DUI Offenders Display Reduced Perception of Intoxication and Heightened Impulsive Choice in Response to Alcohol

Abstract

Driving under the influence (DUI) of alcohol continues to be a major contributor in traffic fatalities. There is growing evidence for heightened trait impulsivity in DUI offenders, but little is known about how impulsivity could interact with alcohol intoxication in a manner that would increase the likelihood of driving while intoxicated. This placebo-controlled study examined the acute effects of 0.65 g/kg alcohol on two facets of impulsivity (impulsive choice and response inhibition), simulated risky driving behavior, and subjective intoxication in a group of 20 DUI offenders and 20 control drivers with no history of DUI. It was predicted that compared with controls, DUI offenders would self-report greater impulsivity, and display greater impulsive choice and driver risk-taking, particularly in response to alcohol. Results showed that alcohol impaired drivers’ inhibitory control and increased their impulsive choice behavior and risky driving behavior. Alcohol selectively increased impulsive choice of DUI offenders, as control drivers showed no alcohol-induced increase in their impulsive choices. Results also showed that, compared with controls, offenders reported feeling less intoxicated and were more willing to drive after drinking. Laboratory studies are beginning to show that DUI offenders differ from non-offenders in their acute responses to alcohol. This study identified two alcohol response characteristics of DUI offenders that indicate their lack of risk awareness during intoxication: heightened impulsivity and reduced subjective intoxication. Strategies and treatments to alter these response characteristics in DUI offenders could enhance their risk awareness during the intoxicated state and possibly reduce risk of DUI recidivism.

Alcohol-related traffic fatality and injury continue to be a major public health problem in the United States. The National Highway and Traffic Safety Administration (NHTSA) reported that in 2016, over 250,000 traffic injuries were alcohol-related, and driving under the influence (DUI) was a factor in one-third of all traffic fatalities (National Center for Statistics and Analysis, 2016). Interventions to reduce recidivism in DUI offenders involve motivational interviewing and education with emphasis on increasing the offender’s perceptions of the risks associated with alcohol use (i.e., risk awareness). Such prevention efforts have had only modest success. Despite considerable economic resources dedicated to these efforts, DUI remains one of the most frequently repeated offenses, with over one-third of offenders being charged with a second DUI offense within five years (National Center for Statistics and Analysis, 2016; FBI, 2006).

Research to improve treatment and prevention seek to identify characteristics of DUI offenders. The research has relied on surveys and personality inventories. Survey studies of DUI drivers consistently provide evidence of poor behavioral regulation. Driving records show that DUI offenders commit more moving violations, such as speeding, and are involved in more accidents compared with the general population (Bishop, 2011; Donovan, Marlatt & Salzberg, 1983; McMillen, Pang, Wells-Parker & Anderson, 1992). Personality inventories of DUI drivers reliably show high levels of impulsivity and sensation-seeking (Ryb, Dischinger, Kufera & Read, 2006; Chalmers, Olenick & Stein, 1993; Hubicka, Kallmen & Hiltunen, 2010). The few laboratory-based assessments of DUI offenders have reported neurocognitive deficits in offenders, especially those with a history of repeat offenses (Bouchard, Brown & Nadeau, 2012; Glass, Chan, & Rentz, 2000; Ouimet et al., 2007). The emerging view is that DUI offenses are not necessarily a consequence of alcoholism or alcohol use disorders, but also can be symptomatic of deficient behavioral regulation characterized by impulsivity and heightened reward sensitivity (Ryb et al., 2006).

Despite growing evidence for heightened trait impulsivity in DUI offenders, little is known about how impulsivity could interact with alcohol intoxication in a manner that would increase the likelihood of driving while intoxicated. A key component of impulsivity is poor inhibitory control which is assessed by stop-signal and go/no-go tasks that measure the ability to suppress pre-potent motoric responses (Fillmore, 2012). Studies using these tasks have found that alcohol impairs the ability to inhibit behavior even at comparatively moderate blood alcohol concentrations (BACs) between 50 mg/100 ml and 80 mg/100 ml (e.g., de Wit, Crean, & Richards, 2000; Fillmore & Vogel-Sprott, 2000; Marczinski & Fillmore, 2003; Fillmore & Weafer, 2011). Our simulated driving studies also show that drinkers who display poorer inhibitory control under alcohol show greater sensitivity to the disruptive effects of alcohol on their driving performance as indicated by increased deviation of lane position, centerline and road edge crossings, and increased steering rate (Fillmore, Blackburn, & Harrison, 2008; Harrison & Fillmore, 2005; Weafer, Camarillo, Fillmore, Milich, & Marczinski, 2008).

