Background: Why This Study Was Needed
For decades, law enforcement officers across the United States have used a set of physical tests — collectively known as Standardized Field Sobriety Tests, or SFSTs — to evaluate whether a driver may be impaired by alcohol. These tests were originally validated in 1981 at a blood alcohol concentration (BAC) of 0.10%, which was then the legal limit in most states.
During the 1990s, a significant shift in policy began to take shape. Many states started lowering their legal BAC limit to 0.08%, and federal pressure was building to make that the national standard. This created an important scientific question: were the same tests still accurate and reliable at the lower threshold? Could officers correctly identify drivers at or above 0.08% BAC using the same methods that had been validated at 0.10%?
To answer that question, the National Highway Traffic Safety Administration (NHTSA) commissioned a formal validation study. Conducted in San Diego, California in 1996 and published in August 1998, the research provides a thorough scientific assessment of whether SFSTs remain effective tools when the legal limit is 0.08% — and even examines performance at the even lower threshold of 0.04%, relevant to zero-tolerance laws for underage drivers.
How the Study Was Conducted
The study was designed to reflect real-world enforcement conditions as closely as possible. Seven officers from the San Diego Police Department’s alcohol enforcement unit participated. All were experienced DWI (Driving While Impaired) specialists who had received NHTSA-approved SFST training from certified instructors and had at least one year of field experience administering the tests. Before the study began, all seven received a four-hour refresher training session led by one of the study’s authors, Dr. Marcelline Burns.
The study took place between May 23 and November 9, 1996. During that period, the participating officers stopped and evaluated motorists using the standard SFST battery. In total, 298 cases were recorded; one was excluded because the driver refused BAC testing, leaving 297 cases in the final analysis.
Who Was Tested?
The sample reflected the realities of traffic enforcement. The overwhelming majority of subjects were male adults, consistent with patterns seen in DWI enforcement nationwide:
- By age: Adults (21 and over) made up 91.9% of cases (273 individuals); youth under 21 accounted for 8.1% (24 individuals).
- By gender: 87.9% were male (261 cases); 12.1% were female (36 cases).
- By outcome: 73.0% resulted in an arrest (217 cases); 21.9% received a warning (65 cases); 5.1% received citations (15 cases).
This breakdown matters for interpreting the results: a large majority of the subjects ultimately had BAC levels above the legal threshold, which means the sample was skewed toward higher-BAC cases. Officers were working with a population where impairment was common — the kind of real-world enforcement scenario the study was designed to replicate.
The Key Modification: Adjusting the Scoring
At the heart of this study was a practical adaptation to the test scoring procedures. The original SFST battery had been calibrated so that certain numbers of observed “clues” (specific observable signs of impairment) pointed to a BAC of 0.10%. To evaluate performance at 0.08%, the researchers made one targeted change: rather than requiring six clues on the Horizontal Gaze Nystagmus (HGN) test to indicate a BAC at or above 0.10%, four HGN clues now indicated a BAC at or above 0.08%. Two HGN clues were designated as the indicator for the even lower 0.04% BAC threshold.
The Three Tests Explained
The SFST battery consists of three separate tests, each measuring a different set of observable behaviors and physiological responses to alcohol:
- Horizontal Gaze Nystagmus (HGN): The officer slowly moves an object (such as a penlight) horizontally in front of the subject’s eyes and observes the movement of the pupils. Alcohol impairs the smooth tracking of eye movement, causing an involuntary jerking called nystagmus. Officers score the number of observed clues, up to six.
- Walk-and-Turn (WAT): The subject is asked to walk nine steps heel-to-toe along a straight line, turn, and walk back the same way. Officers count a set of specific clues indicating difficulty, such as losing balance, stepping off the line, or using arms for balance.
- One Leg Stand (OLS): The subject holds one foot approximately six inches off the ground and counts aloud while balancing. Officers note specific clues such as swaying, using arms, hopping, or putting the foot down.
After administering all three tests, each officer made a binary decision: did they believe the driver’s BAC was at or above 0.08%? They also provided an estimated BAC figure. All subjects were then given a properly administered breath or blood test to obtain the actual, measured BAC — the ground truth against which officer decisions were compared.
