Bringing Order to Life: Temporal Order Effects during the Recall of Important Autobiographical Memories in Young and Old Adults

ABSTRACT

Objective

The present study investigated whether younger and older adults differ in the extent to which they order their important autobiographical memories (AMs) in a chronological way during recall. Moreover, the number of autobiographical memories to be recalled was systematically varied in order to examine whether manipulating the size of the search sample affects the use of a chronological recall strategy.

Methods

Data come from 118 younger adults (M = 23.12, SD = 3.33) and 124 older adults (M = 69.79, SD = 8.94) who recalled either seven or thirteen important AMs.

Results

Using multilevel linear growth models, a temporal order effect for younger and older adults was found. For older adults, the temporal order effect was much more pronounced. The number of recalled events, however, did not affect the temporal order effect.

Conclusion

Our findings point to a higher need for a structured search or a stronger engagement in autobiographical reasoning processes in older age. Furthermore, asking for important AMs encourages participants to provide a comprehensive overview of their life independent of the number of important AMs to be recalled.

Acknowledgments

The authors would like to thank the students from Ulm University for their invaluable help in collecting the data used in the present study.

Disclosure statement

The authors declare that they have no relevant financial or non-financial interests to disclose.

Data availability

The SASR syntax used in the current study is provided in Appendix A3. The dataset is stored at the Mendeley data repository.

Notes

1. Note the German “Hauptschulabschluss”, “Realschulabschluss”, “Fachhochschulreife”, and “Abitur” graduations correspond to approximately 9, 10, 12, and 13 years of formal education, respectively.

2. At the time of data collection, the ethics committee of Ulm University provided a checklist to consider whether a full application for ethical approval was needed. The current study did not meet the suggested criteria. Participants were provided with detailed information about the respective study. Participation was voluntary and required written consent.

3. We decided to ask for 7 AMs because this is the common number when participants are asked for their most important AMs (Zaragoza Scherman, Salgadoa, Shao, & Berntsen, 2017; Rubin, Berntsen, & Hutson, 2009). The decision to ask for 13 important AMs is based on the study procedure where participants were also asked to name 7 or 13 life events of a prototypical life course (life script task, e.g., Berntsen & Rubin, 2002). In order to keep participants’ burden reasonable, we limited the total recall of participants to 20 events (either 7 memories and 13 prototypical life events or 13 memories and 7 prototypical life events). The two tasks were randomized.

4. An anonymous reviewer asked to indicate why the additional measures were not included in the analyses. The present analysis is of data collected in an extensive online study that was designed for several research questions, among them age differences in life goals and values. Therefore, the questionnaires listed above were not included in the present data analyses.

5. Note that the variable output position ranged from 0 to 6 (7-memories condition) and 0 to 12 (13-memories condition), respectively. By scaling output position this way, the intercept captures the (average) age-at-event at the first output position (cf. Hedeker & Gibbons, 2006).

6. A more conservative approach to tackle the problem of substantially different outcome variable variances would be to rank age-at-event values and then use these ranked values as the outcome variable. In additional analyses presented in the supplementary material, we followed this approach. As Table A1 shows, results are similar with respect to significance and explained variance.

7. The total expected number of AMs is calculated as 124 (sample size in the 7-memories condition) × 7 (number of memories) + 118 (sample size in the 13-memories condition) × 13 (number of memories) = 2402.

8. At first glance, it may—as the intraclass correlations show—appear remarkable that the amount of variance between persons was larger in the young group than in the old group. However, if one takes into account that the possible range of age-at-event was much larger in the old group, the Level 1 variance of age-at-event also has the potential to be larger in the old group.

9. Additional analyses showed that in the young group, 26 (21%) of the participants showed an output position effect estimate smaller than zero—implying a backward temporal order effect—, while in the old group, only 2 (2%) participants did.

10. The interaction between sex and output position did not reach significance in either age group, and effect sizes were minimal (results not shown).

11. Note that in relative size, the difference in output position slope was even larger in the young group because the increase in the age-at-event increase was 0.69/0.36 ≈1.9 less in the 13-memories condition compared to the 7-memories condition. The failure of the slope difference to reach significance is due to the signal-to-noise ratio (the output position effect is much smaller in the young group than in the old group), which leads to larger standard errors of parameters.

12. Interactions between experimental condition and sex were also investigated but did not reach statistical significance and were of negligible effect size. The same holds for the triple interactions among experimental condition, output position, and sex.

13. As mentioned above (see Footnote 5), the variable output position ranged from 0 to 6 (7-memories condition) or from 0 to 12 (13-memories condition), respectively. Hence, a division by two in the 13-memories condition led to a maximum output position value of 6—equal to that in the 7-memories condition.

14. Note that calculating the likelihood difference between Models 3 and 4 is not applicable because both models have the same degrees of freedom.

15. We conducted two additional analyses. First, importance of the AM was used as the outcome variable. In an according multilevel growth model, there was no significant linear or quadratic effect of output position in neither the young nor the old group. Moreover, effects were very small. Second, comparable to Nusser & Zimprich (2021), we investigated whether emotional valence of the AM affected the output order. In a mixed-effects autoregressive model, we did not find a significant autoregressive effect in young or old adults. Furthermore, effects were virtually zero.

16. Note that the difference between the explained variance in the young and the old group was similar in the rank-order model and the linear growth model. This implies that the age difference in the temporal order effect is barely affected by the differences in the range of age-at-event between the two age groups and supports the robustness of the age differences found in the present study.