Collective events and individual affect shape autobiographical memory

Abstract

How do collective events shape how we remember our lives? We leveraged advances in natural language processing as well as a rich, longitudinal assessment of 1,000 Americans throughout 2020 to examine how memory is influenced by two prominent factors: surprise and emotion. Autobiographical memory for 2020 displayed a unique signature: There was a substantial bump in March, aligning with pandemic onset and lockdowns, consistent across three memory collections 1 y apart. We further investigated how emotion, using both immediate and retrieved measures, predicted the amount and content of autobiographical memory: Negative affect increased recall across all measures, whereas its more clinical indices, depression and posttraumatic stress disorder, selectively increased nonepisodic recall. Finally, in a separate cohort, we found pandemic news to be better remembered, surprising, and negative, while lockdowns compressed remembered time. Our work connects laboratory findings to the real world and delineates the effects of acute versus clinical signatures of negative emotion on memory.

Keywords: autobiographical memory; collective memory; emotion; surprise; temporal memory.

Fig. 1.

(A) Timeline of real-world events in the United States and COVID-Dynamic administrations across 2020. Orange triangles denote each wave administration (black tick marks depict weekly intervals). The gray curve indicates the daily 7-d average of new, confirmed COVID-19 cases in the United States, black encircled X’s on top of the curve indicate death milestones. The green line shows the monthly unemployment rate. The upper gradient (yellow-red) indicates the daily count of states with active stay-at-home restrictions (peak = 41). The lower gradient (blue-purple) shows the daily count of US antiracism crowd events initiated after the murder of George Floyd. Triangles below the gradients indicate local maxima for the various measures. Events of interest are indicated with vertical blue lines (credit: COVID-Dynamic Study, data are publicly available, see SI Appendix, section 5.1 for resources; 20). (B) Topic modeling for 2020 autobiographical memory distributed by month (memory recalled in December 2020, see SI Appendix, section 2.3 for topic modeling of all sets). Memory entry count (Y axis) represents the number of memory entries, grouped by topic category. Topics included categories related to personal matters (“social,” “occupation,” and “domestic”) as well as real-world events (“covid,” 2020 presidential “election,” Black Lives Matter movement (“BLM”), and “‘covid-onset” which describes the conditions of the early pandemic (e.g., toilet paper shortage); see SI Appendix, Table S2 for keywords associated with each topic. A similar pattern of topics was observed for all 2020 memory collections (but not for 2021 memory, which included different real-world topics, SI Appendix, Fig. S2).

Fig. 2.

Distribution and characteristics of recollections by month. Participants first recalled events as they spontaneously remembered them and then dated each memory by month. (A). Mean across within-participant proportions (including all possible date options) of memory for each month. Memory for March 2020 was greater when recalled that year, as well as one (2020+1) and two (2020+2) years later; this pattern was not observed for memory of 2021. Colors indicate which year and when it was recalled (e.g., “2020+1” is memory for 2020 recalled 1 y later); error bars are SEM across subjects. Note that 2020 and 2021 memory collections occurred in early December of those years (i.e., recency effects are captured by November memory; for those collections, dotted lines between November and December are omitted). See SI Appendix, Fig. S4 for individual-level plots across all date options. (B). Average sentiment score for retrieved memories ranging from −1 (negative) to 1 (positive). Memory for March 2020 was more negative than other months of 2020; this was not the case for 2021 memory. Additionally, more negative sentiment predicted a greater amount of recall for that month, regardless of memory collection. (C). Proportion of participants’ dated recall for their first through third entries by memory collection; colors indicate recall number, black dotted line is the average across first three entries. In addition to the tendency to recall a year sequentially (i.e., higher proportion of January memory as the first entry), March was recalled earlier than other months of 2020; this pattern did not occur for 2021 memory.

Fig. 3.

Likelihood of later retrieving a month and its sentiment given experienced emotion during that month. We administered a series of assessments from April to December of 2020 that evaluated immediate and longer-term affect and tested whether these measures predicted the likelihood of that month being later remembered in autobiographical memory (A–D). We found that negative affect, as assessed by (A) immediate negative affect (39), (B) state and trait-level anxiety (40), (C) experienced stress (41), and (D) depression (42), increased the likelihood of later recalling that month. Shaded regions reflect 95% CIs. (E–H). We also found that all measures predicted the sentiment of autobiographical memories (ranging from −1, negative, to 1, positive) in the expected direction: (E) immediate negative affect led to more negative sentiment for that month, while immediate positive affect led to more positive sentiment; (F) increased state and trait-level anxiety, (G) stress levels, and (H) depression also predicted more negative sentiment for that month. Shaded regions reflect 95% CIs. Marginal distributions represent the variable on the corresponding axis.

Fig. 4.

Individual differences in affect and PTSD predict autobiographical memory. (A and B). We conducted an exploratory and confirmatory factor analysis on all affect measures and found six meaningful factors that varied on valence (positive or negative), arousal (low or high), and relevance to self: negative low-arousal, negative high-arousal, negative self, positive low-arousal, positive high-arousal, and positive self (colors represent each factor). We found that negatively valenced factors predicted greater amount of recall and more negative sentiment in autobiographical memory, whereas positively valenced factors predicted the opposite. (C–D). We also found that PTSD symptoms (memory intrusion, avoidance of trauma-related thoughts, negative cognition, and hyperarousal; 43) led to more recall (although specifically for nonepisodic or external details, see the main manuscript) and more negative sentiment in autobiographical memory (colors represent each symptom, measures were each z-scored for visual comparison). Shaded regions reflect 95% CIs.

Fig. 5.

Collective and affective memory for top 2020 news events and estimated time between them. We asked a new cohort of participants in 2022 to rate the two top 2020 news topics for each month on memorability and remembered affect (ratings ranged from 0 to 100). (A) Surprise ratings for news events, averaged by month (emergency COVID-19 news included in March), yielding a similar pattern to the distribution of 2020 autobiographical memory. (B) The strength of remembered surprise, negative affect, and positive affect were each independently associated with greater memorability of retrieved news items. (C) Participants also estimated how close or far apart pairs of news events were in time (from 0, very close, to 100, very far); some pairs spanned the predominant lockdown period in the United States (Fig. 1A), and some did not. Controlling for actual distance, we found pairs that spanned the lockdown were judged as closer to each other than pairs that did not (there were no differences between pairs that occurred pre- and postlockdown).