The particular correspondence between keys and colors for a particular setting will be referred to as a code.

When punishment is to be given, the closing of the incorrect key also activates a relay. One lead from an inductorium is taken through the contact points of this relay to a small plate strapped firmly against the palm of the subject's left hand. The other lead is taken to a similar plate held against the back of the hand, about an inch behind the knuckles of the middle and fore fingers. The shock is felt through the latter plate.

PROCEDURE

In order that any difference in results with and without shock may be ascribed to the shock, other factors that influence speed and accuracy of reaction must be kept constant or be balanced out. Such factors have been fairly well determined in numerous investigations. Summaries of these have been made by Wundt (3), Ladd and Woodworth (4), Jastrow (5), Henmon (6), and others. In view of these summaries all that need be mentioned here is the method of dealing with the factors.

Many of these factors remain constant when the same apparatus is used with and without punishment. In this class are the nature, size and intensity of the stimulus, the number of diverse stimuli, the amount of difference between the various stimuli, and the nature and amount of reaction to be made. Some factors may be kept constant by care on the part of the experimenter. The brightness of the general field was kept constant by using gray cardboard around the stimulus patch and by keeping the general illumination constant. This cardboard also served as a screen between the subject and apparatus, thus removing to a large extent the distracting influence of the apparatus.

For still other factors the experimenter can approximate constancy by adopting a rigid routine. The routine for a subject's first trial was as follows:

1. A typewritten copy of instructions was given the subject. These instructions, which were formulated after some

preliminary investigations, could be read as often as desired, but no questions concerning them were answered and no information regarding the particular code to be used was given.

2. The stimuli were shown in the manner in which they were to appear; except that they were caused to change by the experimenter's closing of his key for that purpose.

3. The subject was allowed to try the keyboard in order to familiarize himself with the placing of his fingers, the amount of key stroke required, etc.

4. If the subject was to receive shock, the contact plates were strapped to his left hand and he was told that he would receive a shock for errors as defined in his instructions. A few shocks were then given that the subject might know how much of a shock to expect and also that the intensity might be adjusted to a given subject. This adjustment for each subject was required because of the large individual differences in sensitivity to shock. The criterion for equality for various subjects was the flexing of the middle and fore fingers. The actual intensity required for this varied considerably; but it usually elicited the report, "That's about as much as I can stand."

5. All parts of the apparatus, except the circuits for the subject's keys and the shutter which kept the stimuli from being seen, were set in operation.

6. The subject was asked if he would like to go over his instructions again. When he reported that he was ready, the experimenter reminded him, "Five colors with one key for each. Remember that your score is the number of changes you make minus the number of errors."

7. The remaining circuits were closed, the experimenter simultaneously starting a stop watch and saying, "Go."

8. Readings of the two counters were taken each half minute.

9. At the end of the trial, subjects in part A (see below) were given their own scores, but not those of others, while subjects in parts B and C received both their own scores and those of others.

In subsequent trials steps 1, 2, and 3 were omitted and in parts B and C step 6 also.

The use of step 2 equalized the amount of foreknowledge of the stimuli. The arbitrary scoring mentioned in the instructions and again in step 6 tended to cause all individuals to distribute their efforts between speed and accuracy in the same manner. Mentioning the score just before the beginning of each trial and giving the actual score at the end of the trials tended to keep effort at a constantly high level. Also the nature of the task itself tended to call out a constantly high degree of effort.

The remaining factors are not directly under the experimenter's control; but they may be dealt with in special ways. The effects from individual difference and daily fluctuation were theoretically balanced out by the number of subjects used and the number of trials given. The effect from practice was dealt with by dividing the investigation into parts. Preliminary studies had indicated that after 2 minutes, with a particular code, the rate of improvement decreased very sharply, and also that when a number of codes were given, each for 2 minutes, there was considerable improvement from the first to the last. Thus practice was shown to result in two sorts of improvement, improvement in reacting to a specific code and improvement in ability to learn any code. The former will hereafter be referred to as 'specific' learning and the latter as 'general.' The investigation was so planned that a determination of the effect of punishment might be made during 'specific' learning, during 'general' learning, and during little or no evident learning.

Three groups of subjects, designated A, B, and C, were used. Each individual in group A was given ten codes, each for 2 minutes with one-minute intervals. The ten codes, in the order used, are as follows, the color given first being correct for the thumb, the one given second for the fore finger, etc.:

Each subject in groups B and C was given 24 fifteen-minute trials on successive school days. The first four trials were considered practice trials. Group B received code I throughout, while group C was given codes 6-9 for the practice trials and codes 1-5 in rotation for the test trials. Thus, group A performed while both 'specific' and 'general' learning were proceeding rapidly, group B made test records after improvement had practically ceased, and group C made test records after 'general' learning had practically ceased; but, since each trial involved the learning of a new or forgotten code, these records showed 'specific' learning during the first 2 minutes. This part of the C record will be designated Ca and the last part Cb, in which the conditions are practically the same as for group B. The effect of punishment during 'specific' learning is to be determined from the results of part Ca, during 'general' learning from the results of part A in contrast with part Ca, and during little or no evident learning from parts B and Cb.

Group A was composed of sixty individuals, all taking courses in elementary psychology. Groups B and C were composed of 8 and 6 individuals respectively, all graduate students or instructors in psychology. Punishment was administered to alternate individuals of group A and to individuals of groups B and C on alternate days, half of each group receiving punishment on the odd trials and half on the even trials.

Preliminary investigation had shown that the codes differed in difficulty, and thus that another variable was being introduced by using various codes in parts A and C. This merely required that the same codes be used in the same order for both the punished and the unpunished individuals of group A, and that in part C each code be used as often with punishment as without.

RESULTS

The results are given statistically in numbered tables and figures. In the tables the horizontal rows have been designated at the left with symbols or abbreviations. Thus C

is used for the number of correct responses, E for the number of errors, o when no punishment was given, and p when punishment was given. Accordingly, Av. Co., in Table 1-A,

is the average number of correct responses made by individuals without punishment. R dif. is used to indicate the reliability of the difference between records with punishment and without and is obtained by dividing the obtained difference by the sigma of the difference. Per cent. iC indicates the percentage increase in number of correct responses with punishment, and Per cent. iA the same for accuracy. Per cent. iA is in calculation the percentage decrease in inaccuracy, inaccuracy being the ratio of errors to correct responses. In case of decrease in either the number of correct responses or in accuracy the values for Per cent. iC and Per cent. iA become negative.

The results for part A are shown in Tables 1-A and 2-A, the first giving the results for the punished and unpunished halves of the group when the data are tabulated by trials (2 minutes each), and Table 2-A showing the results when the same data are tabulated by half-minutes, the records of each half-minute of all trials of all subjects being grouped.