This study shows that cannabis use status (current vs. former use) and extent of cannabinoid exposure have a differential impact on specific measures of verbal and visuospatial learning. Specifically, select measures of verbal learning in current and former cannabis users were impaired i.e., Long Delay Cued Recall of the CVLT. Instead, visuospatial learning was impaired only in current users i.e., Retroactive Interference and Long Delay Rotated Recall of the BLT. Performance on the CVLT and BLT was not significantly associated with any measures of cannabis exposure.

Poorer CVLT Long Delay Cued Recall performance in current cannabis users was robust to sensitivity analyses which excluded 7 people who did not appear intoxicated but admitted having smoked cannabis recently i.e., <12 h before testing (instead of the required >12 h), and persisted in former users. Therefore, lower Long Delay Cued Recall may reflect the residual effects of chronic cannabis use that are long-lasting and detectable well-beyond prolonged abstinence. This interpretation is consistent with findings that verbal learning alterations in cannabis users are apparent after THC metabolites are no longer detected in urine (9), and in cannabis users with chronic exposure reflected by long duration [i.e., 16 yrs (25) and 23 yrs (24)], dependent and almost daily use (62).

However, these notions are not supported by the lack of robust correlations between Long Delay Cued Recall and any measure of cannabis exposure including abstinence durations and chronicity of use. Therefore, the specific indices of cannabis exposure driving CVLT alterations in these groups of current and former users are unclear. Of note, verbal learning is ascribed by the function of posterior and frontoparietal cortices (41, 46, 63–65) that are high in cannabinoid receptors and thus might be vulnerable to the long lasting effects of repeated impact of cannabinoid exposure via complex neural mechanisms (66).

Surprisingly, most of the CVLT performance differences between current users and controls (i.e., Trial 1, Sum Trials 1–5, List B, Short Delay Free Recall, Long Delay Free Recall) were no longer detectable after sensitivity analyses without 7 people who recently smoked cannabis i.e., <12 h before testing. Therefore, recent cannabis exposure may drive alterations of select components of verbal learning alterations (17, 24, 27, 41, 46, 62). This notion is supported by other study findings that better Ray Auditory Verbal Learning Test performance is associated with longer abstinence duration (67) (List B (27), total words recalled (12, 24). However, we found no significant correlation between better CVLT performance and number of hours from last cannabis use or urinary THC metabolites. Thus, the strength of this finding needs to be verified in future studies that carefully measure abstinence duration (68), as only a few studies of verbal learning to date report how long people abstained from cannabis before testing (25, 35).

Poorer visuospatial learning performance in cannabis users emerged in select BLT measures (Retroactive Interference and Long Delay Rotated Recall). These differences were robust and survived sensitivity analyses without a subgroup of recent users. Our findings are consistent with reports that visuospatial learning alterations in cannabis users affect recall but not acquisition trials (19, 45), but contrast previous meta-analytic findings that failed to find group differences (12).

The discrepancy between our results and those from previous work, might be due to systematic differences between sample characteristics and the tools used to assess visuospatial learning. First, our sample was older (i.e., ~34 years) than the meta-analyzed samples i.e., <26 yrs (12). Interestingly, previous evidence shows that aging affects visuospatial learning (69–71). Therefore, lower visuospatial recall may be due to altered aging processes in cannabis users. However, the lack of significant correlations between age and BLT performance in this study does not support this notion and is to be further tested in future work. Second, the meta-analyzed studies to date used measures of visuospatial learning other than the BLT. Such measures (e.g., accuracy and total scores from the Rey-Osterrieth and Bender Visual-Motor Gestalt Test) might not have been sensitive enough to detect alterations specific to recall rather than acquisition trials (12).

The mechanisms underlying altered BLT in recall trials in current users are unclear. One candidate mechanism is impaired executive functioning (45, 72–74). Indeed, recall—but not acquisition—relies on executive function (75). Also, aging significantly affects the integrity of para-hippocampal and cingulate cortices that are concurrently ascribed to visuospatial learning, and to the residual effects of regular cannabis exposure (19).

