Edited by:
Reviewed by:
*Correspondence:
†These authors have contributed equally to this work.
This article was submitted to Cognitive Science, a section of the journal Frontiers in Psychology.
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
Each of us has a rich set of autobiographical memories that provides us with a coherent story of our lives. These memories are known to be highly structured both thematically and temporally. However, it is not known how we naturally tend to explore the mental timeline of our memories. Here we developed a novel cued retrieval paradigm in order to investigate the temporal element of memory search. We found that, when asked to search for memories in the days immediately surrounding a salient cued event, participants displayed a marked set of temporal biases in their search patterns. Specifically, participants first tended to jump back in time and retrieve memories from the day prior to the cued event. Following this they then transitioned forward in time, and retrieved memories from the day after the cued event. This pattern of results replicated in a second experiment with a much larger group of participants, and a different method of cueing the memories. We argue that this set of temporal biases is consistent with memory search conforming to a temporally ordered narrative structure.
Humans may be unique in their ability to vividly and richly re-experience the important events from their past, an ability which has been referred to as “mental time travel” (
There are two competing hypotheses regarding possible biases in the direction of temporal memory exploration. The first hypothesis comes from studies investigating the retrieval of semantic information over a set of years. Participants were directed to either start at the beginning of the time period and work forward through memory, or start at the end and work backward. Increased accuracy was found when working backward through time (
Both of these conflicting sources of evidence come from studies that have specifically directed participants to explore either forward or backward through time, and therefore none of these studies directly speaks to the issue of how we naturally tend to explore our memories in time. The current study was designed to investigate any biases in the direction of memory exploration in conditions where participants were free to temporally explore in either direction. In order to accomplish this, we developed a novel autobiographical memory testing paradigm. Subjects were required to initially recall a salient personal event and then recall events that occurred in the days immediately around the cued event. Across two experiments, we used this paradigm to investigate possible biases in the temporal direction of memory exploration in order to compare the two competing bias hypotheses against a third null hypothesis of no temporal bias.
In order to determine whether there is a spontaneously elicited temporal direction of autobiographical memory recall, we designed a preliminary study in the form of structured interviews about participants’ autobiographical memories. Because of the great importance of autobiographical memory for the sense of self and the development and maintenance of close interpersonal relationships (
In a structured interview, participants were asked to recall seven personally experienced events. These included: when they last saw a friend from their hometown, a friend from primary school, a friend from high school, a friend from university, a co-worker from their first job after the age of 16, the friend they think they have known for the longest and a family member they only see at family gatherings or occasionally. For all cues, they were prompted to recall events that occurred longer than a week ago. The order of the cues was randomized across subjects.
Twenty females participated. The subjects’ mean age was 26.0 (SD = 4.52) with the range between 19 and 37. All subjects were healthy native English speakers born in the UK. All gave informed written consent and were paid £7.50 per hour for their participation. Testing duration was approximately 20 min. The research was approved by the local research ethics committee.
The test was carried out over the phone in the form of a structured interview and the phone calls were recorded and later transcribed. The time was scheduled ahead with the subjects who were asked to reserve 30–45 min in a peaceful environment.
The subjects were told that the recall task will involve thinking about their network of friends and acquaintances. They were then cued to recall the last occasion in which they had met one of the categories of friends/family listed above (Stimuli and Materials). We refer to this as the
It was emphasized that they should report any specific elements that they clearly remember from the events in the surrounding days and not just assume what may have happened. Subjects were prompted to recall everything they could remember about the surrounding days until they indicated that they could not recall any further events.
The procedure was initially explained to subjects with an example, asking them to recall when they last saw their nearest neighbor, to ensure they understood the task instructions.
To test whether there was a bias in the temporal direction of memory search from the Cued Event, we calculated a bias score. Participants were instructed to recall as many events as they could from the days immediately surrounding the Cued Event, meaning that the number of events recalled varied across trials. The bias score was calculated separately for the first event recalled (
We tested three alternative hypotheses about the data by examining the bias scores (see Figure
Subjects were able to describe details from at least one event that occurred in the days immediately around the Cued Event (Event 1) on 70.7% of trials (SD = 22.35), and just over half recalled a second event (Event 2: mean percentage of trials for which an event was recalled = 54.5%, SD = 27.51). The mean bias score for Event 1 was –0.649 (SD = 0.40) and the mean bias score for Event 2 was 0.537 (SD = 0.57), see Figure
Such temporal structure was apparent in participants’ responses:
and
Our results therefore show a bias to initially recalling the day before when exploring autobiographical memories around a salient event. However, it is possible that there is a bias in the amount of information recalled in the day before compared to the day after, which could explain our results without there being any bias in the natural direction of memory exploration. To test this we examined the mean number of verifiable details recalled for each of the days around the Cued Event (for example, the location, the time, and the people present). There was no difference between the number of details retrieved for the days around the salient event [day before = 1.98 details (SD = 1.14), day after = 1.57 details (SD = 1.47),
In summary, we find that when people search their memory for events on the days around a cued event they show a bias in the direction of memory search, exploring backward in time to the day before. However, following this initial backward search, they then show a bias in progressing forward, to the day following the cued event. To our knowledge this bias in autobiographical memory search has not been previously reported and is consistent with hypothesis 2 in our formulation of possible outcomes (Figure
Having established evidence for an effect, we next sought to: (a) replicate our findings in a larger population, (b) test male subjects, and (c) assess the effects without the interaction with an interviewer. While our interviews were carefully constructed to avoid cueing the subject to recall events from the past or future days we cannot rule out the possibility that subtle social cues may have influenced the subject’s behavior.
