columns indicate the interval given first in presenting a pair for comparison; those at the top of the vertical columns indicate the second interval in the pair. Thus parallel thirds, coming first, were in this table preferred to fifths by C, to minor sevenths by B, to major sevenths and to minor ninths by A. Parallel fifths, coming first, were inferior to thirds by C, better than minor sevenths by B, better than major sevenths by B, and better than minor ninths by A.

The interval under consideration for this particular day was the minor seventh; the method was that of prolongation. After having recorded the standard table, I proceeded to play a passage of parallel sevenths, sustaining each interval exactly one minute. I did this for five minutes, playing thus five intervals of a minor seventh. Immediately after this I asked for new judgments of preference, in which the comparison was made only between the minor sevenths and the other parallels. The results of these judgments, at the end of five minutes, appear in the "first five minute" table below.

When we compare the judgments of the sevenths in the standard table with those after five minutes of prolonging the interval, we find that it has lost one point to the thirds, has gained two from the fifth and three from the minor ninth. This gives it an increase in value of + 4 as compared with the standing of the seventh before prolongation. Repeating the same process for another five minutes, I obtained the results in Table Three. Here the minor seventh has gained two points on the third, six points on the fifth, three on the major seventh, and four on the minor ninth. Its value in this table is accordingly + 15 as compared with that of the standard table.

This process was repeated for a third and a fourth period of five minutes, sometimes for a fifth, but by this time the subjects expressed such general discomfort that it seemed hardly likely that the results could continue to have the same significance, especially as the hour was always nearly up at the end of the fourth trial. The use of a fifth period was therefore abandoned. The only other change in method which took place during the course of the experiments was in regard to the length of the prolongations. During the first few weeks of the experiments I used, ten-minute periods, made up of five two-minute prolongations, and obtained the same type of results as later with the five-minute period, but the subjects indicated a far greater satisfaction in their attitude toward the five-minute length, and it was accordingly adhered to regularly after the first few weeks.

The method of repetition differed from that of prolongation

in that each period, three minutes in this case, was occupied not with sustaining intervals, but with playing repeatedly an entire melody in parallel thirds, fifths, minor or major sevenths, as the case might be. The melody used was, with the exception of one note, within the compass of a sixth, so that the same notes recurred with great frequency. Three minutes sufficed for four repetitions of the entire melody. The first phrase of the melody, eight notes, was used in making the comparison for the tables. The amount of positive or negative change was determined, as before, by counting the difference between A and B one point, between A and C two points, and so on. The difference between an E preference in favor of an interval and one against it was also counted one point.1

The general tables which follow give for each of the four intervals the average amount of change from the standard at the end of each successive period. In the few cases where the periods were longer than five minutes, the result was recorded in the "second five-minute" table. Thus in the case of the first subject, Bridges, the only prolongations of minor sevenths were in ten-minute periods. The figures in each of the five-minute spaces represent an average of from one to four experiments.

I. The subjects were instructed to limit their introspection to the graded expressions of likes and dislikes, unless something of unusual interest occurred. The danger of accompanying paired comparisons by abundant general introspection is shown very clearly in the experiments of T. H. Pear on the major and minor triad." Of his five subjects, the three who introspected least were by far the most self-consistent. His most introspective subject had a coefficient of consistency of 27.5 on a scale where zero was perfect consistency, whereas the coefficient of the least introspective subject was 3.64. It seems almost inevitable that the subject, in attempting to give an account of how he arrives at his judgment, leads himself to judgments that would normally not be made.

114 British Journal of Psych., 1911, p. 57.