This method is so simple that it requires but little training on the part of the experimenter, and but little co-operation on the part of the subject. It is not to be assumed that every reaction obtained by it is a true and immediate association to the corresponding stimulus word; but we have found it sufficient for the purpose of the test if the subject can be induced to give, in response to each stimulus word, any one word other than the stimulus word itself. No attempt is made to determine the exact degree of co-operation in any case. In the early stages of this investigation the reaction time was regularly recorded. The results showed remarkable variations, among both normal and insane subjects. In a series of twenty-five tests, made more recently upon normal subjects, ninety reactions occupied more than ten seconds, and fifty-four of the stimulus words elicited a ten-second response from at least one subject.[1] [Footnote 1: These tests were made by Dr. F. Lyman Wells, of the McLean Hospital, Waverley, Mass., and he has kindly furnished these data.] It is noteworthy that these extremely long intervals occur in connection with reactions of widely differing values. That they are by no means limited to individual reactions is shown in Table III. by a group of selected reactions, all given by normal subjects. TABLE III. Word combination Reaction time Value of in seconds. reaction. comfort--happiness 20 5.0% short--long 11 27.9% smooth--plane 16 2.3% woman--lady 40 4.1% hard--iron 12 2.4% justice--judge 20 9.1% memory--thought 20 8.1% joy--pleasure 18 12.1% It is apparent, even from a superficial examination of the material, that the factors which cause variations of reaction time, both in the normal state and in pathological states, are numerous and complex. It has been the purpose of this study to establish as far as possible strictly objective criteria for distinguishing normal from abnormal associations, and for this reason we have made no attempt to determine by means of introspection the causes of variations of reaction time. It would seem that the importance and magnitude of the problem of association time are such as to demand not merely a crude measurement of the gross reaction time in a large number of cases, but rather a special investigation by such exact methods as have been used by Cattell [1] and others in the analysis of the complex reaction. It would be impracticable for us to employ such methods in a study so extensive as this. [Footnote 1: Mind, Vol. XI, 1886.] In view of these considerations we discontinued the recording of the reaction time.