A. Experimental Design
Ss' driving-related behaviors were examined using a driving simulator (SIM) and a divided attention test (DAT). They were administered alcohol to produce mean test-time BACs from 0.00% to 0.10% for moderate and heavy drinkers and from 0.00% to 0.08% for light drinkers.
Figure 1 outlines the factorial design of the experiment with three factors (age, gender, drinking practice). Each S was tested under a placebo and an alcohol treatment at two sessions separated by a week. Statistically, the alcohol treatment comparisons were nested within each of the cells created by the age X gender X drinking practices factorial design.
The placebo and alcohol treatments were administered at two sessions in counterbalanced order. Half the Ss received the placebo treatment first and the alcohol treatment second, and half the Ss received the alcohol treatment first and the placebo treatment second. Ss in the moderate and heavy drinking categories were tested on the SIM and DAT prior to receiving alcohol and at mean BACs of 0.10%, 0.08%, 0.06%, 0.04%, and 0.02%, and at a final 0.00% on DAT only (Table AP-I-1). Light drinkers were tested on the SIM and DAT prior to receiving alcohol and at mean BACs of 0.08%, 0.06%, 0.04%, and 0.02%, and at a final 0.00% on DAT only. BACs declined approximately 0.01% during a test battery of slightly more than 30 minutes. Since the aim was to test at mean BACs of 0.10%, 0.08%, 0.06%, 0.04%, and 0.02%, Ss were dosed to a BAC 0.01% above the testing level. They were examined repetitively with a breath-sampling instrument, and testing actually began when BACs were 0.005% above the desired mean level. As will be seen in the results, this procedure produced mean BACs extremely close to the desired levels.
The first post-alcohol testing for moderate and heavy drinkers, who were dosed to 0.11%, began when their BACs dropped to 0.105% and for light drinkers, who were dosed to 0.09%, when their BACs dropped to 0.085%. Two consequences should be noted. First, all Ss began their treatments at roughly the same time of day with the result that moderate and heavy drinkers were tested at 0.10% BAC at the same hour that light drinkers were tested at 0.08% BAC. Secondly, and of greater importance, all Ss were tested on the descending blood alcohol curve, because it is extremely difficult to pace alcohol consumption and track a rising alcohol curve for the purpose of behavioral testing at specified BACs (see Moskowitz, Daily, and Henderson, 1974 for such a procedure). As has been well established by the literature on acute tolerance to alcohol, however, Ss exhibit less impairment on a descending than on a rising alcohol curve. This means that during alcohol consumption and absorption, on the ascending limb of the BAC curve, impairment would be greater than what has been shown by this experiment.
During placebo sessions, Ss were tested at times which paralleled the testing times of the alcohol sessions. Based on an assumed mean ethanol clearance rate of 0.017% per hour, a 0.020% decrease in BAC would require approximately 70 minutes. Therefore, at placebo sessions Ss were tested at 70-minute intervals.
B. Regulatory Compliance
The study protocol and informed consent documents were reviewed by the NHTSA Human Use Review Panel (HURP), the SCRI Institutional Review Board (IRB) and the HFN IRB. Conduct of the study was approved by all panels prior to initiating the study. At regular intervals during the course of the study, members of the SCRI IRB were informed of the progress of the study, and they were advised when the study was completed.
C. Pilot Study
Before initiating the main study, pilot studies were performed to clarify two issues. The first issue concerned the advisability of administering an alcohol dose that would produce 0.11% BACs to light drinkers. The second issue was the sensitivity of the SIM response measures using the driving scenarios which had been constructed for this experiment.
1. Light Drinkers
This study was originally designed as a factorial experiment with equal numbers of light, moderate, and heavy drinkers to be dosed to a BAC of 0.11% (0.01% above the desired mean peak BAC). SCRI's prior alcohol experiments have shown repeatedly that moderate and heavy drinkers can reach that level without ill effects. Note that the definitions of drinking categories are derived from a scale from Cahalan, Cisin, and Crossley (1969), which relies on Ss' statements about the quantity and frequency of their drinking. It appeared most unlikely that light drinkers, as defined by their statements, would be able to reach a 0.11% BAC. Although the possible adverse effects of alcohol consumption could have been mitigated somewhat by very slow drinking, the change in time allowed for drinking would have disrupted the session schedule.
A pilot experiment was performed with nine light drinkers (8 females, 1 male) to determine the BAC that could be achieved. Three Ss in each of three age groups (21-24 years, 25-50 years, 51 years and above) participated in one session at the SCRI facility. Based on their height and weight, each was given sufficient alcohol over a 45-minute drinking period to produce a 0.09% peak BAC. They were advised to cease drinking if they began to feel uncomfortable. Eight of the nine completed the drinks and reached 0.09%. The peak level of 0.09% BAC was selected for light drinkers based on their comments and the pilot experiment drinking experience.
2. Simulator (SIM) Measures
In view of the low BACs to be examined in the study, it was assumed that skills performance differences between cell groups in response to equal BACs might be quite small, albeit significant. It was essential, therefore, that the SIM driving scenarios be examined prior to beginning the study to determine sensitivity to alcohol effects.
