Schaffer Online Library of Drug Policy Sign the Resolution
Contents | Feedback | Search
DRCNet Home
| Join DRCNet
DRCNet Library | Schaffer Library | Drugs and Driving 


Calibration of a Portable, Automated, Posture Assessment System Using Graded Blood Alcohol Levels: Comparison With the Standardized Field Sobriety and Cognitive Tests

Robert S. Kennedy, Janet J. Turnage, William P. Dunlap and Julie M. Drexler

Essex Corporation, 1040 Woodcock Road, Suite 227, Orlando, FL 32803, USA


This report describes the fourth in a series of studies assessing the impact of graded dosages of alcohol on cognitive performance readiness. This effort differs from our previous studies in several ways. First, the study used a new implementation of six cognitive performance tests that had previously been shown to be sensitive to blood alcohol concentration (BAC). Second, performance was assessed following each of two successive graded doses to a target level of 0.08% BAC. Third, we compared performance on the cognitive tests to the standardized Field Sobriety Test (FST) developed by the National Highway Traffic Safety Administration and to a portable automated posture assessment system (PAPAS) that is video based and that we are currently developing. Results showed that, by using individual baselines, combined performance measures of individual subjects permits detection of alcohol-induced performance changes at rather low BACs. All six tests of the DELTA battery and all three tests of the standard field sobriety test (FST) battery singly predicted alcohol intoxication. When combined with the standard field sobriety test (FST), the DELTA added unique variance to the prediction of alcohol level. The PAPAS test appears to correlate well with the FST and offers the obvious advantage that in addition to being portable, objective, and automatically scored, it also provides a permanent record of the event.


The complexity of mental and motor skills necessary for operating vehicles safely make them susceptible to impairment by alcohol (Hunt & Witt, 1994). The connection between Blood Alcohol Concentration (BAC) levels and vehicular accidents has been well documented in epidemiological reports which indicate that the risk of being involved in a serious automobile accident increases dramatically as blood alcohol concentration increases (AIDE, 1992, Society of Automotive Engineers, 1986). It is reported that 40 percent of all traffic fatalities (the leading cause of accidental death) are alcohol-related (Zobeck, Stinson, Grant, & Bertolucci, 1993).

Most studies of the acute effects of alcohol report performance decrements (e.g., Wallgren & Barry, 1970a, 1970b), although results are not always consistent as a function of the ability requirements of the task (Levine, Kramer, & Levine, 1975). According to some theories of alcohol and performance, alcohol reduces processing capacity or the efficiency of divided attention, such that alcohol impairment increases for tasks requiring more capacity (Levine et al., 1975). Driving skill performances have been shown to decline at low and very low BACs (Hamilton & Copeman, 1970; Moskowitz, 1973; Moskowitz, Burns, & Williams, 1985), and acute effects of alcohol on piloting have been well documented. Impairment begins at BACs as low as 40mg/dl (0.04% BAC) and increases with dose (Billings, Wick, Gerke, & Chase, 1973; Henry, Flueck, Sanford, Keiser, McNee, Walter, Webster, Hartman, & Lancaster, 1974; Ross & Mundt, 1988).

Tharp, Burns and Moskowitz (1981) developed and field tested a series of psychophysical tests to be used to determine Driving-While-Intoxicated (DWI) arrests. These field sobriety tests (horizontal Gaze Nystagmus [GN], Walk and Turn [W&T], and One Leg Stand [OLS]) were found to be excellent predictors of intoxication in the laboratory and in the field. This battery has now been adopted by the National Highway Transportation Safety Administration (NHTSA, 1983).

Stance and gait have face validity as well as construct validity in the form of neurophysiological pathways, we hypothesized an automated posture test would be useful in driving safety research. We have theorized that head kinematics (position, velocity, and acceleration), measured under controlled conditions before and after exposure to various stimuli (Kennedy & Lilienthal, 1995) may provide a simple, economical and holistic measure of overall CNS integrity, and not just alcohol intoxication. Such a device has been prototyped; it uses video recording and framegrabber technology to measure postural stability (Kennedy, 1993), and early evaluation of the device indicates that it is more sensitive and statistically more powerful than equipment-less floor-based postural tests.

