EVIDENCE LIBRARY
Does 200 mg of Caffeine Make Drivers React Faster? A 100-Driver Trial Found Faster Simulated Braking but No Less Lane Weaving
A randomized 100-driver caffeine trial found faster simulated braking (p=.048), not steadier lane keeping (p=.75). Plasma HPLC results and forensic limits.
Short answer: In a randomized, double-blind experiment involving 100 licensed adults, a single 200 mg caffeine capsule was associated with a slightly shorter reaction time to simulated pedestrians and better accuracy on one computerized shifting-attention test. It was not associated with better lane-position stability during a separate simulated winding-road task. The investigators confirmed caffeine exposure by directly measuring plasma caffeine with high-performance liquid chromatography (HPLC). The experiment did not establish that caffeine prevents a crash, restores an unsafe driver to safe performance, or provides a legal threshold for individual driving ability.
The important scientific distinction is between a treatment group’s average response on a particular artificial task and a conclusion about an identifiable person’s real-world conduct. Here the estimated braking difference was approximately five hundredths of a second. The lane-keeping result, which tested a different part of the driving task, was statistically nonsignificant. Treating the favorable braking finding as a general endorsement of driving after caffeine would ignore the rest of the original experiment.
What was the actual primary study?
Konishi and colleagues published Effect of single caffeine intake on neuropsychological functions in healthy volunteers: A double-blind placebo-controlled study in PLOS ONE in 2018. The original complete primary paper is available at the original research DOI. Its design was a two-arm, parallel-group randomized comparison, not a crossover in which each individual experienced both active caffeine and placebo. The trial enrolled 101 volunteers; one withdrew consent, leaving 100 analyzed participants. All 100 held driver’s licenses. Fifty received caffeine and fifty received placebo.
The investigators intentionally measured several separate constructs: caffeine in a biological specimen, cognitive-test responses, a sudden-braking simulation, and winding-road vehicle-position variability. These endpoints should remain separate in any scientific or forensic discussion. The measured caffeine exposure answers a chemical question. Computerized neurocognitive tests answer specific experimental performance questions. Neither measurement, on its own, establishes an actual traffic event or a legal assessment of one driver.
Who participated, and how was caffeine administered?
The final cohort comprised 50 men and 50 women, ages 22 to 59 years. Assignment was balanced at 50 participants per treatment group: the active group received a capsule containing 200 mg of pure caffeine powder and the control group received an apparently matching capsule containing 200 mg of sugar powder. A pharmacist controlled assignment and sequentially numbered containers; the authors describe the study as double-blind. Participants were instructed to abstain from caffeine for at least three days before testing.
These were healthy Japanese volunteers rather than emergency-department patients, commercial drivers on duty, sleep-deprived transport operators, or people being evaluated after a collision. Their habitual driving frequency and driving history were described at baseline, but the trial did not follow real crashes or naturalistic road behavior. Each of the 100 individuals is one independent participant; the seven braking trials, multiple cognitive subtests and time points are repeated or nested measurements, not additional independent people.
The study did not administer alcohol, tetrahydrocannabinol, methamphetamine, sedatives, or an energy drink. Its result therefore cannot be turned into an experimentally validated rule about caffeine reversing the effects of those substances. Nor did it test repeated caffeine dosing over many hours of wakefulness. It tested one supervised 200 mg capsule in one short observation period.
What chemical specimen was collected?
The researchers drew blood before dosing and again approximately 120 minutes after the assigned capsule. They collected the samples into heparinized tubes, centrifuged them and separated the plasma supernatant. The reported analyte was caffeine in plasma, quantified by an HPLC method using ultraviolet detection and an internal standard. The result is not the concentration of caffeine in unseparated venous whole blood, and it is not a breath measurement, urine screening result, oral-fluid concentration or hair result.
The source reports plasma values in micrograms per milliliter (µg/mL). At baseline the groups had similar means: 0.20 ± 0.60 µg/mL in placebo and 0.23 ± 0.80 µg/mL in caffeine (reported p = .67). After administration, the means were 0.16 ± 0.46 µg/mL in placebo and 10.10 ± 3.29 µg/mL in the caffeine group (reported p < .001). These are group means with standard deviations, not a validated threshold for safe or unsafe driving.
The chemical measurement supports that the experimental treatment produced a measurable difference in plasma exposure. It does not supply a per-person formula converting a caffeine result into braking time. It also does not prove that the concentration obtained at the later blood draw was the concentration prevailing at every earlier cognitive or driving event. Proper analysis must respect the different measurement times.
What happened in the simulated sudden-braking task?
