Narratives bridge the divide between distant events in episodic memory

Many studies suggest that information about past experience, or episodic memory, is divided into discrete units called “events.” Yet we can often remember experiences that span multiple events. Events that occur in close succession might simply be linked because of their proximity to one another, but we can also build links between events that occur farther apart in time. Intuitively, some kind of organizing principle should enable temporally distant events to become bridged in memory. We tested the hypothesis that episodic memory exhibits a narrative-level organization, enabling temporally distant events to be better remembered if they form a coherent narrative. Furthermore, we tested whether post-encoding memory consolidation is necessary to integrate temporally distant events. In three experiments, participants learned and subsequently recalled events from fictional stories, in which pairs of temporally distant events involving side characters (“sideplots”) either formed one coherent narrative or two unrelated narratives. Across participants, we varied whether recall was assessed immediately after learning, or after a delay: 24 hours, 12 hours between morning and evening (“wake”), or 12 hours between evening and morning (“sleep”). Participants recalled more information about coherent than unrelated narrative events, in most delay conditions, including immediate recall and wake conditions, suggesting that post-encoding consolidation was not necessary to integrate temporally distant events into a larger narrative. Furthermore, post hoc modeling across experiments suggested that narrative coherence facilitated recall over and above any effects of sentence-level semantic similarity. This reliable memory benefit for coherent narrative events supports theoretical accounts which propose that narratives provide a high-level architecture for episodic memory. Supplementary Information The online version contains supplementary material available at 10.3758/s13421-021-01178-x.

seconds. Audio was then filtered with a high pass at 80 Hz and low pass at 16kHz, peaks were normalized within and across clips, and finally a compressor was applied to all audio to maximize loudness without a tradeoff in subjective clarity. This approach was undertaken in order to minimize low-level differences in audio across all recorded sentences, and to enable smooth transitions between sentences to be perceived as continuous.
By standardizing all sentences at 5 seconds each, we were able to equalize the length of audio for all stories and all events. Each of the four stories contains 48 sentences at 5 seconds each, and therefore each story elapses for 4 minutes. Within each story, each sideplot event contains 8 sentences (40s each), and these sideplot events are bookended by 8 sentences of the main protagonist story which do not involve one of the four side-characters of interest (40s each). There are therefore 16 main protagonist sentences placed between two sideplot events, however, this is comprised of two separate main protagonist events, because there is a temporal shift in the 9 th of these sentences (e.g. "twenty minutes later," "one hour later"). Previous research has demonstrated that temporal shifts are associated with the perception of boundaries which segment events in perception and in memory (Zacks et al., 2007(Zacks et al., , 2009. Pilot experiments conducted in an independent sample (data not shown) using two alternate sideplot arrangements demonstrated that participants reliably perceive an event boundary at these temporal shifts within our constructed stories, as well as at the starts and ends of side character appearances (black bars in Figure 1). Therefore, we presume that each sideplot event is a distinct event surrounded by two main protagonist events, and that there are two main protagonist events which separate two sideplot events present within a given story. This was in line with our intent to minimize any potential for interactions between sideplot events within a given story. In all, main protagonist and sideplot events are distinct events of equal length (8 sentences/40 seconds each), and the quality and timing of audio clips should not be involved in any observed effects.