We have developed a cognitive test battery (DELTA) which is described in detail elsewhere (Turnage, Kennedy, Smith, Baltzley, & Lane, 1992; Turnage & Kennedy, 1992) and have calibrated and validated the tests in the battery against graded dosages of alcohol (Kennedy, Turnage, Rugotske, & Dunlap, 1994) where much of the theory behind our performance testing has been put forward (cf. also Kennedy, Turnage & Dunlap, 1993a, 1993b). In this work, we have found that roadside sobriety tests generally measure the functional integrity of the vestibular and oculomotor systems, but tests of cognitive performance also degrade in proportion to the alcohol dosage and add unique variance to neurovestibular motor tests of alcohol intoxication.

The present study continues the logical development of the previous findings by using the same battery of cognitive tests and the FST and addressed the following questions: 1) If dosages were lower than in previous studies, but were repeated as might occur with an evening of drinking, would different outcomes occur?; 2) Women drinkers are becoming an increased concern (National Institute on Alcohol Abuse and Alcoholism, 1994). If women were used as subjects, would results be different?; 3) Could an accurate, objective, video-based, portable measure of posture be developed?


The postural test we used was eyes closed, arms folded, heel-to-toe standing which was adapted from the original Fregly-Graybiel battery (Thomley, Kennedy, & Bittner, 1986) and was based on a screening experiment in comparison with 15 other variants. Heel-to-toe turned out to have the best combination of sensitivity, reliability and safety. Subjects wore tennis shoes and were carefully instructed and given a short opportunity to practice heel-to-toe standing (Figure 1).

Figure 1
Positioning of Camera and Subject

For each trial, subjects were asked to effect a stance based on the protocol and were asked to maintain the position for 30 seconds. Position was considered to be "maintained", if the subject did not lift or move either foot or open their eyes. If the position were not maintained the time was recorded from a stopwatch. At the end of each stance, the subjects were scored. Scoring included: 1) one subjective rating of global head motion assessed from the video by two research assistants who after initial training did not communicate with each other regarding the scoring. Over > 500 subject exposures their interrater reliabilities ranged from r = .40 - .90 depending on the range of performances exhibited by the subjects; 2) duration of time standing; 3) a combination 1 and 2; 4) the average velocity in the y plane (Figure 2).

Figure 2
Movement Directions

The approach we decided upon for measuring head movement to characterize postural stability involves video-taping the back of a subject's head using an ordinary video camera/recorder mounted on a tripod or bracket (Figure 1). The subject wears a high contrast target reticle of known dimensions the image of which is critical in postural measurement. The videotape imagery is then captured using a standard framegrabber board mounted in a 486/66megaHz computer. The image of the target reticle is then processed using software developed by Essex to derive head position information.

At the present time the software permits frame by frame change in reticle displacement in vertical and horizontal planes corresponding to z and y according to conventional nomenclature (Hixon, Niven, & Correia, 1966); movement in depth (x) is captured by size change of the reticle (Figure 2). Analysis of the video data is described in detail elsewhere (Kennedy, 1995).


Subjects were requested to not ingest alcohol, other drugs or solid food between the training session and the data collection session. The data collection sessions were held on Saturday between approximately 8:00 AM and 8:00 PM. The sessions were conducted on three weekends with one third of the subjects participating during each weekend. The subjects were dosed with alcohol following the Widmark (1932) equation which expresses the relationship between the distribution of alcohol in the body as a whole compared to the blood (Shipley, 1970).

In the first experimental session, the subjects were given grain alcohol and citrus punch. The amount of alcohol was calculated as follows: ml of grain alcohol = (.13) (.122) (30) (200/190) (weight in pounds). For the second and third sessions, the subjects were given rum and caffeine-free soda. The amount of alcohol was calculated as follows: ml of rum = (.13) (.103) (30) (200/80) (weight in pounds). Female subjects were given 80% of the calculated value, according to the Widmark formula, and individuals who appeared to be obese had their dose reduced by 20%. Testing took place after subjects reached a BAC of .08% and at each .02% BAC interval thereafter until .02% BAC was reached at which time subjects were again dosed with alcohol and the procedure was repeated. Blood Alcohol Concentration was monitored using Intoximeter Model 3000 (1980) and Intoximeter Alcosensor IV, (1991) breath testing units.


We found that all DELTA and FST tests correlated significantly and highly with BAC levels. However, mean percent decrement scores for the DELTA tests did not exceed 5% for most tests until subjects had reached the range of 0.05% to 0.06% BAC. At 0.04% BAC and lower, performance was more variable. Likewise, the mean scores for the FST tests did not exceed the recommended (NHTSA, 1983) two point cut off (for W&T) and the four point cutoff (for GN) until approximately the 0.08% BAC. The mean score for OLS never did reach the standard cutoff value of two points. In general, however, all the FST scores showed a monotonic linear increase as BAC level increased. These results indicate that all DELTA and FST tests accurately predict alcohol intoxication, especially at levels above 0.05% BAC, if one takes a 5% decrement as the cutoff for cognitive impairment. The 5% criterion may be too liberal for certain tasks, however. If the criterion for definition of impairment is lowered to 4% or 3% or even 2%, one would be able to specify impairment at considerably lower BAC. Also, prediction of impairment is more precise at the individual than at the group level.

