Do intuitive ideas of the qualities that should characterize involuntary and voluntary memories affect their classification?

When considering memory properties rated by participants and use them as an indicator of the difference between involuntary and voluntary memories, two issues need to be considered. The first is whether all memories reported by participants as ‘involuntary’ are indeed retrieved without the participant’s intention. The second is to assess whether the ratings of phenomenological characteristics are based on the objective properties of the memories, or on what participants believe the characteristics of involuntary memories should be.

The first is a methodological issue. In experimental psychology, autobiographical memory has most often been investigated using the word-cue method (Crovitz & Schiffman, 1974), where participants are presented with a number of verbal cues and asked to deliberately recall a personal episode in response to a cue (e.g. Barzykowski et al., 2019a, b, c). This word-cue method was also adapted by Schlagman and Kvavilashvili (2008) for the first experimental procedure of studying involuntary memories. In it, participants were asked to perform a boring attentional task while watching short cue phrases on a computer screen, some of which may incidentally trigger involuntary memories. Importantly, participants were specifically instructed to report when unexpectedly autobiographical memories were coming to mind during the task. While this allowed for the recording of involuntary memories under well-controlled experimental conditions, informing participants that they had to report only involuntary memories might have on one hand triggered monitoring processes that might have interfered with the unintentional retrieval, on the other hand the retrieval might have been at least occasionally intentional. The risk then is that this experimental procedure alters the involuntariness of the retrieval by either priming autobiographical memories and/or inducing voluntary processes (for more one the possible effects of different type of instructions see Barzykowski & Niedźwieńska, 2016; Vannucci et al., 2014). Thus, the core of this methodological challenge is to instruct participants to report involuntary memories without changing how these memories are naturally retrieved (see also Barzykowski, 2014; Barzykowski & Niedźwieńska, 2012, for a similar argument). This is especially important given the fact, as already argued by Michael, Garry, and Kirsch (2012), that expectations of a particular outcome, (e.g. the expectation of experiencing involuntary memories) may automatically modify our cognitive processes and behaviour to produce that outcome. So far several solutions have been devised in the attempt to minimize the unwanted influence of this ‘observer effect’ on the involuntariness of retrieval. Among them, for example, (a) the usage of probe-caught methods (e.g. Barzykowski et al., 2019a, b, c; Batool & Mazzoni, 2011; Mazzoni, 2019; Vannucci et al, 2014, 2019; Plimpton et al. 2015), (b) the usage of effort control scales (e.g. Barzykowski & Niedźwieńska, 2016; Barzykowski & Staugaard, 2016, 2018), (c) instructing participants to report any mental content without placing the focus of attention only on retrieval of autobiographical memories (Batool & Mazzoni, 2011; Vannucci et al, 2014), (d) when asking about the phenomenological properties of memories, keeping the rating procedure brief in order to not to interfere with the natural flow of involuntary memory retrieval applied by the usage of the two-step rating procedure described below (Batool & Mazzoni, 2011).

The second issue relates to relying on the participants’ introspection, in general, and on participants’ subjectivity, in particular, in deciding which memory is involuntary or voluntary, and in rating the memory phenomenological characteristics (e.g. vividness, clarity, emotional intensity). Conclusions about autobiographical memory retrieval, in general, and involuntary vs voluntary memories, in particular, which are based on phenomenological characteristics may be especially influenced by participants preconceptions and metacognitive beliefs about remembering their personal past. Given that we do not have any objective indicators of these memory properties yet, we typically “trust” participants’ judgments and their responses. However, there is the need to assess the extent to which these ratings are influenced by the belief participants hold about the nature and characteristics of involuntary memories.

