Can a Smartphone Detect Slower Movement After Cannabis Concentrates When Plasma THC Poorly Predicts It? A 65-User Study

Short answer: In a study of 65 experienced cannabis-concentrate users, smartphone accelerometers detected slower isolated arm movements immediately after self-administered concentrate use, slower leg movements one hour afterward, and temporarily greater postural sway. Finger-tapping speed and balance with the head tilted backward did not show statistically significant overall changes. Changes in measured plasma THC and 11-hydroxy-THC showed only weak associations with one balance measure. The results support a limited, task-specific observation about a frequent-user sample; they do not validate a phone as a roadside impairment instrument or allow a toxicologist to infer any driver’s individual condition from a blood result.

The underlying original experiment is Hitchcock and colleagues, “Acute Effects of Cannabis Concentrate on Motor Control and Speed: Smartphone-Based Mobile Assessment,” Frontiers in Psychiatry (published January 2021), doi:10.3389/fpsyt.2020.623672. Its distinctive contribution is the simultaneous measurement of simple limb acceleration, balance and chemically assayed plasma cannabinoids across three within-person timepoints. Understanding precisely what was measured—and what was not—is essential when the research is cited in a forensic dispute.

What question did the investigators actually test?

The investigators asked whether naturalistic use of potent cannabis concentrates would be followed by measurable changes in basic motor performance among people already accustomed to concentrates, and whether response patterns differed by sex or tracked changes in circulating cannabinoids. These were deliberately simple controlled tasks, designed for portability in a mobile laboratory. The study was not a controlled test of actual vehicle operation, a comprehensive neurological examination, or a validation of a legal THC concentration limit.

That distinction matters because “motor performance” encompasses different processes. Producing a fast jab, raising a leg quickly, tapping a screen with one finger and maintaining upright balance are not interchangeable expressions of one universal intoxication score. A person may exhibit a change in one task but not another. Even a statistically significant average change in a group does not indicate what happened to any particular person without that person’s performance data and an appropriately validated interpretation.

Who participated, and how many independent observations were there?

Seventy-five cannabis-concentrate users consented. Ten were excluded because essential motor testing or qualifying plasma measurements were missing or did not meet prespecified criteria, leaving 65 independent participants—35 men and 30 women. The mean age was 27.88 years, with a standard deviation of 9.49 years. Participants reported using cannabis about 25.83 days per month and concentrates about 17.02 days per month on average. These were experienced users, not a random sample of drivers or people with little cannabis exposure.

Participants had to use concentrates at least four times in the preceding month and have at least a year of general cannabis use. The study also excluded specified other drug-use patterns and particular clinical conditions. That selection reduces some sources of variation but further limits the population to which its results can be generalized. The research team explicitly noted that this sample overlapped substantially with a prior report but was not an independent replication of a new set of users.

Each included participant provided observations at three timepoints: before using the concentrate, immediately afterward, and one hour afterward. The 195 potential visit-time observations are repeated measurements nested within 65 people, not 195 independent participants. Each task also involved multiple movement repetitions and trials. Those technical repetitions improve the precision of an individual’s task summary but cannot be counted as additional human subjects.

Was this a blinded randomized dose-controlled experiment?

No. Researchers randomly assigned participants to purchase a legal-market concentrate labeled about 70% or 90% THC. Participants took the product home, had a period of familiarization, and on the experimental day administered their preferred amount in their own residence before returning to a mobile laboratory. The procedure was naturalistic and could not blind participants to cannabis consumption; federal research restrictions also limited the research team’s handling of the dispensary products. Participants reportedly consumed an average of about 0.13 g concentrate, with large variation (SD about 0.19 g).

The product percentage describes the commercial material’s reported composition, not the amount of THC absorbed by an individual. The route, heat exposure, inhalation pattern, puffing behavior and amount ultimately delivered into the circulation were not standardized. A participant’s pre-use measurement can serve as a within-person reference, but repeating a motor task over time without a placebo condition introduces potential learning, fatigue, anticipation, motivation and time-order effects. Consequently, the temporal findings are informative but do not isolate a THC-specific causal mechanism as cleanly as a placebo-controlled, administered-dose crossover trial could.

How were the cannabinoids measured, and in which physical specimen?

