Page:Memory (1913).djvu/36

ally in case the experimenter is obliged to work alone. For it is impossible to carry on the investigations for any length of time without taking notice of the results. The experimenter must know whether the problem has been properly formulated or whether it needs completion or correction. The fluctuations of the results must be controlled in order that the separate observations may be continued long enough to give to the mean value the certainty necessary for the purpose in hand. Consequently it is unavoidable that, after the observation of the numerical results, suppositions should arise as to general principles which are concealed in them and which occasionally give hints as to their presence. As the investigations are carried further, these suppositions, as well as those present at the beginning, constitute a complicating factor which probably has a definite influence upon the subsequent results. It goes without saying that what I have in mind is not any consciously recognised influence but something similar to that which takes place when one tries to be very unprejudiced or to rid one’s self of a thought and by that very attempt fosters that thought or prejudice. The results are met half way with an anticipatory knowledge, with a kind of expectation. Simply for the experimenter to say to himself that such anticipations must not be allowed to alter the impartial character of the investigation will not by itself bring about that result. On the contrary, they do remain and play a rôle in determining the whole inner attitude. According as the subject notices that these anticipations are confirmed or not confirmed (and in general he notices this during the learning), he will feel, if only in a slight degree, a sort of pleasure or surprise. And would you not expect that, in spite of the greatest conscientiousness, the surprise felt by the subject over especially startling deviations, whether positive or negative, would result, without any volition on his part, in a slight change in attitude? Would he not be likely to exert himself a little more here and to relax a little more there than would have been the case had he had no knowledge or presupposition concerning the probable numerical value of the results? I cannot assert that this is always or even frequently the case, since we are not here concerned with things that can be directly observed, and since numerous results in which such secret warping of the truth might be expected show evident independence