But these two factors--decrease in variability and change in the character of the tests--seem to be far from sufficient to account for the results. The tapping test remains much the same type of process throughout, the only apparent modifications consisting of slight changes in method and perhaps some gradual changes in the muscles. There is certainly no reason for suspecting that tapping and opposites or tapping and discrimination become, as tests, more alike because of frequent repetition. But the increase in correlation is clear in both these cases. Again, it is well established that the discrimination reaction, in the form here used, also tends to become reflex through practice, the conscious discrimination coming only after the correct reaction is made. These experiments called for between 3,075 and 4,100 single discrimination reactions on the part of each observer, which would afford ample time for such a change to show itself. Mere change in the character of the test would then lead us to expect color-naming, opposites, and adding to come more and more to resemble discrimination reaction. But they do not, if the coefficients may be taken as evidence. The coefficients of these tests with discrimination show no tendency to increase, even by the end of the experiment. The assumption of increasing similarity in the character of these pairs of tests would seem gratuitous. Moreover, if there were such increase in similarity, and this be also supposed to account for the higher correlation of color-naming and opposites with adding, coordination and adding should show the same increase in correlation. Just the reverse is actually the case, the correlation of coordination and adding decreasing consistently. Some further factor must then be responsible for the general increase in correlation, aside from decrease in variability (which affects only the first few trials) and progressive qualitative approximation of the tests (which is seen to be inadequate). The doctrine of "general ability" or "general intelligence" at once suggests itself in this connection. If there is such a thing as "general ability" or "general intelligence," we should expect all samplings of that ability to correlate more and more as the measures came to be truer samples. We might indeed expect to find evidences of this general ability only when measuring the "ultimate capacity" of the individuals concerned. The momentary ability revealed in initial trials, or even in the first half-dozen trials, in a given set of tests might well be expected to show only low degrees of correlation. These trials would not be measures of ultimate capacity, but would be largely determined by previous practice, chance variability, momentary attitude and initial method of attack. They would, in short, be samplings only of momentary ability, not of final capacity.