Oculomotor and Pupil Tests to Identify Alcohol Impairment

M. Burns

Southern California Research Institute, 11914 W. Washington Blvd., Los Angeles, CA 90066, USA

ABSTRACT

The ocular motor module (OMM), a video-based image processing computer system, generates visual stimuli for testing ocular motor and pupil function. Subjects (Ss) look into a foam-lined viewport, which excludes all ambient light and allows testing in total darkness. Non-visible infrared light-emitting diodes, to which the video cameras are sensitive, provide images for the computer system. The OMM samples at a rate of 60 samples per sec. Stored data is available for analysis using feature extraction algorithms.

An alcohol-impairment test protocol is based on tests used in Standardized Field Sobriety Tests and Drug Recognition Evaluations. Measures include nystagmus, smooth pursuit, pupil size in darkness, and pupil response to dim light (approx. 8 ft candles) and bright light (approx. 20 ft candles).

Alcohol effects on eye performance were examined with 32 Ss dosed to 0.08% BAC. They were tested predose, at peak BAC, and on the descending BAC curve. Pre- post-dose comparisons are displayed in tables and figures and demonstrate the effects of alcohol on eye signs.

INTRODUCTION

Police officers in most U.S. jurisdictions use standardized sobriety tests to examine drivers suspected of being under the influence of alcohol and other drugs (Burns, 1990, 1988, 1987). Eye signs, including nystagmus, a lack of smooth pursuit movements, and constricted or dilated pupils, are an important component of the testing procedures. Eye measures are sensitive and reliable indices of impairing substances, and officers properly trained in observation methods tend to rely heavily on them as evidence of impairment. The measurement of the eye signs used in sobriety tests now has been instrumented in a computer-based system (EPS-100 Eye Performance System Eye Dynamics, Torrance, California).

An image processing system generates visual stimuli for testing oculomotor and pupil function. Testing requires a subject (S) to look into a viewport, which excludes all ambient light and insures total darkness during 50-sec test trials. Non-visible infrared light-emitting diodes, to which the system's video cameras are sensitive, provide images of the Ss' eyes. The system samples the eye's movements and characteristics 60 times per second, and the data are analyzed with feature extraction algorithms.

The system's alcohol-impairment protocol was examined with 32 healthy volunteers (20 males, 12 females), ages 21 - 50 years, who gave informed consent to participate in a double blind alcohol experiment. Ss' arrived for a test session between 0800 and 1000. Their vital signs were checked and pre-dose breath tests were obtained.

Alcohol doses were calculated to produce expected peak blood alcohol concentrations (BACs) in the range 0.08% to 0.15%. The beverage was orange juice and vodka, and dose amounts ranged from 0.7 to 1.75 oz 80-proof vodka per 25 pounds body weight. Ss consumed the alcohol over periods of 0.5 to 1.5 hour.

Ss are instructed to look into the viewport and follow the light with their eyes. Thus, no training is required, but Ss performed three pre-dose trials to provide baseline data. Post-dose testing began one-half hour after drink completion and was repeated three additional times at hour intervals on the descending BAC curve.

RESULTS

Pre-dose BACs were 0.00%. BACs at each test time appear below (Table 1).

Table 1
BACs at Each Test Time

The eye measurement system yields a Pass/Fail decision at the end of each test. If a BAC of 0.08% or above is taken as the criterion for impairment, the overall results are shown in Table 2.

Table 2
Eye Sign Classifications

Overall, the classifications were 81% correct, but as can be seen in Figure 1, the type and number of errors differed significantly (Chi sq. 43.48, 9 df, p <.01) as a function of the changing BAC distribution over test times. At the first test time when the mean BAC was 0.09%, 11 Ss were incorrrectly classified. It is noteworthy, however, that eight of the errors were for BACs within 0.01% of the criterion. That is, they were within the range, 0.07 to 0.09%.

Figure 1
Identification of Alcohol Impairment with Oculomotor and Pupil Tests

At the second test time when the mean BAC was 0.07%, nine Ss were incorrectly classified. In contrast, at the last two test times when the mean BAC had fallen below 0.05%, most of the subjects were correctly classified. There were only two false positives at each time and a single false negative at the third test time for a S with 0.08% BAC.

Figure 2, which displays the percent correct decisions with BACs arbitrarily grouped as low (< 0.05%), high (>= 0.10%), and mid-range (>= 0.05% <0.10%), again demonstrates that classification accuracy differed by BAC. The high percentage of correct classifications at low BACs, i.e., Correct Rejections, reflects the absence of eye signs. Above 0.10% BAC the signs were sufficiently distinct and reliable to yield 79% correct decisions, and as would be expected, mid-range BACs were the most difficult to classify.

Figure 2
Identification of Alcohol Impairment with Oculomotor and Pupil Tests

CONCLUSION

In summary, the accuracy of a computer-based system, which measures oculomotor and pupil function, was examined with alcohol-dosed Ss. With a measured BAC of 0.08% taken as the criterion for impairment, the system's pass/fail decisions correctly classified Ss on 81% of the trials.

REFERENCES

Burns, M. (1990) Recognition of the drug-impaired driver by examination of behavioral and physiological signs. Proceedings, 34th Annual Meeting Human Factors Society, Orlando.

Burns, M. (1988) The use of horizonal gaze nystagmus as a field sobriety test. Proceedings, 35th International Congress on Alcoholism and Drug Dependence, Oslo.

Burns, M. (1987) Sobriety tests for the presence of drugs. Alcohol, Drugs and Driving, Vol. 3, No. 1, 25-29.