be achieved, but the control tests always gave the same negative result. So the attempt to train this hen to select every third grain from a row of loose grain was as unsuccessful as the training to eat two and to leave the third grain.

One could object that this particular hen was now handicapped in her learning ability after her training had been changed twice, and to investigate the possibility I now repeated some of the simple discrimination experiments described by Katz & Révész (1909). She learned after only twelve trials to discriminate between rice and barley (five loose rice grains between six barley grains, glued on cardboard). Several times she pecked the last glued grain for reasons frequently mentioned before but perfect solutions (all loose grains eaten, none of the fixed ones pecked) were very common. Often three perfect solutions occurred in succession. But as soon as all grains were loose the whole (or nearly the whole) row was eaten at once—thus proving again that these quick ‘successes’ are nothing but a discrimination between loose and fixed grains. Nine days after the last experiment I wanted to see how much of the old alternation training was kept by the bird. A row of loose grains with the original distance of 3 cm. was offered. The result was: all eaten except grain 5, and in a repetition only the grains 4, 5 and 11 were left. On the same day, three more tests with ‘all-loose grains’ were again a complete failure, and 4 days later (48th training day) the results were exactly the same.

Another hen (no. 1) was able to alternate after a very long pause (20 days) without fresh training. The reason for the bad results in the case of hen 2 was probably the antagonism between the two different schemes of training, and it appeared therefore advisable to interpose a complete rest for 2–3 months in order to wipe out, if possible, the memory of the previous forms of training, and to see whether there would be a distinct saving of time in the relearning of the first scheme (simple alternation). (Compare the debate about relearning between Yerkes (1934) and Spence (1939).)

When, after a pause of 3 months, the experiments were resumed on 10 June, 1941, with an all-loose row, all were eaten, and the same happened when the tests were repeated twice. On a training board, however, good results were achieved.

Four days later the first experiment with a freshly stocked graining board gave a perfect solution as also did Exp. 18. I now tried a test with ‘all-loose grains 3 cm. apart. The result was: 3, 5, 7, 9, 11, 13, 12—showing that the alternation was now almost perfectly mastered on the second day of retraining, after an interruption of more than 5 months. (Three training days later a faultless alternation was observed: Exp. 28/1443’.)

It was interesting to see what degree of stability the results of the training had now reached, and the tests were therefore continued on the same day. The hen started with the even ones 2, 4, 6, 8, but now—what had hardly ever happened before—she went back to 7 and ate 7, 10, 12, 14, 13. The next test—all loose again—was similar: 2, 4, 6, 8, 7, 9, 11, 13, 14. One has to realize the position of the grains after 8 and 7 had been eaten: in both cases there was a gap of 12 cm. between the grains 5 and 9—and with grain 9 there started a new regular row (9, 10, 11, 12, 13, 14)..In Exp. 21/1374 the bird commenced with the even ones and in Exp. 22 with the odd ones of the new row, which actually was only the second half of the original series. In both experiments the temptation to eat the last grain, as usual, was irresistible. One can say that here, though the alternation of the whole row was incorrect, the two halves taken separately were correctly treated.

Hen 1 was an energetic and voracious bird, always ‘willing to work ‘during the experiments, but in spite of this fact, her progress was slow compared with cock 4 and hen 2. Even on the nth training day there was no real success, though the fixed grains were distinctly darker than the loose ones. I therefore gave the fixed grains a very dark grey shade by means of a pencil and now she ‘understood’. That is, she associated either the brightness of the grains with edibility, or the darkness with inaccessibility, or both. Only seven experiments were needed before the first perfect selection was achieved, Exp. 10 and later Exps. 19, 21, 34, 36, 37, 39 and 48 of the day gave correct results—all even ones eaten and none of the odd ones touched.

On the 13th training day many experiments gave perfect results, but possibly the training was only to discriminate between different shades. When I now used a freshly stocked training board I expected a failure, but the hen pecked only grain 1 (fixed) and then correctly ate all the even ones (Exp. 20/473). Thus with the help of a striking and unmistakable difference of shade this hen learnt to alternate. After 5 more training days she was able to do it even with a row of all-loose grains.