Beliefs about memory are also referred to as laypeople theories, naïve theories, implicit theories, folk theories, or mindsets; e.g. Zedelius & Schooler, 2017). They are ultimately metacognitive beliefs, where metacognition has been defined as “cognition about cognition” (Flavell & Ross, 1981) or “thinking about thinking” (Yussen, 1985). More specifically, in this context it refers to the “stable knowledge or beliefs about one ‘s own cognitive system, and knowledge about factors that affect the functioning of the system; the regulation and awareness of the current state of cognition, and appraisal of the significance of thought and memories” (Wells, 1995, p. 302). As a result, metacognitive beliefs help people to interpret and understand both their own and other people’s cognitive behaviour (Dweck, Chiu, & Hong, 1995). While in general they may be useful, they may be sometimes also irrational and unreasonable (Palmier-Claus, Dunn, & Lewis, 2011). Metacognitive beliefs may have significant effects on a wide range of phenomena, including emotion maintenance and regulation (e.g. Tajrishi, Mohammadkhani, & Jadidi, 2011), memory performance (e.g. Horhota et al., 2012; Irak & Çapan, 2018), social functioning (Bright et al. 2018), spontaneous thinking (for a review see Morewedge & Kupor, 2018), and well-being (Sellers, Varese, Wells, Morrison, 2017; Østefjells et al., 2017), to name just a few.

Interestingly, the role of people’s lay theories was already extensively studied in mind wandering (for a review see Zedelius & Schooler, 2017), and involuntary memories may constitute the content of at least some of the task-unrelated thoughts studied in mind wandering research (for similar views see Johannessen & Berntsen, 2010; Plimpton et al., 2015; Mazzoni, 2019). For instance, the more control people believe they have over their mind wandering, the less they mind wander, and importantly, this is also true for experimentally induced/manipulated belief (Zedelius & Schooler, 2017; Zedelius, Protzko, & Schooler, 2017, 2020). In addition, in a series of experiments on spontaneous thoughts Morewedge, Giblin, and Norton (2014) provided strong evidence that the perceived lack of control over spontaneous thoughts (which, as a reminder, resides at the heart of the involuntary and voluntary memory distinction) leads people to perceive them as more meaningfully self-insightful. As a consequence, when thoughts appeared to have been retrieved involuntarily rather than voluntarily, people believed them to provide more meaningful and important insight about their own self and had potentially higher impact on their judgements. This series of results would strongly suggest that metacognitive beliefs influence autobiographical memory retrieval, and in particular the rating of involuntary memories characteristics. Involuntary memories might be even more subjected to the influence of beliefs as there is no control over and access to the involuntary memory retrieval.⁴

Strikingly, whereas over the last several years much progress has been made in gaining a better understanding of involuntary and voluntary autobiographical memory retrieval, the role of metacognitive beliefs in creating the distinction between involuntary and voluntary memory retrieval has scarcely been investigated. To the best of our knowledge there is so far only one published study (i.e. Sanson et al., 2020) addressing the possibility that “voluntary” versus “involuntary” retrieval may be indeed an attribution-based process suggesting that it may be also influenced by metacognitive beliefs.⁵

In summary, previous research show that people may ascribe different attributes to the memory retrieval which suggest that some part of the memory retrieval may be an attributional-like process. In addition, there is now also a large body of evidence showing that metacognitive beliefs may influence many aspects of our every-day functioning. However, no prior study investigated their role in the subjective distinction between involuntary and voluntary autobiographical memory retrieval.

The main goal of the present studies was to address the question concerning the possible influence of metacognitive beliefs on phenomenological memory properties. We manipulated retrieval intentionality (voluntary = recall a memory for each cue; involuntary = report any involuntary memory that comes to mind spontaneously). We investigated voluntary autobiographical memories using the word-cue method (Crovitz & Schiffman, 1974), in which an individual is presented with verbal cues and asked to recall a personal memory in response to each cue. To study involuntary memories, we used a modified version of Schlagman and Kvavilashvili’s (2008) experimental design, which allowed us to control the retrieval phase and observe memories retrieved without explicit intention. This an often-employed experimental methodology devised by Schlagman and Kvavilashvili (2008) to elicit involuntary memories under well-controlled experimental conditions (see for example Barzykowski & Staugaard, 2016, 2018; Barzykowski & Niedźwieńska, 2016, 2018a, b; Barzykowski et al., 2019a, b, c; Mazzoni, Vannucci, & Batool, 2014; Vannucci et al., 2014). More precisely, this task employed a two-step procedure of rating the memory content. First, while performing the vigilance task (detecting infrequent target vertical lines in a stream of slides with horizontal lines), participants were instructed to write down any spontaneously occurring memories experienced. In addition, they also rated the retrieved memories on few phenomenological characteristics (Part 1: e.g. vividness, clarity, intensity of emotions). This ‘online’ rating procedure was deliberately brief in order not to interfere with the main vigilance task. Second, after completing the main task, participants were presented with their previously recorded memories and asked to rate the characteristics of the event that each memory referred to using a larger number of questions (Part 2: e.g. importance of the event, personal nature, rehearsal: recalling in the past). Importantly, in the present study we separated these two phases using a filler task between the ‘online’ and post-task rating procedures (see below for more details). Thanks to this two-step procedure it is possible to distinguish between ratings based on characteristics perceived during retrieval (Part 1: The online-rating procedure) and characteristics assigned afterwards, that we hypothesized are based on metacognitive beliefs (Part 2: The post-task rating procedure).

