Memory is an active reconstructive process, and as a result, it is susceptible to distortions and false memories. The study of false memories has been extensively explored in experimental psychology since the adoption of the Deese-Roediger-McDermott paradigm (DRM), which has become a standard method for investigating this phenomenon. In the DRM task, initially developed by Deese (1959) and later modified by Roediger and McDermott (1995), participants are presented with a list of words that are semantically related (e.g., mug, handle, coffee, etc.), converging on a non-presented word that is semantically associated (referred to as a critical lure, such as the word ‘cup’ in our example). Following a delay, participants are asked to recall or recognize these words. In the recognition memory version, participants are required to determine if they remember words presented earlier, including the semantically related critical lure words (i.e., cup) that were never actually presented, among new words that are semantically unrelated to the studied words (e.g., doctor). Typically, participants mistakenly recognize or recall the critical lures as words that were included in the studied list (Roediger and McDermott, 1995). This false memory effect has been extensively documented in a substantial body of literature (see Gallo, 2010, for a review), including its cultural variations (Wang et al., 2021), and is commonly explained by the activation-monitoring theory (Roediger et al., 2001).

This theory is based on the notion of spreading activation in semantic memory networks (Collins and Loftus, 1975). Semantic concepts are represented as nodes interconnected through bidirectional associative links (whose strength is stronger when nodes belong to the same semantic category). Activation of a node spreads along the links to associated nodes within the network. In the DRM task, during encoding, both the presented words and critical lures (i.e., unstudied but semantically related words to the studied ones) are strongly activated through spreading activation from the presented words to the critical lures. During the recall and/or recognition phase, participants engage in a controlled source monitoring process to differentiate between studied and non-studied words. Incorrect retrieval or recognition of unstudied lures can be reduced through a conscious, controlled process of rejecting recollection (Brainerd et al., 2003). Hence, false memories might occur because subjects have insufficient source monitoring that fails to inhibit the indirect spreading activation of the critical lure. This implies that not all subjects are equally predisposed to memory illusion, and the investigation of how individual differences may modulate false memory production is still an open field of research (Winograd et al., 1998; Baird, 2001; Watson et al., 2005). A higher proneness to memory distortion has been observed in subjects with higher tendencies to dissociative experiences (Winograd et al., 1998), higher vividness of imagery (e.g., producing a photograph-like mental picture, Winograd et al., 1998), higher expertise in the domain of the material to be learned (Baird, 2001) and poor sleep (Malloggi et al., 2022a,b, 2023).

However, individual differences based on thought and reasoning have been little studied: for example, it has been observed that false memories are associated with convergent thinking (Dewhurst et al., 2011), categorization ability (Hunt and Chittka, 2014) and negatively correlated with fluid and crystallized measures of intelligence (Zhu et al., 2010). A more prolific line of research has investigated the relationship between Need for Cognition (NFC) (Graham, 2007; Leding, 2011, 2013; Wootan and Leding, 2015; Parker and Dagnall, 2018) and false memories. NFC (Cacioppo and Petty, 1982) can be defined as the tendency to look for challenging cognitive activities and enjoy such mentally effortful tasks. Results support the view that high-NFC individuals are prone to producing more true and false memories than those with low-NFC because of their tendency to greater elaboration of the stimuli. In the context of activation-monitoring theory, greater elaboration leads to stronger connections among nodes in the semantic memory network (Viggiano et al., 2006; Graham, 2007; Zaragoza et al., 2011).

There is also some correlational evidence about the association between false memories and intuitive thinking, which is the tendency to reach decisions quickly based on automatic processes within the dual process theory of thought. Gronchi et al. (2016) found that individuals who rely on intuition are more likely to produce false memories in a recognition memory prose task. This result has been recently replied to by Nichols and Loftus (2019) who employed a recognition DRM-task. However, these studies have critical limitations. First, only recognition tasks were employed. In order to disambiguate different explanations about the relation between thinking and false memories, it is also useful to employ a recall task, as described in Section 3 (see also Gronchi et al., 2016). Second, only correlational evidence has been provided in the studies of Gronchi et al. (2016) and Nichols and Loftus (2019); an experimental design is needed to exclude the possibility that the observed association may be due to confounding variables. Based on the above, we investigated whether and to what extent intuitive thinking influences the production of false memories by using an experimental design employing both recall and recognition tasks.