On the same day tests with unequal distances between the grains were introduced. In contrast to the corresponding experiment with cock 4 (552–4) and hen 2 (589–91), but identical with Exp. 615 of hen 2, the distances varied from 4 to 2 cm. Hen 2 had made one mistake, but hen 1 promptly solved the problem by eating all odd grains in perfect order. It is important to realize the difficulty introduced by the differences in the pecking distances, viz. 8 cm. (grains 1–3) and 4 cm. (grains 9–11, 11 –13). At the same time the gradual diminution of these distances provided an indispensable help as before in the case of other birds. The success of the bird is not lessened by the fact that she ate the grains 14, 12, 10 and 8 on her way back (Exp. 21/629). To exclude any doubt the same scheme was repeated in the following tests. In the first one, grain 3 (now of course loose) was missed by the bird and thrown near grain 6. She therefore ate: 1, 3 (3 → 6), 6, 8, 10, 12, 14, 13, 11, 9. The result of the second test was again perfect (all odd ones and 14, 12, 10, 8, 6 afterwards). When now the normal distance of 3 cm. was doubled by offering a row of loose grains with a distance of 6 cm. the hen started with the grains 2, 4 and then ate all the remaining ones.

On the 19th training day I used a board with distances of 4·5 cm. between two neighbouring grains (distance between fixed grains = 9 cm.). Hen 1 solved the problem perfectly (Exps. 634 and 635). In the next test all fourteen grains were loose—the distance still 4·5 cm.—and again the hen correctly ate all odd ones in the right order. This good result encouraged me to try a 5 cm. spacing, and in order to avoid too long a row unequal distances were now introduced varying from 5 to 2·5 cm. (Exp. 637: 1–2, 2–3 and 3–4 = 5 cm., 4·5 ⁼ 4·5cm.5–6 and 6–7 = 4cm., 7–8 = 3·5 cm., 8–9 and 9 –10 = 3 cm., 10–11, 11–12 and 12–13 = 2·5 cm., 13–14 = 3·5 cm.). Although the pecked distance varied from 10 cm. (1–3). to 5 cm. (11–13), the hen ate all odd ones in perfect order (and 14, 12, 10 on her way back). In a repeat test (638) the result was even better: 1, 3, 5, 7, 9,11, 13,14.

These two experiments seem to contradict the result of Exp. 670 of cock 4, who had failed in a similar trial. One could object that hen 1 had the benefit of fixed odd grains, but this applies only to Exp. 634/635, while in the case of the following three decisive tests of hen 1, all grains were loose. In addition, one could say that the distance in the case of cock 4 was 5 cm., while here it was only 4·5 cm.—but in Exp. 637 the first four grains were 5 cm. apart. The decisive fact is, in my opinion, that in the case of hen 1 the distance of 4·5 cm. was offered in the first experiment of the day, while cock 4 in the corresponding test had eaten immediately before, three times in succession, rows with 3 cm. distance between grains. In order to solve this problem, 10 days later these experiments were repeated with hen 1 (28 January 1941, Exps. 686–690). This day’s tests started with four perfect results. (Even ones eaten in Exps. 1, 2 and 4; odd ones in Exp. 3.) When the distances in Exp. 5/690 were suddenly extended from 3 to 5 cm., the hen ate the whole row of odd and even ones, one grain after another, and thus reacted in exactly the same way as cock 4 in his Exp. 670. Even more significant perhaps is the fact that the reduction of the distance back to 3 cm. resulted in a perfect alternation.

Thus the observation, made repeatedly, that a sudden alteration of more than 50 % of the distances inhibits the alternation was again confirmed.