In the computer driving simulation, volunteers maintained a prescribed speed on a straight two-lane virtual road while human models entered the vehicle’s path. The researchers evaluated how quickly the driver responded to the unexpected appearance by braking. The paper reports seven brake-reaction trials per participant in that task. These were digital scenarios performed in a quiet testing room, not a record of actual pedestrians being struck or avoided on a public road.
The placebo arm’s mean brake reaction time was 0.89 ± 0.12 seconds; the caffeine arm’s mean was 0.84 ± 0.08 seconds. The reported between-group p value was .048. Lower time corresponds to a quicker measured simulator response, so the group comparison favored caffeine on this particular outcome. However, the p value lies close to the authors’ nominal .05 threshold, and the full study reported numerous cognitive and driving comparisons without an evident study-wide multiplicity correction.
Nothing in those two means establishes that every caffeine recipient was quicker than every placebo recipient. Individual distributions overlap, and mean differences do not decide whether a particular driver perceived a real hazard, chose the appropriate maneuver, had adequate stopping distance or avoided injury. The simulation also cannot isolate caffeine’s effects from all possible influences on real traffic scenes, such as lighting, vehicle braking capacity, roadway friction, divided attention and fatigue.
Did caffeine also improve steering or lane stability?
No statistically significant benefit was demonstrated. In a separate winding-road simulation, participants attempted to maintain a lane position while traveling at a prescribed simulated speed. The outcome was standard deviation of lateral position (SDLP), expressed in meters. It summarizes variability in lane placement; a smaller SDLP generally indicates steadier position on that specific task. It is not a chemical measurement and is not interchangeable with time to move one’s foot to the brake.
The mean SDLP was 0.87 ± 0.27 m for placebo and 0.89 ± 0.25 m for caffeine, with p = .75. The study therefore did not demonstrate improved lane stability after a single 200 mg caffeine capsule. A nonsignificant difference does not by itself prove equivalence or prove caffeine can never affect steering under other conditions. It means this trial did not support a beneficial between-arm difference on its particular short simulated course.
This contrast between braking and lane-position outcomes is the central lesson. A treatment can appear favorable on a speeded response while failing to show the same pattern on a continuous-control metric. Any summary saying simply that caffeine “improved driving” removes a crucial qualifier and overstates the actual scope of evidence.
Evidence table 1: Directly measured caffeine and two driving outcomes
| Original measured outcome | Placebo (n = 50) | Caffeine, 200 mg (n = 50) | Reported between-group inference |
|---|---|---|---|
| Baseline plasma caffeine, µg/mL | 0.20 ± 0.60 | 0.23 ± 0.80 | p = .67; not significant |
| Plasma caffeine approximately 120 min after capsule, µg/mL | 0.16 ± 0.46 | 10.10 ± 3.29 | p < .001; greater measured exposure |
| Simulated brake reaction time, seconds | 0.89 ± 0.12 | 0.84 ± 0.08 | p = .048; faster group mean with caffeine |
| Simulated winding-road SDLP, meters | 0.87 ± 0.27 | 0.89 ± 0.25 | p = .75; no significant group difference |
Entries reproduce the primary article’s reported means ± standard deviations and p values. Blood was processed into plasma. The last two rows are simulator behavior, not concentrations, real crashes, or individual legal impairment decisions.
Which computerized attention findings were positive?
Thirty minutes after capsule administration, participants began a series of computerized neurocognitive assessments. The full battery included Symbol Digit Coding, Stroop, Shifting Attention, and the Four Part Continuous Performance Test. These are not synonymous tests: each uses a different combination of instructions, speed demands, response selection, errors and sustained engagement.
On the Shifting Attention Test (SAT), the caffeine arm recorded 55.0 ± 5.7 correct responses versus 52.2 ± 6.4 in placebo (p = .03), and 2.5 ± 2.0 errors versus 3.7 ± 2.6 in placebo (p = .02). Both outcomes favored the caffeine group. Correct-reaction time on the same task was 959.5 ± 133.2 versus 1000.7 ± 127.6 in placebo, but the reported p value was .12, which was not significant. Thus even within the test showing an accuracy benefit, a distinct speed endpoint did not pass the stated significance criterion.
The results should not be converted into a broad proposition that caffeine improved every executive function. The source itself reported no significant between-group effects for the Symbol Digit Coding Test, Stroop Test or Four Part Continuous Performance Test. More correct responses in one shifting-attention assessment do not automatically imply better hazard perception, reduced lane departure risk, or more accurate memory for events.
