Specificity and detail in autobiographical memory: Same or different constructs?

Abstract

Research on autobiographical memory has focused on whether memories are coded as specific (i.e., describe a single event that happened at a particular time and place). Although some theory and research suggests that the amount of detail in autobiographical memories reflects a similar underlying construct as memory specificity, past research has not investigated whether these variables converge. Therefore, the present study compared the proportion of specific memories and the amount of detail embedded in memory responses to cue words. Results demonstrated that memory detail and proportion of specific memories were not correlated with each other and showed different patterns of association with other conceptually relevant variables. When responses to neutral cue words were examined in multiple linear and logistic regression analyses, the proportion of specific memories uniquely predicted less depressive symptoms, low emotional avoidance, lower emotion reactivity, better executive control and lower rumination, whereas the amount of memory detail uniquely predicted the presence of depression diagnosis, as well as greater depressive symptoms, subjective stress, emotion reactivity and rumination. Findings suggest that the ability to retrieve specific memories and the tendency to retrieve detailed personal memories reflect different constructs that have different implications in the development of emotional distress.

Keywords:

• Autobiographical memory
• Memory specificity
• Memory detail
• Emotional distress
• Overgeneral memory

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ Regression analyses using untransformed variables showed a virtually identical pattern of results to regression analyses using transformed variables, in all cases except for only one instance: the fit of the model predicting DASS anxiety among neutral cue words was statistically significant using the untransformed variable (R² = .016, p = .02).

² The average proportion of each type of response to the AMT across all cue valences is as follows: specific memory (M = 0.81, SD = 0.20), extended memory (M = 0.043, SD = 0.076), categoric memory (M = 0.057, SD = 0.104), not a memory (M = 0.017, SD = 0.053), incorrect specific (M = 0.073, SD = 0.140) and same event (M = 0.001, SD = 0.012). Participants retrieved a mean of 3.61 details (SD = 0.97) across all cues.

³ Given recent data suggesting that responses to positive and negative cue words form a single dimension (e.g., Griffith, Kleim, Sumner, & Ehlers, 2012), we reran the regression analyses combining across all cue words. Results were substantively identical to those conducted on the neutral cue words. Among the nine models, those with the following criterion variables were statistically significant: depressive symptoms on PHQ (R² = .014, p = .025), subjective stress (R² = .017, p = .011), emotion reactivity (R² = .021, p = .005), effortful control (R² = .017, p = .019) and rumination (R² = .038, p = .012). The models that predicted PHQ diagnosis, depressive symptoms on DASS, anxiety and emotional avoidance did not reach statistical significance (p = .071, p = .10, p = .42 and p = .071, respectively). As an individual predictor in these combined cue models, memory specificity was significantly associated with more adaptive functioning in terms of lower depressive symptoms on PHQ (β = −0.08, p = .040) and higher effortful control (β = 0.11, p = .007), and was not significantly associated with subjective stress, emotion reactivity and rumination (p = .71, p = .12 and p = .35, respectively). On the other hand, detail was associated with more maladaptive functioning in terms of higher subjective stress (β = 0.12, p = .030), higher emotion reactivity (β = 0.15, p = .009) and higher rumination (β = 0.022, p = .010), and was not significantly associated with depressive symptoms on depressive symptoms on PHQ and effortful control (p = .17 and p = .10, respectively).

⁴ In addition to the reported results, we also recoded our data to assess whether a response included words that referred to the participant's own emotional state experienced during the specific event. Similar to the detail variable, we generated variables that represented an individual's average rate of emotional content embedded in memory responses by averaging binary ratings of the presence versus absence of emotion in each response that was coded as a specific memory. Separate variables were created for overall emotion, positive emotion and negative emotion. Two coders achieved excellent reliability (Kappa's = .93 overall emotion, .86 for positive emotion, 1.00 for negative emotion) on a set of 60 practice items. Overall emotionality was positively correlated with specificity (r = .11, p = .003), but not detail, p > .05. Positive emotionality was not significantly correlated with either detail or specificity. Negative emotionality was positively correlated both with specificity (r = .11, p = .003) and detail (r = .08, p = .049). More importantly, when the emotionality variables were included as covariates in our regression models, the results were virtually identical to our original analyses. Using neutral cue words, the overall model for effortful control (R² = .015, p = .01 to R² = .016, p = .05) and the effect of detail (from β = −0.12, p = .04 to β = −0.10, p = .06) dropped to marginal significance controlling for overall emotionality. When controlling for positive emotionality, the overall model predicting effortful control among neutral cues dropped to marginal significance (R² = .015, p = .03 to R² = .016, p = .05). Finally, when controlling for negative emotionality, the overall model predicting effortful control among neutral cues dropped to marginal significance (R² = .015, p = .03 to R² = .015, p = .06). These analyses suggest that our primary results were not driven by the emotionality of the memory responses.