Proponents of the so-called “standardized” field sobriety tests (SFSTs) have long pointed to field studies which indicate a high correlation between performance on the tests and actual blood alcohol concentrations (BAC). A new study now calls those conclusions into question.
Originally, the National Highway Traffic Safety Administration (NHTSA) paid a private group, the Southern California Research Institute, to conduct studies to find which among the various field sobriety tests used by police were most effective and to develop a standardized 3-test battery. SCRI subsequently reported to NHTSA that a battery of walk-and-turn, one-leg-stand and nystagmus provided a strong correlation with breath test results.
Confronted with questions about those conclusions, NHTSA later commissioned the same researcher who had conducted the original studies, Marcelline Burns, to corroborate the accuracy of her own findings. Burns accompanied a small number of San Diego police officers conducting actual DUI investigations in the field. After administering the SFSTs, the officers were asked to guess whether suspects had blood alcohol concentrations (BAC) over or under .08%. Burns reported a 91% correlation between SFSTs and BAC over-under estimates, arguably validating the battery of tests she had helped create.
A subsequent scientific article called Burns’ conclusions into question. In Hlastala, Polissar and Oberman, “Statistical Evaluation of Standardized Field Sobriety Tests”, 50(3) Journal of Forensic Sciences 1, the raw data used in the validation study were obtained from NHTSA through the Freedom of Information Act. The methodology used was then reviewed and the data subjected to statistical analysis.
The methodology was found to be seriously flawed in a number of respects. For one thing, many of the suspects had very high BACs, making estimates of whether a suspect was over .08% obvious regardless of SFST performance. For another, there was no attempt to isolate the influence of SFST performance from other factors: officers estimated BACs after the field sobriety tests, but they also took into account earlier observations, such as erratic driving, slurred speech, odor of alcohol, flushed face, admissions as to amount of alcohol consumed, etc.
The most glaring defect in Burns’ corroborative study was that “all police officers participating in the study were equipped with NHTSA-approved portable breath testing devices”. In other words, the San Diego officers already had the results of portable breath tests when they were asked to estimate the BACs later obtained at the station!
After reviewing the flawed methodology, the raw data was then statistically analyzed. The conclusions:
If we consider three ranges of MBAC [measured blood alcohol content], 0.00% to 0.04%, 0.04% to 0.08%, and 0.08% to 0.12%, the officers’ EBAC [estimated blood alcohol content] overestimated the MBAC 76%, 67% and 48% of the time, and underestimated it 14%, 26% and 28% of the time.
In other words, officers relying upon field sobriety tests were far more likely to overestimate BACs than underestimate — particularly with those suspects having low BACs.
(T)he utility of the SFST depends very much on how intoxicated an individual is. Accuracy (and specificity) are low when individuals are close to 0.08% MBAC, but if the individuals are quite intoxicated, such as above 0.12%, then accuracy is high.
In other words, in borderline cases involving persons at or under the legal limit, officers were very poor at estimating levels over .08% based upon SFSTs. And it is these cases, of course, that are critical. Suspects with high BACs are relatively easy to single out without the help of field tests; it is for the closer cases, particularly those who are innocent (below .08%), that the SFSTs are designed. And it is with these very cases that the tests apparently fail.
Put another way, accuracy in using field sobriety tests is high when they are not needed — and low when they are.