A phlebotomist obtained venous blood, but investigators then separated plasma from erythrocytes and measured parent delta-9-THC and the active metabolite 11-hydroxy-THC in plasma using validated high-performance liquid chromatography–mass spectrometry. The stated lower quantifiable concentration was approximately 0.32 ng/mL. Across 450 cannabinoid data points, 22 values below that analytical limit were numerically assigned zero for analysis. A value below a method’s quantification limit does not establish that the molecule was chemically absent.

The study’s enrollment/analysis rules required an acute post-use plasma THC measurement of at least 20 ng/mL and an increase over the pre-use measurement. Thus, the final sample deliberately excludes people who had lower or nonqualifying acute responses. That selection is important when judging whether these correlations could be transported to the wider driving population. The study did not report parent THC in forensic venous whole blood, did not assay urine THC-COOH as its exposure variable, and did not construct a validated whole-blood-to-plasma conversion for legal use.

Concentrations obtained from separated plasma must retain that matrix label. Nor is 11-hydroxy-THC chemically the same analyte as parent THC. One cannot add them together and call the result a statutory whole-blood parent-THC concentration or a measured brain concentration. The study’s chemical analyses help describe exposure and explore biological relationships, but the assays are not direct measurements of impaired driving.

How did a smartphone quantify movement and balance?

Researchers used an iPod Touch smartphone-class device and an accelerometer-recording application. For an arm-extension task, the device was attached to the forearm during a brief, fast, isolated forward movement. For the corresponding leg task, it recorded peak acceleration during an upward movement of the lower leg. Peak acceleration, expressed in units of gravitational acceleration (g), was used as a movement-speed-related proxy; it is not an independently measured highway speed or a reaction time in milliseconds.

For finger tapping, the test quantified taps per second in a constrained task. Balance testing used a device at the waist to measure the variation in acceleration while subjects stood in prescribed conditions: eyes open, eyes closed, and eyes closed while tilting the head backward. Larger variability was interpreted as more postural sway under the specified protocol. It is not a direct count of falls or proof that a participant would fail an officer-administered field sobriety test.

Tests were repeated, task procedures were demonstrated, and summaries were calculated using the recorded movement traces. This gives an objective electronic signal rather than relying solely on the participant’s subjective report. However, a portable accelerometer is only as forensically relevant as its protocol, calibration, reproducibility, normative population and validated relationship to the precise issue in dispute.

What changed immediately after use—and what did not?

The sharpest acute finding concerned the arm-extension task. Average peak acceleration fell approximately 15% from pre-use to immediate post-use (p<.001), and the decline remained approximately 16% at one hour versus pre-use (p<.001). There was no significant additional change between the immediate and one-hour measurements (p=.52). This supports a persistent alteration in this isolated movement within the observed time window, not an identified dose–response function or an impairment duration beyond it.

Leg movement showed a different time course. The immediate post-use comparison was nonsignificant (approximately −2%, p=.58), while movement was slower at one hour compared with pre-use (approximately −7%, p=.033) and compared with the immediate timepoint (approximately −6%, p=.026). Finger tapping, despite its intuitive appeal as a phone-based quick test, had no significant overall time effect (p=.100). Calling every phone movement index abnormal would therefore misrepresent the source.

With eyes open, postural sway increased approximately 14% immediately after use (p=.017); the one-hour value differed from pre-use by approximately 8% but was not significant (p=.11). With eyes closed, the acute increase was about 11% (p=.013), but the comparison with baseline at one hour was nonsignificant (p=.36). The more demanding head-back/eyes-closed condition had no meaningful overall time effect (p=.95). These null results are important counterweights to any generalized claim that every balance measure was uniformly impaired.

Evidence table 1: Source-reported changes in six motor outcomes

Were men and women affected differently?

Men generally generated faster absolute arm and leg movements and a higher tapping rate than women. The reported time-by-sex interactions were not significant for arm acceleration (p=.097), leg acceleration (p=.86), tapping (p=.64), eyes-open sway (p=.52), or eyes-closed sway (p=.99). Accordingly, the observed average post-use change was not shown to follow different time trajectories by sex in this sample.

That finding is not proof that the biological response is identical in everyone. The study had 30 women and 35 men, with its own eligibility restrictions and task-specific variability. A nonsignificant interaction can reflect a small true difference, imprecision, or insufficient power; only a suitably designed equivalence investigation could support a firm claim of negligible sex-based variation. Likewise, a higher average baseline acceleration in one sex does not mean a person with lower acceleration is intoxicated.

