Motivational biases in memory for emotions

Abstract

This research examined how individuals' motivations and goals were related to their memory for past emotional experiences. In two studies, participants rated how happy and anxious they felt while completing a challenging anagram task and later recalled their emotions. Bias in memory for emotions was predicted by the combination of participants' general motivation (approach BAS vs. avoidance BIS) and the specific goal they set for the task (approach vs. avoidance). Participants with approach motivation and goals overestimated happiness more, and showed a stronger relation between peak and remembered happiness, than other participants. Participants with avoidance motivation and goals overestimated anxiety more, and showed a stronger relation between peak and remembered anxiety, than other participants. Thus motivational factors known to influence attention to valenced information also predict how emotional experiences will be remembered.

Keywords:

• Motivation
• Memory
• Approach
• Avoidance
• Emotion

Acknowledgements

Support was provided for this study by an American Psychological Association dissertation award and a School of Social Ecology fellowship. Portions of this manuscript were presented at the International Society for Research on Emotions conference, Atlanta, GA.

Thanks to Roxane Silver and Jutta Heckhausen for their support and to Sarah Strader, MaryAnn Profeta, Nicole Navi, Marissa Fortuno, Jenna Tanaka, Cassondra Wiley, and Erika Siegel for assistance with data collection and entry.

Notes

¹Creating dichotomous variables from continuous measures can create bias in estimates and it is generally recommended that researchers avoid this procedure and instead rely on regression analyses (e.g., Cohen, 1983; Maxwell & Delaney, 1993; Tellegen & Lubinski, 1983). The discussion of potential bias has made clear that this procedure lowers the power to detect significant differences and wastes information and can, in cases where multiple variables are dichotomised with a median split, lead to erroneous rejection of the null hypothesis for main effects (Maxwell & Delaney, 1993). We chose to dichotomise measures of approach and avoidance motivation and goals in the present research for several reasons. First, examination of potential bias created by dichotomising continuous variables has focused on median and mean splits. This was not the procedure used in the present investigation. Rather, participants were classified according to whether scores on one variable were higher than scores on another. In addition, our predictions focused on interaction terms and not main effects. Second, the concern that important differences between individuals are ignored by treating participants in each group as if they were the same is less likely to be relevant to approach versus avoidance motivation and goals. The classification of individuals as primarily approach or avoidance is consistent with both theory and empirical findings that only the approach or avoidance system can be activated at a particular point in time (they are mutually inhibitory and the activation in one reduces activation in the other) and that individuals reliably differ in terms of which system they find most motivating (e.g., Gray, 1972; Lang, 1995). Thus, theoretically, individuals who are motivated primarily by one motive or goal are qualitatively different from individuals with other motivation or goals. This theoretical proposition is supported in the present investigation by findings that very few individuals reported being equally motivated by both (across both studies, a total of five participants, or 2.5%, reported equal BAS/BIS scores and these individuals scored perfectly neutral on the scales, indicating that they may not have been responding carefully to the items). Third, there is no alternative analytic procedure that would allow for full consideration of the data. If approach and avoidance motivations and goals were included as continuous measures, other critical information would by necessity be lost or rendered uninterpretable. Depending on the alternative analytic procedure, the consequences of using a different analytic procedure range from only considering remembered emotion rather than bias in remembered emotion to only considering some interactions (e.g., approach goals and motivation vs. avoidance goals and motivation, ignoring those who set goals contrary to their typical motivation) to the need to analyse difference scores as dependent variables (which is accompanied by potential bias), to an inability to interpret the five-way interaction effects that would be the foci of some analyses. While dichotomising continuous predictor variables is not generally recommended, we judged that it was the best analytic strategy in this particular investigation.

²Additional analyses were conducted to ensure that the proposed relationships between motivations and goals were not due to differences in experienced emotion between the first set of unsolvable anagrams and subsequent sets. ANCOVAs were conducted for happiness and anxiety with Set (unsolvable, solvable) as a repeated measure and Motivation (BAS/BIS) and Goal (approach, avoidance) as between-subject factors. These analyses indicated that there was no difference in experienced happiness or anxiety during the unsolvable and solvable sets for participants with various motivations and goals for the task, F(1, 72) = 0.05, ns, and F(1, 72) = 0.40, ns, respectively. Further, Study 2 replicates the findings of Study 1 with manipulated goals.