The effects of episode spacing on adult's reports of a repeated event

ABSTRACT

Witnesses’ reports of repeated events have been the focus of much research; however, the spacing interval between each episode of the event has differed greatly. The aim of the current study was to determine whether spacing interval affects participants’ memory reports. Adults (N = 217) watched one (n = 52) or four videos depicting workplace bullying. The repeated event participants watched the four videos all in one day (n = 55), one per day over four consecutive days (n = 60), or one every three days over 12 days (n = 50). One week after the last (or only) video, participants reported on that video and answered some reflective questions about the procedure. Repeated-event participants also reported on what usually happens across the videos. Single-event participants reported proportionally more accurate information about the target video than repeated-event participants, and spacing interval did not affect repeated event participants’ accuracy. However, accuracy scores were close to ceiling while errors rates were at floor levels, preventing us from drawing strong conclusions. We found some evidence that episode spacing affected participants’ perceptions of their memory performance. Overall, spacing may have a minimal effect on adults’ memory for repeated events, but further research is required.

KEYWORDS:

• Repeated event
• spacing effect
• spacing interval
• particularisation
• investigative interviewing

Research investigating children’s and adults’ memories for episodes of repeated events has important implications for investigative interviewing. Witnesses are required to report on individual episodes of repeated events like sexual abuse, family violence, and workplace bullying with as much detail and accuracy as possible (Fair Work Act, 2009; PPP v. R, 2010; R v. Baker, 2002). Results show that when participants report on a target episode (commonly the last episode), they are proficient at reporting details that did not change across episodes, but they often confuse details that varied across episodes (see Dilevski et al., 2020; Woiwod et al., 2019 for reviews). To date, these studies have adopted vastly different spacing intervals between the presentation of episodes, from experiencing all episodes within 1 h (MacLean et al., 2018) to experiencing them across four weeks (Deck & Paterson, 2020). Deck and Paterson (2020) argued that short spacing prevents the generalisability of results, as repeated event episodes in a forensic context are unlikely to wholly occur within a short period of time. However, little consideration has been given to the effects that the spacing interval between episodes may have on participants’ reports.

Only one study to date has systematically manipulated episode spacing intervals. Price and colleagues (2006) examined 7–8-year-old children’s suggestibility for an episode of a repeated event. Children experienced structured play sessions consisting of eight memory items (e.g., receiving a sticker) that had different instantiations in each episode (e.g., children received a car sticker, an aeroplane sticker, a truck sticker, and a motorcycle sticker across the episodes). Some children experienced only a single episode, while others experienced four episodes that were massed into one session, spaced across four days, or spaced across 10 days. In a later biasing interview, instantiations that were not experienced—but were consistent with the experienced details—were suggested to have occurred in a target episode (e.g., the interviewer told children they received a scooter sticker). A different interviewer then elicited children’s reports of the target episode. Single-event children provided more correct instantiations than children in the three repeated event conditions, which did not differ from each other. In the repeated event conditions, children reported more instantiations from a non-target episode (internal intrusions, e.g., reporting the truck sticker when the motorcycle sticker was received in the target episode) when the episodes were massed into one session compared to when they were spaced over four or 10 days. Further, children who experienced the episodes spaced over four and 10 days reported more suggested instantiations than those in the single condition when specifically asked about each memory item. Price and colleagues’ (2006) findings suggest that spacing episodes across multiple days makes children more suggestible to interviewer-provided details, but less prone to internal intrusions (compared to when all episodes are in one session).

To date, the effects of spacing interval have not been explored in adults. We might expect differences between children’s and adults’ memories because adults’ memory systems are more developed than children’s (Schneider, 2015). For example, with repeated experience adults encode a detailed and complex generic representation in memory of what usually happens at the event (i.e., a script; Schank & Abelson, 1977). While children also encode scripts, they are slower to create them and their scripts are more rudimentary than adults’ (Hudson et al., 1992; Hudson & Mayhew, 2009). In addition, adults are better than children at evaluating the source of a recalled detail; for example, deciding if a particular instantiation occurred in the first or second episode (i.e., a source-monitoring decision; Lindsay et al., 1991; Ryan, 2009). Poor source-monitoring decisions lead to internal intrusion errors, which children make more often than adults (Lindsay et al., 1991; Ryan, 2009). Adults are also more proficient at creating verbatim memory traces that contain the happenings of a particular episode of a repeated event and gist memory traces for the general meaning of events (e.g., the stickers had transport-related pictures; Brainerd & Reyna, 2004; Brainerd et al., 2008; Odegard et al., 2009). Overall, adults’ more developed source monitoring abilities and verbatim traces may mean that spacing has less effect on their reports of internal intrusions and accurate details (compared to children), but adults’ proficiency to extract gist makes them more susceptible to errors involving information consistent with the general meaning of the event (like reporting a non-experienced form of transport; see Brainerd et al., 2008 for review).