Based on the promising results of the first experiment, we adapted the task as a controlled and structured investigation into the organization of autobiographical recall. Instead of describing events in long-form sentences, the participants were only required to think of the first, then the second, and then the third word that came to their minds when thinking about the days surrounding an event.
We were able to collect data from 60 male and 60 female participants. A
MATLAB (2011b, MathWorks) and Psychtoolbox (v3,
The participants’ demographics were first collected. Afterward, instructions were displayed to them, explaining to them that they will be asked to think about the events they experience with people from their life. They were told that they would be given some time to think about each person and then asked to write down the first word that came to mind. They were then told that they would be asked to think about the days
It was emphasized to participants that they should think about the days immediately surrounding the event they had just recalled—critically, however, the words “before” and “after” were never used to avoid biasing them. They were prompted to try their best to recall an event, and spend some time thinking about it even if they could not immediately recall anything. They were instructed to think of a different person if an event had already been brought to mind in a previous trial or if it happened within the past week.
Each memory cue was displayed for 15 s which were counting down on the screen, again preceded by a practice trial. During this time, the participants were asked to recall the displayed event and think about it. Below the event, the following prompt was always displayed: “
They were then asked to type the first word that came to their minds from the day of that event: “
After this, they were asked to recall the first, then the second, then the third word that came to their mind from the days surrounding this event:
“
After they entered the word, they were taken to the next screen, where the same prompts were displayed for the second and third words. This was repeated for each of the nine memory cues. After all trials were completed, they were then told that they will be shown each of the three words they typed, one at a time, and they will have to decide whether each of them is related to the day before or to the day after each event. They were instructed to press the left arrow key if the word described an event on the day prior to the anchor event and the right arrow key if it described an event that occurred after the anchor event. They were instructed to press the “up” arrow key if they typed “none.” They were then told that they will have to provide some context for everything they typed after they finish all trials.
After the participants had indicated the temporal direction for all words, they were finally shown a screen containing all four words that they had entered for the event itself and each of the surrounding events and asked to provide a sentence to describe the context surrounding these words. After they had described each of the events, the experiment was completed.
The three words provided by the participants for the days surrounding each event were coded according to whether they were referring to days before or after the event. Similarly to Experiment 1, if the participant indicated that a word was referring to the days before an event, this was coded with –1 and if it was referring to the days afterward, it was coded with 1. The average of all provided responses was calculated for each of the events, relying on the context provided by the participants. The three words provided by participants did not necessarily refer to three distinct events. As we aimed to examine recall direction for specific events, any words that referred to the same event were collapsed into a single event score. The same analysis as for Experiment 1 then applied.
We were initially only interested in the first event reported by the participants, which was also expressed in the instructions, where an example of words provided was only given for an event on “one of the days surrounding the event.” However, participants commonly provided more than one event. Event 1 was provided in 96.4% of trials and Event 2 in 47.7% of all trials. six females and one male could only recall one event from the days surrounding the Cued Event. Only 8% of all trials included a word referring to a third event occurring in the days surrounding the memory cue. These were not included in the analysis as the statistical power of the test would be insufficient.
The average bias for Event 1 was –0.13 (SD = 0.50) and the average bias for Event 2 was 0.31 (SD = 0.56). The average bias score for Event 1 was –0.164 (SD = 0.51) for females and –0.093 for males (SD = 0.50). For Event 2, the average bias score for females was 0.37 (SD = 0.57) and 0.24 for males (SD = 0.55). An independent samples
Figure
Two one sample
While the initial backward search bias is a clear temporal bias, this might not be the case for the subsequent forward bias for Event 2. It is possible that this bias might be due to limitations in the total number of memories available per day, rather than any genuine temporal biases. For example, if we assume that participants have access to a maximum of one event per day, then the constraints of the task will automatically produce this apparent temporal bias. This is because they are limited to searching memory in the two days immediately surrounding the cued event. As soon as they have retrieved a memory from the day before the cued event, they do not have access to any more events on that day. The only possible remaining memories will be on the day after the cued event, forcing participants to move forward in time to continue searching for a second memory. This pattern of results would appear to follow a temporal search from past to future, but this would merely be an artifact of the underlying memory constraints, and not a true temporal bias. We therefore refer to this alternative account as the “limited memories” hypothesis. We next explored the precise pattern of memory transitions in order to determine whether the apparent bias for Event 2 is genuinely temporal in nature.
