The present field study was designed to assess whether cannabis use increases the susceptibility to false memory formation. To allow differentiation of acute and residual effects of cannabis, we compared true and false recognition memory performance in intoxicated and sober regular consumers of cannabis with non-user controls on the DRM paradigm. Contrary to expectations, cannabis users did not demonstrate an increase in false memory rates for critical lures, relative to controls. However, both intoxicated and sober cannabis users showed elevated false alarm rates in response to new, unrelated items. Moreover, no group differences were detected with regard to true memory performance (hit rates), but the control group demonstrated higher net accuracy in memory performance compared with intoxicated cannabis users.

We found no evidence that cannabis use increases recognition of critical lures. At first instance, this finding seems to conflict with the findings by Riba et al. (2015) who reported elevated susceptibility to false memories in abstinent cannabis consumers. However, a closer inspection of the differences in methodology between the two studies may provide an explanation. In the study by Riba et al., a modified version of the DRM paradigm was used, exposing participants to 20 word lists each containing 4 associated words, which were preceded by an announcement of the list name (e.g., farm animals: horse, hen, sheep, goat; see supplementary materials of Riba et al. 2015). The recognition test then included lure items, which were words categorically related to the presented items (e.g., cow, pig), in addition to old, presented words and new, unrelated items. In contrast, the classic version of the DRM paradigm was employed in our study, where participants were presented with lists of 10 words that all primed identical critical lures, and the lists consisting of the first associates of the critical lure based on association word norms (Stadler et al. 1999). As such, the study by Riba et al. (2015) therefore did not include standard critical lures under the definition of the classical DRM paradigm, but words that were related to previously learned words but were not previously primed in the study phase.

DRM lists and the so-called category lists such as the ones used by Riba et al. thus differ in that DRM lists typically contain multiple different associative relations (synonyms, antonyms, concept relations, etc.) whereas category lists are restricted to only one level of association, that is taxonomy (Knott et al. 2012). Moreover, DRM lists are typically higher in backward associative strength (BAS), a measure of strength of the associative connections from study words to the critical lures. Research has shown that higher BAS leads to higher rates of false memories (Knott et al. 2012; Roediger et al. 2001). Although impossible to compare directly as numbers for BAS are missing in Riba et al.’s study, it can be argued that the two measures of false memory differ in associative strength, as the lures used in Riba et al.’s study can be considered less strongly associatively related to the initially presented lists, compared with the critical lures used in our study. The premise that high BAS tends to elicit high rates of false memories is also mirrored in the fact that in our study, false memory rates were rather high in all groups (65–69%), whereas they were relatively low in the Riba et al. study (20–30%).

On the other hand, a medium-to-large effect was found in cannabis use on false recognition of non-presented, associatively unrelated items, with cannabis users showing elevated false alarm rates. This effect fits well with previous findings reported in the literature of studies using non-DRM word list tasks, where acute cannabis intoxication was found to induce elevated intrusions and false alarms of new, unrelated stimuli (e.g., Hart et al. 2010). True recognition performance in our study was unimpaired by cannabis use, which mirrors other studies such as Riba et al. (2015). As mentioned in the introduction, recognition memory has only inconsistently been found to be impaired by cannabis use, a finding that has been reported both in acute and in long-term studies (Broyd et al. 2016; Solowij and Battisti 2008). However, it was found that net accuracy was highest in the control group, and although pairwise comparisons only detected a statistically significant difference between controls and intoxicated cannabis users, this indicates that cannabis impairs overall recognition accuracy.

Two theoretical frameworks can be used to explain the formation of false memories as elicited by the DRM paradigm: associative-activation theory (AAT; Howe et al. 2009) and fuzzy-trace theory (FTT; Brainerd et al. 2008). According to AAT, processing one word activates a corresponding node (i.e., concept) and spreads activation to surrounding, interconnected nodes within one’s semantic network (i.e., knowledge base). False memories can be produced if spreading activation has automatically activated neighboring but non-presented information, leading to false memories. According to FTT, events are encoded into two types of memory traces: verbatim and gist. The verbatim trace contains item-specific details of an event, while the gist trace captures the underlying meaning of the stimulus. Because verbatim traces fade quickly over time, people rely on gist traces when retrieving memories, thereby enhancing false memory formation (Brainerd and Reyna 2002).

Given that critical lure recognition did not differ in the cannabis use groups relative to controls, it does not appear that cannabis use enhanced activation for the related lure words in memory or produced an over-reliance on gist memory traces. However, given the observed increase in unrelated word recognition and the decrease in net accuracy for the cannabis use groups, it may be the case that cannabis impairs processing of the word lists during encoding (i.e., insufficient processing prevents either extensive activation or strong verbatim and gist trace formation). The finding that light cannabis users had worse net accuracy than the heavy users may also support this suggestion as light users are arguably less tolerant to the effects of THC than the heavy users (Ramaekers et al. 2009), so any cannabis effects would be more pronounced in the light use group, i.e., decreased processing in the light use group.