We have now to look back to Exps. 637–8. Here the distance varied from 5 to 2·5 cm., but perfect alternation was achieved in both cases. It remained to be seen how the bird would behave when the order was reversed. In the following test (639–40) the distances were therefore increased (apart from the first one) (3·5, 2·5, 2·5, 2·5, 3, 3, 3·5, 4, 4, 4·5, 5, 5, 5 cm.). During the first test (all grains loose of course) the hen chose 1, 3, 6, 8, 10, 12, 14, 13, 11, 9 with the pecked distances 6, 8, 6·5, 8, 9·5, 10 cm., but in the second she ‘readjusted’ herself and performed a faultless alternation of the odd grains (pecked distances now 6, 5, 6, 7·5, 8·5, 10 cm.).

I now wished to discover whether this hen would, after 630 experiments, be able to adjust her behaviour on a training board, where, in contrast to the usual arrangement, the even grains were fixed and the odd ones loose. I used an ordinary freshly stocked training board on which the first loose grain was laid in front of the first fixed grain instead of between the first and second fixed grain. There were two possibilities—she would either start with the even ones (now fixed) and would have to abandon them and start afresh (see cock 4, Exp. 641), or she would start with the odd ones (now loose), but in this case she would possibly find trouble in the case of the last ‘irresistible’ grain 14— now fixed. This actually happened and the hen ate (or pecked respectively) very quickly i, 3, 7, 9, 11, 13, 14, 14, 14. It is not clear why grain 5 was left, but the explanation of the quite unusual fact that grain 14 was pecked three times is that this grain had always been loose in previous training experiments and all-loose tests. It will not be misunderstood (see introduction) when I say that the hen ‘felt entitled’ to have the last grain and was consequently very persistent in her effort and afterwards ‘disappointed’ and puzzled. Marold (1939, p. 203) had a similar experience in the case of ‘premature frightening’ of his budgerigars. These birds had learned to eat six grains from a heap of 10–15, and were used to being chased away when they attempted to eat a seventh grain. When Marold hased the bird after it had obtained only four grains it was extremely surprised and did not want to leave the place—it ‘stood its ground’, as it were.

On the 20th training day experiments with a ‘moving lid’ were started. After seven training experiments (correct results in Exps. 3 and 7) the lid was used in Exp. 8/653. The first four grains (two fixed, two loose) were uncovered, and when the hen entered the run the movement of the lid (speed 1 cm./sec.) begun. She pecked and ate only the visible loose grains and went back to cage 1, as the movement of the Ed was too slow. When the speed was increased to 2·15 cm./sec. the hen behaved in the way described in the case of cock 4. She pecked every grain—fixed and loose ones—as soon as they became visible. In later experiments she often pecked only four to eight times and then waited until the whole row was uncovered and visible. She then selected the loose grains which were purposely made somewhat darker and when, without pecking, she had followed the moving lid until the end of the row, she sometimes ate the loose grains backwards: 14, 12, 10, etc. The experiments were repeated for several days with similar results. Thus it was again impossible to find out whether any alternation was possible without the help of the visible ‘pattern’ of the row.

During the following days the hen developed more and more the habit of eating in ‘both directions’, which means that after a successful alternation of the odd or even grains she ate the remaining grains on her way back. This made it sometimes difficult or even impossible to judge the test properly, especially in experiments with rows of unequally spaced grains. It was quite easy to prevent the hen by covering the row with a piece of cardboard as soon as she had finished eating in one direction. The cardboard, just long enough to cover the row, was put alongside the row before the test started, and at the right moment, by means of a wooden handle outside the run, it was pushed over the remaining grains.

Further experiments were again devoted to the problem of ‘readjustment’ to suddenly altered conditions. On the 30th training day, after perfect alternation of all-loose rows with a distance of 3 cm. (pecking distance 6 cm.), the distance was reduced to 2 cm. The first result was now 2, 5, 9, 11, 13, (14) with pecking distances of 6, 8, 4, 4 cm., but in the following tests the bird adjusted herself almost perfectly to the reduced distance of 2 cm. by eating (2), 3, 5, 7, 9, 11, 13 in Exps. 33/814 and 34/815. Then on reversion to the old distance of 3 cm. the alternation was upset, but in the following test, 36/817 (again 3 cm. distance), a secondary spontaneous readjustment effected a faultless selection of all odd grains.