There were no gender differences in performance scores as a function of alcohol dosing. Although females had an overall greater number of correct responses than did males in the acquisition phase of the study, these differences were controlled for in the alcohol phase analysis because each persons score was based on their individual percent decrement from baseline. Although preliminary analyses showed two significant interactions with gender (for W&T and OLS tests), correction for restriction of range due to lower BAC levels for females indicated that the significant results was merely an artifact of the lower BAC levels. Of concern was the fact that female and male BAC levels were significantly different. This occurrence was a direct result of applying the Widmark formula that specifies an 80% dosage level for females on the basis of different amounts of body fat. The females in our population tended to be heavy (average weight = 151 lbs). In the future, it will be necessary to either exclude overweight individuals from such studies or to modify the Widmark equation to more accurately prepare alcohol doses. In general, we suggest that the Widmark equation be reviewed in regard to male-female differences.

Finally, Table 1 shows the correlations between the three field sobriety measures and our four posture assessment measures with the blood alcohol concentrations for 62 separate observations distributed over 11 subjects. As expected, the best measure is gaze stability which correlated with BAC r = .64 (p < .001); and the next best is y Velocity from the PAPAS system which correlated r = .35 (p < .01). Most of the other correlations with BAC were significant at the p < .05, except for the FST standing test. It is noteworthy that the gaze nystagmus test correlates minimally (r = .23; p = .06) with y Velocity and not at all with the other PAPAS measures. Within the FST battery walking and standing scores are correlated significantly (r = .60; p < .001) as are intercorrelations of PAPAS tests even though they represent scores which are independently analyzed (e.g., time, rating, y velocity). This latter implies that similar aspects of the video records of posture are being measured.

Table 1
Correlations Between Field Sobriety Test and Video Based Postural Assessment Tests with Blood Alcohol Concentrations

    Walk Stand Time Rate Aggr yVel BAC
1. Gaze Nystagmus .30 .06 -.00 .02 .03 .23 .64
2. Walk   .60 -.34 .34 .37 .38 .37
3. Stand     -.42 .33 .33 .29 .16
4. Fall Time       -.90 -.97 -.61 -.29
5. Rating         .96 .55 .25
6. Aggregate Rating           .62 .25
7. y Velocity             .35
for N = 62; r(.05) = .24; r(.01) = .30; r (.001) = .40


From the data presented here, the best measure for assessment of alcohol dosage appears to be gaze nystagmus, however additional variance (p < .03) is available if y Velocity is added to a multiple prediction of BAC. This means that aspects of y Velocity add additional information not available from gaze nystagmus alone, and also provides support for the original hypothesis: that head position measurement can be used to index alcohol dosage. While the predictive validity of y Velocity is weaker than gaze nystagmus, there are several reasons why this is an encouraging finding for a first attempt: 1) gaze nystagmus correlations may have profited by implicit knowledge by the troopers who were also members of the research team; 2) gaze nystagmus is only useful for alcohol and some drugs and posture has an opportunity for more general indices of driver performance; 3) our algorithm for y Velocity is still being developed and will be augmented by x and z; and 4) improved algorithms should be forthcoming in future studies.


AIDE (1992). Drunk driving takes its toll. San Antonio, TX: USAA Casualty Insurance Co., pp. 6-11.

Billings, C. E., Wick, R. L., Gerke, R. J., & Chase, R. C. (1973). Effects of ethel alcohol on pilot performance. Aerospace Medicine, 44(4), 379-382.

Hamilton, P., & Copeman, A. (1970). The effect of alcohol and noise on components of a tracking and monitoring task. British Journal of Psychology, 61: 149-156.

Henry, P., Flueck, J., Sanford, J., Keiser, H., McNee, R., Walter, W., Webster, K., Hartman, B., & Lancaster, M. (1974). Assessment of performance in a Link GAT-1 flight simulator at three alcohol dose levels. Aviation, Space, and Environmental Medicine, 45, 33-44.

