EVIDENCE LIBRARY
Does Blood THC Predict Lane Weaving the Morning After Cannabis? A 130-Driver Simulator Study
A 130-driver primary study found no significant corrected simulator lane differences or THC-driving correlations 12–15 hours after smoking. See the limits.
Short answer: In a 2026 observational study, 65 adults who regularly smoked cannabis did not perform significantly worse than 65 matched nonrecent users on verbal recall or Trail Making tests 12–15 hours after their last reported smoking session. But the cannabis group reported more residual drug effects, and among those 65 cannabis users, higher measured whole-blood THC and use of higher-potency products were associated with poorer results on selected cognitive tests. A non-significant group comparison and a significant within-group association answer different scientific questions. Neither establishes whether a particular individual was impaired or fit to drive the next morning.
The primary source is Matheson and colleagues’ 2026 original study, “Cognitive performance and subjective effects the morning after last use of smoked cannabis by adults who use cannabis frequently: An observational study”, published in Drug and Alcohol Dependence Reports. The distinctions below follow the original Methods, Results, Tables 1–5, and Discussion rather than an abstract alone.
What did the researchers actually find the morning after cannabis?
The investigators found no statistically significant difference on any of eight reported cognitive outcome variables between their 65-person cannabis group and 65-person control group after adjustment for sex, age, and education. For example, the adjusted delayed verbal-recall score was 9.28 words in the cannabis group and 9.55 in controls (between-group p = .60). Trail Making Part A times were 39.87 and 43.18 seconds, respectively (p = .39). Those differences were not demonstrated to be effects of residual cannabis.
At the same session, the cannabis group gave higher ratings of feeling drug effects. Its adjusted 0–100 visual-analogue rating for “I feel this effect” was 13.78, versus 1.24 among controls (difference 12.54 points; p < .01). Within cannabis users, whole-blood THC was inversely correlated with delayed verbal recall (r = −.40, p < .01). These observations can coexist without contradiction: a comparison of two heterogeneous groups tests one question, while a correlation among exposed people tests another.
Who participated, and when were they examined?
The researchers enrolled 130 healthy adults, including 65 regular cannabis smokers and 65 controls, matched by age and sex. Each group comprised 32 women and 33 men; the average age in each group was approximately 30 years. People in the cannabis group reported smoking cannabis on average four to seven days a week. Controls reported no cannabis use during the preceding month. This was a between-person observational comparison, not random assignment to cannabis versus placebo.
Cannabis-group participants obtained their own legally purchased pre-rolled cannabis cigarette, smoked at home in the evening, documented their starting and stopping times, and attended a single laboratory session the following day. The self-reported interval since smoking averaged 13 hours (standard deviation 0.9 hours; range 12.2–15.4 hours). The researchers checked the product packaging and weighed the remaining portion to estimate consumption. They did not directly supervise or independently time the smoking, and the product dose was not standardized across participants.
The study excluded several confounding medical and drug-use conditions, required a test-day breath-alcohol reading of zero, and administered a urine drug screen. Caffeine consumption was not standardized. Sleep measures were collected, but no design can remove every difference in prior cannabis exposure, education, social circumstances, or habitual cognitive performance.
What did the cognitive tests measure—and what did they omit?
The verbal free-recall task presented a 12-word list across three learning trials. Researchers scored initial recall, cumulative words recalled, improvement across trials, delayed recall, and the fraction retained after the delay. These are observable laboratory measures of selected verbal-memory processes, not a comprehensive evaluation of witness perception, encoding under stress, or the truth of an eyewitness account.
The paper also used the paper-and-pencil Trail Making Test. In Part A, the participant connected numbered targets in order; Part B alternated numbers and letters. Completion times captured visual scanning, sequencing, motor speed, and elements of mental flexibility. A slower completion time is a test-performance observation; it is not synonymous with unsafe driving or a clinically diagnosed cognitive disorder. This particular article did not report a fresh road-driving experiment. A related driving-simulator publication drew on the larger research program but addressed a different endpoint and must not be substituted for these cognitive results.
What were the adjusted cannabis-versus-control results?
Table 1 reconstructs selected comparisons from the original paper’s adjusted estimates. The sample denominator was 65 people per group; the p values test group differences, not whether a particular person’s performance changed from the previous night.
| Actual measured task outcome | Cannabis group, adjusted mean | Control group, adjusted mean | Group comparison |
|---|---|---|---|
| Initial verbal recall, words | 9.49 | 9.63 | p = .74; not significant |
| Total words across learning trials | 26.61 | 26.95 | p = .75; not significant |
| Delayed verbal recall, words | 9.28 | 9.55 | p = .60; not significant |
| Word-retention fraction | 0.87 | 0.88 | p = .68; not significant |
| Trail Making A, seconds | 39.87 | 43.18 | p = .39; not significant |
| Trail Making B, seconds | 57.49 | 62.05 | p = .34; not significant |
The study’s other two reported cognitive outcomes—verbal learning score and the Part B minus Part A completion-time difference—were likewise nonsignificant (p = .86 and .80). Thus, “there was no demonstrated average difference in these specified tests” is defensible. “Cannabis had been proven harmless,” “there was no residual THC,” and “every participant was unimpaired” are not.