Whilst no study has systematically manipulated the spacing interval of repeated event episodes with an adult sample, plentiful research has explored the effects of spacing interval on adults’ ability to learn cognitive stimuli like word pairs and foreign vocabulary (e.g., Cepeda et al., 2006; Delaney et al., 2017). Results show that adult learning improves as the spacing intervals between repeated stimuli increases, particularly when compared to the presentation of stimuli together in one massed session (the spacing effect; Ebbinghaus, 1885; see Smith & Scarf, 2017, for a review). Two types of theories explain this effect. Encoding variability theories propose that increased spacing intervals between stimuli causes contextual differences during encoding (such as different testing conditions and locations), whereas stimuli presented together in one massed session overlap in contextual features and are encoded similarly to each other. More contextual variety at encoding increases the number of retrieval pathways, which—in turn—increase the likelihood of successful retrieval (Crowder, 1976; Melton, 1970). Deficient processing theories propose that stimuli receive less processing when presented in one massed session because attention is divided between them in working memory, and the memory traces for repeated stimuli may quickly decay. Presenting stimuli at larger spacing intervals increases the attention—and therefore quality of processing—for each individual stimulus, which enhances consolidation of the stimuli in long-term memory to facilitate later retrieval (Dellarosa & Bourne, 1985; Koval, 2019).

The spacing effect literature has focused on cognitive stimuli like word pairs, whereas the repeated event literature has focused on personally experienced events, such as activity sessions. Compared to cognitive stimuli, experienced events are more complex and engaging, are intermingled with personal interpretations and connections in memory, and are supported by a script containing typically occurring details and temporal order of what usually happens (Conway, 2005; Hudson et al., 1992). Further, memory for experienced events becomes more general over time (e.g., remembering the gist of events, or what usually occurs; Brainerd & Reyna, 2004; Myles-Worsley et al., 1986; Sekeres et al., 2016). Overall, these differences may mean that spacing effects are less pronounced for experienced events.

Whilst the spacing of experienced repeated event episodes has not been considered with adult participants previously, a body of research has examined how adults naturally segment an experienced event into smaller units. Event segmentation theory proposes that humans continuously parse live events into sub-units, typically establishing boundaries between units at points of change in the event (such as a scene change; Zacks et al., 2007). Research on event segmentation has reliably found that events are better recalled if they are segmented rather than considered as one continuous event (Flores et al., 2017), and that spacing the units by inserting breaks or pauses between them can further improve memory for the event (Boltz, 1992; Schwan et al., 2000). Thus, we might expect larger spacing intervals between repeated event episodes to improve memory for adult participants.

The current study examined the effects of episode spacing on adults’ memory reports of a repeated event. Adult participants watched one (single) or four (repeated) videos showing workplace bullying incidents. The repeated videos were shown all in one day (massed), spaced across four days (spaced-short), or spaced across 12 days (spaced-long). One week after the final (or only) video, participants recalled what happened in the target (last or only) episode. Those in the repeated event conditions also recalled what usually happened across the series. Given that best-practice interviewing should avoid suggestive questions (e.g., Risan et al., 2020), we elicited adults’ memory reports using only non-suggestive prompts (unlike Price and colleagues’ (2006) study with children). Finally, since witnesses’ perceptions of their memory reports influence jurors’ assessments of witnesses’ testimony (Tenney et al., 2007), our participants were asked reflective questions about their experience of reporting on the event, including rating their perceived accuracy of their report.

We made three predictions. Based on Price et al.’s (2006) experiments with children, we first predicted that single-event participants should provide a higher proportion of accurate details from the target episode than repeated-event participants. Second, we predicted that participants in the massed condition should have a higher proportion of internal intrusions than participants in the two spaced conditions. We were also interested in reports of non-experienced details, because adults have previously been shown to be more susceptible to errors that are consistent with the gist of experienced word lists after experiencing spaced rather than massed lists (Dubuisson et al., 2012). Since spacing the episodes should increase participants’ abilities to extract their gist, our third prediction was that participants in the two spaced conditions should have a higher proportion of external intrusion errors (details that did not occur but are consistent with the gist of the videos) than participants in the massed condition. We examined participants’ reports about what usually happened because we were interested to see if spacing condition affected general memory reports, and we had participants reflect on their experience of recalling the event to investigate their insights into their own memory accuracy. We did not make formal hypotheses for these variables.