Did a larger plasma THC increase predict worse motor control?

Not reliably enough for individual classification. The investigators compared changes in motor outcomes with changes in plasma parent THC and plasma 11-hydroxy-THC from before use to immediately afterward. For eyes-open sway, they found weak positive correlations with plasma THC change (r=.247, p=.049) and with 11-hydroxy-THC change (r=.296, p=.017). These correlations describe statistical covariation between two change scores. They do not specify a sensitivity, specificity, decision threshold, or probability that any individual person is impaired at an asserted concentration.

The investigators also studied whether different motor outcomes changed together. Acute eyes-open and eyes-closed sway changes correlated modestly (r=.381, p=.002), as did acute arm and leg acceleration changes (r=.348, p=.005). In contrast, acute arm change was not associated with eyes-open sway change (p=.88) or eyes-closed sway change (p=.70). Even different phone measures did not cohere into a single universal scale. These exploratory multiple associations were not established as an externally validated diagnostic rule.

The arm effect was observed as a time comparison. A weak correlation for a separate balance outcome does not bridge that evidentiary gap. Importantly, a plasma THC value does not convert to a legal whole-blood threshold simply because both are reported in ng/mL.

Evidence table 2: Associations that constrain forensic interpretation

What does a statistically significant result establish?

A p-value for a repeated-measures comparison addresses the evidence against a statistical null model under specific assumptions. It does not identify the molecule that caused the change, prove an individual’s impairment, state the magnitude of a safety risk, or validate a device for police screening. The study’s original analysis also made a number of comparisons across tasks, timepoints and correlations without a blanket correction of every inference across the full analysis family. Isolated p-values near .05 should therefore be treated as exploratory, especially when multiple related questions have been tested.

The reported percentage change is an average, not a fixed effect that should be mechanically applied to each individual. A 15% mean arm acceleration decrease does not permit a reconstruction of a particular defendant’s movement, reaction to hazards, braking, or collision avoidance. Conversely, a nonsignificant tapping result does not prove that every user’s fine motor control was unaffected. Sound forensic interpretation retains the full pattern: positive findings, null findings, their measurement units, and the scope of uncertainty.

Can the results support a roadside smartphone impairment test?

They show that a portable accelerometer can record laboratory motor signals that change over a relatively short observation window after naturalistic concentrate use. That is a feasibility and group-research finding, not a clinical or roadside validation. To classify a driver, a test would need standardized acquisition and device calibration, normative reference distributions, interexaminer reproducibility, prespecified thresholds, external validation, error rates in realistic populations, and evidence of a relationship to the safety-critical function at issue.

Those requirements are especially important for a simple ballistic movement. Device position, limb length, strength, mobility restrictions, pain, prior injury, age, fatigue, hearing or comprehension of instructions and testing practice can affect recorded acceleration. Balance may be influenced by vestibular disorders, medications, footwear, musculoskeletal function, vision and the testing surface. A motor-data signal is objective in the sense that it is electronically recorded; its cause and forensic significance remain interpretive questions.

There is also a contextual mismatch. Driving comprises divided attention, hazard detection, real-time choices, pedal operation and vehicle control in traffic. Isolated upper- and lower-limb accelerations may be relevant to motor science but were not established here as substitute endpoints for actual road driving or crash avoidance.

What alternative explanations and study limitations matter?

The absence of a placebo arm is central. An immediate pre/post comparison cannot fully distinguish effects of THC from repeated-task fatigue, expectancy, timing, travel between home and laboratory, or altered effort. The administered amount and inhalation pattern were chosen by participants, and one hour of follow-up provides no information about recovery beyond that hour. Participants were selected frequent concentrate users whose qualifying acute plasma THC response met the investigators’ threshold; effects in occasional consumers, people using lower-strength products, and populations with medical conditions remain unknown.

Sample attrition and analytic rules matter too. Ten of 75 people did not enter the analyzed sample. The 20-ng/mL acute plasma selection threshold narrows the investigated exposure range by design. Plasma rather than whole-blood analysis and inclusion of an active metabolite must be respected rather than concealed. The authors’ supplementary motor-battery procedures are referenced by the paper, but details not reported in the complete original article should not be invented to fill any evidentiary gap.

