Abstract
Imagination inflation occurs when participants increase their certainty that they have experienced an event after they imagine the event occurring. Two experiments (with a total of 291 participants) examined the effects of imagining events in the future on participants’ certainty they had experienced those events in the past. Participants rated their certainty in having experienced events and then imagined experiencing some of those events either in the future or in the past. One or two weeks later, participants completed certainty ratings a second time and completed some individual difference measures. In both imagination conditions (future and past), certainty ratings increased more for imagined events than for control events. Autobiographical memory specificity and self-concept clarity did not significantly predict this effect. These findings suggest that imagining events in the future makes people more certain that they have happened in the past.
Keywords
Imagination inflation occurs when individuals increase their certainty that they have experienced an event after they imagine the event occurring. Garry, Manning, Loftus, and Sherman (1996) and Hyman and Pentland (1996) were among the first to study imagination inflation. Garry et al. (1996) had participants rate their confidence that they had experienced 40 events, and 2 weeks later, they imagined experiencing 4 of those 40 events. They found that confidence ratings increased more frequently for events participants imagined than for events they did not imagine. In a study by Hyman and Pentland (1996), some participants underwent three guided imagery interviews. In the interviews, participants tried to remember some events that their parents reported had occurred and one event that their parents did not report. Other participants served as a control and were told to think about the events instead of going through the guided imagery task. Those who imagined events that were not reported by their parents were more likely to remember that event as having occurred. In both of these studies, the act of imagining childhood events made participants more likely to believe those events occurred.
Some studies have examined individual differences in susceptibility to imagination inflation. These studies have found that factors such as hypnotic suggestibility, dissociativity, imagery ability, locus of control, and extroversion may be predictive of imagination inflation (Heaps & Nash, 1999; Horselenberg et al., 2000; Paddock et al., 1998, 1999; Paddock, Terranova, Kwok, & Halpern, 2000). These studies have demonstrated that certain characteristics of participants are related to their susceptibility to imagination inflation and this study examined additional predictors of imagination inflation.
The imagination exercises in imagination inflation tasks depend on episodic counterfactual thinking. Participants engage in episodic counterfactual thinking when they imagine a childhood event that could have occurred but did not (De Brigard, Addis, Ford, Schacter, & Giovanello, 2013). Thus, episodic counterfactual thinking is critical to imagination inflation. In this study, we examined the effect of participants imagining events in the future on their certainty that the events occurred in the past. Imagining experiencing an event in the future requires episodic future thinking: projecting oneself into the future to preexperience an event (Atance & O’Neill, 2001). Schacter (2012) argued that imagining the future is an adaptive process: It allows people to “try out” different ways in which events may occur to determine the best course of action. Because imagining the future relies on a constructive memory system, imagined future events are sometimes mistaken for experienced events. In addition, episodic future thinking and episodic counterfactual thinking recruit a similar network of brain structures (Schacter, Addis, & Szpunar, 2017; Schacter, Benoit, De Brigard, & Szpunar, 2015; Schacter, Benoit, & Szpunar, 2017). In another line of research with the Deese-Roediger-McDermott (DRM) paradigm, Dewhurst, Anderson, Grace, and van Esch (2016) found that participants who rated words by their relevance to a future scenario had higher levels of false memories than participants in other conditions. The primary purpose of this study, then, was to determine if imagining events in the future would lead to increases in certainty that they have happened in the past.
A secondary purpose of this study was to examine possible predictors of susceptibility to imagination inflation. Dewhurst, Anderson, Berry, and Garner (2017) examined the relationship between autobiographical memory specificity and susceptibility to false memories in the DRM paradigm. Their participants completed the Sentence Completion for Events from the Past Test (SCEPT) and DRM tasks with negatively valenced and neutral critical lures. The SCEPT has participants respond to sentence stems. The number of responses that include specific autobiographical memories is used as one’s autobiographical memory specificity score. Participants with higher SCEPT scores are believed to have more specific autobiographical retrieval, whereas those with lower SCEPT scores have more general retrieval. Dewhurst et al. (2017) found that participants’ number of specific memories on the SCEPT was positively correlated with false recognition of both types of critical lures. Raes et al. (2006) on the other hand found that among participants with major depressive disorder, those with lower autobiographical memory specificity made more source memory errors. Dewhurst et al. (2017) explained these seemingly discrepant findings by pointing out differences in the tasks used in their study and in Raes et al.’s (2006) study. Dewhurst et al. (2017) also recommended that future research examine autobiographical memory specificity in more naturalistic memory paradigms, including the imagination inflation paradigm. Thus, the secondary purpose of this study was to examine the relationship between autobiographical memory specificity and susceptibility to imagination inflation.