Method

Participants

An a-priori power analysis showed that 176 participants were required to detect medium effect sizes (${\eta }_p^2$ = .06) with power of .80. A medium effect size was chosen to reflect the medium-large effects detected for spacing intervals in previous work with children (e.g., Price et al., 2006). We over-recruited due to expected drop-out: 230 participants were recruited through Prolific. However, 13 participants reported that they did not remember watching the videos at all and were excluded from the sample. The final sample consisted of 217 participants (18–62 years, M = 31.84, SD = 10.80; n_female = 122, n_male= 94, n_missing= 1).

Participants experienced a single event (n = 52) or a repeated event with four episodes presented in one massed session (massed; n = 55), spaced over 4 days (spaced–short; n = 60), or spaced over 12 days (spaced–long; n = 50). There were no age or gender differences across the four conditions, ps ≥ .56. Assignment of participants to event conditions was not random, as each condition was advertised one-at-a-time. Participants signed up to a condition, and when one condition was full the next condition was advertised, etc. Once participants signed up to a condition, they were excluded from seeing advertisements for future conditions. This procedure allowed the study advertisements to accurately reflect expected time involvement and reimbursement, which differed across conditions.

Materials

Videos

Participants watched either one or four workplace bullying videos (see also Dilevski et al., in press). Videos were 1–2 min long and focused on an incident of bullying in an office setting between the same two workers (the perpetrator, Rick, and the victim, Dan). Some details varied across the videos while others remained stable (see Table 1 for a summary). For a description of the narrative for each video, see the coding manual available at: https://osf.io/6e2dc/?view_only=592607266537414ebe911e5d53d7620c . Repeated-event participants watched the four videos in the same order¹, and the last video was the target video. Participants in the single event condition only watched the target video without viewing the three preceding videos.

Table 1. Key details that varied and remained stable across the video series.

	Video 1	Video 2	Video 3	Video 4
Details that varied across the videos
Victim’s attire on top of shirt	Navy blue tie only	Purple paisley tie and a dark vest	Purple paisley tie only	Navy blue tie and a black suit jacket
Bully’s tie design	Diamond pattern in black and gold	Deep red with small black circles	Bright red	Silver/grey with purple floral pattern
Bully’s shirt colour	Red	Green	Blue	Purple
Number of witnesses	Two	One	Three	Zero
Setting	Office cubicle	Office cubicle	Meeting room	Big boss’s office and toilet
Beginning of video	Victim and colleague are celebrating completing work on the “Anderson file”	Victim is missing from his desk. Colleague attempts to discourage the bully from playing a prank on the victim.	Boss is telling a joke to a group of three workers sitting around a table	The victim is outside a big boss’s office. He enters and reports the bullying. The boss dismisses the victim.
Catalyst for bullying	Bully accuses victim of not pulling his weight at work	Victim is out at the toilet. He later is upset about losing work and tells the bully that “it isn’t particularly funny”	The bully claims the victim is late to the meeting. Victim states that he didn’t receive the memo for a new meeting time.	Bully sees victim exit the big boss’s office and assumes that victim has been complaining about the bullying.
Bullying acts	Bully excludes the victim from an office social event and gives him extra work to complete on the weekend	Bully unplugs the victim’s computer (losing his work) and laughs at him	Bully criticises the victim in front of other workers, dismisses his presentation and tells him to write a summary of it	Bully calls the victim a “two-faced maggot” and a “pathetic piece of shit”. Bully verbally threatens to make victim’s life difficult and give him a bad reference
Other workers’ responses to the bullying	Colleague leaves prior to the bullying	Colleague tells victim to “lighten up” about the bullying	Colleague trips the victim in front of a group of workers	Big boss is too busy to listen to the victim about the bullying
Details that remained the same across the videos
Victim pant colour	Black	Black	Black	Black
Victim shirt colour	White	White	White	White
Bully pant colour	Black	Black	Black	Black
Bully’s actions at end of video	Bully leaves scene	Bully leaves scene	Bully leaves scene	Bully leaves scene
Victim’s action at end of video	Victim remains at the scene	Victim remains at the scene	Victim remains at the scene	Victim remains at the scene

Memory report

Participants answered questions online to provide their memory reports (full questionnaire available online). For repeated-event participants, the questionnaire contained three key tasks: 1) a free recall report of the target video, 2) a free recall report of what usually happens across the video series (i.e., their script), and 3) reflective questions about their perceptions of recalling the videos. Single-event participants completed tasks 1 and 3 only. The tasks are described in more detail below.

Reporting the target video

Participants described everything that happened in the target video in a text box. They were told to describe everything from the beginning to the end of the target video in as much detail as possible, and not to guess what happened but to write everything they remembered. For repeated-event participants, instructions clearly identified “the last video” as the target.