Similar observations were made on offering the grains for the first time not in a straight line, but in the form of a sine curve (39 cm. long) of total amplitude 2 cm. and wave length 24 cm.

The experiments started on the 32nd training day with normal test rows (straight line, 3 cm. distance) which gave nearly perfect results. The first curved line (Exp. 5/871) gave the expected result (eaten 1, 3—then all grains from 4 to 14), although the alteration of the distances, compared with the previously eaten straight line, was negligible—only about 3·1 cm. compared with 3·0 cm.—which was probably not even perceived. Two tests with straight lines gave (Exp. 6) a nearly and Exp. 7/873 a perfect selection of all odd grains. The succeeding tests with rows in a sine curve showed a slight improvement, and the last two experiments of the day (21–22) showed again a nearly perfect alternation (1, 2, 3, 5, 7, 9, 11, 13, 14—leaving 4, 6, 8, 10, 12). ‘Spontaneous adjustment’ to a rather different condition was therefore demonstrated.

On the 33rd training day the effect of ‘shifting’ of the row was again examined. At first the row (of course all grains loose) was moved on the teak board 21 cm. to the left, and the hen kept the alternation. Even at 50 cm. a row was eaten in nearly perfect alternation: 2, 1, 3, 5, 7, 9, 11, 13, 14. In the next test the rows of grains were placed not on the teak board, but on the concrete of the table—in a line parallel and only 12 cm. away from the usual position—but now the hen ate nearly all without hesitation. A repetition gave the same result. When now in the next test (14/909) the row was on the teak board again the alternation returned at once (1, 3, 5, 7, 9, 11, 12, 13, 14), at least for the first six pecking reactions.

On the 34th training day ‘memory tests’ were carried out. The hen had not been used for a full week. The first test of the day (1/918) with an all-loose row with normal distances (3 cm.), gave a satisfactory, though not perfect, result: 1, 4, 6, 8, 10, 12, 13, and then the hen was stopped to prevent her from eating backwards. The next four tests, however, gave completely negative results.

Twelve days later the results were again completely negative. But there was still the possibility that some of the training was left. So 3 days later I started again with an allloose test and the hen performed a perfect alternation (all odd ones eaten, after grain 13 stopped, Exp. 1/938). Repetitions of this test on the same day were less successful. Test 938, however, is remarkable, as such a perfect result had not been observed for more than a month—especially as there had been in the meantime no real training—and no training at all since the negative results of the 34th and 35 th training day. These last observations show how fallacious it would be to conclude from a few adverse results that the bird is not able to solve the task. A perfect alternating selection is most unlikely to be achieved by pure chance, but if the hen is ‘willing’ to do it, she can master this problem even after several weeks without any new training.

It is unnecessary to report in detail the experiments with the other birds, as they yielded no different results. Hen 3, as already stated, was used for checking the results of previous papers, and so was cock 5. The latter learnt very quickly—after only twenty trials at the second day of training—to select without mistake every second grain of rice from a light board (weight about 360 g.). This was used as he several times pecked loose the grains glued to cardboard—a difficulty of which W. Kohler (1915) had previously noted. Exps. 22, 25, 26, 30–33, 39, 42-43 were again all faultless; but when in Exps. 45/71 all grains were loose, the whdle row, as I expected, was eaten. The same bird was used for the problem of selecting every third grain. He trained himself, as hen 2 had done before, to secondary cues. When these were cut out, he failed.

Cock 10 showed no difference from other birds in his progress of learning.