Interestingly, we found for the first time that both current and former users showed similar Learning Curves across the verbal and visuospatial domains, suggesting the employment of common learning strategies across both domains (73). This similarity was apparent, given the similar structure of the BLT and the CVLT (i.e., both tests start with 5 Learning Trials, followed by an Interference trial and Delayed Recall trials). The overlapping learning curves between current and former users (see Figure 1), suggest that alteration of learning curves commences during regular use and does not recover after prolonged abstinence.

There are several limitations to this study. First, we examined a small sample of former users, and our findings require validation in larger samples. Second, our group of current users were not matched to controls for level of education, IQ and severity of sub-clinical psychopathology symptoms. Nevertheless, premorbid IQ was matched between groups, and we controlled for current IQ i.e., by using IQ as a covariate in all group comparisons (years of education were not associated with any performance measures and were thus not included as a covariate).

We also minimized potential confounding impact of severe mental health conditions on cognitive performance, by screening for any diagnoses of psychopathology. However, we cannot rule out that sub-clinical psychopathology symptoms in our sample, which are shown to exacerbate cognitive deficits (15), may have driven our findings on cognitive alterations. Since sub-clinical psychopathology symptoms as measured in this study did not exert a significant effect on performance, they were therefore not included as covariates in the analyses. On the other hand, our sample is representative of cannabis users within the general community, where higher (although not diagnosable) psychopathology symptoms and worse cognitive outcomes have been consistently reported (14, 76–79).

Third, correlational analyses were run in current users and former users including males and females, and not in different sexes separately, therefore precluding a detailed understanding of possible sex differences in the emerging alterations. Our strategy mitigated type-1 errors, as we have run a substantial number of correlational analyses in current and former cannabis users across all cognitive variables examined. Fourth, the number of abstinent users (n = 12) was small for statistical analyses, and findings pertaining to this group require replication in larger samples. Further, the cross-sectional study design and the lack of information on the sample' resilience levels, prevented to determine if abstinence was the reason for better performance compared to current usage, or the consequence of a latent factor such as resilience that also leads to better performance.

Our findings might be accounted for in the design of cognitive interventions aimed to support abstinence in cannabis users. For example, if a clinical practitioner knew that a patient uses cannabis regularly, and that regular cannabis use is associated with impaired Long Delay Cued Recall, the practitioner may implement strategies to ensure that their client recalls information/instructions critical for engaging with the treatment (e.g., repeating or asking the client to repeat the instruction/information, sharing written instructions). Also, knowing that regular cannabis users have worse performance on Retroactive Interference and LD Rotate Recall, could indicate to a practitioner that visuospatial information relevant for the treatment (e.g., the location of appointment/testing/treatment sites) may not be retained when learning occurs; and may prompt the practitioner to implement strategies to boost recall of visuospatial information relevant for the patient to attend treatment (e.g., sharing a map or sending a text reminder with information on appointment/testing/treatment sites).

In conclusion, our findings suggest that current cannabis use is associated with chronic residual effects on select aspects of verbal learning (Long Delay Cued Recall) and of visuospatial learning (Retroactive Interference and Long Delay Rotated Recall). Prolonged abstinence in former users was associated with altered verbal learning but intact visuospatial learning, suggesting that abstinence has a different impact on distinct domains of learning. Our findings warrant the conduct of future work that systematically tracks how learning performance in chronic cannabis users is affected during intoxication to map acute-on-chronic effects, residual effects beyond intoxication, and those that remain with prolonged abstinence, to track how the residual-on-chronic effects dissipate over time. Additional studies with careful assessment of cannabis use indices and exposure to THC and other cannabinoids, should examine how these findings extend to cannabis user groups who experience worse mental health outcomes.