It is less suited to asserting a drug-specific causal percentage, mapping a plasma THC measurement into legal whole blood, determining time since use, or reconstructing one person’s decision-making at the wheel.

How should an expert evaluate this study in litigation?

A defensible assessment first identifies the proposition to be proved. If the proposition is that select electronically recorded motor outcomes changed after self-administered concentrates in 65 selected frequent users, the original results provide evidence with recognizable limitations. If the proposition is that a driver with a particular whole-blood THC value was impaired during a specific incident, the study supplies neither the direct specimen nor the validated individual prediction required. These are materially different claims, and the evidence should not be expanded from the first to the second without additional support.

The expert should state the correct independent sample size, identify the physical test matrix as separated plasma, keep parent THC apart from 11-OH-THC, disclose how below-quantification data were treated, distinguish each motor protocol, and quote the nonsignificant tapping and head-back balance findings alongside the significant arm/leg/sway observations. Interpretation should account for repeated measures, the self-administered naturalistic exposure, multiple testing, practice or fatigue, and the absence of actual driving outcomes.

Related analytical context on this site includes why biological test matrices cannot be casually interchanged, specimen collection and preanalytical considerations, and the relationship between screening and confirmatory testing. These are general contextual resources, not substitutes for the study’s original research methods.

Seven frequently asked questions

1. Did a smartphone actually measure THC intoxication?

No. The smartphone-class device measured acceleration, taps and balance-related motion during structured tasks. THC and 11-OH-THC were measured separately in laboratory-analyzed plasma. The phone did not chemically analyze cannabinoids or establish a validated intoxication score.

2. Did every motor test get worse after concentrate use?

No. Arm acceleration declined immediately and at one hour; leg acceleration declined significantly by one hour; eyes-open and eyes-closed sway rose acutely. The overall tests for finger-tapping rate and head-back/eyes-closed sway were nonsignificant. Treating them as uniformly abnormal would omit important original findings.

3. Could a plasma THC level be used as the legal whole-blood THC concentration?

No. Plasma was physically separated from cells before analysis. The paper did not report a directly measured forensic whole-blood parent-THC concentration or validate an individual conversion. Even a properly assayed whole-blood number would still require separate evidence to interpret actual individual impairment.

4. Did the two plasma cannabinoid correlations predict impairment well?

No clinical or forensic prediction performance was validated. The reported eyes-open sway correlations were weak (r=.247 for THC change and r=.296 for 11-OH-THC change). They do not supply individual classification accuracy, a threshold, a last-use clock or a driving-performance estimate.

5. Were the observed effects different between men and women?

The study observed some baseline differences in absolute movement speed but found no significant time-by-sex interaction for the reported motor tasks. That means it did not demonstrate distinct post-use trajectories by sex; it does not prove biological equivalence or guarantee identical responses in each individual.

6. Does a nonsignificant finger-tapping result prove no cannabis effect?

No. Finger tapping had no statistically significant overall time effect in this experiment (p=.100), which is not a formal equivalence finding. It does, however, defeat an unsupported claim that the study demonstrated significant worsening of all smartphone motor outcomes.

7. Can these results show that cannabis caused a particular crash?

No. Researchers did not observe road traffic, vehicle collisions, hazard response or the actions of a defendant in any specific case. Establishing incident causation would require independent case evidence and methods appropriate to the actual behavior and timeline, rather than extrapolation from group mean accelerations.

Bottom line: Objective measurements require restrained interpretation

The valuable result is not that smartphones diagnose cannabis impairment. It is that accelerometer-based tests detected specific, differently timed motor and balance changes following self-administered cannabis concentrates in a selected experienced-user group, while other tasks were null and plasma cannabinoid changes offered weak associations with one balance endpoint. Forensic accuracy requires naming the sampled matrix, retaining the actual denominators, acknowledging both significant and nonsignificant findings, and refusing to turn an exploratory group result into an individualized legal conclusion.

Primary source: Hitchcock LN and colleagues. Acute Effects of Cannabis Concentrate on Motor Control and Speed: Smartphone-Based Mobile Assessment. Frontiers in Psychiatry, 2021, 11:623672. https://doi.org/10.3389/fpsyt.2020.623672.