Abstract
Original article: Akhtar, S., Justice, L. V., Morrison, C. M., and Conway, M. A. (2018). Fictional first memories. Psychological Science, 29, 1612–1619. doi:10.1177/0956797618778831
The authors accidentally missorted the data file in the Supplemental Material for this article, such that the various fields were not properly aligned. The authors thank David Pillemer for alerting them to this error. In the course of correcting that sorting error, the authors detected 128 cases (1.9%) in which the reported age at time of occurrence seemed at odds with the description of the event. In about 85% of those cases, the respondent mentioned one age in the event description but entered a different age when directly queried (e.g., “About the age of 3 yrs having an injury to my hand, crying . . .” dated as having occurred at age 5 years). Most of the remainder were judged to be discrepant because the nature of the described event did not seem to fit with the reported age (e.g., “A rocking sensation in a cot next to my parents bed . . .” dated as having occurred at age 6 years). The authors speculate that these reported ages of occurrence were typographical errors, and they judged that the best course of action would be to delete those respondents’ data. Deleting those cases led to many trivial changes in the statistics reported in the article, which are now being updated; none of these changes affect the substance of the findings.
To correct the missorting error as well as a mistake regarding how respondents were asked to report their age at the time of the event occurrence, the authors are providing new Supplemental Material for this article. This Corrigendum lists the corrections being made to the article itself.
The following changes are being made to the abstract (p. 1612): The beginning of the first sentence is being changed to “In a large-scale survey, 6,313 respondents . . .” The second sentence is being changed to “In good agreement with many other studies, where mean age at encoding of earliest memories is usually found to fall somewhere in the first half of the 4th year of life, the mean age at encoding here was 3.15 years.” The fourth sentence is being changed to “However, we found that 2,479 first memories (nearly 40% of the entire sample) dated to an age at encoding of 2 years and younger, with 887 (14.1%) dating to 1 year and younger.”
In the introduction (p. 1612), the first sentence of the first paragraph is being changed to “In many studies of the recall of earliest memories, the first memory is found to date to the 4th year of life, typically about 3 years 4 months . . .” The opening of the second sentence of the second paragraph is being changed to “Howe, Courage, and Edison (2003), in their review of this research, concluded that the processes underlying the ability to form autobiographical memories are functional by the 4th year of life . . .”
The following changes are being made to the Results section:
In the first paragraph (pp. 1613–1614), the opening of third sentence is being updated to “Further, 39 memories reportedly encoded over the age of 10 years were not used because . . .” A new sentence is being added after the third sentence, and further changes are being made to the following sentences, so the remainder of the paragraph will read as follows: A further 128 memories were removed from the database as the reported age at encoding did not match with the age reported in the memory description. Thus, a total of 6,313 memories were used, and of these, 4,029 were from female respondents (63.8%; mean age = 46.65 years, 95% confidence interval, CI = [46.07, 47.23]) and 2,284 were from male respondents (36.2%; mean age = 44.98 years, 95% CI = [44.23, 45.73]). Of the respondents, 86.3% (5,448) were UK nationals, and the remaining 13.7% (865) resided in other parts of the world. Figure 1 shows the distributions of memories across age groups of respondents and shows clearly that memories were sampled across the life span.
The second sentence of the second paragraph (p. 1614) is being changed to the following: What is immediately evident in Figure 2 is that there were a large number of unexpectedly early memories, with 39.3% (2,479) of the sample having what we term improbably early memories, dating to 2 years and younger (M = 1.64, 95% CI = [1.62, 1.66]); 52.6% (3,322) reporting what we term probable memories, falling between age at encoding of 2 and 5 years (M = 3.63, 95% CI = [3.61, 3.66]); and the remaining 8.1% (512) reporting an age at encoding of 6 or more years (M = 7.29, 95% CI = [7.17, 7.41]), which we term improbably late memories.
The second sentence of third paragraph (p. 1614) is being updated as follows: Despite this unexpected distribution, the overall mean age at encoding of the whole sample was 3.15 years (95% CI = [3.11, 3.19]), which compares favorably with previous findings of the mean age of the earliest memory that place it in the first half of the 4th year of life.
The third, fourth, and fifth sentences of the fourth paragraph (p. 1614) are being changed to read as follows: The sample was therefore split into two new groups: a younger group comprising respondents within the 11– 15, 16–20, and 21–25 age groups (n = 1,189), similar to the majority of participants sampled in Rubin’s (2000) study, and an older group comprised of all remaining respondents (n = 5,124). The mean age at encoding was 3.38 (95% CI = [3.28, 3.48]) for the younger group and 3.10 (95% CI = [3.05, 3.14]) for the older group. These means were reliably different, t(1753) = 5.06, p < .001, d = 0.17, 95% CI = [0.10, 0.23]), showing that the older group had reliably earlier first memories than the younger group.
In addition, a number of the percentages and the examples in Table 1 (p. 1615) are being updated, and several values in Figures 1 and 2 (both p. 1614) are also being changed.
Percentage of Memories Within Each Semantic Category Across Memory Types
Percentage of respondents across age groups.
Frequency of age at encoding grouped by memory type.
In the first paragraph of the Memory Content section (p. 1615), the third sentence and the first half of the fourth sentence are being updated as follows: A Poisson regression with planned comparisons (early vs. probable and early vs. late) found no reliable difference in word count between improbably early memories (M = 69.17, 95% CI = [66.99, 71.35]) and probable memories (M = 67.97, 95% CI = [66.04, 69.91]; p = .16, b = 0.007, 95% CI = [−0.003, 0.016]), but improbably early memories had a reliably higher word count than improbably late memories (M = 66.11, 95% CI = [61.45, 70.78]; p < .001, b = 0.049, 95% CI = [0.034, 0.063]). Although reliably different, memories across all three categories had negligible differences in word count (~3 words) . . .
In the Discussion section (p. 1616), the second sentence is being updated as follows: Yet the main finding of the present survey of earliest memories, the largest such survey ever conducted, is that 2,479 (39.3% of the entire sample) of the earliest memories dated to when respondents were 2 years of age or younger, with, astonishingly, 887 (14.1%) dating to 1 year or younger.
Since initial publication of the article, the authors have also decided that it is not appropriate to include the respondents’ event descriptions on the open Web because doing so might compromise the anonymity of the data. The first sentence of the Open Practices section (p. 1618) is thus being replaced with the following two sentences to clarify where the full data can be obtained: Deidentified data have been made publicly available via the Open Science Framework and can be accessed at https://osf.io/2ezsm/. The full data set with event descriptions can be requested from M. A. Conway (
Finally, Note 2 (p. 1618) is being replaced with the following: Age at encoding was binned into response categories from 0 to 1, 2, 3, and so on. Thus, responses of “1 year” (the shortest option) could plausibly indicate an age up to 1 day short of turning 2 years old, and reports of “2 years” could mean anything including and above 2 years old to 1 day short of the 3rd birthday.