The puzzle-box experiment.
Place a hungry young cat in a strange cage, with a bit of fish lying just outside, and you are sure to get action. The cat extends his paw between the slats but cannot reach the fish; he pushes his nose between the slats but cannot get through; he bites the slats, claws at anything small, shakes anything loose, and tries every part of the cage. Coming to the button that fastens [{309}] the door, he attacks that also, and sooner or later turns the button, gets out, and eats the fish. The experimenter, having noted the time occupied in this first trial, replaces the cat, still hungry, in the cage, and another bit of fish outside. Same business, but perhaps somewhat quicker escape. More trials, perhaps on a series of days, give gradually decreasing times of escape. The useless reactions are gradually eliminated, till finally the cat, on being placed in the cage, goes instantly to the door, turns the button, goes out and starts to eat, requiring but a second or two for the whole complex reaction. Perhaps 15 or 20 trials have been required to reach this stage of prompt, unerring response. The course of improvement is rather irregular, with ups and downs, but with no sudden shift from the varied reaction of the first trial to the fixed reaction of the last. The learning process has been gradual.
This is the typical instance of learning by "trial and error", which can be defined as varied reaction with gradual elimination of the unsuccessful responses and fixation of the successful one. It is also a case of the substitute response. At first, the cat responds to the situation by reaching or pushing straight towards the food, but it learns to substitute for this most instinctive response the less direct response of going to another part of the cage and turning a button.
The cat in this experiment is evidently trying to get out of the cage and reach the food. The situation of being confined in a cage while hungry arouses an impulse or tendency to get out; but this tendency, unable at once to reach its goal, is dammed up, and remains as an inner directive force, facilitating reactions that are in the line of escape and inhibiting other reactions. When the successful response is hit upon, and the door opened, the dammed-up energy is discharged into this response; and, by repetition, [{310}] the successful response becomes closely attached to the escape-tendency, so as to occur promptly whenever the tendency is aroused.
There is no evidence that the cat reasons his way out of the cage. His behavior is impulsive, not deliberative. There is not even any evidence that the cat clearly observes how he gets out. If he made a clean-cut observation of the manner of escape, his time for escaping should thereupon take a sudden drop, instead of falling off gradually and irregularly from trial to trial, as it does fall off. Trial and error learning is learning by doing, and not by reasoning or observing. The cat learns to get out by getting out, not by seeing how to get out.
Summary of Animal Learning
Let us take account of stock at this point, before passing to human learning, and attempt to generalize what we have observed in animals of the process of learning.
(1) Elimination of a response, which means detachment of a response from the stimulus that originally aroused it, occurs in three main cases:
(a) Elimination occurs most quickly when the response brings actual pain; the animal makes the avoiding reaction to the pain and quickly comes to make this response to the place where the pain occurred; and thus the positive reaction to this place is eliminated.
(b) Elimination occurs more gradually when the response, without resulting in actual pain, brings failure or delay in reaching a goal towards which the animal is tending. The positive response of entering and exploring a blind alley grows weaker and weaker, till the blind alley is neglected altogether.
(c) Elimination of a response also occurs, slowly, through negative adaptation to a stimulus that is harmless and also useless.