The two experiments of this study examined whether imagining events in the future leads to an increase in certainty ratings that those events have occurred. Both experiments also examined the relationship between autobiographical memory specificity and imagination inflation. Experiment 1 used a sample of college students and Experiment 2 used an online sample to examine the generalizability of the results.
Experiment 1
In Experiment 1, participants completed a Life Events Inventory (LEI) by rating their certainty that they had experienced 40 events. The participants then completed a guided imagination exercise that instructed them to imagine experiencing events either in the past or in the future. The imagined events were included in the LEI. Two weeks later, they completed an autobiographical memory specificity measure and they completed the LEI a second time. We included both positive and negative events in the imagination exercises because some studies have shown that imagination inflation is influenced by event valence (e.g., Bays, Zabrucky, & Foley, 2015; Sharman & Barnier, 2008). We examined the change in ratings for the imagined events relative to control events that were not imagined. Based on previous imagination inflation studies (e.g., Garry et al., 1996), we predicted that participants who imagined events in the past would have increased certainty ratings for those events. More specifically, we predicted that the increase for the imagined events would be greater than the increase for control events. Because of the functional and neurological similarities between episodic counterfactual thinking and episodic future thinking (Schacter et al., 2015), we predicted that imagining events in the future would lead to similar imagination inflation: Again, that the increase in certainty for the imagined events would be greater than the increase for control events. We also examined the relationship between specific memories in the autobiographical memory test and the magnitude of imagination inflation across participants. Based on the study of Dewhurst et al. (2017), we predicted a positive correlation between memory specificity and imagination inflation.
Method
Participants and design
Participants were undergraduate students enrolled in introductory psychology courses at a medium-sized public university in the United States. Participants received credit toward the completion of a research requirement in those courses. A total of 96 participants completed the first session. Twenty of them did not return for the second session and data from three additional participants were removed before conducting any analyses because they failed an attention check question (described in the Materials). The final sample consisted of 73 participants (20 men and 52 women, and 1 who identified as agender), who ranged in age from 18 to 31 (M = 19.71, SD = 2.15) years.
The minimum sample size was based on an a priori power analysis calculated for the within-subject effect of imagining events on certainty ratings. This effect has varied considerably in size in previous studies (see Brewin & Andrews, 2017, for review). To be conservative, we predicted a medium effect size (Cohen’s d = .50) for this variable. Using G*Power (Faul, Erdfelder, Lang, & Buchner, 2007), we determined that a minimum of 34 participants was needed in each imagination instruction condition to obtain power of .80, with α = .05 (two tailed), to detect the effect of imagining events in each imagination instruction condition. We stopped data collection after we had a minimum of 34 participants in each condition. We ended up with 34 in the past condition and 39 in the future condition.
Experiment 1 had a 2 (imagination instruction: past or future) × 2 (event: imagined or not imagined) × 2 (valence: positive or negative) mixed-model design. Imagination instruction was a between-subject variable, whereas event and valence were within-subject variables. The dependent variable was the change in confidence ratings across the two sessions.
Materials
The materials consisted of an LEI, an imagination exercise, and the Autobiographical Memory Test (AMT; Williams & Broadbent, 1986). The LEI consisted of 40 events. We selected 38 items from norms provided by Fields and Brown (2015). Eight of the 38 items we selected served as critical events in this study. The other 30 were filler items. The critical events were chosen based on four criteria from the Fields and Brown (2015) norms. First, the event must have had a mean occurrence rating of less than 3.5 (on a scale of 1 to 6) so that there would be room to inflate confidence ratings after imagination exercises. Second, the event must have had a plausibility rating greater than 3.0 (on a scale of 1 to 6) so that participants might find experiencing the event plausible. Third, because in the imagination exercises some participants would imagine experiencing the events in the future, we selected events that adults might experience. Finally, because we wanted some events to be negative and some to be positive, we selected four events with desirability ratings lower than 2.5 and four events with desirability ratings greater than 4.0 (again, on a scale of 1 to 6). The negative events were “having a prized possession stolen,” “pretending to be someone else online,” “having a skin sample taken by a nurse,” and “breaking something at a friend’s house and not telling him or her about it.” The positive events were “being given a new pet as a gift,” “having a surprise birthday party,” “finding a lost wallet,” and “being in the background of a news segment on TV.” In addition to the 8 critical items and 30 filler items, 2 items were included in the LEI as attention check items. These items, taken from Pezdek, Blandon-Gitlin, and Gabbay (2006), were “winning a million dollars” and “playing for the LA Lakers.” Participants rated all 40 events on a 6-point scale from 1 (definitely did not happen) to 6 (definitely did happen). We excluded participants (n = 3) who rated either of the attention check items greater than 1.
