Figure 1.
Graphical appearance of the Cue-Recalled Autobiographical Memory (CRAM) test user interface.
Each panel represents a separate part of the CRAM test. (A) Subjects first label 30 memories recalled upon presentation of 7 words stochastically sampled from their natural language usage frequency. (B) Each memory is then dated into one of 10 temporal bins, based on the subject's age at the time of the event, the date of the event, and/or the time lapsed from the event. (C) Finally, participants score the content of 10 memories by counting the number of elements recalled from the event for each of eight distinct features (People … Details). Every feature is accompanied by a brief definition, schematically illustrated here by a few dotted lines underneath (see section B of the Supporting Information S1 for their full text).
Figure 2.
Temporal distributions of memories collected with the CRAM test.
(A) Consistent with previous findings, the CRAM temporal distribution of autobiographical memories shows a retention effect in the most recent time bins and a power decay toward remote bins. Data re-plotted (with permission) from previous studies, i.e. Jansari & Parkin [20], and Rubin & Shulkind [11], are adapted by converting memory age to youth. The term youth reflects the age of the subject at the time of the recalled episode relative to their age at the time of study participation, and is grouped into 10 bins from the most remote (0) to the most recent (9) episodes. Temporal distributions are also equivalent between the Excel and web-based CRAM test formats (inset). (B) Temporal distributions, plotted as mean ± standard error across individual subjects to enable statistical comparison, are not significantly different between genders and between native and non-native English speakers (inset).
Table 1.
Total content from all 1424 scored memories reported as the sum of elements from all features of each memory.
Figure 3.
Total content increases for memories retrieved from the first to the last tenth of life. Boxes represent the mean, 25th, and 75th percentiles, while whiskers indicate 95% confidence intervals. Scored memories are divided into remote (bins 0–7) and recent (bin 8 and 9) time intervals, which account for 45.4%, and 54.6% of all dated memories, respectively. The total content of recent memories is 31% greater than that of remote memories (p<0.0001). The ratio between the two, i.e. the memory content lost to temporal degradation, is shown in the inset as the average (dotted line) and for two subject partitions as mean ± standard error.
Figure 4.
Feature composition of memory content.
(A) The count of elements in each of the eight features is reported as mean ± standard error over all scored memories. More elements pertaining to Places and People are recalled than those pertaining to Contexts and Episodes. The dashed line represents the average number of elements (2.5). (B) The relative feature composition of recent and remote memories is similar (the slices are ordered by the recent rank). Although Feelings shifts from 4th to 6th rank from the remote to the recent distribution, the actual value change is very modest. (C) The content ratio (remote/recent) across all features was similar between genders with the only exception of Details, for which the decay was significantly greater in males (light hatching) than in females (dark hatching). (D) Recall composition of elements and youth in a typical AM. (E) For all features except one, the variability in the number of elements within subjects is greater than that between subjects. However, variability of total content is similar within and between subjects.
Table 2.
Feature Cross-Correlation.
Figure 5.
Frequency and duration of autobiographical memories.
(A) The Spontaneous Probability of Autobiographical Memories (SPAM) was sampled on a participant-by-participant basis by dividing the number of memories reported by the total phone call prompts. (B) The mean and standard deviation of AM duration was computed for each subject over an average number of memories per subject of 48.5 (the standard deviation of this number was 29.3, range 3–133). Mean durations are directly and significantly correlated with the standard deviation. The slope of the best fitting line indicates a coefficient of variation of 0.65 with a 95% confidence interval of ±0.07 (dashed lines). (C) The number of AMs experienced per hour was calculated for every subject by multiplying his/her sampling probability by 3600 and dividing the result by the same individual's mean duration in seconds.
Table 3.
Quantification of the spontaneous occurrence of autobiographical memories.
Figure 6.
Retrieval probability of memories and features across youth.
Integration of CRAM and SPAM data allows a number of quantitative estimates. (A) The average time elapsing between two memory recalls (e.g., more than 3 hours for the most remote AMs, but less than 10 minutes, see inset, for the most recent AMs). (B) The average recall rate of total content (e.g., more than 3 total elements per minute from the most recent tenth of life, but less than 3 per hour from the most remote tenth of life, see inset). (C) The number of elements for each feature recalled in one hour from each life period (the linear fit in the log scale, indicative of a power function, represents the average across features). (D) Momentary probability of recalling at least one element of a particular feature from a distinct bin, grouped in quartiles.