Reporting what usually happens

Repeated-event participants described everything that usually happens in the videos in a text box (i.e., they described their event script). They were told to describe everything in as much detail as possible, and not to guess what happened but to write everything they remembered. Instructions specified that this question pertained to all the videos.

Reflective questions

Participants reflected on their experience of recalling the videos. First, participants rated how (1) accurately they believed that they recalled the target video, and (2) how difficult it was to remember the target video, from 1 (Not at all) to 6 (Extremely). Second, repeated-event participants completed the same ratings for their reports of what usually happens. Third, all participants were asked a qualitative question: “Is there anything else that would have helped you remember the video better?” and provided free-text space to respond.

Procedure

Ethics approval was granted by Deakin University. The study was advertised on Prolific to subscribers living in Australia and aged 18–65 years. To ensure that participants did not attempt to memorise or make records about the video/s, the study was advertised as an examination of people’s emotional responses to watching workplace bullying interactions. Participants provided informed consent and demographic information (e.g., age, gender). The single event condition watched the target video, then the session concluded for the day. The massed condition watched all four videos, then the session concluded for the day. The two spaced conditions watched the first video in the workplace bullying series, then the session concluded for the day. The spaced–short condition then watched the remaining videos at a rate of one-per-day for the next three days. The spaced–long condition watched the remaining videos on every third day (i.e., they watched the second video three days after the first video, then watched the third video three days after that, etc.). All participants (including the single event) always completed a filler task for 1 min after each video to avoid them rehearsing any details. In the filler task, participants were presented with pairs of numbers and had to judge whether the numbers were the same.

One week after watching the final (or only) video, participants completed the online questionnaire. A one-week delay was selected to replicate the delay in Price and colleagues’ (2006, Experiment 2) study. Participants were paid £10.00/hr for participating, which took an average of 20.32 min (single condition) and 38.13 min (repeated conditions).

Coding

Memory reports

There was a wide range in the length of participants’ reports about the target episode (range = 0–219 words; M = 65.88, SD = 37.34) and what usually happens (range = 7–526 words; M = 74.75, SD = 71.65). There was no difference in word counts for either prompt across spacing conditions, Fs < .96, ps > .41. Participants who provided responses of 20 words or fewer in their reports on the target episode (n = 17) or reports on what usually happens (n = 9) were excluded from analysis as their responses were too insufficient to code (e.g., P154: “I cannot remember the details, but I know it was about the man being bullied and harassed”; P176: “Staff member being bullied for no reason”). If a participant provided a sufficient response (> 20 words) to one prompt but not the other, their sufficient response was coded and analysed but their insufficient response was not. The coding schemes for responses to both prompts considered each individual clause in the response, so that everything participants reported was coded (rather than being limited to a pre-determined list of memory items; Gross & Hayne, 1998; Patterson & Pipe, 2009). A clause was defined as a group of words that included a subject and an implicit or explicit verb (e.g., “a mean boss picking on one of his employees”, “the HR person dismissed the victim”). The categorisation and scoring of clauses differed for reports of the target episode and what usually happens.

Target video reports

In addition to the 17 excluded participants noted above, five participants did not respond to the prompt about the target video. The remaining 195 participants provided an average of 9.85 clauses (SD = 4.45). Coders categorised each clause as accurate (it accurately described the target video), an internal intrusion (it contained information from a non-target video), or an external intrusion (it contained information that did not occur in any video). Of note, all external intrusions were consistent with the details presented in the target video (e.g., “the bully threatened to fire the victim” never occurred but the bully did threaten the victim with a poor reference in the target video). Because clauses are small units, typically a few words, an individual clause never fell into more than one category. Participants’ scores were converted into proportions of the total number of reported clauses (see also Danby et al., 2022; Sharman et al., 2022). For example, if a participant provided 10 accurate clauses, one internal intrusion, and two external intrusions, they received proportional scores of .77 (10/13), .08 (1/13), and .15 (2/13) respectively.

Eleven participants—all from the repeated event conditions—provided a report of a video that was not the target video (they all provided a coherent and clear report of the third video). This occurred evenly across the repeated event conditions (ns: massed = 2/50, spaced–short = 3/55, spaced–long = 6/42), zs < 1.74, ps > .08. For these participants, the accuracy of their clauses was marked against the video they recalled (i.e., a clause was categorised as accurate if it described a detail from the reported non-target video and an internal intrusion if it described a detail from any other video; see Sharman et al.’s (2022) scoring procedures). This procedure meant that participants were not overly penalised for confusing which video was the last one, reflecting that they had likely bound the details of one video together but just confused which was the last (Lee et al., 2015; Mammarella & Fairfield, 2008).