The introduction to the imagination exercises was adapted from Garry et al. (1996) and the instructions for the exercises were adapted from Bays et al. (2015). Both of these are available in the Appendix. To introduce the exercises, we told participants that they would imagine four events and then answer questions about them. Participants then began to imagine the first of four events. After the first event, participants were prompted to answer three questions about the event on a computer. These questions required contextual details of the imagined event, including how the participant felt when the event occurred. For example, for the event, being given a new pet as a gift, participants were asked, “What type of pet were you [will you be] given,” “Who gave [will give] the pet to you,” and “How did [will you] feel when you received [receive] the pet?” Participants repeated this process by imagining the second event, answering three questions about it, and then proceeding to the third and fourth events.
In the AMT, we asked participants to write about a specific time that they felt each of 10 specific emotions. Participants wrote about a specific time in which they felt happy, sorry, safe, angry, interested, clumsy, successful, hurt, surprised, and lonely. We instructed participants that the events should be something that happened at a particular time on a particular day and that each emotion should have a unique event.
Procedure
The experiment was conducted in two sessions. In the first session, participants provided informed consent, were assigned to one of the two between-subject conditions, completed the LEI, and completed the imagination exercises. Of the eight critical events, each participant completed imagination exercises for two positive and two negative events. The other two positive and two negative events served as control events, and the imagined and control events were counterbalanced across participants.
Two weeks after the first session, participants completed the second session. In the second session, participants completed the AMT and then completed the LEI for a second time. After the LEI, participants were thanked and debriefed. Both sessions occurred in a small lab and participants were run either alone or with one other participant assigned to the same condition.
Results
We included event valence as a factor because some studies have shown event valence affects imagination inflation (e.g., Bays et al., 2015). In a three-way analysis of variance (ANOVA) that included event valence, it did not have a significant main effect on confidence rating changes, and it did not interact with any other variables. Therefore, we combined the two valence conditions in the following analyses.
We conducted a two-way mixed-model ANOVA with imagination instruction (past or future) as the between-subject variable and event (imagined or not) as the within-subject variable. The dependent variable was the change in ratings from the first LEI to the second LEI. Positive numbers indicate an increase in ratings from the first to the second LEI. The means and 95% CIs for each condition are provided in Table 1. There was a significant effect of whether events were imagined or not, F(1, 71) = 13.07, p = .001, η2p = .16, with ratings increasing more when they were imagined than when they were not. There was no significant effect of the imagining condition, F(1, 71) = 0.48, p = .490, η2p = .01, and these two factors did not significantly interact, F(1, 71) = 1.02, p = .316, η2p = .01.
Mean Increases in Certainty Ratings From the First LEI to the Second LEI for Imagined and Control Events and for Participants Who Imagined Events Occurring in the Past and Those Who Imagined Events Occurring in the Future.
The main effect of imagining events and the lack of an interaction between the two factors suggested that there were imagination inflation effects in both the past and future conditions. To verify this and to directly test our first two hypotheses, we conducted simple effects tests by comparing the difference in confidence ratings for imagined and control events in each condition. Among those who imagined events in the past, the change in ratings was greater for imagined events than for control events, t(33) = 2.72, p = .010, d = 0.47, supporting our first hypothesis. This difference was also significant for those who imagined events in the future, t(38) = 2.33, p = .025, d = 0.37, supporting our second hypothesis.
Imagination inflation studies often report how many participants experience imagination inflation by reporting how many participants have ratings that increase more from the first LEI to the second LEI for imagined events than for control events. In the total sample of 73 participants, 21 of them (28.8%) showed imagination deflation, 10 of them (13.7%) showed no effect, and 42 of them (57.5%) showed imagination inflation. Of the 34 participants who imagined events in the past, 9 of them (26.5%) showed imagination deflation, 5 of them (14.7%) showed no effect, and 20 of them (58.8%) showed imagination inflation. Of the 39 participants who imagined events in the future, 12 of them (30.8%) showed imagination deflation, 5 of them (12.8%) showed no effect, and 22 of them (56.4%) showed imagination inflation. Hence, most participants in both imagination instruction conditions experienced imagination inflation.