Taken together, these findings indicate that heightened sensitivity to the disinhibiting effects of alcohol could contribute to the display of impulsive and other inappropriate behaviors, such as driving while intoxicated, thus being characteristic of DUI offenders. We tested this possibility by comparing sensitivity to the disinhibiting effects of a controlled dose of alcohol between a group of DUI offenders and demographically-comparable control drivers with no history of DUI (Van Dyke & Fillmore, 2014a). Results showed that DUI offenders self-reported greater levels of impulsivity than did controls, but no group difference was observed in the degree to which alcohol impaired inhibitory control or driving performance.

It is unclear why the groups did not differ in their sensitivity to the disinhibiting effect of alcohol. One reason might have to do with the multifaceted nature of impulsivity. The study focused on one specific aspect of impulsivity, poor response inhibition measured as the ability to momentarily suppress a prepotent (i.e., instigated) behavioral response. However, another important aspect of impulsivity involves dysregulation in responding to reinforcers in the environment. Impulsivity is also characterized by a failure to delay immediate rewards in favor of greater rewards in the long-term. Such sub-optimal decision-making or impulsive choice is measured by delayed discounting tasks that measure the individual’s preference for sooner, but smaller rewards (impulsive choices) versus delayed, but larger rewards (non-impulsive choices) (e.g., Bickel & Marsch, 2001; Johnson & Bickel, 2002). Rewards in discounting tasks are typically hypothetical sums of money (e.g., $5 today or $100 in one year). However, in experiential discounting models, participants receive actual monetary rewards and experience the actual delays in reward delivery (Dougherty, Mathias & Marsh, 2003; Dougherty, Mathias, Marsh, & Jagar, 2005). Studies using experiential discounting tasks have shown that alcohol increases impulsive choice such that intoxicated participants display a greater preference for immediate but smaller rewards over larger but delayed rewards (e.g., Reynolds, Richards, & de Wit, 2006).

Such impulsive choice behavior while intoxicated could play an important role in DUI behavior. Deciding to drive after drinking yields the immediate reward of convenience for the drinker, to travel home, to the next bar, or elsewhere, and the decision to drive after drinking could reflect a failure to delay or forego that immediate reward. Indeed, some recent research supports this notion. There is some evidence for the role of impulsive choice in drinking and driving behavior. McCarthy et al. (2012) examined college students who reported recent histories of driving within 2 hours of having three or more drinks. The researchers reported that these “drink drivers” became more impulsive in response to a moderate dose of alcohol compared with a control group with no history of drinking and driving. To date, no research has tested the possibility that DUI offenders display heightened choice impulsivity in response to alcohol. However, such evidence would be important as it would show how a specific behavioral mechanism of impulsivity contributes in situ to the decision to drive after drinking in DUI offenders.

It is also important to examine alcohol effects on specific driving behaviors that are characteristic of impulsivity in DUI offenders. Our prior studies of simulated driving in DUI offenders examined the acute impairing effects of alcohol on measures that emphasized driver skills, such as lane position, standard deviation, and braking time (Van Dyke & Fillmore, 2014a, 2014b). DUI offenders and control drivers were similarly impaired by alcohol. However, it is now well recognized that alcohol likely contributes to motor vehicle injuries and fatalities by its joint effects of impaired driver skill and increased risk-taking behavior. Indeed, driver simulation studies of alcohol have found that intoxicated drivers are more willing to choose risky traffic lanes over less-risky options (Burian, Hensberry, & Liguori., 2003; Burian, Liguori, & Robinson, 2002), maneuver through narrower gaps (Cohen, Dearnaley, & Hansel, 1958), and underestimate potential collision time with oncoming traffic (Leung & Starmer, 2005). Risk-taking is often measured by proxemics, indicated by instances where drivers maneuver their vehicle closer to other vehicles (e.g., tailgating) on the road. This behavior is quantified by determining drivers’ time-to-collision (TTC). TTC is a time-related safety margin measure determined by the bumper-to-bumper distance between the driver’s vehicle and other vehicles on the road, divided by the closing speed of the vehicles (Taieb-Maimon & Shinar, 2001; Zhang & Kaber, 2013). Thus, TTC provides a measure of the time (in seconds) it would take for a collision to occur between two or more vehicles on the roadway (Zhang, Antonsson, & Grote, 2006). Risky driving is evidenced by lower TTC values (less time to a potential collision). Research indicates that greater risk-taking, as measured by TTC, is associated with increased risk for motor vehicle collisions (e.g., Hayward, 1972; Ranney, 1994; Summala, 1985, 1988; Wilde & Murdoch, 1982). Recent studies in our laboratory showed that following alcohol, drivers decreased their TTC by driving closer to other vehicles on the roadway relative to placebo (Laude & Fillmore, 2015; Van Dyke & Fillmore, 2017). We have also shown that drivers with poorer inhibitory control tend to display the greatest level of risk-taking in driving simulation (Laude & Fillmore, 2015). To date, no research has examined the degree to which alcohol increases driver risk-taking in DUI offenders. To the extent that such risk-taking is mediated by impulsivity, offenders could display heightened sensitivity to alcohol effects on risk-taking behavior during driving.