Jungle fowl were extremely shy and suspicious. They had to be trapped each time in a shifting cage for the experiments, and it took several days before they dared to eat in the run. They never ate the whole row at a time, but after eating three to five grains looked round as if to see if danger were imminent. The cock (6) learned more quickly than the domesticated birds to distinguish loose and fixed grains by shade, and then made very few mistakes. A board freshly stocked with grain, however, gave a completely negative result (23/43). It is, of course, quite impossible to draw any conclusions about his learning ability from this fact, as it was much too early to expect more. No essential differences in learning from domestic fowl were observed.

The experiments described in this paper show that fowls are able to perform an alternation, i.e. to select every second of a number of equidistant loose grains arranged in a straight or curved line. In addition, the distance between the grains can be varied to a certain extent. The learning of the alternation, however, is effected in a different way than was previously assumed. Earlier papers, dealing with the same problem, maintained that birds were able to alternate after only 15–30 training experiments, and even report the ability to alternate in an irregular way, e.g. to select grains 2, 8, 4, 10, 6 instead of 2, 4, 6, 8, 10. This would be convincing if these experiments were supported by control tests in which all grains were loose, but there are no reports on such tests in these papers. This is all the more surprising as their authors have stressed the fact ‘that apart from the glueing method control tests are necessary in which all grains should be loose, and only if the experiments are successful under these conditions can one guarantee the results’ (Révész, 1922, p. 471). We are therefore bound to assume that these birds learned merely to discriminate between loose and fixed grains. Consequently, the conclusions drawn by Révész (1922, p. 470), especially his assumption that the distances between the grains were irrelevant for the ability to select every second grain, are without foundation. Nevertheless, the results of Révész’s experiments have been reported as established facts in many papers dealing with similar problems, though other authors have already severely criticized his technique (Bingham, 1922; Warden & Baar, 1929).

In contrast to previous workers I found that 500–600 training experiments were necessary for the birds to learn a true alternation in rows in which all grains were loose. The real difficulty was to make the birds ‘understand’ what was actually required from them, and in order to achieve this the temporary use of crude secondary cues was often indispensable. This detour training has also been found necessary in experiments dealing with similar problems, e.g. the ‘figural help’ given to birds in experiments by Arndt (1939) and O. Koehler and his collaborators (1935–6). In my experiments differences in shade proved to be more effective than figural help.

Many readers will be surprised to realize that the presence of easily recognizable secondary cues in training experiments never yielded 100% correct solutions for any length of time. Apart from lack of attention, there is certainly often a lack of ‘interest’ in the correct performance of alternation. The discrimination of loose and fixed grains was no doubt often very easy for the bird, but in spite of this fact it did not act accordingly as it was of so little importance for it. Neither was there any punishment in the case of a negative choice—except that a maximum of seven grains were pecked in vain—nor any reward in the case of a correct choice—except that a maximum of seven grains were available. This would not particularly influence the bird, however, as it was allowed to eat four to six times this amount between every two experiments. In addition, the bird had to be kept between two physiological extremes: when it was very hungry it was clearly inclined to peck at every grain lest it should miss one, and when it was not hungry it would not always trouble to find the few loose grains. From this point of view it is certainly surprising that the birds were able to learn the somewhat unnatural task of performing an alternation.

This is all the more clear if we look at the results of the very few corresponding investigations carried out with other animals. Though neither the alternation in a temporal maze nor in the ‘multiple-choice’ apparatus can be strictly compared with the experiments described in this paper, it is significant that, in contrast to many successful experiments in the spatial maze, the alternation tests in the temporal maze (which alone represent a true alternation) gave very poor results (Hunter, 1928; Gellerman, 1931; Karn, 1938). The alternation in ‘multiple-choice’ experiments was, apart from the hitherto unsolved middle-box problem,^* the most difficult task-for all the animals investigated (Coburn & Yerkes, 1915; Yerkes & Coburn, 1915; Yerkes, 1916, 1934, etc.). It has been achieved only by two pigs, one of two monkeys and one of four chimpanzees.

