to relevant words seem usually, though not invariably, more irregular as well as longer than those for words referring to the other object.
When the results were turned over to M. F. W., it occurred to her that it would be interesting to see how far correct judgments could be based on the nature of the associations alone, leaving the times out of account altogether. She accordingly went over all the associations, with the times concealed, and tried to discover in each experiment which of the two objects had been seen by the observer. The aim of this was not so much to test the practicability of ignoring reaction times,—for there would be no advantage in leaving them out of account,—as to study the nature of the considerations that enable one to base a judgment on the character of the associations, a factor that always has to be taken into account. The total number of experimental results examined was fifty-two; in thirty-four of these a correct judgment was reached; in sixteen, a wrong judgment, and in two cases it was impossible to decide. The number of correct judgments that might be' accounted for by chance was twenty-five. Evidently reaction-time alone is a much safer guide than the character of the associations alone.
What, now, are the peculiarities of the associations made which lead one to infer that a particular object has or has not been seen? After examining the associated words in each experiment, M. F. W. recorded her judgment and the reasons for it; and a careful study of these reasons seems to show that the types of self-betrayal fall into two classes according as the observer does or does not try to avoid significant associations. When there is no effort to avoid, we have, first, the cases where a relevant stimulus word suggests a word that would not naturally occur to a person who had not seen the object. This may be called the most naïve type; it would be the natural method of self-betrayal if there were no effort to conceal. It proved to be rare in these experiments, and to be an uncertain indication. Only one correct judgment was based upon it, where the word 'red' suggested 'ink,' as it would not have been very likely to do unless the red ink bottle had been seen. On the other hand, in two cases an incorrect judgment was reached on the same basis; once the object was wrongly judged to have been 'snake' because 'coil' suggested 'snake,' and another time it was judged to have been the ink bottle because 'fluid' suggested 'ink.' Needless to say, the judgments in both these cases were merely tentative; the unsatisfactory nature of the evidence was recognized, and it was influential only because no other basis for judgment presented itself. More convincing were the cases where a relevant word gave an association different from that which would have occurred had the object to which the stimulus word referred been seen. Thus in two cases correct judgments were reached because the stimulus word 'write' was perceived as 'right,' which would hardly have been the case had the ink bottle been seen. This form of self-betrayal is classed with the preceding because in neither case need there be any effort to conceal on the observer's part. Under this same head should come the cases where the significant associations are given with abnormal frequency. Thus in one series where the watch had been seen, the only reaction words given were 'watch' and 'time,' repeated again and again; in another case where the cards had been seen 'cards' was given five times as a reaction word, and in still another, where the object was the ink bottle, 'ink' was given six times. On the other hand this kind of evidence twice led to a wrong conclusion, 'snake' being given seven times and again five times when it was not the object seen. If this form of supposed self-betrayal could be relied upon, it would be because it indi-