Reports of what usually happens

The 156 repeated-event participants with sufficient reports provided an average of 8.44 clauses (SD = 6.97), which either (1) generally described what typically occurred, or (2) described each video individually. Thus, coders categorised each clause as accurate general (it accurately described what usually happened across the video series; e.g., “the bully was mean to the victim”), accurate specific (it accurately described one but not all videos; e.g., “the bully confronted him in the bathroom one time”), or external intrusion (it contained information that did not occur in any videos). Again, all external intrusions were consistent with the details presented in at least one of the videos (no external intrusions were inconsistent with the videos), and again no clauses fell into more than one category. Participants’ scores were converted into proportions of the total number of clauses provided.

Reflective questions

Likert scales

Participants’ ratings of their perceived accuracy and difficulty when reporting the target video and (for repeated-event participants only) what usually happened were analysed even if they had provided insufficient responses to the free recall questions above. This was because we did not wish to exclude participants who may have found the written memory reports difficult. However, the five participants who did not provide a response about the target video at all were excluded from analyses of Likert data because they could not logically rate their own accuracy to a question they had not answered. Thus, ratings for 212 participants were analysed.

Qualitative question

Of the 217 participants, 66 did not provide a response of anything that would have helped them remember the target episode better or they reported that there was nothing that would help. A further nine participants did not answer the question properly (i.e., they provided a detail they did/did not remember well or a reason why they think they remembered the videos well/poorly) and were excluded from the coding. The remaining 142 participants’ responses fell into four categories: 1) behaviours or strategies designed to help memory or reporting of details (memory support), 2) changes to the video content or presentation (video changes), 3) changes to the timing of the spacing interval between the videos (spacing), and 4) changes to the delay between the last/only video and the online questionnaire (delay). See Table 2 for example responses and descriptions.

Table 2. Qualitative question response categories.

Response category	Description	Example responses
Memory support	Engaging in strategies or behaviours that enhance memory performance would improve memory of the videos.	If I had taken notes straight after watching them. (P152) Watching them twice probably. (P139) Writing down what took place, or a prompt to remember certain things before viewing the video. (P106)
Video changes	Changes to the presentation or content of the videos would improve memory of the videos.	Perhaps subtitles and names for each main character involved being displayed. (P80) I would likely have been able to differentiate videos better if they were set in different locations. (P198)
Spacing	Adjusting the spacing interval between the four videos would improve memory.	If I had seen [the videos] all one after another on the same day. (P153) Perhaps if it was spaced out differently (closer together in time) it would've been much easier to recall. (P108)
Delay	Reducing the delay between the final video and the memory report would improve memory.	Watching [the videos] more recently, a week is a long time! (P107) A shorter period between the videos and having to recall them. (P199)

Interrater reliability

Coders were blind to participant conditions and hypotheses. One researcher was responsible for coding participants’ reports of the target video and their responses to the qualitative question. A second researcher coded 15% of responses about the target video, and 35% of responses to the qualitative question for interrater reliability, which showed strong similarity in the categorisation of clauses (kappa = .81) and of qualitative responses (kappa = .95). Two coders together coded participants’ reports of their event scripts; they double-coded 30% of the sample with good reliability (kappa = .76)

Results

Our dataset is available at: https://osf.io/6e2dc/?view_only=592607266537414ebe911e5d53d7620c.

Target video reports

We examined the effects of spacing condition (single, massed, spaced-short, spaced-long) on participants’ proportional accuracy, internal and external intrusion scores. Proportional scores were nonnormal and non-homogenous, so an adjusted Welch’s F test was used (see Blanca et al., 2017; Delacre et al., 2019; Schmider et al., 2010). The main effect of condition on accuracy was significant, F(3, 101) = 3.59, p = .02, ω² = .07. Planned contrasts showed that single-event participants had a higher proportion accuracy than repeated-event participants, t(141) = 3.10, p = .002, d = .52. Bonferroni-corrected post-hoc tests showed that the three repeated event conditions did not significantly differ from each other², ps > .08. Table 3 provides group means.

Table 3. Proportional scores for reports of the target video and what usually happens in the videos.

	Target video free recall reports (n = 195)						What usually happens (n = 156)
	Accurate clauses		External intrusions		Internal intrusions*		Accurate general		Accurate specific		External intrusions
	M	SD	M	SD	Yes n	No n	M	SD	M	SD	M	SD
Single event	.97	.06	.03	.07	0	48	–	–	–	–	–	–
Repeated massed	.90	.18	.05	.09	6	44	.54	.32	.37	.30	.09	.16
Repeated spaced-short	.96	.09	.04	.09	1	54	.49	.33	.41	.32	.10	.14
Repeated spaced-long	.94	.08	.05	.07	3	39	.50	.30	.37	.28	.12	.14

*Internal intrusions were analysed dichotomously (yes, no) due to a low number of participants making any internal intrusions.