The training of the fowls here described was actually not to alternate, but simply to eat the grains in the position of the even numbers of a series, as the odd ones—with the few exceptions detailed above—were always fixed during the training. Thus it is remarkable that the birds learnt to alternate independently of whether they started with the first or second grain in the row. Consequently, in successful tests with all-loose rows the birds ate all odd or all even grains, and there was no proof of a predilection for either of them, even if they started afresh in the middle of an interrupted row where a gap produced, as it were, a new row of grains (e.g. Exp. 1374/76 of hen 2).

From a theoretical point of view it is necessary to realize that these experiments differ entirely in method from those in which a simple discrimination between two ‘stimuli’ is achieved by the animal. In the latter case it is of the utmost importance to avoid helping the animal by means of minute movements given unwittingly by the experimenter. This danger proved non-existent, as the pecking reactions took place so quickly that it was often difficult for the observer to follow the movements of the birds, and it was obviously quite impossible to give any help in selecting the right grains. This was checked by observers who had never seen the birds before and did not know what was expected from them—but the results remained unchanged.

One difficulty of experimental procedure should be mentioned. While in simple discrimination tests the results are either positive or negative, it is obvious that here a great number of intermediate solutions are possible. Only such tests could be considered as entirely successful in which the birds had eaten all the loose grains and had pecked none of the fixed grains during training experiments, or had performed a true alternation in a crucial test in which all fourteen grains were loose. This shows the difficulty in judging the results, and the even greater difficulty in estimating the progress of learning. Using arbitrarily a criterion of five or ten perfect tests in succession would be a great mistake and might lead one to underestimate the possible achievements of the birds.

Another difficulty for the experimenter was to find how soon and how often tests with all-loose rows could be offered between the training experiments with partly fixed grains. It is of course quite impossible to work according to a fixed plan—in fact nothing could be more detrimental to the successful progress of the experiments. It is often a matter of intuition to decide whether on a certain day the birds will ‘stand’ some more tests without endangering the training. In the beginning of the experiments I was frequently overcautious in thinking it necessary to return to the training as soon as some tests with all loose grains gave unsatisfactory or irregular results. Later I realized that in such cases the training is not destroyed, but merely ‘suspended’ by the bird, and very often the alternation returned without any fresh training. Gradually I learned which conditions were likely to induce the bird to abandon the training habit. For example, doubling all the (equal) distances between the grains (all loose of course) had in every case the effect of making the bird abandon the alternation at once. This, however, could be prevented by gradually introducing the extension of the distances.

The use of rows with unequally spaced grains showed that sudden alterations suspended the alternation, while extensions or reductions of the distances, gradually introduced in one and the same row, made no difference. The fact that the pecking distance in one and the same row could be doubled in this way seems to me very remarkable. If a bird is able to perform a perfect alternation in a row of loose grains, the correct pecking distances of which vary from 8 to 4 cm., it shows without any doubt that the pecking distance does not play any decisive role. Of course this differs from Révész’s (1922, p. 470) untenable claim that the distances between the grains in the row have no influence whatever on the ‘quality’ of the alternation.

In the case of an alteration in the usual position of the row of grains the results were as follows: a shifting of the whole row by 30 cm. was the maximum, though often smaller distances adversely affected the results of the tests. Again, when introduced gradually, much larger distances did not upset the alternation..

The original aim to eliminate any ‘pattern’ effect by means of a moving lid which made only parts of the row of grains visible proved impracticable as the birds were unable to resist the ‘temptation’ of a moving grain. Actually the grains were stationary, but their increasing distance from the moving lid was sufficient to produce the effect of a real movement. Benner (1938) had a similar experience during experiments with fowls on the discrimination between real grains, and grains projected by means of an epidiascope on a frosted glass pane from below. The birds learned to recognize the difference in a short time, and after 10 days did not look at the projected grains at all. But when Benner now moved the real grains so that the images on the frosted pane moved as well, the birds pecked them again. Possibly a method employing a punishing device in the case of a negative choice would be necessary in this type of experiment, but the disadvantages of such a method are obvious.