We then explored the relationship between autobiographical memory specificity and the magnitude of imagination inflation. Due to computer error, two participants’ AMT scores were not recorded so the following analyses are based on 71 participants. Participants’ AMT responses were coded into one of four categories: Responses were coded as specific memories when participants described an event that lasted for 1 day or less, as nonspecific memories when participants described an event that lasted for more than 1 day, as categorical memories when participants described an event the occurred repeatedly, and as no memories when participants did not describe an event or wrote about something that was not a memory. Initially, two coders coded all participants’ responses. They agreed on 83.9% (596 of 710) of their codes (Cohen’s kappa = .53). Disagreements were coded by two additional coders to determine a consensus.
The number of specific memories in the AMT ranged from 1 to 10 (M = 7.42, SD = 1.89). We first examined the correlation between specific memories in the AMT with the overall magnitude of imagination inflation (the increase in confidence for imagined items minus the increase in confidence for control items) with all 71 participants who had AMT scores. This correlation was not significant, r(69) = .19, p = .112. We then examined this correlation in both imagination instruction conditions. This correlation was not significant in the future condition, r(36) = .15, p =.380, or in the past condition, r(31) = .26, p = .143. Thus, our third hypothesis, that these correlations would be positive, was not supported.
Our final analysis explored increases in certainty ratings as a function of the initial rating. Specifically, we wanted to know if those events that received uncertain ratings in the middle of the 6-point scale show larger increases than those at either extreme (which have more certain ratings). This analysis focused only on those events that were later imagined. There were a total of 292 ratings for the four imagined events (73 participants × 4 ratings). Of those, 135 events received a rating of 1, 31 received a rating of 2, 16 received a rating of 3, 21 received a rating of 4, 17 received a rating of 5, and 72 received a rating of 6. The participants appeared fairly certain in their responding with 206 of 292 ratings being at the extreme ends of the scale. Those events with ratings at the low end of the scale appeared most susceptible to imagination inflation. Events that received a rating of 1 increased by a mean of 1.31 (95% CI [0.99, 1.63]) and those that received a rating of 2 increased by a mean of 1.52 (95% CI [0.84, 2.19]). Events that received initial ratings in the middle of the scale tended to stay consistent. Events that received an initial rating of a 3 increased by a mean of 0.31 (95% CI [−0.57, 1.20]) and those that received a rating of 4 increased by a mean of −0.05 (95% CI [−0.80, 0.71]). Events that received high initial ratings tended to decrease. Those events that received an initial rating of 5 increased by a mean of −0.47 (95% CI [−1.42, 0.48]) and those that received a rating of 6 increased by a mean of −1.00 (95% CI [−1.40, −0.60]). Thus, imagination inflation appeared most robust for events that participants are fairly sure they had never experienced.
Discussion
We found evidence for imagination inflation in Experiment 1. Events that were imagined by participants showed an increase in certainty ratings, whereas events that were not imagined did not. This occurred for participants who imagined experiencing events in the past and for participants who imagined experiencing events in the future, supporting our first two hypotheses. The novel contribution was the latter effect demonstrating that imagining events in the future increased participants’ certainty that they occurred in the past.
Our third hypothesis was that there would be a positive correlation between autobiographical memory specificity and imagination inflation. That is, participants with more specific autobiographical memories would be more susceptible to imagination inflation. This hypothesis was not supported. The correlation was positive but not significantly different from zero. It is important to note that our power analysis was based on our first two hypotheses and the sample from Experiment 1 was underpowered to detect small correlations. We ran a second experiment to examine this relationship in a larger sample.
Experiment 2
The purpose of Experiment 2 was to replicate and extend the findings from Experiment 1. We made several changes from our first experiment. To be consistent with Dewhurst et al. (2017), we measured autobiographical memory with the SCEPT rather than the AMT. We also used a larger, more diverse sample (in terms of age and geographic location) by recruiting from Mechanical Turk (MTurk) rather than a college sample. This also allowed us to examine if imagination inflation effects could be studied online as no study, to our knowledge, has included an online imagination inflation manipulation. The larger sample also allowed us to have sufficient power to detect the relationship between autobiographical memory specificity and imagination inflation. Furthermore, in addition to autobiographical memory specificity, we examined a related potential individual differences predictor of imagination inflation. Self-concept clarity (SCC) refers to the clarity and stability of an individual’s self-concept (Campbell et al., 1996). Because imagination inflation involves participants’ autobiographical memory for events they have experienced, concepts related to autobiographical memory may predict imagination inflation. SCC has been shown to relate to some aspects of autobiographical memory retrieval, and older age is related to greater SCC (Campbell et al., 1996; Diehl & Hay, 2011; Fuentes & Desrocher, 2012; Lodi-Smith & Roberts, 2010).