Hixon, W. C., Niven, J. I., & Correia, M. J. (1966). Kinematices nomenclature for physiological accelerations. Monograph 14, Naval Aerospace Medical Institute, Naval Aerospace Medical Center, Pensacola, FL. 8 August, 1966.

Hunt, W. A., & Witt, E. D. (1994). Behavioral effects of alcohol ingestion: Implications for drug testing. Toxic Substances Journal, 13, 41-49.

Intoximeter, Inc. (1980). Intoximeter 3000 Supervisor's Manual. St. Louis, MO, Intoximeter, Inc.

Intoximeter, Inc. (1991). ALCO-SENSOR IV Operator's Manual. St. Louis, MO, Intoximeter, Inc.

Kennedy, R. S., & Lilienthal, M. G. (1995). Implications of balance disturbances following exposure to virtual reality systems. Proceedings of the Virtual Reality Annual International Symposium '95, 35-39.

Kennedy, R. S. (1993). Device for measuring head position as a measure of postural stability. Final Report No. 9260166, National Science Foundation, Washington, DC.

Kennedy, R. S., Turnage, J. J., & Dunlap, W. P. (1993a). The use of dose equivalency as a risk assessment index in behavioral neurotoxicology. Neurotoxicology and Teratology, 14, 167-175.

Kennedy, R. S., Turnage, J. J., & Dunlap, W. P. (1993b). Dose equivalency: A metric to index driver-related performance deficits across different drugs based on alcohol equivalency levels. Proceedings of the 12th International Conference on Alcohol, Drugs and Traffic Safety (691-700). Cologne, Germany.

Kennedy, R. S., Turnage, J. J., Rugotzke, G. G., & Dunlap, W. P. (1994). Indexing cognitive tests to alcohol dosage and comparison to standardized field sobriety tests. Journal of Studies on Alcohol, 55(5), 615-628.

Levine, J. M., Kramer, G. G., & Levine, E. N. (1975). Effects of alcohol on human performance: An integration of research findings based on an abilities classification. Journal of Applied Psychology, 60(3), 285-293.

Moskowitz, H. (1973). Laboratory studies of the effects of alcohol on some variables related to driving. Journal of Safety Research, 5: 185-199.

Moskowitz, H., Burns, M. M., & Williams, R. F. (1985). Skills performance at low blood alcohol levels. Journal of Studies on Alcohol, 46(6), 482-485.

National Highway Traffic Safety Administration (1983). Field evaluation of a behavioral test battery for DWI (driving while intoxicated). Washington, DC.

National Institute on Alcohol Abuse and Alcoholism (1994). Alcohol Health & Research World (Special Focus): Women and Alcohol, 18(3).

Ross, L., & Mundt, J. (1988). Pilots' attitudes toward alcohol use and flying. Aerospace Medicine, 59, 913-919.

Shipley, C. V. (1970). Chemical tests manual for Kentucky. Eastern Kentucky University: Traffic Safety Institute.

Society of Automotive Engineers, Inc. (1986). Alcohol, accidents, and injuries (P-173). Warrendale, PA.

Tharp, V., Burns, M., & Moskowitz, H. (1981). Development and field test of psychophysical tests for DWI arrest. DOT Final Report, ODT HS 805 864: Washington, DC.

Thomley, K. E., Kennedy, R. S., & Bittner, A. C., Jr. (1986). Development of postural equilibrium tests for examining environmental effects. Perceptual and Motor Skills, 63, 555-564.

Turnage, J. J., & Kennedy, R. S. (1992). The development and use of a computerized human performance test battery for repeated-measures applications. Human Performance, 5(4), 265-301.

Turnage, J. J., Kennedy, R. S., Smith, M. G., Baltzley, D. R., & Lane, N. E. (1992). Development of microcomputer-based mental acuity tests. Ergonomics, 33, 1271-1295.

Wallgren, H., & Barry, H., III. (1970a). Actions of Alcohol, Vol. I, New York: Elsevier Publishing.

Wallgren, H., & Barry, H., III. (1970b). Actions of Alcohol, Vol. II, New York: Elsevier Publishing.

Widmark, E. M. P. (1932). Principles and applications of mediolegal alcohol determination. Berlin: Urban & Schwartzenberg.

Zobeck, T. S., Stinson, F. S., Grant, B. F., & Bertolucci, D. (1993). Surveillance Report #26: Trends in alcohol-related fatal traffic crashes, United States: 1979-91. Rockville, MD: National Institute on Alcohol Abuse and Alcoholism, Division of Biometry and Epidemiology.