How Accurately Did Officers Identify Impaired Drivers?
Decision Accuracy at the 0.08% Threshold
When officers made their binary arrest-or-no-arrest decision based on the three-test battery, the results were consistently strong. Out of 297 cases, officers made the correct call in 91% of cases overall. When only accounting for cases where subjects’ measured BAC clearly fell on one side of the 0.08% line — excluding borderline cases — the accuracy rate reached as high as 94%.
Looking at the two types of decisions separately reveals where the battery is strongest:
| Decision Type | Correct Cases | Accuracy Rate |
|---|---|---|
| Arrest decisions (“Yes, at or above 0.08%”) | 210 correct out of 234 | 90% |
| Non-arrest decisions (“No, below 0.08%”) | 59 correct out of 63 | 94% |
| Overall combined accuracy | 269 correct out of 297 | 91% |
The slight edge in non-arrest accuracy (94% vs 90%) is worth noting: officers were somewhat more reliably correct when deciding to release a driver than when deciding to arrest. Both rates, however, represent a high degree of accuracy for a field-administered test conducted under real-world conditions.
How Accurate Were Officers’ BAC Estimates?
Beyond the yes/no decision, officers were also asked to estimate the subject’s actual BAC level. The study compared these estimates to the measured BAC values and found the difference to be remarkably small. The table below shows the mean estimated and measured BAC across demographic subgroups:
| Group | Mean Estimated BAC | Mean Measured BAC | Difference |
|---|---|---|---|
| Overall (all 297) | 0.117% | 0.122% | 0.005% |
| Adults (21+) | 0.120% | 0.125% | 0.005% |
| Youth (under 21) | 0.083% | 0.090% | 0.007% |
| Male | 0.117% | 0.121% | 0.004% |
| Female | 0.117% | 0.129% | 0.012% |
The researchers described the overall 0.005% gap between estimated and actual BAC as “operationally irrelevant” — a difference so small it falls within the margin of error of evidentiary breath-testing equipment. Across all demographic groups, officers consistently underestimated BAC slightly, but the gap never exceeded 0.012%. The slightly larger gap for female subjects (0.012%) was the most notable variation, though still quite small in practical terms.
How Each Test Performed on Its Own
The study also evaluated each test individually, not just as part of the combined battery. This is important for understanding what each test contributes and where the combined approach adds the most value. The table below shows the accuracy of each individual test in predicting whether the subject’s BAC was at or above 0.08%:
| Test | Overall Accuracy | “Yes” Decision Accuracy | “No” Decision Accuracy |
|---|---|---|---|
| Horizontal Gaze Nystagmus (HGN) | 88% | 87% | 93% |
| One Leg Stand (OLS) | 83% | 86% | 73% |
| Walk-and-Turn (WAT) | 79% | 82% | 69% |
All three tests performed meaningfully above chance levels, but the HGN test showed the strongest overall accuracy both individually and in combination with the others. The Walk-and-Turn and One Leg Stand were particularly less reliable when officers decided a driver was not impaired (lower “No” decision accuracy), though the combined battery significantly improved upon any individual test. This is precisely why the SFST battery is designed to use all three tests together: each one adds information that the others may miss.
How Strongly Did Tests Correlate With Actual BAC?
Beyond simple accuracy, the study measured how closely each test’s results tracked with the actual measured BAC — a statistical analysis known as correlation. A correlation of 1.0 would be perfect; a correlation of 0 would mean no relationship at all. The analysis covered 261 cases for which complete data on all three tests was available. All correlations were statistically significant at p = .005, meaning the results are extremely unlikely to be due to chance:
| Test or Battery | Correlation With Measured BAC |
|---|---|
| All Three Tests Combined | r = 0.69 |
| Horizontal Gaze Nystagmus (HGN) | r = 0.65 |
| Walk-and-Turn (WAT) | r = 0.61 |
| One Leg Stand (OLS) | r = 0.45 |
The combined three-test battery showed the highest correlation (r = 0.69) with actual BAC, confirming that using all three tests together produces a more reliable result than relying on any single test alone. The HGN test individually had the strongest correlation of the three (r = 0.65), while One Leg Stand had the weakest (r = 0.45). The authors note that for field conditions, a correlation in the range of 0.65 to 0.70 is considered very high — a standard this battery meets or exceeds.
