My experiments on discrimination were of the following general type: I got the animal into the habit of reacting to a certain signal (a sound, movement, posture, visual presentation or what not) by some well-defined act. In the cases to be described this act was to come down from his customary positions about the top of the cage, to a place at the bottom. I then would give him a bit of food. When this habit was wholly or partly formed, I would begin to mix with that signal another signal enough like it so that the animal would respond in the same manner. In the cases where I gave this signal I would not feed him. I could then determine whether the animal did discriminate or not, and his progress toward perfect discrimination in case he did. If an animal responds indiscriminately to both signals (that is, does not learn to disregard the ‘no food’ signal) it is well to test him by using two somewhat similar signals, after one of which you feed him at one place and after the other of which you feed him at a different place.
If the animal profits by his training by acquiring ideas of the two signals and associates with them ideas of ‘food’ and ‘no food,’ ‘go down’ and ‘stay still,’ and uses these ideas to control his conduct, he will, we have a right to expect, change suddenly from total failure to differentiate the signals to total success. He will or won’t have the ideas, and will behave accordingly. The same result could, of course, be brought about by very rapid association of the new signal with the act of keeping still, a very rapid inhibition of the act of going down in response to it by virtue of the lack of any pleasure from doing so.
For convenience I shall call the signals after which food was given _yes_ signals and those after which food was not given _no_ signals. Signals not described in the text are shown in Fig. 29, below. The progress of the monkeys in discriminating is shown by Figs. 30 and 31, on pages 199 and 201. In Figs. 30 and 31 every millimeter along the horizontal or base line represents 10 trials with the signal. The heights of the black surface represent the percentages of _wrong_ responses, 10 mm. meaning 100 per cent of incorrect responses. Thus the first figure of the set, Left hand, _a_, presents the following record: First 10 trials, all wrong; of next 10, 7 wrong; of next 10, 6 wrong; of next 10, 7; of the next, 9; of the next, 9; of the next, 4; of the next, none; of the next, 3; of the next, 2, and then 70 trials without an error.
[Illustration: FIG. 29.]
I will describe some of the experiments in detail and then discuss the graphic presentation of them all.