Did people using higher-potency cannabis perform differently?
Within the cannabis group, 17 people reported using infused pre-rolls, which contain added cannabinoid concentrates or distillates, while 48 used noninfused pre-rolls. Their adjusted delayed-recall estimates differed: 8.10 words for infused-product users versus 9.96 for noninfused-product users, with a reported difference of 1.87 words, t(57) = 2.82, p < .01, and Cohen’s d = 0.84. The retained-word fractions were 0.78 versus 0.92, also reported as statistically different. Other recall variables and Trail Making outcomes were not significantly different between these product subgroups.
Product-label THC percentage was inversely correlated with delayed recall (r = −.36, p < .01) and retention percentage (r = −.28, p = .02). Yet the investigators did not find significant associations between those cognitive outcomes and their estimate of THC inhaled from product weight and labeling. Label potency, adding concentrates, estimated consumed product mass, inhaled THC, and absorbed biological dose are different measurements and should not be used interchangeably.
Because users selected their own products, these subgroup results are associations, not randomized proof that concentrates caused a specified loss of words. The small 17-person infused subgroup also limits confident extrapolation to specific cannabis formulations, newer products, and forensic individuals.
What were the measured whole-blood and oral-fluid cannabinoid levels?
Blood was obtained from participants and analyzed as whole blood for THC and selected cannabinoids; oral fluid was collected separately with Quantisal devices. Samples were frozen at −80 °C and shipped to a laboratory for quantification. These are not plasma or serum measurements, and no conversion from oral-fluid THC to a legal blood concentration was validated. The reported lower limit of quantification was 0.2 ng/mL for the assayed cannabinoid measures; for analysis, below-quantification values were replaced with 0.1 ng/mL. This substitution is a modeling convention, not a physical observation of 0.1 ng/mL.
Within the cannabis group, mean (standard deviation) measured whole-blood values were THC 2.75 (3.97) ng/mL, 11-hydroxy-THC 6.87 (15.48) ng/mL, THC-COOH 28.83 (44.78) ng/mL, and CBD 0.26 (0.97) ng/mL. Mean oral-fluid THC was 31.24 (75.12) ng/mL. The large standard deviations show substantial variation; means cannot describe any single person’s level. Only three participants had CBD above the analytical quantification limit, so the authors did not calculate CBD-performance correlations.
Whole-blood THC represents a direct analyte concentration at the next-day collection; THC-COOH is a carboxy metabolite rather than THC itself. Oral fluid samples a different biological compartment and can behave differently from blood. For guidance on why a specimen result must remain linked to its matrix, see the site’s existing biological-matrices explanation.
Which cannabinoid concentrations correlated with cognition?
The original Table 4 reports correlations within the 65-person cannabis group, not impairment-classification accuracy, causal effect estimates, or comparisons with the 65 controls. Table 2 preserves several notable findings and also identifies the separate outcome to which each coefficient pertains.
| Analyte and actual specimen | Observed cognitive association among cannabis users | Study statistic |
|---|---|---|
| THC, whole blood | Fewer words on delayed recall | r = −.40; p < .01 |
| THC-COOH, whole blood | Fewer words on delayed recall | r = −.29; p = .02 |
| THC, whole blood | Longer Trail Making A completion time | r = .46; p < .01 |
| THC, whole blood | Longer Trail Making B completion time | r = .31; p = .01 |
| 11-hydroxy-THC, whole blood | Longer Trail Making B completion time | r = .27; p = .03 |
| THC, oral fluid | Longer Trail Making A completion time | r = .35; p < .01 |
Several other analyte–task pairings were not significant. For example, oral-fluid THC was not significantly associated with delayed verbal recall in Table 4 (r = −.16, p = .21). Whole-blood 11-hydroxy-THC likewise was not significantly associated with delayed recall (r = −.10, p = .45). A positive association of THC with a timed task means the task took longer as concentration increased in this study; it does not mean every person above a particular concentration was impaired.
The researchers tested multiple cognitive endpoints, subjective endpoints, products, and analyte correlations. Individual nominally significant p values should be interpreted against that analytical multiplicity and the study’s exploratory observational design; they are not externally validated decision thresholds for driving, employment, testimony, or criminal responsibility.
Why did the cannabis group feel effects when their group cognitive scores did not differ?
Self-reported drug sensations and cognitive task scores measure different constructs. Participants rated how strongly they felt a drug effect, a good or bad effect, dizziness, drowsiness, and other experiences. Compared with controls, the cannabis group reported more on most visual-analogue scales and scored higher on the Addiction Research Center Inventory’s marijuana subscale. For example, adjusted “good effects” ratings were 35.60 versus 4.37 (p < .01), and “drowsy” ratings were 22.90 versus 7.71 (p < .01).
These are self-reports, not brain concentration measurements or direct evidence of unsafe behavior. Conversely, an average cognitive null does not invalidate subjective symptoms. The authors’ race-and-education sensitivity analysis left most conclusions unchanged, but some subjective contrasts changed: the “positive mood” difference was no longer significant in that sensitivity model (p = .19). This cautions against describing every questionnaire difference as robust to plausible adjustment.