To examine the SIM scenarios, six females and two males attended two training sessions to learn to drive the simulator. After training, these pilot Ss, who ranged in age from 23 to 68 years, were tested on the SIM in a single day. They were first tested prior to receiving alcohol and then on a declining alcohol curve at 0.02% intervals. Five of the Ss were light drinkers and began post-dose trials at 0.085% BAC. Three moderate drinkers began post-dose trials at 0.105% BAC.
Ss in the pilot test exhibited performance impairment at all active BACs, in comparison with the initial, pre-alcohol test. Therefore, the simulator scenarios were considered sufficiently sensitive to be used in the main study.
D. Apparatus
As previously discussed, a simulator and a divided attention task were selected for this study. A literature review by Moskowitz and Fiorentino (2000) identifies these as the most sensitive of currently-available tasks for the examination of low BAC effects. They were selected not only because they are sensitive to alcohol but also because of their relevance to driving.
- Reaction time to peripheral signals (sec)
- Incorrect responses to peripheral signals (number)
- Speed deviation (mph)
- Lane position deviation (ft)
- Collisions (number)
- Times over speed limit (number)
- Response time (sec) with a Maximum Allowable Response Time of 10 sec
- Incorrect responses (number)
- Tracking error (cm)
1. Driving Simulator (SIM)
The driving simulator was constructed by Systems Technology, Inc. of California with scenarios and secondary tasks developed by SCRI personnel. SIM is a computer-based system, which uses three video monitors in a horizontal arc presenting a 110o angle view of the driving scene. The image responds to input from the steering wheel, accelerator and brake, and there is appropriate visual and sound feedback. A concurrent secondary task requires the detection of visual signals in peripheral vision. The driving scenarios include rural, suburban and urban segments. The total travel distance is 63,000 feet, and the drive typically requires 18 to 20 minutes. The simulator provides a variety of response measures from which a representative subset was selected for this experiment.
The rural segment is a straight one-lane road with shallow curves. Periodic cross traffic tests Ss' perception of speed and distance, and wind gusts increase steering difficulty. Since this segment lacks confounding variables, speed and lane position can be measured.
The suburban segment markedly increases driving demands. Ss drive a three-lane expressway at 55 mph, slowing to 45 mph for posted curves. They make frequent lane changes to pass other vehicles and to avoid cross traffic, entering traffic, and stalled cars.
In the urban segment, the driver attempts to maintain the posted 45 mph speed limit and slows to 25 mph for curves on a two-lane roadway through a city with 11 signal-controlled intersections. Pedestrians enter and cross in walkways at the signals.
The secondary task mirrors the information-processing demands and the dual-task nature of actual driving. Ss monitor the periphery of the visual display and respond to signals which appear at the extreme right and left. The signals are a left arrow, a right arrow, and a horn. Ss respond to a left or right arrow with the corresponding turn signal and to the horn sound with the horn button. In total, 72 signals occur at random intervals during the drive. Measures include response time and number of incorrect responses.
The following SIM response measures were analyzed:
2. Divided Attention Test (DAT)
Stephens and Michaels (1963) characterized driving as a time-shared activity between a visual search-and-recognition task and a tracking task. The DAT used in this study is conceived as an analogue of the time-sharing and information-processing demands of driving.
The DAT shares the SIM hardware and requires Ss to allocate attention to multiple sources of information on three video monitors. The concurrent structure and task demands prevent parallel processing of information, and attention must be alternated between tracking and visual search.
A one-dimensional pursuit tracking task appears on the center screen. As a red ball moves horizontally in response to a forcing function, Ss use a joystick to try to keep a white cross superimposed on the ball. The distance between the ball and cross is recorded as tracking error. As Ss perform the continuous tracking task, they also monitor four arrays, each containing six numbers in a 2 X 3 pattern, which appear above and below center in left and right peripheral vision. The numbers change continually, and Ss' task is to detect the appearance of the number "2". Response requires selection of the button on a 4-button response pad which corresponds to a target's position. For example, a correct response to a target in the upper left array is made with the upper left button. A 12-minute trial presents two targets at each of 24 positions.
Recorded measures for DAT include:
E. Subjects
A total of 168 Ss participated in this study. These Ss were divided into 24 cells, defined by the four categories of age, the two categories of gender, and the three categories of drinking practices. Thus, there were seven Ss in each cell.
1. Gender
Eighty-four men and 84 women participated in the study. The average age of the men was 34 years 11 months, and the average age of the women was 33 years 2 months. See Table AP-I-2 for other characteristics of the Ss, including mean height, weight, and age for each of the classifications.
2. Age
Four age groups each contained 21 men and 21 women. These were youthful drivers (19-20 years), young adult drivers (21-24 years), adult drivers (25-50 years), and older drivers (51-69 years). The mean age for the four groups were 19 years 8 months, 22 years 5 months, 32 years 8 months, and 61 years 7 months, respectively.
