scores (standard time) of both Alpha and the Otis Tests and also with the times in the Otis Tests (all .90), and that the correlation is high though lower with the times of Alpha (.70). We are inclined, therefore, to conclude that reaction time is a very important factor in intelligence as the tests test it, both because of the relation of these reaction times to the important speed-factor as analyzed out of performance in the tests, and also because of their direct relation to the intelligence scores.

TABLE VII

RANK CORRELATIONS BETWEEN TIMES AND SCORES OF TABLE VI, Q.V. For More EXPLICIT INDICATION OF THE NATURE OF THE ITEMS

In drawing this conclusion we wish, however, to make two reservations. In the first place, we should like to see this experiment repeated with more subjects and more reaction times. We do not trust the application of statistical method unreservedly where the number of cases is small. We were surprised at the finding and only the termination of the academic year prevented us from going further. In the second place, we wish to point out that we are emphasizing facts of central tendency and not of variability. We can not explain why the rank order of A, B, and C is different for the three sets of times except to say that these three subjects were relatively close together in all three series and that their order is reversed by chance variability. It is possible that an increase in the number of cases might yield consistency even

here. On the other hand, it is possible that different timeelements, not common to the different situations, inhere in the variability and still require isolation, since they act concurrently with the central tendency the existence of which is indicated in this section of our paper.

CONCLUSIONS

Our results are set forth numerically in Tables VI. and VII. The upper left section of Table VII. indicates that there is a high correlation between score in an intelligence test, speed in an intelligence test, and speed in a simple reaction.

The upper right and lower left sections of Table VII. show that a great lengthening of the time limits for a standard test destroys these correlations, and we have noted in the paper that this result occurs because the faster subjects do not have an opportunity to utilize the additional time allowed. The table does not indicate that the "slow but accurate" subject has any advantage over his more rapid rival in any case where time is at all important. We have observed in the paper that he may have an advantage if time is of no value, but that he does not always have an advantage even then.

In general, we have concluded that speed of reaction is an important, and probably the most important, factor in individual differences in the intelligent act.12 We find that these differences in speed are not due to gross distractions or irrelevant acts, but inhere in a single item of an intelligence test and probably in so simple an act as the muscular reaction.

In view of this fact we note that intelligence assumes in psychology the rôle of 'power' in mechanics. It would seem to be primarily, although as the tests test it not entirely, mental power. 'Power,' it will be observed, is work done against time; the unit of power is unit work in unit time.13

The conception of intelligence as mental power would seem to give new meaning to intelligence as the 'common factor' or as 'general ability.' Spearman's conception of the common factor demands that we conceive all mental acts as involving

12 Cf. Thorndike's analysis, note 7, p. 73, supra.

13 Cf. note 3, p. 72, supra.

general ability and special abilities.14 Hence we can test for general ability only as we test so many special acts that the special abilities may be supposed to cancel one another, leaving exposed the net influence of general ability. Power is just such a general factor in mechanical operations. By itself it has no meaning. It is ability to do work against time, and it is exhibited only as it is utilized in some special machine or in some natural mechanism.

It is needless to point out that, if the relation of intelligence (as the tests have tested it) to reaction time of any sort can finally be established, great consequences, both practical and scientific, would follow. Practically we should have a precise and economical way of testing a fundamental and socially important individual difference. Such tests would not yield all the relationships that intelligence tests now do, because current intelligence tests also involve many educational and perhaps other psychological factors. Nevertheless, what we lost in predictive and diagnostic value we should make up in precision and in scientific intelligibility.

Scientifically we should have come already to some knowledge of the nature of intelligence and should be prepared to proceed further. Since we have failed in this study to localize the speed factor by bracketing it approximately at some level of complexity, it might be well to seek to determine next the rate of conduction of the nervous impulse in individuals of different intelligence. We make this proposal, not because we ourselves contemplate it with gravity, but because we think such a result would be no more surprising than would be the final establishment of the relation of reaction time to intelligence. After all we really know very little about the rate of conduction of the nervous impulse, and physiologists have recently discovered that what was supposed to be a constant value may occur at different rates in different fibers of the same nerve; a single stimulus may set up three or even four separate impulses of which the fastest may have a rate three times that of the slowest, as in the case of the sciatic nerve of the frog recently reported by Erlanger and Gasser.15

14 B. Hart and C. Spearman, Brit. J. Psychol., 1912, 5, 51-79; Spearman, Psychol. Rev., 1914, 21, 101-114.

Note. We take advantage of our proof-sheets to note that Weidensall has already indicated that speed may in some cases be the primary factor in so-called differences of intelligence, 16 and that Bernstein, on the other hand, is said to have found no correlation between quickness and intelligence.17 Unfortunately Bernstein's monograph is not immediately available to us, so that we remain in ignorance of the method and of the observational meaning of the terms of the title.

15 J. Erlanger and H. S. Gasser, Amer. J. Physiol., 1924, 70, 624–666. The rate of the fastest fibre was 47 and of the slowest 14 m./sec. In the dog's saphenous nerve they got differences of the order of the difference between 86 and 46 m./sec.

16 J. Weidensall, The mentality of the criminal woman, Educ. Psychol. Monog., 1916, 14, 279 ff.

17 E. Bernstein, Quickness and intelligence, Brit. J. Psychol. Monog., 1924.

FURTHER OBSERVATIONS ON THE SPEED OF

RETINAL IMPRESSION 1

BY PERCY W. COBB

Lighting Research Laboratory, Nela Park, Cleveland, Ohio

The earlier communications (1, 2) implied in this title. treat of the measurement of the duration of visual stimulation necessary, under various conditions, to determine an appropriate response. The general plan has been to establish certain conditions as to the test-object and the surroundings of the field upon which it was seen and then to measure by a suitable method, in which time was the variable, the value of the time-threshold. The plan involved a limited number of subjects and a comparatively long term of experimentation with each, yielding generally ten series for each subject under each set of conditions. The serial group-method was used, each series consisting of ten exposures at each value of a series of graded times.

In the work now to be described a different plan was used. Whereas the earlier work yielded only a few time values well above 1000, most of the averages were much lower, chiefly around 300 or lower; and it was desired to learn more completely what occurred when the threshold-time of exposure was of longer duration. It was therefore decided to work with a small number of fixed times of exposure, to use brightness as the experimental variable within the series, and to use testobjects, various as to size and contrast, whose visibility thresholds at the durations chosen should occur within the range of brightness which was of especial interest, namely, from about I to 100 ml. As a matter of fact, under some of

1 The first portion (I.) of the experimental work was carried out at the Laboratory of Pure Science, Nela Research Laboratories, under the directorship of the late Dr. Ernest F. Nichols. Credit is due Mrs. Marion Smith Steel for carrying out the greater part of it. The second portion (II.) was done at the Lighting Research Laboratory, under the directorship of Mr. M. Luckiesh, the details being carried out by Mr. Frank K. Moss.