I have reproduced this bookshelf drama at length because it is a typical instance of how animals learn by experience. We must especially note the gradual nature of the process. Having succeeded once, the kitten does not therefore succeed the next time without further trial and failure. What happens is that the amount of tentative groping gradually diminishes, until at last the end can be attained directly without let or hindrance. The clever tricks of animals, which excite surprise by their resemblance to behaviour distinctly human, seem to be acquired in the same way. There is a sad lack of systematic observation of the process by which animals come to do such things as opening a door by lifting a latch. But in the few instances in which the successive steps have been examined, it has been found that the perfect result only emerged after successive trials and failures involving much tentative groping, and that to succeed once by no means entailed immediate success the next time. This view has been corroborated by a series of experiments conducted by Mr. Thorndike of Columbia University, on dogs, cats, and chicks. "The method was to put the animals when hungry in enclosures from which they could escape (and so obtain food) by operating some simple mechanism, e.g. by turning a wooden button that held the door, pulling a loop attached to the bolt, or pressing down a lever. Thus one readily sees what sort of things the animals can learn to do and just how they learn to do them. Not only were the actions of the animals in effecting escape observed, but also in every case an accurate record was kept of the times taken to escape in the successive trials. The first time that a cat is put into such an enclosure, some minutes generally elapse before its instinctive struggles hit upon the proper movement, while after enough trials it will make the right movement immediately upon being put into the box. The time records [plotted down in curves] show exactly the method and the rate of progress from the former to the latter condition of affairs. . . . What happens in all these cases is this: The animal on being put into the box, and so confronted with the situation 'confinement with food outside,' bursts forth into the instinctive activities which have in the course of nature been connected with such a situation. It tries to squeeze through any openings, claws and bites at the walls confining it, puts its paws through and claws at things outside, trying to pull itself out. It may rush around, doing all this with extraordinary vehemence and persistence. If these impulsive activities fail to include any movement which succeeds in opening the door, the animal finally stops them and remains quietly in the box. If in their course the animal does accidentally work the mechanism (claw the button round, for instance), and thus win freedom and food, the resulting pleasure will stamp in the act, and when again put in the box the animal will be likely to do it sooner. This continues; all the squeezings and bitings and clawings which do not hit the vital point of the mechanism, and so do not result in any pleasure, get stamped out, while the particular impulse, which made the successful clawing or biting, gets stamped in, until finally it alone is connected with the senseimpression of the box's interior, and it is done at once when the animal is shut in. . . . Although it was of the utmost importance to them to get out of the various boxes, and it was therefore certain that they would use to the full their mental powers, none of the animals gave any sign of the possession of powers of inference, comparison or generalisation. Moreover, certain of the experiments seem to take the ground from beneath the feet of those who credit reason to animals. For it was found that acts {e.g. opening doors by depressing thumblatches and turning buttons) which these theorisers have declared incapable of performance by mere accident certainly can be so done. It is, therefore, unnecessary to invoke reasoning to account for these and similar successes with mechanical contrivances, and the argument based on them falls to the ground. Moreover, besides destroying the value of the evidence which has been offered for the presence of reason in animals, the timerecords give us positive evidence that the subjects of these experiments could not reason. For the slopes of the curves are gradual. Surely if a cat made the movement from an inference that it would open the door, it ought, when again put in, to make the movement immediately. If its first success was due to an inference, all trials after the first should take a minimum time. And if there were any slightest rudiment of a reasoning faculty, even if no real power of inference, the cat ought at least some time in the course of ten or twenty successful trials to realise that turning that button means getting out, and henceforth make the movement from a decision, not a mere impulse. There ought, that is, to be a sudden change from the long, irregular times of impulsive activity to a regular minimum time. The change is as a fact very gradual.*

Finally, experiments made in another connexion show that these animals could not learn to perform even the simplest acts by seeing another do them or by being put through them by the experimenter. They were thus unable to infer that since another by pulling a string obtained fish, they might, or that since fish were gained when I pushed round a bar with their paws it would be gained if they pushed it round themselves."+

* "Thus the successive times taken by one cat in a certain box were (in seconds) 160, 30, 90, 60, 15, 28, 20, 30, 22, 11, 15, 20, 12, 10, 14, 8, 8, 5, 10, 8, 6, 6, 7." The animals "would, in the case of some difficult associations, happen to do the thing six or seven times, but after long periods of promiscuous scrabbling, and then forever after would fail to do it." (Psychological Review, vol. v., No. 5, p. 552.)

+Science, new series, vol. vii., No. 181 (June 17,1898), pp. 818, 820821; an abstract of the original paper in Monograph Supplement, No. 8, of the Psychological Review.

The best examples in adult human beings of this gradual emergence of the right way, and gradual disappearance of wrong ways of doing a thing, are presented by the process of acquiring bodily or other dexterity merely through practice. The cook who can concoct a sauce in a peculiarly felicitous way, but cannot teach anybody else to mix the ingredients in the right proportions, must have attained success mainly by mere tentative groping without the aid of definite comparison or general ideas. Mr. Thorndike points out that association, as it existed in his animals, "is not homologous with anything in human association except such conscious connexions as a man feels in playing tennis or billiards. The essential thing in it is not the idea, but the impulse. That this sort of human associations is homologous with the animal sort is borne out by the fact that they are, like the latter, formed gradually by the stamping in of successes due to trial and error, and are not able to be formed by imitation or by one's being put through them." *