The same facts which made it impossible to exclude the training to a certain pattern prevented me from performing any decisive tests on the effectiveness of rhythm in these experiments. That a temporal rhythm alone is not decisive is proved by several observations—perhaps in the most convincing way by the tests of hen 2 mentioned above (p. 147). On the other hand, it is obvious that the temporal sequence of the pecking movements and especially its uniform speed has a certain influence on the correct alternation, as an interruption often, though not always, produced a transition from the even to the odd grains, or vice versa, and caused other mistakes. The observations of hen 2 during the fresh training to leave every third grain after she had learned alternation were highly illuminating. They cast an interesting sidelight on the debate between Bierens de Haan (1935) and Koehler & Wachholtz (1936) over the question whether a bird is capable of performing a ‘swallowing rhythm’ instead of a ‘pecking rhythm’. An alternation can be effected by two quite different training actions: either to eat, or to leave every second grain. It does not matter that the results of both are identical in the case of a simple alternation. For this bird the eating of every second grain was decisive, as hen 2 made use of the alternation only after a successful pecking, and thus the unexpected effect— eating grains 2, 5, 8, 11, 14—resulted.

1. By glueing the odd numbers of a row of grains to heavy teak boards fowls were trained to perform a true alternation, i.e. to peck every second grain from a standard row of which all fourteen grains were loose. This was achieved after 500–600 training experiments in 17–19 training days. No punishment was used in this method.

2. Claims by other authors that a training to a true alternation can be achieved in a much shorter time could not be confirmed. As such results are not supported by control tests in which all grains were loose, it is probable that in these cases the birds have learnt only to discriminate between fixed, and loose grains. This also applies to the reported training to select every third one from a series of grains.

3. Although the actual training was effected by preventing the birds from eating the odd grains of the row, the result was nevertheless independent of whether the birds started with the first or the second grain. Rows of odd grains and rows of even grains were eaten without any preference.

4. A sudden increase of the distances between the single grains induced the birds in all cases to abandon the alternation, but it returned without any fresh training when the usual distances were used again. A gradual extension or reduction of the distances, however, did not influence the alternation.

5. A similar result was obtained when the distances between the grains in one and the same row were unequal. Here, too, the alternation was kept when the distances increased or decreased gradually. In this case the correct pecking distances could have the ratio 1:2 at both ends of the same row thus proving again a high degree of independence of the absolute pecking distance.

6. In most cases a modification of the experimental conditions (e.g. a different colour of the ground surface, or a curved line of grains instead of a straight line) gave negative results, but a ‘spontaneous readjustment’ resulting in alternation took place frequently without any fresh training.

7. The intended use of a moving lid covering parts of the row of grains in order to avoid a training to a figural pattern or to a temporal rhythm proved impracticable as in this case the birds did not resist pecking at every single grain as soon as it became visible. Apparently the moving of the lid relative to the actually motionless grains had the same effect as really moving objects which have proved to be highly attractive to many birds.

8. The pecking rhythm had no decisive influence on the alternation, but any change of the uniform speed of the pecking movements definitely facilitated mistakes.

9. After the successful training to alternation, boards with one fixed and two loose grains (1, 4, 7, 10, 13 fixed) were used for the unsuccessful training to leave every third grain untouched. The result was that the loose grains 3, 6, 9, 12 were not pecked (see p. 147). This proved that only the successful pecking acts were ‘counted’ by the bird.

10. Relearning of alternation after an interruption of more than 5 months resulted in a considerable saving of time: instead of 19 training days (550 experiments) alternation was almost perfectly mastered again by one bird at the second training day (29 experiments) and completely at the fifth training day (99 experiments).

This investigation was carried out with the aid of a grant from the Society for the Protection of Science and Learning, to which the writer here wishes to express his thanks. he is also greatly indebted to Prof. Edward Hindle for his hospitality in the Zoology Department of the University of Glasgow, and for every possible help in carrying out the experimental work.