Only 10 participants made internal intrusions (5.1%); they were all in repeated event conditions. Specifically, five participants made one intrusion, three made two intrusions, one made three intrusions, and one participant³ made 11 intrusions. Due to the infrequent internal intrusions, an F-test was not appropriate. Instead, we dichotomously coded whether participants made any internal intrusions (see Table 3 for distribution) and conducted a 3 (condition: massed, spaced-short, spaced-long) x 2 (intrusion: yes, no) contingency table. Due to small expected cell counts, a Fisher-Freeman-Halton exact test was used (Conover, 1980). The test was not significant, p = .11.

External intrusions were made by 54 of the 195 participants (27.7%). Specifically, 42 participants reported one external intrusion, nine reported two, and three reported three intrusions. The F-test examining the effects of spacing condition on external intrusions was not significant, F(3, 105.13) = 0.59, p = .62.

Reports of what usually happens

Three one-way ANOVAs were conducted to examine the effects of spacing condition (massed, spaced-short, spaced-long) on the proportional accurate general, accurate specific, and external intrusion scores. None of the ANOVAs were significant, Fs < .59, ps > .55 (see Table 3 for means).

Reflective questions

Likert ratings

We conducted two one-way ANOVAs to explore whether event condition (including the single event) affected participants’ perceived accuracy and difficulty ratings about reporting the target video. The ANOVAs were non-significant, Fs ≤ 2.03, ps ≥ .11, η² s ≤ .03.

Next, we compared repeated-event participants’ perceived accuracy and difficulty for reports of the target episode versus reports of what usually happens. We conducted two 3 (condition: massed, spaced-short, spaced-long) x 2 (prompt rated: target episode, event script) mixed ANOVAs. See Table 4 for group means. The ANOVA for perceived accuracy showed a significant between-subjects effect of condition, F(2, 157) = 3.12, p = .047, ${\eta }_p^2$= .04. Bonferroni-corrected post-hoc comparisons showed no significant difference between conditions (ps > .05), but the difference closest to significance was the massed condition perceiving themselves more accurate than the spaced-long condition, p = .065. There was a significant within-subjects effect of prompt, F(1, 157) = 25.45, p < .001, ${\eta }_p^2$= .14. Pairwise comparisons with Bonferroni corrections showed that participants perceived themselves to be more accurate reporting on what usually happens than reporting on the target video, p < .001. The interaction was not significant, F(2, 157) = 0.11, p = .89, ${\eta }_p^2$= .001.

Table 4. Likert ratings of perceived accuracy and question difficulty.

	Perceived accuracy				Prompt difficulty
	Target		Usually		Target		Usually
	M	SD	M	SD	M	SD	M	SD
Massed (n = 53)	4.04	1.19	4.49	1.10	3.42	1.22	2.83	1.34
Spaced-short (n = 58)	3.95	1.10	4.45	1.10	3.40	1.14	2.64	1.20
Spaced-long (n = 49)	3.53	1.21	4.10	1.30	3.76	1.30	3.24	1.52

The ANOVA for perceived difficulty had no significant between-subjects effect of condition, F(2, 157) = 2.95, p = .06, ${\eta }_p^2$= .04. However, there was a within-subjects effect of prompt, F(1, 157) = 29.31, p < .001, ${\eta }_p^2$= .16. Pairwise comparisons with Bonferroni corrections showed that participants rated the target episode prompt more difficult to answer than what usually happens, p < .001. The interaction was not significant, F(2, 157) = 0.43, p = .65, ${\eta }_p^2$= .005.

Qualitative question

Of the 142 participants who provided a response about what would help them to remember the videos better, most commented on memory support (n = 66) or video changes (n = 53); fewer commented on the spacing interval (n = 14) or delay to recalling the video (n = 9). Strategies for memory support included ideas such as re-watching the videos, taking notes while watching them, or being warned before watching them that there would be a memory test. Video change responses suggested adding subtitles, adding numbers or labels to videos or characters, or adding more unique details to each video. All spacing responses indicated that an interval between each video was perceived to be detrimental to memory recall of the target video (i.e., a shorter interval would make recall easier); only participants in the two spaced event conditions made these responses (spaced-short n = 5, spaced-long n = 9). All delay responses indicated that reducing the delay would improve recall.

