A Comparison of Forward Blocking and Reduced Overshadowing in Human Causal Learning

Article excerpt

In this study, we directly compared forward blocking with reduced overshadowing in a human causal learning study using an A+, B- (first learning stage), AX+, BY+, KL+ (second learning stage) design. The results showed that reduced overshadowing was significantly stronger than forward blocking. These results are problematic for at least some associative learning models but were predicted on the basis of higher order reasoning accounts of cue competition in human causal learning.

Learning causal relationships is an essential skill in daily life because it allows us to anticipate and control events in the environment. For example, if a person learns that each time he walks in the sun he will get a headache, he will avoid walking in the sun or he will protect his head with a cap. Yet despite the importance of learning causal relationships, there is still much controversy about the underlying processes.

In human causal learning (HCL) studies, participants are presented a number of trials on which causal cues and outcomes are either present or absent. After they have observed some or all of the information, the participants are asked to make a judgment about the causal relation between each cause and the outcome. In one standard design, participants are first exposed to A+ (cue A followed by the outcome) and B- (cue B not followed by the outcome) trials. In a second learning stage, the participants see AX+ (cue A and cue X presented together and followed by the outcome) and BY+ trials. Several studies (e.g., Aitken, Larkin, & Dickinson, 2001; Dickinson & Burke, 1996) have found that the causal ratings for cue X at the end of the second learning stage are lower than the causal ratings for cue Y. This result is referred to as forward cue competition.

Forward cue competition, however, confounds two different cue competition effects-namely, forward blocking and reduced overshadowing. When a target cue X is paired together with an alternative cue A and followed by the outcome (AX + ), causal ratings for cue X are lower when the AX+ trials are preceded by A+ trials than when no A+ trials are presented. This effect is referred to as forward blocking (see, e.g., Dickinson, Shanks, & Evenden, 1984; Shanks, 1985). Similarly, when a target cue Y is paired together with an alternative cue B and followed by the outcome, causal ratings for cue Y will be higher when BY+ trials are preceded by B- trials than when no B -trials are presented. This effect is referred to as reduced overshadowing (e.g., De Houwer, Beckers, & Glautier, 2002).

The aim of this study was to compare the relative contributions of forward blocking and reduced overshadowing to forward cue competition. This can be achieved by adding KL+ control trials to the second learning stage of a forward cue competition design (that normally consists of AX+ and BY+ trials only). Just like the target cues X and Y, the control cues K and L are always presented in compound with another cue. The difference between the control and target cues is that the control cues are paired with another cue that has not been presented before, whereas the target cues are paired with a cue previously presented on its own and followed by the outcome (forward blocking) or not followed by the outcome (reduced overshadowing). Within this design, forward blocking can be calculated by subtracting the causal rating for X from the mean causal rating for K and L. Similarly, reduced overshadowing can be calculated by subtracting the mean causal rating for K and L from the causal rating for cue Y.

There are some indications from previous studies that reduced overshadowing may be stronger than forward blocking. Larkin, Aitken, and Dickinson (1998) found that unovershadowing was stronger than backward blocking (but see Wasserman & Berglan, 1998, for conflicting results). Both unovershadowing (also known as release from overshadowing) and backward blocking refer to cue competition effects similar to reduced overshadowing and forward blocking, respectively, but the order of the presented events is reversed; that is, AX+ (BY + ) trials precede the A+ (B-) trials in a backward blocking (unovershadowing) design. …