Why people rehearse their memories: Frequency of use and relations to the intensity of emotions associated with autobiographical memories

Abstract

People may choose to rehearse their autobiographical memories in silence or to disclose their memories with other people. This paper focuses on five types of memory rehearsal: involuntary rehearsal, rehearsal to maintain an event memory, rehearsal to re-experience the emotion of an event, rehearsal to understand an event, or rehearsal for social communication. A total of 337 participants recalled event memories, provided estimates of how often each event was rehearsed and for what reason, and rated the affective characteristics of the events. Rehearsal frequency was highest for social communication and lowest for rehearsals aimed at understanding events. For many rehearsal types, rehearsal was more frequent for positive than negative events. Frequently rehearsed events tended to show less affective fading. The pattern changed when events were socially rehearsed. For positive events, increased social rehearsal was related to a reduction in affective fading. For negative events, increased social rehearsal was associated with increased affective fading.

Keywords:

• Autobiographical memory
• Rehearsal
• Emotion
• Fading affect bias

Acknowledgements

We would like to thank many undergraduate research assistants who helped collect and tabulate the data on this extended research project: Lynn Wootton, Joshua Stimple, T. Rene Forst, Scott Powell, and Lakisha Pinson.

Notes

¹We conducted a pilot study to verify that these rehearsal types are frequently used. In this pilot study participants took part in an hour-long discussion of their personal memories and their patterns of event rehearsal. Three researchers asked each participant about specific event memories and the strategies they employed to retrieve their memories. During this discussion, participants were asked to write down the various ways in which they retrieved and thought about their personal memories. Two undergraduate assistants who were blind to the research hypotheses were instructed to code the participants’ responses into as many unique categories as possible (Cohen's κ = 0.83). All disputes were settled by discussion. The assistants then compiled a list of the rehearsal types that were recorded by at least half of the participants. All five of the rehearsal types described in the present article were reported by at least half of the participants.

²Our autobiographical memory studies often generate more data than can be conveniently described in a single paper. Furthermore, because of the variety of data that are collected, a single data set can often address a variety of research questions. This characterises the data set used in the present article. For example, elements of this data set were used in a paper by Walker et al. (2003a); Table 1 was also depicted in Skowronski and Walker (2004). For purposes of clarity, a few results concerning fading affect and valence differences in fading affect are repeated from those earlier articles. However, the results that are reported in the present article, especially the results from the regression analyses, go substantially beyond the results described in either of those two articles.

³An additional between-participants variable that was used in all analyses concerned was the emotional state of the participant (dysphoric vs non-dysphoric). However, because results concerning this variable are extensively discussed in Walker et al. (2003a), in this article we omit discussion of effects concerning this variable. Moreover, to simplify presentation, we omit discussion of effects involving the replication variable. Only two such effects emerged: a main effect for replication, F(2, 329) = 6.82, MSE=.68, p<.01; and a Rehearsal Type×Gender×Replication interaction, F(7, 7549) = 3.73, MSE=.17, p<.01. Examination of the means for these effects suggests that they did not substantially qualify interpretation of other effects reported.

⁴We were concerned that the results obtained in these experiments may have been distorted by the transformation of the rehearsal estimate data necessitated by the unusually high estimates provided by some participants. We replicated the study using frequency estimates reported on bounded response scales, which eliminates these outliers. Analysis of the data from this sample yielded results that were quite similar to those reported in the present paper.

⁵The mean values that are reported are least-squares means that reflect adjustments made because of other variables entered into the regression models. These adjustments tended to enhance the differences that were present in the raw means.

⁶Preliminary analyses were conducted to ensure that the effects reported in the article were not caused by differences in the initial affect ratings provided by participants.

⁷We actually conducted multiple sets of analyses to examine how rehearsal predicted fading affect. These included models reflecting separate examinations of the extent to which each rehearsal type predicted fading affect without controlling for the other rehearsal types. Other analyses simultaneously predicted fading affect using the variables of initial event affect, initial event extremity, and each rehearsal type. Yet another analysis used a five-level rehearsal type variable as a predictor of fading affect, simultaneously entered into a model with initial event affect and initial event extremity. It is obviously not possible to report all model variants. We report the results that we believe most accurately reflect the data. Readers are invited to contact us to obtain results from the non-reported analyses.