Can this experiment distinguish residual drug effects from long-term cannabis-use differences?
Not definitively. The same participants were not tested repeatedly under randomized THC and placebo conditions. The people who regularly smoked cannabis could have had pre-existing differences in cognitive ability or adaptive strategies. The groups differed in race and education despite matching on age and sex, and unmeasured factors remain possible. The cannabis users’ consumption time and dose were reported rather than directly supervised. These limits prevent an unequivocal causal attribution of observed correlations to THC still acting the next morning.
Likewise, a null average group difference does not prove the absence of every residual pharmacological effect. Statistical nonsignificance is not a formal equivalence finding and does not establish that any possible difference must be clinically trivial. The study tested two families of cognitive tasks at one morning session after self-selected smoked products; it did not evaluate every real-world task, alternative routes such as edibles, infrequent users, or dangerous incident outcomes.
What does the study mean for forensic blood THC interpretation?
The paper supplies empirical evidence that some frequent users retain quantifiable THC in whole blood about 12–15 hours after reported smoking. It does not demonstrate that a blood THC result alone dates the smoking event, reconstructs the dose, predicts the person’s exact delayed-recall score, establishes a legal impairment threshold, or proves that intoxication caused an incident.
For a case involving a blood specimen, an analyst would need its collection time, actual matrix, identification and quantification method, analytical calibration and uncertainty, specimen storage, possible competing substances, and a separately supported account of conduct. The difference between a screening result and a compound-specific quantitative assay matters; the site’s confirmatory-testing guide addresses this distinction. Collection and preservation are independently relevant, as summarized in its specimen-collection overview.
The presence of THC-COOH should not be described as a direct measurement of contemporaneous psychoactive THC. Nor should oral-fluid THC be silently substituted for a blood concentration. No legal per-se cut-off, crash risk ratio, individual test sensitivity for impairment, or back-extrapolation rule was established by this study.
How should a lawyer or expert examine a next-day cannabis allegation?
The practical sequence begins by specifying which proposition is being advanced: recent use, residual drug exposure, a subjective sensation, a measured cognitive-task deficit, actual driving behavior, or causation of a specific event. Those propositions require different evidence. A court should not be asked to treat them as interchangeable simply because the same drug is mentioned.
Next, identify whether a reported cannabinoid figure arose from whole blood, plasma, serum, or oral fluid; inspect the laboratory method, quantification range, time, preservation, and possible analytical issues. Assess the individual’s actual conduct independently rather than attaching a universal response to a population-level mean or correlation coefficient. If the evidentiary issue concerns memory, identify precisely which component—attention at the scene, initial encoding, later recall, recognition, or testimony consistency—was measured and which remains untested.
Finally, present both sides of the original research: the nonsignificant cannabis-versus-control cognitive comparisons and the significant selected within-user blood THC/potency associations. Suppressing either side produces a misleading account. Neither result is an individualized medical or forensic determination.
Frequently asked questions
Was the average cannabis group cognitively impaired the next morning?
Not as established by this paper’s group comparisons. Eight reported verbal recall and Trail Making outcomes showed no statistically significant cannabis-versus-control differences. That is narrower than proving no individual experienced impairment.
Did the authors directly measure THC in the blood?
Yes. They quantified cannabinoids in whole blood, separately from oral-fluid THC. Those measurements were not modeled blood-equivalent results derived from breath testing.
Does a whole-blood THC level of 2.75 ng/mL identify a person’s impairment?
No. 2.75 ng/mL was the cannabis group’s mean whole-blood THC concentration, with a 3.97-ng/mL standard deviation. The study established neither an individual impairment classification nor a legal threshold at that value.
Were participants randomly assigned to different cannabis doses?
No. Cannabis users chose legally purchased products and smoked at home. Potency, infused-product status, and actual exposure were not randomized or directly controlled.
Did infused joints cause worse memory?
The 17 infused-product users had lower delayed-recall and retention estimates than the 48 noninfused users. The design supports association, not a conclusive causal finding about infused products.
Was oral-fluid THC a reliable indicator of delayed verbal recall?
Not in the reported within-cannabis-group correlation: oral-fluid THC and delayed recall had r = −.16 and p = .21. Oral-fluid THC correlated instead with certain Trail Making times, a distinct endpoint.
Does the study prove frequent users can safely drive 13 hours after smoking?
No. This paper assessed selected cognitive tests and subjective effects at a single next-day visit, not individual on-road safety. Safety and event causation require evidence not supplied by these group and correlation analyses.
Primary study and interpretation boundary
Matheson J, Zakala C, Zhao S, et al. (2026). Cognitive performance and subjective effects the morning after last use of smoked cannabis by adults who use cannabis frequently: An observational study. Drug and Alcohol Dependence Reports 19:100432. doi:10.1016/j.dadr.2026.100432. The researchers reported an observational matched-group comparison, direct whole-blood and oral-fluid cannabinoid measurements, verbal memory and Trail Making performance, and separate subjective drug-effect scores. This original experiment does not supply a validated individual impairment test or causal reconstruction of any particular event.