Blocking and backward blocking involve learned inattention

Abstract

Four experiments examine blocking of associative learning by human participants in a disease diagnosis procedure. The results indicate that after a cue is blocked, subsequent learning about the cue is attenuated. This attenuated learning after blocking is obtained for both standard blocking and for backward blocking. Attenuated learning after blocking cannot be accounted for by theories such as the Rescorla-Wagner model that rely on lack of learning about a redundant cue, nor can it be accounted for by extensions of the Rescorla-Wagner model designed to address backward blocking that encode absent cues with negative values. The results are predicted by the hypothesis that people learn not to attend to the blocked cue.When a compound of two cues, A and B, is paired with an outcome, both cues will typically accrue moderate associative strength with the outcome. If, however, the pairing of the compound with the outcome is preceded by an earlier phase oftraining in which cue A by itself is paired with the outcome, then cue B does not accrue much associative strength with the outcome. Apparently the previous learning about cue A has blocked (i.e., prevented) learning about cue B (Kamin, 1969).There are two traditional theories ofblocking that have vied for 30 years. One theory, proposed by Kamin (1969) and formalized in the classic Rescorla-Wagner (RW; Rescorla & Wagner, 1972) model, argues that blocking is caused by lack of learning. According to this approach, in the second phase oftraining, there is lack ofsurprise about the outcome because cue A already predicts the outcome. To the extent that there is no predictive error, there is no change in associative weight from cue B to the outcome. The blocked cue remains essentially unaffected. A primary motivation of the RW model was accounting for blocking (cf. Miller, Barnet, & Grahame, 1995;Siegel & Allan, 1996), and the model continues to be cited as the standard explanation of blocking (e.g., Domjan, 1998, pp. 107-110).A second theory of blocking, suggested by Sutherland and Mackintosh (1971) and extended by Mackintosh (1975), argues that participants do learn something about the reThis research was supported in part by NIMH FIRST Award I-R29-MH51572. For assistance administering the experiments, the authors thank James Cortright, Christy Doherty, Jacob Hall, Scott Harris, Erin DeMien, Angie McKillip, Heather Shelton, and Elyse Weiss. For helpful comments on earlier drafts of this article, the authors thank Michael Fragassi, Mark Johansen, Teresa Treat, John Wixted, Pete Wood, Michael Young, and two anonymous reviewers. Correspondence should be addressed to J. K. Kruschke, Department of Psychology, Indiana University, 1101 E. 10th St., Bloomington, IN 47405-7007 (e-mail: kruschke@indiana.edu). dundant cue B. Specifically, participants learn to suppress attention to it because it predicts no change in reinforcement. Thus, this theory asserts that blocking is caused not solely by lack oflearning but by learned inattention to the blocked cue. Mackintosh and Turner (1...

Attentional control of associative learning—A possible role of the central cholinergic system

“…Kruschke (2001) proposed a model (EXIT) that "provides a framework wherein Mackintosh's (1975) formulas for attention learning and for association learning derive from the same motivation, gradient descent on error" (Kruschke, 2001). He was able to use his model to fit human data from an earlier behavioral study (Kruschke and Blair, 2000). The architecture of the EXIT model allows it to simulate the control of attention over associative learning to an exceptional degree.…”

“…We used the Leabra framework for activation dynamics and learning (O'Reilly and Munakata, 2000;O'Reilly, 1998), which provides a coherent This model is capable of simulating human performance in the above mentioned fictitious diagnosis task (Kruschke and Blair, 2000). Error-driven weight changes are determined by the generalized recirculation algorithm (GeneRec, O'Reilly, 1996), which is a central component of the Leabra framework.…”

“…Interestingly, Kamin (1968) described the blocking effect as "attention-like processes in classical conditioning". Kruschke and Blair (2000) demonstrated the control of attention over associative learning in a behavioral study. They extended a standard blocking experiment to demonstrate attenuated learning for blocked stimuli in a fictitious diagnosis task.…”

“…Among other stimulus sets, the combination BI was presented during this test. Participants chose disease 6 associated with symptom I over disease 1 (58.8% to 15.0%; Kruschke and Blair, 2000). This reflects blocking of symptom B by A, so that no strong B-1 association had developed.…”

Blocking in category learning.

“…It is not the case that a list of vehicles is a tone, so the task did not involve hierarchical classification. Similar prediction tasks have been used in many other studies on blocking, for example, Chapman and Robbins (1990) used a stock market prediction paradigm; Kruschke and Blair (2000) used symptoms that predict diseases; and Shanks (1985) used different weapons that cause tanks to explode. Without the act of classification, participants might have considered inferences a less likely prospect, thereby reducing the incentive to learn more features.…”