Participants completed an LEI and imagination exercises for three events during their first session. One week later, they completed the SCEPT, the SCC scale, and the LEI a second time. We predicted that participants’ certainty ratings would increase more for imagined events than for control events in both the past and future conditions. We also predicted a positive correlation between imagination inflation and autobiographical memory specificity as measured by the SCEPT. In addition, because higher SCC is related to better autobiographical memory retrieval, we predicted that SCC would be negatively correlated with imagination inflation. Finally, we explored the relationship between SCC and participants’ age.
Before commencing data collection, we preregistered our data collection and analysis plans on aspredicted.org.
Method
Participants and design
We conducted two power analyses based on the results of Experiment 1. First, the size of the imagination inflation in Experiment 1 was d = .47 for those in the past condition and d = .37 for those in the future condition. Using the smaller of these effects, a power analysis conducted with G*Power (Faul et al., 2007) indicated that we needed 60 participants in each condition to have power of .80 to detect this effect (with two-tailed α = .05). Next, the overall correlation between autobiographical memory specificity and imagination inflation was r = .19 in Experiment 1. To have power of .80 to detect this relationship, we needed a total sample of 215 participants. That is the number we used for Experiment 2.
We had a total of 399 participants complete the first session on MTurk using Turk Prime (Litman, Robinson, & Abberbock, 2017). Of them, 294 passed the attention and memory check questions. We invited those 294 people to complete Session 2. Of those 294 people, 228 completed Session 2 and 218 of them passed the attention check questions. Thus, our final sample size for Experiment 2 was 218, with 105 in the past condition and 113 in the future condition.
All participants were MTurk workers residing in the United States. There were 80 men and 137 women, and 1 participant who chose “decline to state” for gender. The participants ranged in age from 18 to 77 (M = 39.26, SD = 12.99) years. Unlike with the college sample of Experiment 1, education level varied in Experiment 2. The highest level of education selected by 15 participants was high school, for 42 participants it was trade, technical, or vocational school, for 7 participants it was associate degree, for 23 participants it was bachelor’s degree, for 95 participants it was master’s degree, for 31 participants it was professional degree, for 4 participants it was doctoral degree, and 1 participant declined to state his or her education.
The design of Experiment 2 was similar to that of Experiment 1. Experiment 2 had a 2 (imagination instruction: past or future) × 2 (event: imagined or not imagined) mixed-model design with imagination instruction as a between-subject factor and event as a within-subject variables. Because valence had no significant effect in Experiment 1, it was omitted from the design of Experiment 2. The dependent variable was the change in confidence ratings across the two sessions.
Materials
The materials for Experiment 2 were similar to those of Experiment 1. They included a 40-item LEI, an imagination exercise, a measure of autobiographical memory specificity, and a measure of SCC. The LEI consisted of the same 40 items as Experiment 1: 38 taken from Fields and Brown (2015) and two attention check items taken from Pezdek et al. (2006). Participants responded on a scale from 1 (definitely sure the event did not happen) to 8 (definitely sure the event did happen) for each event.
To increase the generalizability of our findings, we used different critical items in Experiment 2. The six critical items were “cut yourself so bad that you had to get stitches,” “been startled in a movie and spilled your popcorn,” “found a large bill on the ground,” “hit your head and got checked for concussion,” “accidentally switched coats with someone,” and “met the mayor.” All six events had mean occurrence ratings less than 3.0 and mean plausibility ratings greater than 3.0 based on Field and Brown’s (2015) norms. These six events were assigned to two sets with each set containing one negative, one neutral, and one positive event. The first three in the aforementioned list were in one set and the last three were in the other set. The negative events had mean desirability ratings less than 2.0, the neutral events had mean desirability ratings between 2.0 and 4.0, and the positive events had mean desirability ratings greater than 4.0. One set served in the imagination condition for each participant and the other set served in the control condition. These were counterbalanced across participants. The wording of the imagination exercises was identical to that of Experiment 1 and there were three questions after each imagination exercise that required the participant to report details of the imagined event. After the third imagination exercise, there was a memory question asking participants which events they had just imagined.
To assess autobiographical memory specificity in a manner consistent with Dewhurst et al. (2017), we used the SCEPT (Raes, Hermans, Williams, & Eelen, 2007). The SCEPT may be a better measure of autobiographical memory specificity in nonclinical populations than the AMT (Raes et al., 2007). The SCEPT consists of 11 sentence stems that participants complete. For example, participants read the stem, “I still remember well how . . . ” and they have to write something to complete the sentence. Participants are instructed that they can complete the stems anyway they want to but each sentence must be on a different topic.
To assess SCC, we used the SCC scale (Campbell et al., 1996). This scale contains 12 statements that participants respond to on a 1 (strongly disagree) to 5 (strongly agree) scale and it is intended to measure how stable and clear one’s self concept is. An example statement is, “My beliefs about myself often conflict with one another.”
Procedure
Participants completed two online sessions. In the first session, participants completed the LEI and then completed imagination exercises for three events. All participants who passed the attention check questions in the LEI and the memory check question after the imagination exercises were invited to complete the second session exactly 1 week after they completed the first session. The second session contained the SCEPT, the SCC Scale, and the second LEI. The LEI had the same attention check questions in it and anyone who failed either question was not included in the final sample.
Results
We conducted a two-way mixed-model ANOVA with imagination instruction (past or future) as the between-subject variable and event (imagined or not) as the within-subject variable. The dependent variable was the change in ratings from the first LEI to the second LEI. Positive numbers indicate an increase in certainty ratings from the first to the second LEI. The means and 95% CIs for each condition are provided in Table 2. There was a significant effect of whether events were imagined or not, F(1, 216) = 33.44, p < .001, η2p = .13, with ratings increasing more when they were imagined than when they were not. There was no significant effect of the imagining condition, F(1, 216) = 1.82, p = .179, η2p = .01, and these two factors did not significantly interact, F(1, 216) = 0.90, p = .345, η2p = .00. These results replicated those from Experiment 1.
Mean Increases in Certainty Ratings From the First LEI to the Second LEI for Imagined and Control Events and for Participants Who Imagined Events Occurring in the Past and Those Who Imagined Events Occurring in the Future.
As in Experiment 1, the main effect of imagining events and the lack of an interaction between the two factors suggested that there were imagination inflation effects in both the past and future conditions. To verify this, we examined imagination inflation in each condition. The change in ratings was greater for imagined events than for control events for those who imagined events in the past, t(104) = 5.31, p < .001, d = .52, and for those who imagined events in the future, t(112) = 3.17, p = .002, d = .30, supporting our first two hypotheses. Again, these findings replicated those from Experiment 1.
Next, we examined how many people showed imagination inflation. In the total sample of 218 participants, 60 of them (27.5%) showed imagination deflation, 31 of them (14.2%) showed no effect, and 127 of them (58.3%) showed imagination inflation. Of the 105 participants who imagined events in the past, 25 of them (23.8%) showed imagination deflation, 18 of them (17.1%) showed no effect, and 62 of them (59.0%) showed imagination inflation. Of the 113 participants who imagined events in the future, 35 of them (31.0%) showed imagination deflation, 13 of them (11.5%) showed no effect, and 65 of them (57.5%) showed imagination inflation. Consistent with Experiment 1, most participants experienced imagination inflation in both imagination instruction conditions of Experiment 2.
We then explored the relationship between autobiographical memory specificity and the magnitude of imagination inflation. Participants’ SCEPT responses were coded into two categories: Responses were coded as specific memories when participants described an event that lasted for 1 day or less, as nonspecific memories when participants described an event that lasted for more than 1 day, occurred repeatedly, or anything else that could not be coded as a specific memory. There were four coders; each of them coded half of the participants’ response, and so all responses were coded by two coders. They agreed on 82.8% (1,985 of 2,398) of their codes (Cohen’s kappa = .48). Disagreements were coded by an additional coder to determine a final coding.
The number of specific memories in the SCEPT ranged from 1 to 6 (M = 2.04, SD = 1.50). We first examined the correlation between specific memories in the SCEPT with the overall magnitude of imagination inflation (the increase in confidence for imagined items minus the increase in confidence for control items) with all 218 participants. This correlation was not significant, r(216) = .07, p = .283. We then examined this correlation in both imagination instruction conditions. This correlation coefficient was not significant in the future condition, r(111) = .09, p =.331, or in the past condition, r(103) = .05, p =.648. These findings replicated the failure to find significant relationships in Experiment 1.