A 4 (event condition) x 4 (response) contingency table explored response differences across event conditions. Due to small expected cell counts, a Fisher-Freeman-Halton exact test was used. Responses were significantly related to condition, p < .001. Spacing responses were made by the spaced-long condition more often than expected, and by the single and massed conditions less often. Compared to expected values, memory support responses were more often made by the single condition, video change responses were more often made by the massed condition, and delay responses were more often made by the spaced-short condition. Table 5 shows response distributions.

Table 5. Distribution of qualitative question responses (n = 142).

	Memory support	Video changes	Delay	Spacing
Single event	24	9	1	0
Repeated event conditions
Massed	12	20	2	0
Spaced – short	19	12	5	5
Spaced – long	11	12	1	9
Total	66	53	9	14

Discussion

The current study examined whether the interval between episodes of a repeated event affected adults’ memory reports. Consistent with our first hypothesis, we found that repeated-event participants were less accurate in reporting on the target video than single-event participants. This finding is consistent with Price et al,’s (2006) study of children as well as the repeated event literature more generally (see Dilevski et al., 2020; Woiwod et al., 2019 for reviews). When accuracy is defined narrowly (i.e., as the ability to attribute details to the correct episode; Price et al., 2006; Woiwod et al., 2019), single event witnesses tend to be more accurate than repeated event witnesses.

Like Price et al. (2006), the current study also found no differences in accurate details reported between the repeated event conditions (i.e., across the massed and two spaced conditions). These findings contrast the spacing effect literature, which has shown that learning improves when the intervals between repetitions increase (Ebbinghaus, 1885; Smith & Scarf, 2017). It is possible that, because events are more complex and integrated than cognitive stimuli (Conway, 2005; Hudson et al., 1992), spacing does not further improve retrieval. Price and colleagues (2006) found reduced spacing effects after a one-week delay (Experiment 2) compared to a one-day delay (Experiment 1). Memory for experienced events becomes more general over time (Brainerd & Reyna, 2004; Sekeres et al., 2016). Participants in the current study experienced a one-week delay before providing their memory reports, which may have increased their reliance on general memory and reduced any effects of episode spacing. Consistent with this explanation, we also found no effect of spacing on repeated-event participants’ reports about “what usually happens”, which taps into their scripts of the general event structure and temporal order. It is possible that a shorter delay (less than one week) might have resulted in more effects of spacing condition in the current study.

Contrary to our second and third hypotheses, we did not find any evidence that spacing affected rates of memory errors when reporting on a target episode: both internal and external intrusions were comparable across spacing conditions. While Price et al. (2006) found an effect of episode spacing on internal intrusions, this study sampled children, who are still developing their source-monitoring abilities so are more prone to internal intrusion errors than adults (Lindsay et al., 1991; Ryan, 2009). Indeed, our adult participants reported very few internal intrusions regardless of spacing condition. Consequently, a floor effect may explain why event spacing did not influence internal intrusion reports.

External intrusions were the most common type of error our sample made about the target episode (though reported errors were still quite low). Of note, all external intrusions about the target episode were consistent with what occurred in the target video. Reporting non-presented details that are consistent with presented content has commonly been found in memory studies (e.g., reporting seeing details beyond the boundaries of a presented photograph in Munger & Multhaup, 2016; reporting seeing unpresented segments of a video showing someone making a sandwich in Gerrie et al., 2006). These errors occur when participants generate a non-presented detail internally (such as imagining it or drawing on their prior knowledge) and erroneously attribute it to the observed stimuli. Accurately monitoring what details were witnessed versus generated internally becomes more difficult when there are delays after an event (Sussman, 2001). Reporting on the target video after a one-week delay might have influenced our participants’ commission of external intrusions more so than spacing interval. Even with immediate recall, the spacing effect has previously been shown to be small for the commission of errors that are consistent with the theme of presented stimuli (relative to its size for reporting accurate details, Dubuisson et al., 2012). It is also possible that external intrusions may have been too low in the current study to detect a spacing effect, especially if the spacing effect is small for stimuli-consistent errors.

Interestingly, while spacing had no overall effect on repeated event participants’ memory reports, it influenced their perceived accuracy. That is, there was a significant effect of spacing condition on participants’ perceived accuracy ratings. While the post-hoc tests were nonsignificant, the effect seemed to be driven by the massed condition rating themselves as more accurate than the spaced-long condition. Previous studies have also shown that learners perceive themselves as better at performing skills learned after one massed learning session, rather than spacing learning over several sessions (Dunning et al., 2004; Simon & Bjork, 2001). Similarly, 14 participants in the spaced-short and spaced-long conditions believed that reducing spacing (e.g., presenting all videos in a massed session) would improve their memory performance. Dunning et al. (2004) asserted that adults often mistake learning a skill quickly for competence performing it, leading to increased beliefs of performance in massed conditions (without necessarily demonstrating increased performance).