Empirical evidence has supported the long-standing philosophical distinction of two different systems of thought, known as the dual process theory of thought, for the last 30 years. The first system is fast, automatic, relatively effortless, and based on associative processes. The second system is slow, controlled, requires a certain amount of effort, and is based on deliberation (Sloman, 1996, 2014; Epstein and Pacini, 1999; Lieberman, 2003; Stanovich, 2004; Kahneman and Frederick, 2005; Evans, 2006; Pennycook et al., 2018). The dual process theory of thought is a general label that includes several specific theories that differ in some respects as to the interaction between the two systems (De Neys and Glumicic, 2008; Thompson, 2013; Sloman, 2014). Depending on the specifics that different authors emphasized, the two systems have been labeled in different ways: System 1 vs. System 2 (Stanovich, 1999; Stanovich, 2004; Kahneman and Frederick, 2005); intuition vs. deliberation (Sloman, 2014); intuitive vs. analytical (Epstein et al., 1996); associative vs. rule-based thinking (Sloman, 1996); and fast vs. slow thinking (Kahneman, 2011). Following Evans (2006), there are two main views about how the two systems may interact. The Default-Interventionist (DI) models sustain that System 1 generates responses by default and System 2 may or may not serially intervene (Evans and Stanovich, 2013). On the contrary, Parallel models claim that the two processes occur simultaneously with continuous monitoring of potential conflicts (Denes-Raj and Epstein, 1994; Sloman, 1996). In both cases, System 2 may suppress the response activated by System 1. A third, more recent, view sustains a Hybrid Two-Stage model (De Neys and Glumicic, 2008; Thompson, 2013; Newell et al., 2015). According to this model, a “shallow analytic monitoring process” is continuously active to detect potential conflicts between the responses produced by the two systems. When an actual conflict is found, an “optional deeper processing stage” is activated to inhibit the System 1 (intuitive) response.

Overall, dual process theories of thought reflect the general two-system views (stimulus-driven vs. goal-driven) that are recurrent across different psychology domains. In particular, the inhibitory mechanism hypothesized by dual process-theories of thought appears to be similar to inhibitory control within attention in perceptual tasks (Giovannelli et al., 2016; Gavazzi et al., 2017, 2019; Benedetti et al., 2020). Although the pioneering research of Kahneman, which paved the way to the dual-process distinction has its roots in the attention domain (Kahneman, 1973, 2011), the relationship between inhibition in thinking and inhibition in cognitive control is still an open problem (Dorigoni et al., 2022).

Another line of research (Shenhav et al., 2012; Zemla et al., 2016; Bronstein et al., 2019; Pennycook and Rand, 2019; Gronchi and Zemla, 2021) has focused on a construct based on the dual process theory of thought called cognitive reflection. It is defined as “the ability or disposition to resist reporting the response that first comes to mind” (Frederick, 2005, p. 35), and it is measured with the Cognitive Reflection Test (CRT; Frederick, 2005). The standard version of CRT is composed of three questions. Each item has an intuitive but wrong answer. Thus, the right answer requires the inhibition of the natural wrong answer and the intervention of deliberative processes in order to find the right way to correctly solve the problem. For example, one question on CRT states: A bat and a ball cost $1.10 in total. The bat costs $1.00 more than the ball. How much does the ball cost? The intuitive and automatic response is 10 cents, the difference between $1.10 and $1.00. This is the first, but potentially erroneous, response that comes to mind. Indeed, with a little reflection, an individual may realize that a response of 10 cents implies a cost for the bat of $1.10 (1$ more than the ball), for a grand total of $1.20. This clear contradiction, if noted, should induce the reader to think more carefully about the problem, concluding that the right answer is 5 cents: in this case, the bat costs $1.05 with a grand total of $1.10 as stated initially. Several ways to compute the CRT score have been proposed (Böckenholt, 2012; Sinayev and Peters, 2015; Thomson and Oppenheimer, 2016; Erceg and Bubić, 2017) in order to capture different aspects of the CRT questions. The most common calculation methods involve summing the number of correct answers (yielding a score that reflects both the use of System 2 and numerical abilities, see Sinayev and Peters, 2015) or the number of intuitive answers. More recently (Sinayev and Peters, 2015; Dorigoni et al., 2022), it has also been proposed to sum non-intuitive answers (both correct and incorrect ones) in order to measure the tendency to inhibit the initial responses that come to mind.

Within the dual process theory of thought literature, the CRT has been widely used to investigate how cognitive reflection is related to several kinds of beliefs, including paranormal and religious beliefs (Pennycook et al., 2012, 2016; Shenhav et al., 2012; Bahçekapili and Yilmaz, 2017), apparently impossible mental magic effects (Gronchi et al., 2017; Gronchi and Zemla, 2021), and fake news (Bronstein et al., 2019; Pennycook and Rand, 2019).

It is important that Frederick (2005) gave a broad definition of cognitive reflection, whereby the CRT scores reflect a mix of dispositions and abilities. Since the contribution of Frederick (2005), extensive literature has investigated if the construct of cognitive reflection can be interpreted mainly in terms of styles, abilities, or both (Campitelli and Gerrans, 2014; Stagnaro et al., 2018; Otero et al., 2022).

Also, research has developed means to induce the use of a particular thinking process. For example, adding a time constraint when performing a task (Finucane et al., 2000) makes the use of intuitive thinking more likely. Other studies (Shenhav et al., 2012) have observed that asking participants to write a paragraph about a situation where they “use their intuition/first response that comes to mind” or “careful reasoning” while solving a problem can induce the use, respectively, of intuitive or analytical thinking.

In a previous study (Gronchi et al., 2016), we employed a modified version of the DRM task, specifically the recognition test of prose memory, to investigate the hypothesis that intuitive thinking could lead to the generation of false memories. Our findings revealed a relationship between the number of intuitive responses on the CRT and the number of lures encountered. However, we were only able to provide correlational evidence linking intuitive thinking and the presence of lures based solely on recognition memory. Furthermore, we did not measure potential confounding variables.

In Gronchi et al. (2016), we hypothesized two (not mutually exclusive) possible explanations for this correlation: (i) more intuitive participants may produce more false memories because they are less effective in inhibiting lure words that come to mind (in the same way the intuitive response in the CRT is not hindered); (ii) more intuitive participants are more prone to use a familiarity heuristic. That is, given a set of potential responses (written or internally generated), they rely more on feelings of familiarity to determine if something is a real memory or not. An attempt to disambiguate these two explanations may be based on the comparison between recognition and recall. In a recall task, responses must be generated internally: according to the activation-monitoring hypothesis, both target words and lures are activated in the semantic networks, and in the monitoring phase, false memories may be generated if not sufficiently inhibited. Also, the role of inhibition in a DRM recall task has been observed by Colombel et al. (2016). In a recognition task, possible responses are already available under the eyes of the individuals. Although monitoring and inhibition play a role in this case, there is evidence that familiarity is involved in false memories in recognition tasks (Arndt, 2010; Hanczakowski et al., 2013) but not in recall tasks (Schwartz and Metcalfe, 1992; Benjamin, 2005; Malmberg, 2008).

For these reasons, we assume that the role of inhibition mechanisms should be stronger in a recall task and that the use of the familiarity heuristic should be more effective in a recognition task. Another way to differentiate the role of inhibition with the general use of a heuristic consists in employing different ways to compute the CRT score: the score that measures inhibitory control (based on the sum of non-intuitive answers, both correct and wrong, see Dorigoni et al., 2022) can contribute to highlighting the possible different roles in false memory production. In the following studies, we test the hypothesis that intuitive thinking leads one to produce more false memories. First, in a correlational study, we investigated the relation between false memories obtained with a DRM task (both recognition and recall) and the CRT, measuring several potential confounding variables (such as NFC, numerical abilities, and working memory capacity). Then, in an experimental study, we prompted either analytical or intuitive thinking [using the manipulation proposed by Shenhav et al. (2012)] to verify if the latter induces more false memories in a standard DRM task differentiating between recognition and recall.

In our previous contribution (Gronchi et al., 2016), we found correlational evidence linking intuitive thinking, as measured by the CRT, with the number of lures encountered in recognition memory. This was achieved by utilizing a non-standard variant of the DRM task, specifically, a prose memory test. Prior to conducting an experimental study (see Study 2), our aim was to replicate these findings using a standard DRM task (Roediger and McDermott, 1995; Stadler et al., 1999), testing both recall and recognition memories. In addition, we included several other measures usually associated with the CRT (numeracy skills, working memory capacity). NFC was also incorporated because it is correlated with both the CRT and the production of false memories (Graham, 2007; Leding, 2011, 2013; Wootan and Leding, 2015). To the best of our knowledge, NFC is the only thinking-related variable that has been investigated in terms of individual differences in memory accuracy.

Also, we computed two CRT-based scores: CRT-deliberative (the sum of correct responses) and CRT-inhibition (the sum of non-intuitive responses, both correct and wrong) to differentiate the general use of analytical thinking from the inhibitory component of intuitive thinking. According to previous literature, we expect to observe the following associations:

This paper aims to investigate whether intuitive thinking may induce a higher rate of false memory. Here we complement the correlational evidence observed in Study 1 and previous works (Gronchi et al., 2016; Nichols and Loftus, 2019) manipulating the use of the two systems of thought. Thus, we experimentally primed intuitive or analytical thinking using the approach of Shenhav et al. (2012) and then measured false memory production in a DRM task in both recall and recognition tasks. As detailed in Section 3, we put forth two possible explanations for the association between intuition and false memories. First, more intuitive individuals may be less likely to inhibit false memories that they internally generate. A second explanation (that may coexist with the first one) is that more intuitive individuals may produce more false memories because they rely more on feelings of familiarity to determine if something is a real memory or not. The comparison of recall and recognition DRM tasks can provide evidence supporting either or both explanations: the familiarity heuristic is expected to have a stronger influence in the recognition task (Arndt, 2010; Hanczakowski et al., 2013), but not in the recall task (Benjamin, 2005; Malmberg, 2008), while the inhibitory mechanism should operate conversely. Considering the observed correlation between the inhibitory component of cognitive reflection and false memory production in Study 1, our prediction was that intuitive thinkers would exhibit a higher rate of false memories in the recall task, where the inhibition of lures plays a more prominent role.

The reconstructive nature of memory and the possible formation of non-veridical memories are hallmarks of memory research (Bartlett, 1932; Hemmer and Steyvers, 2009; Kiat and Belli, 2017; Baddeley et al., 2020; Frisoni et al., 2021). However, there is little research on how general thought processes may influence the formation of memories (Nichols and Loftus, 2019). In two studies, we found evidence of a relationship between cognitive reflection and false memory production, in line with previous research (Gronchi et al., 2016; Nichols and Loftus, 2019).

In a previous exploratory work (Gronchi et al., 2016), relying on the activation-monitoring explanation of false memory formation, we hypothesized that intuitive thinking could lead to more false memories due to inadequate inhibition of strongly activated lures or because of increased use of the familiarity heuristic to distinguish the memory to recollect. Exploiting the possibility to distinguish between a general tendency to use deliberation and the inhibitory component of cognitive reflection (Erceg and Bubić, 2017), in Study 1, we investigated the correlations among DRM task measures with these two components of the CRT. We found that only the inhibitory component was associated with false memory production in both recognition and recall tasks. Through manipulation of cognitive reflection obtained by priming either intuitive or analytical thinking before a DRM task, in Study 2, we confirm the centrality of the lack of inhibition in false memory production. Indeed, according to the activation-monitoring hypothesis, in the recall task, the lack of inhibition should play a central role in false memory production [as confirmed by Colombel et al. (2016)], whereas familiarity is strongly involved in the recognition task but not the recall task (Schwartz and Metcalfe, 1992; Benjamin, 2005; Malmberg, 2008; Arndt, 2010; Hanczakowski et al., 2013). On the assumption that the role inhibition component is stronger in the recall tasks whereas the familiarity heuristic is primarily involved in recognition tasks, the difference in recollected lures observed only in the recall task observed in Study 2 suggests again the centrality of a lack of inhibition in the production of false memories. As previously discussed, cognitive reflection can be thought of in terms of a mix of ability and style (Frederick, 2005). Our experiments are not able to differentiate between a lack of inhibition due to a decrease in this ability or a less marked disposition. This remains an open question for future studies.

In the recall of the DRM task of Study 1, the difference in lure production was not associated with other potential confounding variables related to numeracy, working memory capacity, or the need for cognition. Differently from expectations, we did not replicate the association between the need for cognition and false memories observed by several studies (Graham, 2007; Leding, 2011, 2013; Wootan and Leding, 2015; Parker and Dagnall, 2018). A possible explanation is that the sample of Study 1 did not guarantee enough statistical power to detect such an effect.

Although the activation monitoring hypothesis remains a credible theory, this is not the only possible explanation for false memory formation. Indeed, the Fuzzy-Trace memory theory (Reyna and Brainerd, 1995) has become increasingly central in literature on false memory (Reyna and Brainerd, 1998; Brainerd and Reyna, 2002; Reyna et al., 2016; Bialer et al., 2021). According to this theory, there are two types of memory processes. Verbatim trace represents the details of the stimulus (recollection), whereas gist trace represents the stimulus’ semantic features rather than its surface details. The false memories phenomenon could occur due to a dissociation between verbatim and gist processes in information retrieval. While true memory is sustained by both verbatim and gist processes, false memory is sustained only by gist processes that include the critical lures along with the studied items (Reyna and Brainerd, 1995; Brainerd and Reyna, 2002). Fuzzy Trace memory theory shares the same principles as the Fuzzy Trace theory of reasoning (Brainerd and Reyna, 2001), which represents an alternative account of reasoning processes compared to the dual process theory of thought (Reyna and Brainerd, 2011). A recent paper by Thompson et al. (2021) discussed the common ground and differences between the two theories. Specifically, they argue that the Fuzzy-trace theory is focused on the representations (verbatim vs. gist), while the dual-process theory of thought concentrates on the process. Keeping into account that the hypothesis that drove this work was at the process level (the relation between activation-monitoring and the dual process theory of thought distinction), our results are generally compatible with the Fuzzy Trace memory theory interpretation of false memory generation. Indeed, Reyna (2012) discussed the role of inhibition in incoherent responses when gist and verbatim representations conflict.

Overall, our work suggests that intuitive and analytical thinking can have a significant impact on false memory production, presumably because of the inhibitory component of cognitive reflection. Also, this contribution may represent a first step in deepening the relationship between higher-level thought processes and memory formation.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The study-performed according to the Declaration of Helsinki-is part of a set of behavioral and non-invasive study on face and word recognition memory processing, which were approved by the Research Committee of the University of Florence (protocol number 17245_2012). The studies were conducted in accordance with the local legislation and institutional requirements. The participants provided their written informed consent to participate in this study.

GGr and SR conceived the study. GGr, SR, GGa, FG, and MV contributed to the design of the study. GGr, SR, GGa, and FG performed the experiments. SR organized the database. GGr and GGa performed the statistical analysis. GGr wrote the first draft of the manuscript. SR, FG, GGa, and MV wrote sections of it. All authors contributed to manuscript revision, read, and approved the submitted version.