According to AAT, false memories or false alarms depend on the strength of association. DRM lists with high BAS may result in a stronger spreading activation, leading both controls and cannabis users alike to be certain that they remember the critical lure. They might receive a feeling of familiarity when being exposed to the critical lure. On the other hand, if there is no association as in the present study, or a lesser degree of association as in Riba et al.’s study, the level of uncertainty is greater, and thus individuals who are acutely intoxicated or have residual levels may exhibit a tendency toward more liberal responding. Multiple explanations can be put forward for this liberal responding. According to Ranganathan and D’Souza (2006), cannabinoids may induce increased intrusions due to increased mental activity, leading to irrelevant associations. In line with this idea, a recent animal study with cannabinoid receptor type 1 (CB1) knockout mice showed that hippocampal CB1 receptor activation increases the formation of incidental associations (Busquets-Garcia et al. 2018). Alternatively, cannabis use has been associated with increased impulsivity in decision-making (e.g., Metrik et al. 2012; Ramaekers et al. 2009). When making decisions under conditions of uncertainty, this may play out as a lowered decision threshold resulting in greater liberal acceptance of new information.

The current study has several strengths and novelties. Previous studies have been useful in illuminating the role of cannabis in false memory production but have several drawbacks: Specifically, in Riba et al. (2015), the participants were abstinent for at least 4 weeks, making it difficult to determine acute intoxication effects, whereas in Ballard et al. (2012), the participants received a specific dose that may not account for individual differences in tolerance levels. While the Vredeveldt et al. (2018) study advanced on this design by testing participants who chose their own cannabis dose, false memory was not measured directly using a method known to successfully lead to reliable levels of false memories. These drawbacks have been addressed in the current design. In a between-subjects design, we compared groups of acutely intoxicated individuals who use cannabis regularly, sober individuals who use regularly, and individuals without a history of cannabis use. This allowed the distinction between acute and residual effects of cannabis use. Moreover, the study was conducted in a naturalistic setting, maximizing ecological validity, as in a coffee shop, people are more likely to consume a dose specific to their tolerance levels.

However, the study is not without limitations. There was an unequal distribution of sex across groups, with the two cannabis groups consisting largely of male participants, while the control group had a higher proportion of female participants. Although this mirrors findings of cannabis use being more prevalent in males (e.g., Cuttler et al. 2016), it might pose a confounding factor, especially since when sex was included as a covariate in the analyses, no statistically significant group difference was detected for net accuracy anymore. It should be noted though that previous research gives no reason to expect sex effects in DRM performance (Bauste and Ferraro 2004; Seamon et al. 2002). A study by Dewhurst et al. (2012) found a sex difference but only with regard to negative stimuli while there was no difference for neutral lists. Nevertheless, future studies may need to account for sex differences by recruiting a more balanced sample.

Furthermore, as encoding and retrieval occurred in the same session (approximately 10 min apart), it is not clear whether cannabis impacts the encoding or retrieval of experiences. The previously described studies by Ballard et al. (2012) and Doss et al. (2018) separately examined the effects of THC on encoding and recognition testing 2 days apart, meaning encoding occurred during intoxication and retrieval while sober, and vice versa. Doss et al. found that THC at retrieval increased false memory effects whereas Ballard et al., if anything, found reduced false memory effects of THC during encoding. Future studies could investigate the issue of different memory stages and cannabis further by separating the encoding and testing phases with a longer time interval and varying the timing of intoxication. Future studies should also include an additional word category consisting of related but not critical lures, similar as in the study by Riba et al. (2015), to see whether the results converge. Measures of recall rather than just recognition memory, and metacognitive measures such as assessments of confidence would allow for a more comprehensive understanding of the effects of cannabis on multiple memory processes.

Finally, this study has important implications for legal, forensic as well as clinical settings. If cannabis users, intoxicated or sober, have a greater tendency for liberal responding when uncertain, this may have consequences in such settings. When presented with new, irrelevant information, they might be more likely to accept this new information as true or familiar, resulting in erroneous reporting. Even though DRM false memory seems far removed from autobiographical memory for a prolonged event such as a crime, the paradigm preserves an essential property of everyday false memories, namely that they arise from meaning relations (Brainerd 2013). Spontaneous false memories such as those in the DRM can arise in and have been relevant to legal cases (Brackmann et al. 2016; Howe et al. 2017; Otgaar et al. 2019), underlining the importance of the current study.