particular reaction or series of reactions, nor to let up at any time and neglect to react as quickly as possible. The reaction movement was always made in very much the same way. The subject placed the middle and forefinger of his right hand on the key at the warning signal, in case they were not already there, and held the key down until the reaction. If the subject at any reaction was not ready or distracted in any way, he signalled at the time to the experimenter and the reaction time obtained was thrown out, whether long or short. The subjects were instructed not to count, as I was afraid they might adopt this method of estimating the longer intervals. Just what difference it would have made, if they had, I do not know. All the subjects reported that they avoided counting.

The subjects were informed of the general nature of the experiment, and knew also the duration of the intervals used and the order in which they were to occur. They were informed that each interval would be repeated 25 times before the next longer one was taken up. They were not called upon to introspect more than to be sure that they followed the instructions. I did not wish them to try to introspectively describe the "action" consciousness. This was not my problem, and I did not want to require anything of the subjects which might distract them from the main task in hand, which was to react as quickly as possible every time the stimulus sound occurred.

The sittings lasted about an hour or sometimes a little longer, and occurred at the same time upon immediately successive days for each subject. Each "Series" in Table I refers to one hour's results of one subject. Each hour's work was as follows: After the control of the chronoscope, 25 reactions were taken with a I sec. preparatory interval; then 25 with a 2 sec. interval; then 25 with a 4 sec. interval, and so on up by steps of 4 secs. through a 24 sec. interval, and in some cases through a 32 sec. interval. The return to a I sec. interval, as though to repeat the whole experiment, was never made at the same sitting.

In presenting the results in Table I, I have followed the most widely used procedure, and given the average and the mean variation. In taking the averages no results were thrown out merely

because they deviated greatly from the average. I may have included in this way a number of mistakes, but the only readings which I threw out (and I threw them all out) were those where I had noted possible error or mistake at the time, or those which were accompanied by a signal from the subject that he was distracted or not ready. In first working up my results

TABLE I

Effect of Variation in Preparatory Interval, Regular Procedure.

N, for each interval of each series 25. Total number of reactions = 3,825.

I applied Chauvenet's criterion for the rejection of extreme cases, but found it too laborious in view of the fact that its application did not seem to change the general trend of the results. It may be that I have included a number of prematurely begun reactions in the case of the 1 and 2 sec. intervals, but as these intervals were repeated only 25 times, I doubt if many such occurred. Of course a reaction which occurred before the stimulus would not be measured by the apparatus used.

The mean variation is given separately for each day's results. The "Series," each of which represents the data of one day's sitting, are arranged for each subject from top to bottom in the order taken. The average reaction time in thousandths of a second is given first, and the mean variation from each average given immediately below that average in italics. In the lines headed at the left, "Average," is given the average of all the average reaction times of each day; that is, the grand or total average. In the lines below these, headed "Av. M. V.," is given the average of the mean variations of the daily averages. The intervals are given in seconds in the first line of the table.

The data of Table I are represented graphically in Figure 2. The duration of the preparatory intervals is represented by the abscissae and the reaction time by the ordinates. Only the grand average reaction-time for each interval, for each of the three subjects is plotted.

Since the general course of the curve is somewhat similar for each of the three subjects, I have plotted a curve, Figure 3, which represents in graphic form the data for the three subjects regarded as constituting a single group, allowing equal weight to the results obtained with each subject. The numerical data represented graphically in Figure 3 are given in the bottom line of Table I.

That the averages and mean variations given in Table I truly represent the general tendency is shown by comparison with the full distributions of the last 100 reactions to each interval for subject Vs, presented in Table II.

Subject Vs is chosen as typical, and only the last four sittings 'See Merriman, Method of Least Squares, 1903, 166.

with each interval are represented, so as to have data which do not represent too widely differing stages of practice. Even then, it must be remembered that each of these four sittings occurred on different days, and since the subject reacts uniformly faster or slower on any given day, a table which mixes the distributions obtained on different days gives a much less narrow range of reaction times for each interval than a table showing the same number of reactions all made the same day. On the

FIGURE 2

Graphic Representation of Effect of Variation in Preparatory Interval, Regular Procedure, for Each of Three Subjects.

The durations of the preparatory intervals are represented as abscissae; the reaction times, as ordinates. SZ, VS and WW stand for the three different subjects, for each of which the curve is plotted separately.

other hand, the average is probably all the more reliable, as any peculiarities in one day's results would tend to be eliminated by averaging with results obtained the other days. It is because the results for each day represent different stages of practice that I have not given in Table I the M. V. for the whole number of reactions, but instead, the average of the M.V.'s obtained on separate days.

The conclusions to be drawn from the data of Table I are quite simple, and clearly expressed in the curves of Figures 2 and 3. It must be constantly remembered that the conclusions hold only for the particular procedure used. Data presented later

FIGURE 3

The Curve Obtained by Combining the Three Curves of Figure 2.

The durations of the preparatory intervals are represented as abscissae; the reaction times, as ordinates. The formula for the above experimental curve is y=A+B. log x, in which y equals reaction time and r the duration of the interval, and A and B are constants.