Testing for Sequential Dependencies

Publisher Website
Google Scholar
Abstract
Certain psychophyslcal procedures assume that successive responses are statistically independent. A method for testing the validity of this assumption using the up-down procedure is described. Two identical up-down strategies are run concurrently. Observations for each strategy are interleaved accord-ing to a rule designed to produce a large difference between the resulting estimates should a sequential dependency exist. The null hypothesis that no dependency exists is postulated, and the 2 estimates are compared for differences significantly larger than could have occurred by chance. The experiment was designed to test the possibility that responses of the same type are more likely to follow in sequence than would occur with statistically independent judgements. The masked threshold for a 250-Hz tone in wide-band noise was measured for 5 subjects on 3 separate occasions. Two of the 5 subjects showed a small but consistent difference between the 2 concurrent estimates, Indicating a slight tendency for like responses to follow in sequence. The technique of interleaving by rule may also be of value in measuring hysteresis effects and in detection-theory experiments, where it may be possible to discriminate between sequential effects influencing the subject''s sensitivity and those affecting the subject''s choice of criterion.

This publication has 0 references indexed in Scilit: