Introduction
False memory (FM) research has been abundant during the past three or four decades.
FM has been demonstrated on varying tasks such as verbal reports of incidents being affected by language manipulation (Loftus and Palmer, 1974), semantic word-list recall tests (Roediger and McDermott, 1995), and the creation or implantation of childhood false memories via verbal and pictorial stimuli (e.g. Gary and Wade, 2005). Research on FM is highly important due both to the information it provides about the mechanisms of human memory, and the real-world implications it has (e.g. for eye-witness reports and child sex abuse cases). Reported rates of FM vary extensively between studies (Scoboria, Wade, Lindsay et al., 2017), but a comprehensive review estimated the rate of complete full false memory recall at 15% in memory implantation studies (Brewin and Andrews, 2017).
Research to date has uncovered numerous effects on FM, including, but not limited to, emotional valence, personality and presentation factors. Brainerd, Reynard and Ceci (2008) reviewed the extensive data on the relationship between false memory and age. The authors concluded that while the FM rate increases with age, it is only for memories that are rooted in meaningful connections between events. This is consistent with the notion that frequency of similar prior events induces higher false recall rates due to priming or some similar mechanism. Brainerd et al. (2008) further concluded that age increases in FM are generally accompanied by ‘net declines in the accuracy of memory.’
Attempts to explain the variations of false FMs and the numerous effects upon them under a number of unifying principles have been made, culminating in a theory put forward by Brainerd and Reyna (2002) called Fuzzy-Trace-Theory (FTT), which is a dual-process theory of memory. Central to FTT are two parallel (as opposed to sequential) processes that occur during encoding: verbatim or specific traces which maintain precise features of an event; and gist traces, which preserve the general meaning but lack the specific detail of verbatim encoding. The theory also contends that verbatim and gist elements are retrieved by separate pathways and are therefore influenced by different cues. While FM is supported by gist processes but inhibited by verbatim processing, both types of processing support actual memories. This theory is well supported by the literature and is generally favoured over the implicit association theory of false memory (e.g. Cabeza and Lennartson, 2005).
In a similar vein Mazzoni and Vanucci (2007) attempt to explain FM by appealing to a reconstructive process coupled with metacognitive aspects. While the majority of the supporting literature for FTT comes from studies using word lists (e.g. the Deese-Roediger-McDermott paradigm [DRM]), Mazzoni and Vanucci describe principles explaining findings of investigations of hindsight bias, the misinformation effect and childhood false memories. Drawing on elements of FTT, the authors emphasise the role of reconstructive processes in the creation of false childhood memories. Plausibility of the hypothetical event is an intuitive and important element in producing FM in this view. Importantly, imagining an event increases both perceived familiarity and plausibility of an event. However, the authors also contend that ‘the absence of episodic representation might trigger a greater need for reconstructive processes’ and therefore the more sensory information that a subject is required to imagine, the more difficult that may be. Persistent guided visual imagery has been effective in eliciting false autobiographical memories, but the effects of adding auditory imagery or olfactory imagery to the hypothetical event has not been investigated.
The aforementioned explanatory frameworks of FM seem to do little in the way of predicting what the effects of adding multiple sensory imagery will be on the recall of FM. Furthermore, the recruitment of multiple sensory modalities has been shown to increase net memory accuracy. Given that it improves true memory recall, but that vivid imagery increases plausibility of events, the additive effects of sensory imagery in different modalities could reasonably be projected to have either positive or negative influence on false recall.
Additionally, D’Argembeau and Van der Linden (2006) found that participants with a higher capacity for visual imagery experienced a higher level of detail (both visual and in other modalities) when remembering past events. Therefore, the hypothesis being explored presently is that there will be a pattern of either increased or reduced false recall of childhood false autobiographical events due to the effects of additional sensory information, mediated to some extent by the influence of the visual imagery capacity of participants.
Methods
Participants
80 undergraduates (40 females, 40 males) from the UCD School of Computer Science participated in the present study. The mean age was 20 (SD = 2.1). It was established via communication with parents of each participant that the subject had not experienced the false event that was to be presented during the study. Subjects were assessed on imagery capacity. 64 participants were within the low to moderate range and were thus classified as having low to moderate imagery capacity (LMIC). These participants were then randomly assigned to one of the four experimental groups described below. The remaining participants were classified as subjects with high imagery capacity (HIC) and were subsequently assigned to fill the remaining slots in each of the four groups. In total, each group contained 16 participants with LMIC and 4 participants with HIC.
Materials
Participants were assessed on imagery capacity using the Vividness of Visual Imagery Questionnaire (VVIQ), which contains 16 items that refer to various situations the subject is probed to imagine. Subjects then rate the vividness of imagery on a five-point scale.
The study of FM has been questioned in terms of the methods used to investigate the various phenomena involved. Some query the generalizability of the DRM task to real world autobiographical memory (e.g. Wilkinson and Hyman, 1998), which has been popular due to its robustness and controllability. In the present study, we created a hypothetical event. Following Garry and Wade (2005) this consisted of a childhood hot air balloon ride. This allowed for varying versions of the story to incorporate imagery instructions that could include auditory and olfactory imagery instructions in the relevant conditions. The parents of participants supplied two actual childhood events for each participant that were not remembered by the participant. These were subsequently scripted in order to match the level of detail and emotional valence of the false event. The scripts detailed the unfolding of the events as well as other salient details such as the location, year and people present or involved.
Design
The present study implements a 2×4 factorial design. There were two independent variables. The first independent variable was the imagery vividness capacity of participants; the second independent variable was the level of sensory detail (number of sensory modalities involved in the imagery of the stimulus) provided to them. There were two levels of participant imagery vividness capacity, and four levels of imagery modality. Each of the four modality groups (control; visual imagery only; visual + auditory imagery; visual + auditory + olfactory imagery) were therefore further split by imagery vividness capacity of participants, essentially resulting in 8 groups. Given that imagery vividness capacity is naturally occurring and cannot be manipulated, while the experimental groups that are based on the level of sensory detail provided is controlled, this is a quasi-experimental exploratory design. The dependent variable is the rate of false recall.
Procedure
The first stage of the study assessed the imagery vividness capacity of participants using the VVIQ described above. Participants were asked to imagine hypothetical events and then rated the vividness of the imagery. The VVIQ had strong internal reliability (Cronbach’s alpha = .84). The scores were recorded for each participant. These scores were used to classify participants as LMIC or HIC, and participants were then assigned to one of the experimental groups as above.
Participants were told that the study was a straightforward test of memory. For participants in the control group (G0), a written description of the childhood hot air balloon ride was presented and read aloud by one of the research team. The participants were not instructed to specifically imagine any of the sensory details of the event. The three actual events provided by parents were presented in an identical fashion.
For participants in the visual imagery group (G1), the researcher read aloud a script of the event, highlighting various visual details (e.g. the colour and pattern of the balloon) and instructed the participant to imagine the details carefully. They were also presented the two actual events in an identical fashion.
Participants in the visual and auditory condition (G2) received the same instructions as those in G1 but with additional auditory details (e.g. participants were instructed to imagine the sounds of the hot air balloon mechanism, the bird utterances in the sky and the voice of their parents pointing out views below).
Participants in the visual and auditory and olfactory imagery condition (G3) received the same instructions as those in G2, but with additional olfactory sensory details (e.g. the smell of banana sandwiches that their parents had brought, the smell of the burner above). Participants in all conditions were presented the actual events that they had experienced with the same level of sensory detail that the false event was accompanied by.
Results
Considering the results of previous research on false memory implantations, it is expected that recall for the true events will be close to ceiling in all conditions, with no significant differences between groups.
Overall, it is reasonable to assume that participants in the control group will be significantly less likely to remember the false event that those in any of the three experimental conditions. This is due to the lack of sensory detail provided. As described in the introduction, familiarity and vividness within the reconstructive or gist processing has been strongly implicated in the recall of FM. It is expected that this will be replicated here. Therefore, if the control group here does not differ significantly, the previous research will need to be reconsidered.
The differences between the experimental groups will hopefully shed more light on the nature of the effects of imagery vividness, sensory detail provided and the reconstructive or gist processes involved. For example, if participants with HIC experience more false recall across groups, it can be cautiously concluded that imagery vividness has an important role. However, it will be the central point of this study to examine how this interacts with the amount of sensory detail provided. For instance, if the results show that in G2 and G3, HIC participants experience more FM than their LMIC counterparts within those groups, but there are no significant overall differences between the groups in terms of false recall, then that would indicate that the reconstructive process is not aided by the olfactory sensory detail. This has two possible explanations, however one of these can reasonably be ruled out. Firstly, it may be that olfactory information is simply unhelpful in creating FM. This scenario is unlikely however, given that prior research demonstrates a link between olfaction and memory. In light of this previous work, it the finding would appear to be an artefact of the ordering of the adding of modalities, i.e. olfaction was the last sense to be added; had vision instead been in the place of olfaction, then this explanation would have forced a conclusion that visual guided imagery has no impact. But, previous research has conclusively shown that this is not the case. Therefore, it is more likely that this latest addition to the sensory detail has no further impact only because the amount or additional distinct types (regardless of what the type may actually be) of sensory detail has already reached its maximum impact. The cognitive demands of adding additional detail when there is already a large amount present, is too great.
Alternatively, if participants in G3 prove less likely to experience false recall than participants in G2 but more likely than those in G1, then it may indicate that sensory detail is helpful to a point, and rather than merely levelling off as in the scenario just described above, it instead becomes detrimental to the recovery of FM. The interpretation employed here is that while the HIC participants’ superior imagery can still enhance the familiarity of the event and thereby increase the effectiveness of gist processing, in G3 the requirements for additional sensory detail ceased to aid false recall. Instead, these requirements begin to inhibit false recall due to an apparent overload on cognitive demand for gist processing. Though this is ‘unhelpful’ for recalling false memories, it is a generally positive effect in that it enhances net accuracy of actual memory. This seems plausible given the finding of Brainerd et al. (2008) that increases in FM are accompanied by net decreases in memory accuracy when combined with other research supporting FTT. However it cannot be reasoned conclusively without carrying out this study, as the other hypothetical results are also quite plausible.
Possible limitations of the current study include uneven sample sizes of HIC and LMIC participants. LMIC participants are more plentiful in the population from which the sample was obtained. Additionally, it would be exceedingly difficult to control fully for emotional valence between true and false events across the entire sample. Previous research indicated that FM recall is more likely to be elicited if the participant themselves was not responsible for reading the stimuli. In the present study, the researcher orally presented stimuli in each condition. This may result in a higher rate of FM than some studies employing self-presentation.