To accomplish this, we made use of the fact that there were many trials where Event 1 was recalled from the day after the cued event. This property is important, because the limited memories account would predict a symmetrical pattern of memory transitions. In other words, if the only factor driving the apparent transition from past to future is a limit in the number of accessible memories, then we ought to see an equally strong effect in the opposite direction, from future to past, on those trials where Event 1 was retrieved from the day after the cued event. In contrast, if there is a genuinely temporal explanation for the bias, then there should be a significantly greater probability of transitioning from the past to the future, then from the future to the past.
To test these two hypotheses, the probability of forward and backward transitions was calculated. A
Forward transition probability was calculated as the number of forward transition trials divided by the total number of trials where the first recalled event was before the anchor, and a second event was provided. Backward transition probability was calculated as the number of backward transitions divided by the total number of trials where the first recalled event was after the Cued Event, and a second event was provided. For example, if a participant showed a forward transition (“before” to “after”) on four trials, and showed no transition (“before” and “before”) on three trials, their forward transition probability would be 4/(4 + 3) = 0.57. This would indicate that if a participant initially recalls an event before the Cued Event.=, they have a likelihood of 57% of transitioning through time to events after the Cued Event. If the overall trend is indeed that participants prefer to move forward in time, the
The mean forward transition probability was 0.629 (SD = 0.311) and the mean proportion of backward transitions was 0.425 (SD = 0.351). A paired samples
We next considered the possibility that affective valence of a recalled event might have influenced the direction of recall (
Finally, we compared the main results from experiments 1 and 2 in order to determine whether there were any significant differences in effect size between the two. An independent samples
These results replicate those of Experiment 1, and demonstrate a temporal bias in memory exploration, such that participants tend to initially search back in time until they find one memory, then proceed forward in time until they find another memory. They further show that the forward transition between memories is a genuinely temporal bias, rather than being due to other explanations such as memory limitations. Finally, these results demonstrate that the temporal bias effects hold across two very different means of cueing the memories, and are not simply an artifact of interview-based memory cueing.
The two experiments reported here provide, to the best of our knowledge, the first systematic investigation of the order of recall of events in the days immediately surrounding a single cued event. The results suggest that there is a significant tendency to chronologically structure autobiographical recall such that we prefer to first recall events before a salient memory anchor and then mentally travel forward in time and continue to recall events after it. This initial backward bias, followed by a forward transition tendency, was observed regardless of the number of details participants provided for each of the surrounding days (Experiment 1) or the valence of the details (Experiment 2). Thus the effect may be assumed to exist separately from the vividness or richness of details and the emotional context of the memory.
We propose that this pattern of results is consistent with the idea that autobiographical memories are linked together into a coherent narrative structure (
This pattern of results is consistent with that reported by
One possible limitation of the current study was that the retrieved memories were not externally validated as real memories. While spontaneous confabulation is rare in non-clinical populations and tends to be restricted to patients with frontal lobe damage (
One important question raised by these results, and others in the autobiographical memory literature, is how we encode the temporal information of new memories. Given that there is a clear temporal structure among our autobiographical memories, and a clear bias in the way that we explore these memories, there must be some mechanism by which this temporal information is stored. Research on temporal coding of sequence recall has produced compelling evidence for models such as the Temporal Context Model (
According to the Temporal Context Model, the retrieval of one item from a series facilitates the recall of the subsequent item. Thus, one interpretation of this model is that it would predict an initial recall of the day after the cued event. This is not what we observed. However, the focus of the Temporal Context Model is temporal coding over short periods of time, and it has not yet been adapted to explain the existence of temporal structure over much longer periods of time. We propose that the cued event acts as an anchor which activates the representations of events on the surrounding days. The cue “days immediately surrounding the cued event” leads subjects to set their retrieval orientation to the start of the experiences to be recalled. Once the first day is recalled, recall spreads forward to the day after. This forward driven recall may occur via mechanisms described in the Temporal Context Model (
In summary, we provide the first evidence of a bias in the temporal direction of autobiographical memory exploration, such that we preferentially recall the days before a cued event and then proceed by recalling the days after the initial autobiographical anchor. This temporal bias is consistent with a narrative structuring of memories, even under circumstances that minimized any social benefits of narrative structuring. We suggest that these results are therefore consistent with an automatic structuring of memories into a coherent narrative, consistent with current theories of autobiographical memory (
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
We thank Eva Berlot, Yuhua Guo, and Alice Wickersham for help with piloting the computer-based test, Helena Gellersen, Sarah Jeffs, and Ronda Embick for help with data collection, and Catherine O’Shea for prompting us to explore this research topic. This research was funded by a Wellcome Trust Project Grant and a James S. McDonnell Foundation Scholar Award to HJS.