Evidence table 2: Selected cognitive benefits and real null findings
| Source Table 3 computerized endpoint | Placebo (n = 50) | Caffeine (n = 50) | Reported p and interpretation |
|---|---|---|---|
| Shifting Attention correct responses, count | 52.2 ± 6.4 | 55.0 ± 5.7 | .03; more correct with caffeine |
| Shifting Attention errors, count | 3.7 ± 2.6 | 2.5 ± 2.0 | .02; fewer errors with caffeine |
| Shifting Attention correct response time, study-reported latency units | 1000.7 ± 127.6 | 959.5 ± 133.2 | .12; not significant |
| Symbol Digit Coding correct responses, count | 66.1 ± 9.7 | 67.2 ± 11.2 | .74; not significant |
| Stroop simple reaction time, study-reported latency units | 320.3 ± 45.3 | 308.0 ± 40.2 | .09; not significant |
| Continuous Performance Test Part 4 omission errors, count | 2.9 ± 2.2 | 2.6 ± 2.3 | .36; not significant |
These are selected published outcomes rather than all endpoints in the battery. No adjustment for the entire set of reported comparisons is established by this table. “Not significant” describes the authors’ test results, not proven equivalence or a measured null effect in any real collision.
Why do the sampling times matter?
The capsule was administered before both cognitive and driving tasks. Computerized cognitive tests began at approximately 30 minutes after ingestion; the next part of the test sequence involved simulated driving. The investigators reported plasma collection at baseline and 120 minutes. The tasks and the plasma sample were not all obtained at the same instant, and the publication did not provide a validated time-resolved concentration-response model for every volunteer.
Drug concentrations are snapshots of a compartment at a time, subject to absorption, distribution and elimination. A plasma measurement at 120 minutes cannot simply be copied backward into a claimed precise concentration at the first shifting-attention response. Nor is the prescribed 200 mg capsule dose an observed mass delivered to any specific central nervous system receptor. The measured postdose plasma distribution establishes exposure, but it does not calibrate a driving impairment classifier.
In a real forensic case, the analyst would also need to know whether the laboratory tested whole blood, serum or plasma, whether the person consumed other medications, the sample collection time, analytical method validation, and the relevant actual conduct. This experiment offers no universal conversion from caffeine plasma µg/mL to a driver’s stopping decision, and no validated legal cutoff derived from its results.
Was the statistical evidence strong enough for a safety claim?
The authors used a nominal two-sided p < .05 rule for their comparisons. The braking result, p = .048, just met that rule. The SAT accuracy findings, p = .03 and p = .02, also met it. Many other comparisons—among reaction times, coding, Stroop, sustained attention and lane tracking—did not. When numerous outcomes are tested, some nominal p values below .05 can appear by chance. The original report does not provide a global multiple-testing adjustment or independently confirmed effect for the braking difference.
A statistically significant mean difference is not automatically a practically decisive difference. The response gap in the simulator was 0.05 seconds on average, with variability within both groups. Its usefulness depends on actual travel conditions and whether the simulated hazard reasonably represents the real-world question. The study did not randomize people into actual crashes and cannot measure real reduction in collisions. The p value is a statement about the study’s statistical model and null hypothesis, not the probability that a named driver was safe.
Conversely, p = .75 for SDLP is not evidence that caffeine always has no effect on steering. Proper wording is more modest: this particular controlled trial failed to demonstrate a difference in short-duration lane variability. A rigorous article should retain both the significant and nonsignificant evidence instead of citing only the favorable rows.
Could ordinary coffee or energy drinks yield the same result?
That comparison was not directly tested. The intervention was a single capsule of pure caffeine, not a measured cup of brewed coffee, decaffeinated coffee, tea or a commercially prepared energy drink. Food, other constituents, habitual consumption, fatigue and dose all differ in those settings. The study’s 200 mg experimental dose cannot determine the exact contents of someone’s beverage or establish whether that beverage would reproduce these group means.
The authors requested a three-day caffeine washout, but caffeine withdrawal was not controlled with dedicated symptom or performance measures. The article includes baseline daily intake descriptions yet also lists incomplete assessment of habitual caffeine use among its limitations. Both facts should be recorded when assessing generalizability. An individual who is accustomed to caffeine, experiencing withdrawal, or severely sleep-deprived may not respond exactly like the average participant in this sample.
A separate trial involving repeated caffeine during prolonged sleep deprivation would pose a different scientific question. This experiment did not investigate recovery from 77 hours awake, and its results should not be repurposed to imply rescue from sustained sleep-related lapses.
What would be required to evaluate an actual driving event?
At minimum, a case-specific analysis would need the event chronology, contemporaneous behavioral observations, the relevant road geometry and speeds, the vehicle and braking information, any toxicology report with matrix and collection time, and the question the evidence is offered to answer. A video showing when a hazard became perceptible addresses different facts from an HPLC plasma caffeine result. A simulator’s mean BRT does not establish a real driver’s perception point or stopping distance.