To assess the role of metacognitive belief, besides expecting a greater effect of belief in the post-task rating procedure, we also manipulated the presence/absence of information about the retrieval intentionality at the time of the phenomenological characteristic rating of the memories. For the rating, participants were provided with either (1) separated lists of the memories they had reported, clearly and correctly labelled as involuntary or voluntary; or (2) one mixed list containing in random order both involuntary and voluntary memories, which were not labelled. Ratings were requested after a delay in order to minimize the possibility that participants remembered the intentionality of memory retrieval. Participants who received the labelled separate lists were in the “specified memory origin” group. The participants who received the mixed list were in the “unspecified memory origin” group. In general, we argue that any metacognitive beliefs about involuntary and voluntary memories should be especially strongly activated when people are informed of whether the memory was retrieved voluntarily or involuntarily (‘specified memory origin’ group) and minimized when people are not informed of the voluntary or involuntary nature of retrieval (‘unspecified memory origin’ group). Therefore, if beliefs about the nature of retrieval play a role, then we should observe stronger differences in phenomenological characteristics between involuntary and voluntary memories in the ‘specified memory origin’ group compared to the ‘unspecified memory origin’ group. The straightforward idea along these lines is that since in previous studies involuntary and voluntary memories were always clearly labelled and significant differences were found in phenomenological characteristics, we do not know whether these observed differences were due to actual differences in memory characteristics, or to the belief about which qualities should characterize involuntary and voluntary memories. If this is true that beliefs play a major role, then involuntary and voluntary memories should not be rated so differently when their origin is not highlighted at all. Our study is thus the first to examine the phenomenological characteristics of involuntary and voluntary memories when minimizing the information about the origin (voluntarily or voluntarily retrieved) of a memory, and thus to minimizes the potential role of metacognitive belief.

Overall, we expected that memories retrieved involuntarily (i.e. during the involuntary retrieval phase) would be in general more accessible compared to voluntary memories, and therefore more quickly retrieved. Phenomenologically, we expected to replicate the results as observed in previous studies (e.g. Barzykowski & Staugaard, 2016, 2018; Barzykowski et al., 2019a, b, c), showing that involuntary memories would be perceived more effortless, vivid and clear, accompanied by physiological sensation, emotional intensity, relevance to a person’s current life situation. We also expected, and this is the main hypothesis, the difference between phenomenological characteristics of involuntary and voluntary memories to be significantly lower in the unspecified memory group than in the specified memory group. This result was expected in particular in the post-task rating.

We employed a mixed design. Task (involuntary memory retrieval vs. voluntary memory retrieval) was the within-subjects factor. We called memories reported during the involuntary task ‘involuntary memories’ and memories reported during in the voluntary task ‘voluntary memories’. In other words, we assumed the two tasks to differ substantially in retrieval intentionality and operationalized the retrieval intentionality as the conscious decision to retrieve a memory. The other factor in the design (between-subjects) was information about the origin of the memories for the phenomenological characteristics rating (specified memory origin group vs. unspecified memory origin group). While participants in the specified memory origin condition were explicitly informed about the type of memories to be rated and told whether the memory was from the voluntary or involuntary task, participants in the unspecified memory origin group rated memories presented in a random order in a mixed list, without being told whether voluntary or involuntary. We expected participants in this group did not pay explicit attention to the nature (voluntary or involuntary) of the memories. We assessed the influence of this information on the phenomenological characteristics reported by participants.

A total of 60 participants (44 females, M_age = 23.42, SD = 4.28, range 19–39 years; one participant did not indicate age) were recruited and randomly assigned to the two experimental groups (specified memory origin and unspecified memory origin). Participants were tested in groups of two to twelve in a laboratory with separate computer stations. Five participants did not report any involuntary autobiographical memory, and additional seven had less than 50% of correct responses on the vertical lines task, and their results were excluded from the analysis. The final sample consisted of 48 participants with 24 participants in the specified memory origin group (18 females, M_age = 23.66, SD = 4.99, range 20–39 years), and 24 participants in the unspecified memory origin group (19 females, M_age = 23.19, SD = 3.08, range 19–32 years). They participated in return for a 20 PLN (ca. 5 USD).

To investigate the role of metacognitive beliefs on the attribution of phenomenological characteristics to autobiographical memories, we manipulated the information given to participants when they were asked to rate the memories they had retrieved in an involuntary task and in a voluntary task. For half of the participants, the memories were presented in two separate blocks, clearly labelled as involuntary or voluntary. For the other half memories were presented in a mixed list without any labelling. Labelling the memories as voluntary and involuntary would activate and facilitate any existing metacognitive beliefs about involuntary and voluntary memories. If reporting phenomenological characteristics depends on metacognitive beliefs, then a main effect of information about the memory origin should be obtained on the characteristics assigned to the memories. For example, in the specified origin condition one might expect involuntary memories to be rated as more effortlessly retrieved, more clear, vivid, personal, important, to name just a few of possible expected differences.

In our overall results, we were able, first, to replicate the following well-known findings from previous studies (e.g. Barzykowski & Staugaard, 2016, 2018; Barzykowski et al., 2019a, b, c; Berntsen, Staugaard, & Sørensen, 2013; Schlagman & Kvavilashvili, 2008): (1) despite the fact that there were many more cues in the involuntary conditions, more voluntary memories were observed, (2) voluntary memories were retrieved significantly more slowly than involuntary memories, (3) involuntary memories compared to voluntary memories showed a range of indicators suggesting increased accessibility, including higher physiological impact, emotional intensity, vividness, clarity, specificity, personal nature, importance, positivity, unusualness, rehearsal and recency.

However, this pattern of results did not vary between the two groups, those who received information about the origin of the memory (voluntary vs involuntary task) and those who did not. This suggests that such subjectively attributed characteristics may reflect truly perceived properties of memories, without being influenced by any pre-existing metacognitive beliefs. While these findings are in line with the results of previous studies, they also support the hypothesis that involuntary retrieval (i.e. no intention and no expectancy that autobiographical memory will come to mind) favours highly accessible memories (threshold hypothesis, see Barzykowski & Staugaard, 2016, 2018; Barzykowski et al., 2019a, b, c). Therefore, as argued by Barzykowski et al. (2019a; b, c), Barzykowski and Staugaard (2016, 2018) involuntary memories may be an example of a memory content that, because of its phenomenological properties (e.g. emotional intensity, personal relevance, vividness, clarity, unusualness), are especially good at drawing one’s memory-related attention. As a result, they may pass the awareness threshold more easily and thus may more likely be reported. Attributions of phenomenological characteristics might then reflect the way people perceive their qualities during retrieval, rather than depending on preconceptions on how involuntary and voluntary memories should differ.

However, while we did not observe any main effects of the memory origin on phenomenological characteristics, we did find an interaction between the memory task (i.e. involuntary vs voluntary) and memory origin (i.e. specified vs unspecified origin). More precisely, involuntary memories recorded in the specified memory condition were rated with the lowest effort ratings compared to involuntary and voluntary memories across all conditions. This suggest that participants may have a preconception about how effortless involuntary memories should be, because they may believe that involuntary memories should pop up spontaneously and automatically.

It should be noted that, contrary to the hypothesis of belief influence, providing participants with information about the origin of the memory attenuated (but not eliminated) differences between involuntary and voluntary memories in terms of specificity, clarity and personal relevance. As a result, when participants did not know the exact origin of a memory, IAMs were rated as more specific compared to voluntary memories, more clear (clarity of voluntary memories was decreased) and, although shy of statistical significance, more of personal nature. These results can support the claim that the phenomenological characteristics of the two types of memories overall do not depend on a metacognitive belief on how voluntary and involuntary memories should be.

When discussing the results, it could be argued that instructing participants in the involuntary memory recoding phase to report only involuntary memories (i.e. engaging participants into monitoring their stream of awareness looking only for autobiographical memories) might have increased the similarity between involuntary and voluntary memories (for a similar argument see Batool & Mazzoni, 2011; Barzykowski & Niedźwieńska, 2016; Barzykowski & Staugaard, 2016, 2018; Barzykowski et al., 2019a, b, c; Vannucci et al, 2014). This could have influenced the phenomenological differences between involuntary and voluntary memories as well as other measures such as mental effort and reaction time. For this reason, we repeated the experiment while instructing participants, similarly to previous studies (e.g. Barzykowski & Staugaard, 2016, 2018, unrestricted conditions; Barzykowski & Staugaard, 2016, 2018; Barzykowski et al., 2019a, b, c, unrestricted conditions; Barzykowski & Niedźwieńska, 2018a, b; Mace & Unlu, 2020; Vannucci et al., 2014, unrestricted conditions), to report any spontaneously occurring thought during the involuntary memory recording phase.

A total of 60 participants (46 females, M_age = 23.78, SD = 3.96, range 19–36 years; two participants did not indicate their age) were recruited and randomly assigned to the two experimental groups: the specified memory origin and the unspecified memory origin. Participants were tested in groups of two to twelve in a laboratory with separate computer stations. Eight participants did not report any involuntary autobiographical memory, two had less than 50% of correct responses on the vertical lines task, and additional three participants guessed the true goal of the study (i.e. that we were specifically interested in memories), and their results were excluded from the analysis. Therefore, the final sample consisted of 47 participants with 22 participants in the specified memory origin condition (18 females, M_age = 23.80, SD = 4.02, range 19–35 years), and 25 participants in the unspecified memory origin condition (19 females, M_age = 23.20, SD = 2.74, range 20–30 years). They participated in return for a 20 PLN (ca. 5 USD).

While instructing participants to report any involuntarily retrieved mental content (i.e. not asking participants to monitor their flux of awareness looking only for memories), we replicated the findings of Study 1 demonstrating an increased accessibility of involuntary memories compared to voluntary memories (lower retrieval latencies and higher ratings of most of the phenomenological characteristics). Most pertinent, also in experiment 2 we did not observe any robust effects of being informed about the origin of the memory, although when participants knew the exact origin of memory they were more prone to rate all memories as more unusual and identity oriented compared to the unspecified memory group. Importantly, this effect was similar for involuntary and voluntary memories, suggesting that they indeed differ from each other in the perceived characteristics, rather than in memory-related beliefs.

Across two studies we were able to replicate the main findings reported in the literature (e.g. Barzykowski & Staugaard, 2016, 2018; Barzykowski et al., 2019a, b, c; Berntsen, Staugaard, & Sørensen, 2013; Schlagman & Kvavilashvili, 2008) demonstrating robust differences between involuntary and voluntary memories. More precisely, the presence of retrieval intentionality was mainly and directly manifested by the higher number of voluntary memories that took also longer to be retrieved compared to involuntary memories. In addition, consistently across both studies, involuntary memories were rated as retrieved with less effort and showed several indicators of increased accessibility.

These findings provide additional empirical support to the hypothesis that because of certain phenomenological properties (e.g. vividness, emotional intensity, clarity, personal relevance, unusualness), some mental contents may be especially good at drawing one’s memory-related attention and, thus, they may pass the awareness threshold more easily and are more likely to be reported (i.e. threshold hypothesis, Barzykowski & Staugaard, 2016, 2018; Barzykowski et al., 2019a, b, c; Barzykowski et al., 2020).

When looking at the possible effects of the information about the memory origin, we expected to observe stronger differences between involuntary and voluntary memories in the ‘specified memory origin’ group, compared to the ‘unspecified memory origin’ group. We argued that if only metacognitive beliefs play a role, differences should be then minimized in the unspecified memory origin group and maximized in the specified memory origin group. In Study 1 where participants in the involuntary memory phase were instructed to report only spontaneously occurring memory, memories reported in the specified memory origin did not differ from memories in unspecified memory origin (i.e. main effects of the memory origin on phenomenological characteristics were not statistically significant). However, there was a significant interaction showing that knowing the origin of a memory made people rate involuntary memories as more effortlessly retrieved (compared to not knowing the origin of the memory). This result might be due to the information given in the specified memory origin condition, emphasizing that involuntary memories are retrieved automatically. If due to mentioning automaticity in retrieval, the effect also shows that the rating of memory characteristics can be relatively easily influenced, possibly by activating beliefs about characteristics of involuntary memories.

Finally, contrary to the hypothesis of belief influence, providing information about the origin of the memory actually attenuated (although not eliminated) differences between involuntary and voluntary memories in terms of specificity, clarity and personal relevance. Overall, then, these results cannot support the claim that the phenomenological characteristics of the two types of memories depend on the metacognitive belief on how voluntary and involuntary memories should be. Although there was an influence on perceived effort, the well-known robust differences between involuntary and voluntary memories observed in previous studies (e.g. Barzykowski & Staugaard, 2016, 2018; Barzykowski et al., 2019a, b, c; Schlagman & Kvavilashvili, 2008) were not abolished and/or undermined by the knowledge about the memory origin.

The same pattern of results was obtained in Study 2, in which participants in the involuntary memory phase were instructed to report any spontaneously occurring mental content. More precisely, while we did not observe any significant interaction, knowing the exact origin of a memory made participants more prone to rate memories in general as more unusual and identity oriented compared to the unspecified memory condition. This effect was similar for involuntary and voluntary memories, indicating that beliefs affected similarly both types of memories. Finally, it is important to highlight that we observed actually few effects of the memory origin across two studies, which supports the idea that the manipulating the presence/absence of information about the retrieval intentionality was generally effective.

Taking all these findings together, it can be argued that the differences between phenomenological characteristics of involuntary and voluntary memories seems then to reflect a genuine difference in qualities, that people perceive during retrieval, and are not due to memory-related beliefs.

While the main goal of the present study was to examine the role of metacognitive beliefs on the phenomenological memory properties, the overarching goal was also to reflect on possible factors influencing involuntary and voluntary memory retrieval. Over the years, substantial knowledge has been gained about the way memories relating to our personal past are retrieved. The distinction between involuntary and voluntary memory retrieval was a main breakthrough (due to seminal work by Ebbinghaus, 1885 and pioneering studies by Berntsen, 1996, 1998), which led to substantial investigations (e.g. Barzykowski & Niedźwieńska, 2016, 2018a, b; Barzykowski & Staugaard, 2016, 2018; Barzykowski et al., 2019a, b, c; Berntsen, 1998; Berntsen et al., 2013; Mace, 2005; Mace & Unlu, 2020; Mazzoni, 2019; Mazzoni et al., 2014; Plimpton, Patel, & Kvavilashvili, 2015; Schlagman & Kvavilashvili, 2008; Staugaard, & Berntsen, 2014; Vannucci et al., 2014, 2015). At the same time, different types of voluntary memories (directly and generatively retrieved) were identified (Conway & Pleydell-Pearce, 2000; Harris et al., 2015; Haque & Conway, 2001; Rubin & Berntsen, 2009; Uzer et al., 2012) which differ in terms of how automatically and effortlessly they are retrieved. We underline the fact that this distinction already represents an attenuation of the strong division between involuntary and voluntary memories. Recently, Barzykowski and colleagues (Barzykowski & Staugaard, 2016, 2018; Barzykowski et al., 2019a, b, c) went even further and addressed the possibility that the retrieval of voluntary direct memories and involuntary memories might be relatively similar. In other words, involuntary retrieval shares several processes and stages with voluntary direct retrieval, with differences pertaining mostly to the intentionality of the initial search. The authors proposed the threshold hypothesis to explain the complex interplay between different factors operating at several levels of memory retrieval. For additional clarity, let us unfold the act of a given autobiographical memory retrieval into the following stages (Wilckens, Erickson, & Wheeler, 2012)⁸: (1) pre-retrieval stage, (2) retrieval stage, (3) post-retrieval stage, (4) retrieval outcome report stage. (1) Various processes, including the creation of expectations, during the pre-retrieval stage can boost or impair the retrieval stage. For example, one may be placed in a “retrieval mode” in which “the cognitive system is prepared for or expects memory construction and recollection” (Conway, 2001, p. 1379). During this phase retrieval intentionality can also be explicitly formed influencing the subsequent retrieval process. The effect of priming might also occur in this phase, as it changes the cue-item discriminability, which can be defined as “how easily a given cue isolates an item” (Rubin, 1995, p. 151 as cited in Berntsen, 2009, p. 107), and, for some memories, it enhances the likelihood that the memory will enter a person’s awareness. (2) The retrieval stage relates to the forming and developing of an autobiographical memory, but without explicit self-reflection; namely, a given memory might have been formed but one may not be explicitly aware of having a memory yet (something that refers to an experiential level of consciousness; Baird et al. 2013). Importantly, during this stage a memory is triggered by and/or accessed via a given cue, and it may be either reconstructed, directly retrieved, voluntarily searched or involuntarily recalled, depending on the memory pre-retrieval and retrieval processes involved. (3) After the memory is formed, during the post-retrieval stage people may realize (i.e. by monitoring their stream of awareness, and/or mental contents) to have a memory in mind. Thus, this stage relates to the ability to, for example, extract autobiographical content from the stream of consciousness to explicitly become aware of having a memory that is autobiographical (this is the level of meta-awareness). One may also control the retrieval process by directing or modifying it, depending on the goal (e.g. if the goal of the retrieval is or is not achieved in the voluntary retrieval). At this stage, then one is fully aware that an autobiographical memory was actually retrieved, and in voluntary retrieval, if the memory content meets the given criteria then the search may be terminated. (4) In the last stage, the retrieved memory may be shared with others and reported by giving a verbal account of the content. This approach then focuses on two main difference between involuntary and voluntary memories, the first being intentionality during the pre-retrieval phase, the second being metacognitive monitoring and control processes and beliefs (see Koriat & Goldsmith, 1996; Mazzoni & Kirsch, 2002), which may play a role in all four phases.

Previous studies provide support to the notion that these stages may serve as independent, albeit interconnected, retrieval phases. For example, Baird and colleagues (Baird et al., 2013) demonstrated that the participant’s ability to monitor their stream of awareness and extract content of thoughts from it (including autobiographical contents) may be impaired by cognitively demanding tasks. As a result, it can occur that even when a memory is successfully retrieved (retrieval stage), one might not be aware of having it in mind (post-retrieval stage).⁹ As a result, involuntary memories are less frequently observed under highly demanding activities (e.g. Barzykowski & Niedźwieńska, 2018b; Vannucci et al, 2015, 2019).

It may be argued that meta-cognitive beliefs and preconceptions about different types of memories may operate during pre-retrieval, post-retrieval and retrieval outcome report stages. For instance, as argued by Vannucci et al. (2014), having a clear expectation of what an autobiographical memory should be, one may decide in the pre-retrieval phase to look only for this type of memories; similarly, in the post-retrieval phase memories are selected that meet more closely some given criteria. There is also the possibility that some pre-existing beliefs typical of the post-retrieval phase might retroactively influence the pre-retrieval phase, by priming a certain type of memories, thus increasing their possibility to be retrieved. Similarly, but with opposite effects, participants beliefs about what a memory should be can influence negatively its report. For example, the memory plausibility criteria influence what people report as memories (e.g. it is possible, by changing the belief of the plausibility of demonic possession can change their rate of report, Mazzoni et al, 2001; Mazzoni, 2007). One can also hypothesize that when reporting the properties of a given memory, one may boost or downplay its properties, depending on the idea about how these memories should be (retrieval outcome reporting phase).

While the memory retrieval as a whole should not be treated as an attribution-based process, some aspects should be. For example, processing its outcomes, reflecting on the memory properties, and eventually reporting them may indeed take such an attributional form. This accords well with the Whittlesea’s approach (e.g. Whittlesea, 1997; Whittlesea & Leboe, 2000) positing that subjective judgements and evaluations are rather independent and separate from retrieval (production) and can influence how this retrieval proceeds (Mazzoni & Hanczakowski, 2011). For this reason, one should take into account the role of attributions when studying such a transient form of remembering as involuntary memories. Although in the current study we found that overt, belief/based attributional processes have only a minimal role in the retrieval of involuntary memories, dividing the retrieval act into smaller sub-phases opens up a set of interesting questions, also for future research on autobiographical memory retrieval.

When discussing the results of the present studies, some limitations may be taken into account. For example, while engaging our participants into the involuntary and voluntary memory retrieval phase we did not counterbalance the order of these conditions. As a result, participants first performed the vigilance task (i.e. involuntary memory retrieval phase) and then they were instructed to recall autobiographical memories in response to each cue presented on the screen. This was mainly because we wanted to keep our participants unaware of the true goal of our study; namely, the autobiographical memory retrieval. Engaging participants into the voluntary memory retrieval before the involuntary memory phase might have maximized the risk that participants during the involuntary retrieval phase would continue voluntarily recalling memories in response to cues presented on the screen.

One may also argue that the delay (i.e. 40 min) between the online-rating procedure of the involuntary retrieval phase and post-rating procedure was insufficient to make participants rely on metacognitive beliefs in the condition in which they had been informed about the voluntary or involuntary origin of the memory. With a longer delay the memory for the origin of the memory would have faded more, leaving more room for an effect of metacognitive beliefs (participants would have to rely more on the intuitive ideas of the qualities that should characterize involuntary and voluntary memories). Thus, it would be interesting to extend the interval to a few days.

It may be also argued that engaging participants in the online-rating procedure might also have an anchoring effect on participants and thus limit how different people would then rate their memories during the post-task rating procedure. We however feel this to be rather unlikely, given the number of rating questions per each mental content and the delay period. In addition, this would be especially unlikely in Study 2, where during the involuntary rating procedure participants rated any involuntarily occurring mental contents and then, during the post-task rating procedure answered questions in relation only to autobiographical memories. Since there were more additional mental contents, remembering each memory origin was even more difficult in Study 2 than in Study 1.

It is also worth highlighting that in our studies we did not explicitly ask our participants about their metacognitive beliefs of involuntary and voluntary memories and retrieval. Thus, there is still a need for more direct measure of metacognitive beliefs (e.g. Carciofo et al., 2017; Zedelius et al., 2020). For example, it would be interesting to examine what pre-existing beliefs participants may actually have of different types of memories and then investigate how these beliefs do actually influence the memory retrieval. Finally, as already mentioned in the introduction, the rationale of the study was that if only metacognitive beliefs influence the rating of memory characteristics, then we should observe larger differences between involuntary and voluntary memories when participants are highly focused on this distinction. However, one may argue that it would be important to verify first which beliefs people hold about the characteristics of involuntary and voluntary memories. While there are no published studies on this matter yet, recently Sanson, Staugaard, & Barzykowski, in preparation asked two groups of participants to rate the phenomenological characteristics of typical involuntary or voluntary memory. Although participants were not instructed to recall their own memory but only to rate general characteristics, involuntary compared to voluntary memories were rated as less effortful, less voluntary, more interrupting the ongoing activity, more mood changing, less positive and more negative, more often accompanied by physical reactions, more unusual, and with more intense emotions. These findings clearly demonstrate that people may indeed have some sort of layperson understanding of what involuntary and voluntary memories should be. Therefore, we argue that such pre-existing metacognitive beliefs should show an effect on the rating of characteristics of memories people retrieve. Future studies will address the relationship between pre-existing beliefs and the rating of memory characteristics.

In the present studies we investigated the possible role of metacognitive beliefs on the phenomenological characteristics of involuntary and voluntary autobiographical memories. Across two studies, our results clearly and consistently demonstrated that the phenomenology of the memories in general, and the phenomenological differences between involuntary and voluntary memories, in particular, are not strongly and relevantly influenced by metacognitive beliefs. Involuntary memories were rated as very different from voluntary memories, replicating previous findings showing the increased accessibility of involuntary memories. Taking these results together, the two studies suggest that involuntary memories might indeed have intrinsic characteristics that make them phenomenologically different from voluntary memories, differences which are not due to people’s understanding of what involuntary and voluntary memories should be.