3. Drinking Practices
Ss were classified as light, moderate, and heavy drinkers, with 56 Ss in each category. They were categorized by the Cahalan, Cisin and Crossley (1969) Quantity-Frequency-Variability scale. During the period of S selection, it was decided to exclude any applicant near the borderline of a category. This decision was based on the fact that the categorizations rely on self-reports of alcohol use, and the literature suggests that self-reports have considerable variability. Since comparisons were to be made between drinking categories, it was not advisable to include Ss who might be incorrectly characterized as a result of the variability of their responses. For that reason, whenever applicants' responses placed them at the border of light and moderate or at the border of moderate and heavy, they were not included. It should be noted that recruitment of volunteers for an alcohol experiment does not attract extremely light drinkers.
F. Procedures
The procedures described in the following sections were followed for the experiments in Los Angeles and in Toronto.
1. Subject Recruitment and Screening
Applicants responded to newspaper ads and were interviewed first by telephone and then in-person. They were screened in terms of health history, current health status, and use of alcohol and other drugs. Pregnancy, chronic disease, or evidence of substance abuse resulted in exclusion. Those applicants who met study criteria were enrolled to fill age, gender and drinking-practices cells as illustrated in Figure 1.
2. Training Sessions
To learn to drive the SIM and perform DAT, Ss attended two 4-hour training sessions during the week prior to their first treatment session. The training sessions were separated by at least one day. Instructions, demonstrations, practice trials, and feedback proceeded by a standard protocol.
a. SIM Training
A Research Assistant (RA) demonstrated basic operation of the SIM and observed a S's first drive through rural, suburban, and urban scenarios. The RA provided instruction as needed. Following the introductory drive, the RA instructed and demonstrated the secondary task. In a second drive, the S both drove the SIM and performed the secondary task. The RA continued to provide instruction.
b. DAT Training
The RA first instructed and demonstrated only the DAT tracking task, and the S performed a 6-min trial of tracking alone. The RA then instructed and demonstrated the visual search task, and the S performed a 6-min trial of visual search alone. The initial training ended with a 12-min trial of the combined task.
c. Practice Test Batteries
After Ss had been trained on both SIM and DAT, they were required to rest for 30 minutes. They then performed the entire test battery without instruction or feedback. At the conclusion of the second battery, the RA discussed the S's performance with him or her, providing positive reinforcement for good scores and noting areas needing improvement.
Ss were given three DAT trials and four SIM drives on both training days. At the end of the second training day, Ss' scores were reviewed to determine whether criterion performance levels had been achieved. No S required an additional training session.
3. Experimental Test Sessions
Each S was tested at two sessions, which were separated by one week. Half the Ss received placebo at the first session and alcohol at the second session; the other half of the Ss received treatments in the reverse order.
a. Alcohol and Placebo Beverage Administration
Ss' were tested on the SIM and DAT prior to being given alcohol, at the highest BAC for their drinking classification, and at 0.02% BAC intervals as their alcohol levels decreased. They were tested on DAT when their BACs returned to zero. Note that testing occurred only on the descending limb of the BAC curve.
Calculations of alcohol doses were based on the amount of body water into which the alcohol would be distributed. Body water was estimated as a percentage of a S's body weight taking into account gender and age (Frisch, 1988), and the estimate was adjusted for frame size and body composition. To insure that alcohol was not administered to pregnant women, urine specimens were obtained from women of childbearing age and were tested for pregnancy prior to treatment administration.
The beverage was one part 80 proof vodka and 1.5 part orange juice for moderate and heavy drinkers, who received the beverage as three equal drinks at 10-minute intervals. The dilution was one part vodka to two parts orange juice for light drinkers, who received three equal drinks at 15-min intervals. Ss were instructed to pace each drink evenly over the entire drinking period, and they were monitored by an RA who periodically advised them of the time remaining to complete each drink.
The placebo beverage was identical to the alcohol beverage except that water was substituted for vodka. Vodka (10 ml) was floated on top of the beverage, and the edge of each cup was swabbed with vodka to produce an initial odor and taste of alcohol. Administration procedures were identical to those described for the alcohol beverage.
At both alcohol and placebo sessions, breath specimens were obtained with an Intoxilyzer 5000 for BAC measurements beginning 30 minutes after the end of drinking. If an initial BAC was lower than the target, breath sampling was repeated at 10-minute intervals until the target was reached or until successive tests showed that the BAC had begun to decline. In the latter case, a booster dose was given. A breath specimen was obtained at the conclusion of the first test battery, and BAC monitoring continued in the manner described above. The Intoxilyzer display of the measurements was shielded from the S's view.
b. Performance Testing
Table AP-I-1 displays performance testing schedules. At alcohol sessions, the batteries (SIM and DAT) were initiated within +/- 0.005% of the target BAC and were repeated at 0.02% intervals with the final DAT beginning at 0.00%. Testing was initiated at placebo sessions after obtaining the first breath specimen. Timing of subsequent test batteries allowed sufficient time for a 0.02% decrease with the interval calculated at a 0.017% per hour metabolism rate.