Scores of the SCC scale ranged from 1.42 to 5.00 (M = 3.71, SD = 0.83). SCC scores did not significantly correlate with imagination inflation, r(216) = .04, p = .590. Consistent with some previous studies of SCC, SCC scores were positively correlated with participants’ age, r(216) = .21, p = .002. For completeness, we also examined the correlation between SCEPT and SCC scores and found that it was not significant, r(216) = .06, p = .408.
As we did in Experiment 1, we explored increases in certainty ratings as a function of initial ratings, focusing only on those events that were later imagined. There were a total of 654 ratings for the three imagined events (218 participants × 3 ratings). Of those, 285 events received a rating of 1, 60 received a rating of 2, 24 received a rating of 3, 32 received a rating of 4, 48 received a rating of 5, 25 received a rating of 6, 29 received a rating of 7, and 151 received a rating of 8. Participants again appeared fairly certain in their responding with 436 of 654 ratings being at the extreme of the scale. As was the case in Experiment 1, those events with ratings at the low end of the scale appeared most susceptible to imagination inflation. Events that received a rating of 1 increased by a mean of 1.22 (95% CI [0.98, 1.47]), and those that received a rating of 2 increased by a mean of 1.07 (95% CI [0.52, 1.61]). Events that received initial ratings in the middle of the scale did not change as much. Events that received an initial rating of a 3 increased by a mean of 0.33 (95% CI [−0.56, 1.22]), events that received an initial rating of 4 increased by a mean of −0.22 (95% CI [−0.90, 0.46]), those that received a rating of 5 increased by a mean of 0.35 (95% CI [−0.15, 0.86]), and those that received a rating of 6 increased by a mean of −0.08 (95% CI [−0.75, 0.59]). Events that received high initial ratings tended to decrease. Those events that received an initial rating of 7 increased by a mean of −0.59 (95% CI [−1.24, 0.07]) and those that received a rating of 8 increased by a mean of −0.66 (95% CI [−0.90, −0.41]). Thus, imagination inflation appeared most robust for events that participants are fairly sure they had never experienced.
Discussion
The results of Experiment 2 were nearly identical to those of Experiment 1. We found imagination inflation in both the past and future conditions and the effect did not differ between the two conditions. We also did not find any significant relationship between autobiographical memory specificity, SCC, and imagination inflation. We expand on these findings in the General Discussion.
General Discussion
In the two experiments in this study, participants rated their certainty of experiencing events, imagined some of them, and then in a follow session they rated their certainty again. There were two imagination conditions. As in other imagination inflation studies, some participants imagined experiencing events in the past. Unique to this study, other participants imagined experiencing events in the future. Compared to control events that were not imagined, certainty ratings increased after imagination, and this occurred in the past and future conditions. Experiment 1 used a college sample and Experiment 2 used MTurk. We are unaware of any previous studies that examined imagination inflation in an online study. Overall, this study was the first, to our knowledge, to show that future thinking can lead participants to become more certain that they had experienced events in the past and the first to show that imagination inflation can be studied online.
The experiments in this study also included measures of autobiographical memory specificity. We used the AMT in Experiment 1 and the SCEPT in Experiment 2. Because Dewhurst et al. (2017) found that more specific autobiographical recollections predicted susceptibility to the DRM effect, we predicted that autobiographical memory specificity would predict susceptibility to imagination inflation. Neither experiment in this study found a significant relationship between autobiographical memory specificity and imagination inflation. However, recent studies have shown that susceptibility to one type of false memory tends to be unrelated or poorly related to other types of false memories (Bernstein, Scoboria, Desjarlais, & Soucie, 2018; Calvillo & Parong, 2016; Patihis, Frenda, & Loftus, 2018). These studies have concluded that separate cognitive processes are involved in different types of false memories. Therefore, it may be the case that individual difference predictors of the DRM effect are not significant predictors of imagination inflation. Because SCC appears to influence aspects of autobiographical memory retrieval, we added the SCC scale as an individual difference measure in Experiment 2. SCC scale scores did not predict imagination inflation, but we replicated some previous studies that showed that SCC increases with age (Campbell et al., 1996; Diehl & Hay, 2011; Fuentes & Desrocher, 2012; Lodi-Smith & Roberts, 2010).