When All Three Tests Agreed
Among the 188 cases where the subject’s measured BAC was at or above 0.08%, all three tests produced consistent results — pointing in the same direction — in 86% of cases (162 out of 188). In the 14% of cases where there was some disagreement between tests (26 cases), the tests’ involvement was far from equal:
- The HGN test was involved in only 4 of the 26 discrepancies.
- The Walk-and-Turn test was involved in 16 discrepancies.
- The One Leg Stand test was involved in 15 discrepancies.
This pattern further reinforces HGN’s role as the most consistent and reliable indicator in the battery. While all three tests contribute to the overall picture, the HGN result was the least likely to be “the odd one out” when tests disagreed.
Examining the Errors: False Positives and False Negatives
No test is perfect, and the study examined the cases where officers got it wrong. There were two types of errors:
False Positives: 24 Cases
These were cases where the officer decided the driver was at or above 0.08% BAC, but the measured BAC turned out to be lower. There were 24 such cases — about 8% of the total sample. But the details matter. The study examined each of these 24 cases individually and found that many of them were either borderline by measurement or explainable by the evidence the officer observed:
- In 9 of the 24 cases, the measured BAC was at or close to 0.07% — within the normal margin of error of breath-testing equipment. Six of these cases measured exactly 0.070%, and three measured between 0.074% and 0.078%.
- In 16 cases, the officer’s estimate was consistent with the HGN clues they actually observed, suggesting the estimate was reasonable given what they saw.
- In 7 cases, the officer’s estimate was inconsistent with their HGN observations.
- One case involved a juvenile whose measured BAC was 0.069% — a result that would still be relevant under zero-tolerance laws for underage drivers.
- One case involved a subject found to be a psychiatric patient, which explained unusual performance on the tests despite a relatively low BAC.
The researchers calculated that if the borderline cases (those within the margin of error of testing equipment) and the juvenile zero-tolerance case were counted as correct decisions, the overall accuracy rate rises to 94%. This underscores that many so-called “errors” occurred in genuinely ambiguous situations where the SFST results were internally consistent — the measured BAC simply fell just under the threshold.
False Negatives: 4 Cases
These were cases where the officer decided the driver was below 0.08% BAC, but the measured BAC was actually at or above that level. There were only four such cases in the entire study — fewer than 2% of all 297 cases:
- In two cases, officers observed only two HGN clues even though the subject’s actual BAC was 0.10% or 0.12% — suggesting the subjects may not have displayed the expected physiological response, or the clues were missed.
- In the other two cases, officers recorded four HGN clues — which under the modified scoring should have indicated a BAC at or above 0.08% — but nonetheless estimated a lower BAC, apparently not fully applying the updated scoring guidelines.
The very low number of false negatives (just 4 cases out of 297) is significant from a public safety perspective. These were instances where an impaired driver may not have been arrested despite being above the legal limit. The much larger false positive count (24 cases), while still a small proportion, is arguably less serious from a public safety standpoint — those individuals received additional scrutiny rather than being waved through.
Could Officers Detect Even Lower BAC Levels?
An important secondary question in this study was whether the modified test battery could help officers distinguish drivers at the 0.04% BAC threshold — relevant for zero-tolerance laws applicable to drivers under 21 years old or those in certain regulated occupations.
The results were encouraging but more mixed than the 0.08% findings. The analysis focused on 83 cases where the subject’s BAC fell in the range between 0.04% and 0.08%:
- When officers estimated that a motorist’s BAC was between 0.04% and 0.08%, their arrest decisions were 94% accurate (51 correct out of 54 decisions).
- However, accuracy dropped significantly for cases where officers estimated BAC below 0.04%. In those cases, officers were correct only 52% of the time (15 correct out of 29 decisions) — barely better than chance.
