Table 1.
Experiment 1 participant demographic information by test environment.
Fig 1.
Task schematic used for Experiment 1 and online testing in Experiment 2.
Includes stimulus images reproduced from the Konklab image repository (https://konklab.fas.harvard.edu) under a CC BY license and reproduced with permission from Konklab.
Table 2.
Memory performance by test environment.
Fig 2.
Model fit to response errors in the (left) in-person and (right) online studies.
Fig 3.
Memory performance by test environment.
Plots (A-D) present the median as the central line, the first and third quartiles as the box edges, and the lines indicate the minimum and maximum values (within the range of the first and third quartiles multiplied by 1.5 times the interquartile range) for (A) mnemonic discrimination, (B) mean absolute localisation error, (C) semi-parametric retrieval success, and (D) semi-parametric precision. Plots (E-F) show the distribution of standardised group differences in (E) retrieval success and (F) precision (K) generated using a permutation procedure. The red star indicates the standardised actual difference in the model parameters (Online – In Person).
Table 3.
Experiment 2 participant demographic information.
Table 4.
Non-responder analysis.
Fig 4.
Top: Likert scale of 0-5 where 0 = never and 5 = daily activity. Bottom: Scoring procedure consistent with [23]. Higher scores indicate more engagement. After applying Bonferroni correction (α = 0.003), differences in playing a musical instrument and developing an artistic pastime were no longer significant. Abbreviations: PA, Physical Activity (mild, moderate, vigorous).
Fig 5.
Marginal effects of precision model displaying three-way interaction effect.
Age and precision values have been transformed to z-scores. Shading represents 95% confidence interval. A negative relationship between age and precision was observed for female participants at both time-points. In male participants, there was no marked age effect on precision at T1, however, at T2 there was a positive age effect on precision with older male participants performing better than younger male participants.
Fig 6.
Marginal effects of retrieval success model displaying three-way interaction effect.
All continuous values have been converted to z-scores. Bars represent 95% confidence interval. Results suggest that the lifestyle score associated with each time point (typical, pandemic) had a greater impact on memory performance at each time-point in participants with a family history of dementia. In participants with a family history of dementia, time-point had a more negative effect on retrieval success when participants had a higher lifestyle score, indicating higher lifestyle engagement.
Fig 7.
Marginal effects of change in retrieval success model displaying three-way interaction between the time between sessions, lifestyle change, and depression at T2.
All values have been converted to z-scores. Shaded areas represent 95% confidence interval. Higher scores on the depression scale indicate a greater degree of depressive symptoms. Results suggest that lifestyle change did not have a substantial influence on the change in retrieval success, with increased time, in participants with higher depression scores (+1 SD). In contrast, for participants with lower depression scores (−1 SD), the predicted change in retrieval success differed substantially with lifestyle change, such that a longer time between T1 and T2 had a more negative effect on retrieval success in individuals with greater reduction in lifestyle engagement.