Abstract

Analyses of raw data from H. Schmidt\'s (1969) automated precognition experiments were undertaken to determine if the results could be attributed to Ss capitalizing on local biases in the target sequences, as J. E. Alcock has suggested. Global nonrandomness was refuted by Good\'s Generalized Serial Test. A computer program was developed to identify successive blocks of trials for which the singlet target frequencies were significantly diverse. There was a strong tendency in these biased blocks for target and response frequencies to match on miss trials. Weaker effects in the same direction were found for doublets. When expected hits in all blocks were adjusted for this matching bias (MB), it was found that the bias could not account for all the significance. When the criterion for biased blocks was liberalized, the MB effect could account for all the significance in the high-aim files but not the low-aim files, and it was absent in 1 of the 3. Two control tests using new !random targets for miss trials gave chance results. The MB effect is shown to be at least partly attributable to subject response biases to preceding targets, rejecting Alcock\'s hypothesis.