- Overall combined accuracy across all decisions in this lower BAC range was 80%.
This suggests that the SFST battery, even with modified scoring, becomes less reliable as a tool for making very fine BAC distinctions at very low concentrations. It performs well for the primary enforcement purpose of identifying drivers at or above 0.08%, and has reasonable utility at 0.04% for arrest decisions, but it is not designed to detect very low-level alcohol consumption with precision.
What Officers and Prosecutors Said
Beyond the statistical results, the study assessed whether the SFST battery with its modified scoring was practically acceptable to the people who use it — both law enforcement officers in the field and the prosecutors who must present evidence in court. The interviews revealed a picture of strong buy-in from both groups.
Officer Acceptance
All seven participating officers reported that the SFST battery was administrable under all reasonable environmental conditions encountered in the field and valid for distinguishing drivers at or above 0.08% BAC and potentially useful at even lower thresholds. The officers described the tests as:
- Essential for their professional duties in DWI enforcement.
- Administrable under all reasonable environmental conditions encountered in the field.
- Valid for distinguishing drivers at or above 0.08% BAC, and potentially useful at even lower thresholds.
Prosecutor Acceptance
Prosecutors from the San Diego City Attorney’s Office — the attorneys who handled the DWI cases arising from this study — offered a particularly telling endorsement. Their key points:
- None of the 298 DWI arrests during the study period was negatively affected by SFST evidence, nor by the exclusion of other, non-standardized sobriety tests.
- The SFST battery had been challenged in court by defense attorneys, but those challenges had been unsuccessful because the battery was developed and validated through systematic scientific research.
- By contrast, prosecutors reported that a non-validated test used by another local law enforcement agency had successfully been challenged in court — precisely because it lacked the scientific backing of the SFST battery.
- Prosecutors preferred a standardized battery consistently applied across all agencies, making the scientific basis of the evidence easier to defend.
The conclusion the prosecutors reached — that SFSTs offer a legally defensible, scientifically grounded foundation for DWI enforcement — gave practical weight to the study’s quantitative findings.
An Important Note on What These Tests Measure
The study makes a careful distinction worth highlighting for general readers. The SFSTs are designed to measure whether a driver’s BAC is at or above a legal threshold — they are not designed to directly measure driving ability or predict exactly how impaired a person will be on the road. The HGN test, for example, might seem to have no obvious connection to operating a vehicle, since it involves watching a person’s eye movements rather than observing driving behavior. The study’s authors acknowledge this “face validity” concern but note it does not undermine the test’s scientific standing: under DWI statutes, the relevant legal question is whether a driver’s BAC exceeds a legal limit, and the HGN test is a scientifically valid tool for estimating that — which is its appropriate purpose.
As the authors note, driving is a complex activity that cannot be measured at roadside. What can be measured at roadside is BAC, and the SFST battery is designed precisely to do that — not to substitute for a full driving evaluation.
Conclusions
The San Diego SFST validation study provided, in the authors’ words, “clear evidence of the validity” of the Standardized Field Sobriety Test battery for accurately discriminating between drivers who are above and below the 0.08% BAC threshold, when a slightly modified scoring procedure is applied. The modifications required were minimal — primarily adjusting the number of HGN clues that indicate the 0.08% threshold from four (at 0.10%) to four (now indicating 0.08%) — and the resulting accuracy was high, with officers making the correct decision in 91% of cases overall.
The study results also “strongly suggest” that the SFST battery can support accurate enforcement at the 0.04% BAC threshold as well, though performance at that lower level is somewhat less reliable, particularly for distinguishing BAC levels below 0.04%.
Both law enforcement officers and prosecutors found the battery “fully acceptable” as a practical tool for establishing probable cause in DWI cases. The research concluded that with appropriate training and the modified scoring procedures described in the study, SFSTs remain valid, defensible, and useful enforcement tools as legal BAC limits move lower.
Read the full original research report: Validation of the Standardized Field Sobriety Test Battery at BACs Below 0.10 Percent — Anacapa Sciences, Inc., August 1998, commissioned by NHTSA.
