Contextual assimilation and its effect on the division of attention between nonverbal signals

Subjects performed a recognition task in which a low or a high frequency signal was presented simultaneously with a dim or a bright intensity one. The independent variable was the degree of association between the signals on the two dimensions. Unknown to the subjects four levels were tested in separate conditions, 1.0, 0.8, 0.5 and 0.2. Increasing the positive association improved overall sensitivity as measured by d' on both dimensions, but left contingent sensitivity unchanged. Changing the association did not affect overall bias, as measured by Beta. However, holding context constant by analysis produced significant differences between individual measures of contingent bias in the direction of assimilation, for example “low” responses tended to accompany dim signals and “dim” responses. The differences were unaffected by the changes in association. The results are interpreted as favouring models of contextual assimilation which assume that biasing at some level of processing depends upon the mutually contingent interaction of discrete events rather than on memory and expectation of the relative frequency of conjoint events. Further support for this conclusion comes from previous studies which indicate that assimilation occurs with separate responses but not with unitary ones. The implications of the findings for research on the division of attention are discussed, particularly the need to control variability in measured bias to establish a true change in the allocation of attention. Two generalized models of contextual assimilation are proposed, one based on “object” and the other on “lexical” assimilation. The possible application of these models to synaesthesia is considered.