Repeated-event participants further judged themselves to be more accurate when reporting what usually happened than when reporting the target video. This judgment is consistent with script theory, which states that an event script supports retrieval of the typically occurring details from a repeated event (Hudson et al., 1992; Schank & Abelson, 1977). Previously, Danby et al. (2022) surveyed laypeople about their understanding of event memory, finding that lay adults expected the general or consistent details from a repeated event to be better remembered than changing ones. Similarly, our results for repeated event participants suggest that they felt they performed more accurately when recalling the general details rather than details unique to the target episode. The single event condition’s perceived accuracy could only be considered for reports of the target video (not reports of what usually happened); their perceived accuracy did not significantly differ from the repeated event conditions. In Danby et al.’s survey of lay adults, participants expected repeated events to be recalled worse than single events. However, people generally have inflated expectations for their own performance on cognitive tasks (Kruger & Dunning, 1998; Metcalfe, 1998), so we may have not detected a difference between single and repeated event conditions’ perceived accuracy because our participants were rating their own memory (rather than rating memory for a fictional a vignette in Danby et al., 2022). The ability to accurately judge ones’ own memory performance has important implications when it comes to how eyewitnesses are perceived. Research has shown that confidence-accuracy calibration impacts credibility assessment made by legal decision makers, with witnesses with better calibration being judged more credible (Tenney et al., 2007).

This study has limitations. First, we collected memory reports online using free text boxes. This meant that we could not probe participants with follow-up questions, so some responses were very short and had to be excluded from the analyses. It is also possible that the high accuracy scores in the current study reflect participants’ provision of the details that they were most confident about, and there was no subsequent prompting for more detail. Replication of our study with face-to-face interviews that include follow-up prompts is needed to address this limitation. Further, the videos were quite different from each other, even though they all depicted workplace bullying between the same employees. For example, they differed in the locations in which the events took place, and the type of bullying employed (verbal, physical). While ecologically valid, these differences may have made the episodes less likely to be confused with one another and thus may have resulted in lower rates of internal intrusions across all conditions. Future research could utilise videos that are highly structured and similar to one another to test this notion. Last, participant assignment was non-random so that recruitment advertising and payment could accurately reflect participant requirements for each condition. It is possible that participants willing to commit to longer conditions may have increased motivation compared to shorter conditions. Future research would benefit from random assignment to conditions.

Despite these limitations, this study makes an important contribution to the literature. The findings suggest that the spacing effect may not be evident for adult recollections of a repeated event when tested after a one-week delay, at least in response to a free recall question and when episodes are very different from each other. This has important methodological implications because it suggests that stimuli for repeated event studies could potentially be shown to participants in one session, rather than across the course of multiple weeks, without any effect on memory. Consequently, research examining repeated event memory may be logistically easier to conduct in the future. However, given this is the first time spacing has been manipulated within an adult sample, replication with a range of events (such as live events rather than videos) and retention intervals (such as delays longer or shorter than one week) is needed. Overall, our findings provide a starting point for future investigations of the effect of spacing of experienced events on adults’ memory performance.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 Video order was not counterbalanced in the current study since Dilevski and colleagues (in press) found no effect of counterbalance order on adult participants’ memory reports.

2 We further examined the lack of statistically significant differences in participants’ memory reports across spacing conditions using Bayesian one-way ANOVAs conducted in JASP (version 0.16.2). Bayes factors were examined and compared to conventions suggested by Lee and Wagenmakers (2013) and Jeffreys (1961) for interpretation. For reports of the target video, we found anecdotal evidence for the null hypothesis of no difference across the repeated event conditions’ accuracy scores (BF₁₀= 0.59) and strong evidence for the null hypothesis across all conditions’ external intrusion scores (BF₁₀ = .05). For reports of what usually happens, we found strong evidence for the null hypothesis of no difference across all conditions for accurate-general and accurate-specific scores (BF₁₀  = .09 and BF₁₀  = .08 respectively), and moderate evidence for the null hypothesis on external intrusion scores (BF₁₀ = .11).

3 This participant was in a repeated event condition and clearly identified that they were reporting on the last video only but reported happenings from all four videos in depth. Accordingly, details that were from the last video were coded as accurate, but details from the other three videos were coded as internal intrusions, leading to a large internal intrusion score. Analyses were run including and excluding this participant, and the pattern of results was the same either way. Accordingly, we present results including this participant.