There are several explanations for imagination inflation that could explain the results of this study. Garry and Polaschek (2000) highlighted two explanations for imagination inflation: source confusion and familiarity. Source confusion occurs when the source of participants’ memory for a recently imagined event is mistaken for an event that was actually experienced. According to the familiarity account, imagining events makes them more familiar and this increase in familiarity leads participants to believe they must have occurred.
Brewin and Andrews (2017) reviewed two additional explanations for imagination inflation. First, because initial ratings for to-be-imagined events tend to be low, their increase on the second LEI may simply reflect regression to the mean. Second, imagining an event may lead to the retrieval of a true memory rather than the creation of a false memory. Any combination of these four accounts could explain the results of this study, but the fact that we counterbalanced imagined and control events casts doubt on the regression to the mean explanation. To elaborate, Pezdek and Eddy (2001) argued that imagination inflation occurs because of regression toward the mean. They found that events with low initial ratings tended to increase and those with high initial ratings tended to decrease. We found the same patterns with events that were imagined in this study. However, if low initial ratings for events regress toward the mean, this should occur equally for imagined and control events (particularly because we counterbalanced which events were imagined and which were control), yielding no imagination inflation effect. In this study, imagined events’ certainty ratings increased significantly more than control events, suggesting that regression toward the mean is an insufficient explanation.
This study had several strengths and limitations. Like most imagination inflation studies that employ the LEI, we focused on a small number of events for imagination exercises. In Experiment 1, we had eight critical events of which participants imagined four, and in Experiment 2, we had six critical events. The effects reported in this study may be dependent on the selected events. Different critical events would likely alter the effects. However, we used different critical events in Experiments 1 and 2 and found similar effects of imagination inflation suggesting that our findings are generalizable. Because some studies have reported that event valence affects imagination inflation (e.g., Bays et al., 2015; Sharman & Barnier, 2008), we had participants imagine a combination of positive and negative events (and neutral events in Experiment 2). Event valence did not affect imagination inflation and it did not interact with the other two variables in Experiment 1. Furthermore, the interval between the two sessions would likely affect the results of this study (Goff & Roediger, 1998). However, we had a 2-week interval in Experiment 1 and a 1-week interval in Experiment 2 and found similar results. Finally, the specific imagination instructions in this study would likely affect the results. Bays et al. (2015) found that some instructions were more successful in showing imagination inflation than others. The generalizability of the results of this study should be examined in future studies by manipulating some of these methodological details.
The novel contribution of this study was that we had some participants imagine experiencing events in the future and found that imagination inflated their certainty of experiencing these events in the past. These results are consistent with theories that claim that episodic future thinking relies on constructive memory processes (e.g., Schacter, 2012). The similarity in imagination inflation between the past and future imagination conditions in this study is consistent with claims of similarity between episodic counterfactual thinking and episodic future thinking (Schacter et al., 2015). Future studies should examine the effects of future thinking on other types of false memories. Dewhurst et al. (2016) showed that future thinking led to increased false memory in the DRM paradigm and this study showed that future thinking leads to imagination inflation. The effects of future thinking should be examined in other paradigms, such as the misinformation paradigm.
Appendix
The introduction and guided imagery instructions used in this study.
You are going to imagine four events. Each time, you will read a brief description of the event, you will be given a few moments to picture the event, and then you will be given more details to imagine. Try to picture each event as clearly and completely as you can. Then you will answer some questions about the event you pictured. It may help you to form a more complete mental picture if you include familiar places, people, and things in the imagined event. Also, close your eyes if that helps your imagination.
Imagine yourself at some point in the [past/future]. It may be a few weeks [ago/from now], a few months [ago/from now], or a few years [ago/from now]. Imagine that you {EVENT}. Begin by thinking about the setting you [were/will be] in when the event [took/will take] place. Really cast yourself [back/forward] in time, imagine that you are there, and pretend that it is really happening. Look around yourself and think about what you see. As you are casting your mind [back/forward] in time, think about as much detail as you can. Think about sights, sounds, smells, and time of day. Focus on every detail that comes to mind, even if it seems irrelevant. If you cannot remember all of an object, person, or event, then focus on the fragments as they come to you.
Once you have finished imagining this event, please respond to a few questions about the event. After the questions, please write a description in the space below including everything that you imagined pertaining to this event. Spend up to 2 minutes answering questions and writing about the event.
Footnotes
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.