Category relevance attenuates overshadowing in human predictive learning

Abstract

In situations in which multiple predictors anticipate the presence or absence of an outcome, cues compete to anticipate the outcome, resulting in a loss of associative strength compared to control conditions without additional cues. Critically, there are multiple factors modulating the magnitude and direction of such competition, although in some scenarios the effect of these factors remains unexplored. We sought to assess whether the relative salience of the elements in a compound of cues modulates the magnitude of the overshadowing effect in human predictive learning. Two separable categories (i.e., colors and symbols) were used in a predictive learning task. In Experiment 1, different groups of participants were granted with different time of exposure to a compound of cues belonging to different categories (color and symbol) to evaluate potential differences in the magnitude of overshadowing. Furthermore, we used posttest questionnaires to assess whether participants used either only one or both categories during training, and assessed if this impacted the magnitude of overshadowing. In general, overshadowing was not modulated by the time of exposition, except in the case of very short time of exposition with prominent learning about the most salient category. In Experiment 2, the relative salience of a category was biased via prior experience either with a biconditional discrimination or attending only the relevant category (either color or symbol). The previously relevant category was less prone to overshadowing, but not the alternative one. Results are discussed in light of attentional and configural theories of associative learning. (PsycInfo Database Record (c) 2023 APA, all rights reserved).

Figure 1. Stimuli and Procedure of Experiments 1 and 2

Note. Panel A represents the stimuli used during the task. Panel B symbolizes the procedure of the task. ITI = intertrial interval. See the online article for the color version of this figure.

Figure 2. Results From the Training Phase

Note. Proportion of participants who predicted the shock outcome in blocks of two trials. Each figure represents the response in the presence of the color cue alone (left), the symbol cue alone (middle), and the compound of color + symbol (right). Star symbols symbolize Group 1s, gray symbols Group 3s, and white symbols Group 9s. Solid lines represent the reinforced cues and dashed lines represent the nonreinforced cues. Error bars are SEM.

Figure 3. Overshadowing Index in Experiment 1

Note. Panels A and B display the overshadowing index of groups trained with a 1s, 3s, and 9s stimulus duration: Panel A displays the indexes for the color cues whereas Panel B displays the indexes for the symbol cues. Panels C and D display the overshadowing index of the participants according to their declared processing type (Both: using color and symbol; color; and symbol). Panel C displays the data corresponding to the color category whereas Panel D refers to the indexes of the symbol cues. Note that the overshadowing index was calculated by subtracting the predictive rating of the target cue (trained in compound with a competitor) from the ratings of the control cue (trained by itself); a value of 0 represents similar ratings to the control and target cues, that is, no overshadowing. Positive and negative values are indicative of overshadowing depending on the type of reinforcement. * p < .005. ** p < .001. n.s. p > .05. Error bars are SEM.

Figure 4. Pretraining Discrimination in Experiment 2

Note. Proportion of participants who predicted the shock outcome in blocks of two trials. Star symbols symbolize group Biconditional, gray symbols group Color-Relevant and white symbols group Symbol-Relevant. Solid lines represent the reinforced cues and dashed line represents the nonreinforced cues. Error bars are SEM.

Figure 5. Overshadowing Discrimination in Experiment 2

Note. Proportion of participants who predicted the shock outcome in blocks of two trials. Each figure represents the response in the presence of the color cue alone (left), the symbol cue alone (middle), and the compound of color + symbol (right). Star symbols symbolize group Biconditional, gray symbols group Color-Relevant and white symbols group Symbol-Relevant. Solid lines represent the reinforced cues and dashed line represents the nonreinforced cues. Error bars are SEM.

Figure 6. Overshadowing Index in Experiment 2

Note. Panel A displays the overshadowing indexes for the color cues whereas Panel B displays the indexes for the symbol cues. Note that the overshadowing index is calculated by subtracting the predictive rating of the target cue (trained in compound with a competitor) from the ratings of the control cue (trained by itself); a value of 0 represents similar ratings to the control and target cues, that is, no overshadowing. Positive values are indicative of overshadowing. * p < .005. ** p < .001. n.s. p > .05. Error bars are SEM.