Laboratory interpretation should begin by reviewing which biological matrix was actually analyzed and whether the analytical method answers the intended question. The study measured caffeine in centrifuged plasma. It did not collect a legal venous whole-blood concentration suitable for direct interchange with plasma results. A precise number in one matrix cannot be transplanted into another without validated support.
The evidentiary difference between a chemical assay and a performance test is equally important. Analytical confirmation and its limitations address compound identity and concentration; an independently validated behavioral method would be needed to connect measured exposure to performance. Even a sound population experiment does not decide the cause of a specific motor vehicle collision.
How should an expert describe this study in a report?
A defensible characterization is that, in 100 healthy licensed volunteers randomized 1:1, a supervised single 200 mg caffeine capsule generated higher mean plasma caffeine and was associated with faster simulated emergency braking and more accurate shifting-attention responses than placebo, while short-course simulated lane variability and most computerized outcomes did not significantly differ. That description preserves the dose, measured matrix, sample size, controlled endpoints and limits of generalization.
An expert should not claim that plasma caffeine of approximately 10 µg/mL guarantees safe driving, that the group mean predicts a specific person’s response, or that the result validates a claimed actual collision avoidance. The postdose plasma value is descriptive of this cohort; it is not a confirmed legal impairment boundary or proof of the chemical concentration at every task time. If a clinical timing question arises, sample timing and therapeutic monitoring principles provide useful context but do not transform these observations into a new patient-specific model.
Finally, the study should not be presented as evidence that caffeine uniformly counters intoxication from other drugs. There was no alcohol, THC, or stimulant coadministration arm. The absence of such experiments is a limitation of scope, not a statistically proven null interaction.
Frequently asked questions about caffeine, braking, and lane control
1. Did 200 mg of caffeine significantly shorten simulated brake reaction time?
Yes, the caffeine group averaged 0.84 seconds compared with 0.89 seconds for placebo, with a nominal between-group p = .048. This was a computer-simulated pedestrian hazard in a controlled room, not an observed reduction in real accidents. The experiment did not prove that every caffeine-treated individual reacts faster.
2. Did the same study show less lane weaving?
No significant difference was observed for standard deviation of lateral position: placebo 0.87 m, caffeine 0.89 m, p = .75. This was a separate short winding-road simulation, so its negative result must not be replaced with the favorable braking result.
3. Was caffeine measured in whole blood or plasma?
The authors centrifuged collected blood and quantified caffeine in separated plasma using HPLC with ultraviolet detection. The reported postdose mean was 10.10 µg/mL in the caffeine group. It should not be mislabeled as an unseparated venous whole-blood concentration or breath-alcohol-style measurement.
4. Do seven simulator brake trials mean 700 independent drivers?
No. There were 100 independent people, 50 per randomized group. Repeated simulated braking exposures within a person are nested observations. Treating seven trials per person as seven distinct participants would misstate the biological sample and the independent randomization units.
5. Were all cognitive and reaction-time tests improved?
No. The Shifting Attention Test showed more correct responses and fewer errors with caffeine, but its correct-response latency was nonsignificant. The Symbol Digit Coding, Stroop and Four Part Continuous Performance tests did not demonstrate overall between-group differences reported as significant.
6. Does a favorable p value establish individual safe driving?
No. A p value from a group comparison is not a safety certificate or an impairment cutoff. The study did not collect on-road collisions, real hazard responses or enough person-specific validation to classify an individual driver. The braking p = .048 also warrants caution alongside numerous tested outcomes.
7. Can the finding be applied to caffeine plus alcohol or THC?
Not on the basis of this paper. The randomized intervention compared caffeine alone with placebo. No alcohol, cannabis or combined-intoxication arm was tested. Separate original experimental evidence and the specific event record would be needed before drawing any conclusion about mixed exposures.
Original scientific source and bottom line
Primary study: Konishi Y, Hori H, Ide K, Katsuki A, Atake K, Igata R, and colleagues (2018), Effect of single caffeine intake on neuropsychological functions in healthy volunteers: A double-blind placebo-controlled study, PLOS ONE 13(10):e0202247. Read the original peer-reviewed research paper for the full randomization description, sample preparation, instrument methods, numerical results, and authors’ limitations.
Bottom line: This 100-person study supports a narrow group-level finding of faster simulator braking and better shifting-attention accuracy after one 200 mg caffeine capsule, accompanied by directly measured plasma caffeine exposure. It did not demonstrate less lane weaving, a broad improvement across cognitive tests, a validated whole-blood caffeine driving cutoff, or a reduction in actual collisions. A precise forensic account should say exactly which effect was measured and exactly which conclusion remains untested.