Familiarization tasks
Three brief tasks aimed to familiarize participants with ten characters from the stories, in order to increase the likelihood that character names and relationships could be used as successful recall cues. Two characters were the main protagonists (Charles, Karen), four characters were CN or UN side-characters (Beatrice, Melvin, Sandra, Johnny), and four additional characters had supporting roles within the main protagonist stories (i.e. main plots, not sideplots). Participants were instructed to "get to know these characters," because they would appear in the upcoming stories. In the first task, character names were presented alongside faces (Bainbridge et al., 2013) that were selected for high memorability and diversity. Each of 10 name-face pairs appeared on the center of a black background for 4 seconds, followed by a 1s fixation screen, and were repeated for a total of 20 trials. In the second task, participants were presented with character-character relationships for the same ten characters, and were instructed to familiarize themselves with these relationships. For each of 20 trials (2 repetitions per pair, 4s per trial, 1s fixation), character names and faces were presented side-by-side, with their relationship displayed on the right side of the screen (e.g. "best friend"). The third task was identical to the second task, except there was a response component. On 20/30 trials, relationships were correct, whereas on 10/30 trials, relationships were incorrect. Participants were required to indicate whether the relationship was correct or not (3s response window), after which they were shown the correct relationship (2s), which was displayed in green if previously correct, or in red if previously incorrect. In all, these tasks were designed to orient participants' attention to characters, their relations, and the potential for these characters to be involved in situations within the upcoming stories, with the aim of successfully evoking subsequent recall by cuing with character names.

Sleep questionnaires (Experiment 2 only)
Standard sleep questionnaires were collected from Experiment 2 participants (Broughton et al., 1982;Hoddes et al., 1972Hoddes et al., , 1973Horne & Ostberg, 1976;Johns, 1991;Terman & Terman, 2005). The Epworth Sleepiness Scale (ESS), Morningness-Eveningness Questionnaire (MEQ), and the Stanford Sleepiness Scale (SSS) were administered at the beginning or end of each session (i.e., before or after stimulus presentation or recall). The SSS was administered before and after encoding (SSS-Encoding) and before and after retrieval (SSS-Retrieval). Participants were additionally asked to report hours of sleep for three nights preceding the experiment. As depicted in Table S1, none of these measures significantly differed between Wake and Sleep groups in Experiment 2.

Epworth Sleepiness Scale (ESS).
The ESS assesses daytime sleepiness by asking participants to rate, on a 4-point scale, their usual chances of dozing off or falling asleep while engaged in eight different activities (e.g., watching television, sitting, lying down, etc.). It ranges from 0 to 24, with higher numbers indicating higher daytime sleepiness.

Morningness-Eveningness Questionnaire (MEQ):
Consisting of 19 items, the MEQ assesses individual differences in morningness and eveningness-the degree to which participants are active and alert at certain times of day. Scale items query preferences in sleep and waking times, and subjective "peak" times at which respondents feel their best.

Stanford Sleepiness Scale (SSS):
The SSS evaluates sleepiness at specific moments in time. Consisting of only one item, the scale requires respondents to select one of seven statements best representing their level of perceived sleepiness. This was administered before and after encoding, and before and after retrieval- Table S1 presents the averages across these two sets of measurements (SSS (Encoding), SSS (Retrieval)).

Average Hours of Sleep:
This is the mean daily sleep reported by participants for the three days prior to experiment Session 1.

Outlier analysis (Experiment 1 only)
As described in the Experiment 1 Results, four subjects within the Delayed Recall group were identified as outliers in analyses of main plot recall. We investigated the impact of these subjects on sideplot recall findings. On average, sideplot recall in these subjects (Coherent Narratives: X =15.7 details/cue, SD=7.9 details/cue, min=4.0 details/cue, max=27.5 details/cue; Unrelated Narratives: X =13.1 details/cue, SD=7.6 details/cue, min = 4.5 details/cue, max=25.5 details/cue) was numerically higher than for other Delayed Recall subjects (Coherent Narratives: X =10.4 details/cue, SD=9.3 details/cue, min=0 details/cue, max=32.0 details/cue; Unrelated Narratives: X =7.3 details/cue, SD=7.1 details/cue, min = 0 details/cue, max=31.5 details/cue). Therefore, we tested whether removing these outlier subjects would change the pattern of findings for sideplot recall. We performed a 2 x 2 ANOVA incorporating a within-subjects factor within the Immediate Recall group. Thus, removing main plot outlier subjects from the sideplot recall analysis did not change the overall pattern of findings.