The purpose of the present study was to examine the effects of alcohol on impulsive choice and driver risk-taking in a group of DUI offenders. DUI offenders were compared to “control” drivers with no DUI history. Laboratory tasks examined impulsive choice and simulated driving performance following 0.65 g/kg alcohol that produces a peak blood alcohol concentration of approximately 80 mg/100 ml, and following a placebo beverage. It was predicted that alcohol would increase impulsive choice and increase driver risk-taking with regard to DUI offenders. It was also predicted that compared with controls, DUI offenders would self-report greater impulsivity and display greater impulsive choice and driver risk-taking, particularly in response to alcohol.

Methods

Participants

Forty adults between the ages of 21 and 34 participated in the study. Volunteers consisted of 20 DUI offenders and 20 controls with no prior DUI arrests. Each group was comprised of 15 male and 5 female subjects. This ratio was chosen based on recent estimates indicating the ratio of male to female DUI offenders is 3:1 in the United States (e.g., NHTSA, 2015). The samples were not matched on demographic characteristics or cultural background. Volunteers had to meet a strict set of intake criteria, primarily driven by safety concerns for alcohol administration. Online postings and fliers placed around the greater Lexington community advertised for the recruitment of individuals for studies on the effects of alcohol on behavioral and mental performance. Some of the advertisements directly targeted individuals arrested for DUI. All DUI offenders were required to have at least one alcohol-related DUI conviction in the past five years, whereas control subjects had no prior DUI convictions or license revocations. All DUI offenders were at least 21 years or older at the time of the DUI conviction. Convictions were verified by State District Court Record Reporting Systems (e.g., Courtnet©). Two DUI offenders had two DUI convictions and all others were first-time offenders. Interested individuals called the laboratory and completed a telephone screening during which information on demographics, drinking habits, drug use, and physical and mental health was gathered. Individuals reporting history of psychiatric disorder, CNS injury, or head trauma were excluded from participation. All volunteers were current consumers of alcohol, but individuals were excluded if their current alcohol use met criteria for a severe alcohol use disorder on the Structured Clinical Interview for DSM-V (SCID-V). Many of these individuals are physically dependent and thus experiencing adverse health consequences from their excessive drinking. It is unethical to contribute to these problems by administering alcohol to these individuals. Individuals consuming fewer than two standard drinks per month were also excluded from participation. All volunteers were required to have held a valid driver’s license for at least the past three years and drive on a weekly basis. The use of any psychoactive prescription medication and recent use of amphetamines (including methylphenidate), barbiturates, benzodiazepines, cocaine, opiates, and tetrahydrocannabinol (THC) was assessed by means of urine analysis. Any volunteer testing positive for the presence of any of these drugs (except THC) during the sessions was excluded from participation. In the event a participant tested positive for THC, the participants were asked to self-report the last time of marijuana use. If the time of last use was greater than 24 hours prior to the session, the session continued as normal. If participants reported using marijuana in the past 24 hours, attempts were made to reschedule the session to a later date. No female volunteers who were pregnant participated in the research, as determined by self-report and urine human chorionic gonadotrophin levels (Icon25 Hcg Urine Test, Beckman Coulter). Women who were breast feeding were also excluded from participation. All volunteers had a minimum high school level of education. The University of Kentucky Medical Institutional Review Board approved the study (IRB Protocol 12–0737-F1V, Behavioral Dysregulation and Alcohol Sensitivity in Risky Drivers). Less than 1 in 3 volunteers in either group failed to pass screening. Recruitment continued until we obtained 20 study-eligible volunteers with 15:5 male-female sex ratio in each group. All study volunteers provided informed consent prior to participation and received a base payment of $115 (before task-specific monetary bonuses) for their participation.

Apparatus and Materials

Choice Impulsivity.

A two-choice impulsivity paradigm (TCIP; Dougherty et al., 2003) was used to assess participants’ ability to delay responding for immediate rewards in favor of delayed rewards. Participants responded to one of two images (i.e., circle or square) on a computer screen by clicking on the image of their choice using the computer’s mouse. The circle was associated with a short time delay (i.e., 5 seconds) and the square was paired with the long time delay (i.e., 15 seconds). After making a response, participants experienced the respective time delay in real time before proceeding to the next trial. After the delay, the reward (i.e., $0.05 or $0.15) appeared on the screen and was added to the participant’s “bank”, which kept a running total of task earnings and was visible on the computer screen at all times during the task. Impulsive choices were indicated by a greater number of responses to the short-delay reward compared with the long-delay reward. The measure of interest was the proportion of total responses to the short-delay reward (i.e., impulsive responding) relative to the long-delay reward (i.e., non-impulsive responding) across 50 test trials. The TCIP required approximately 12 minutes to complete.

Response Inhibition.

A cued go/no-go reaction time task was used to measure participants’ response inhibition to no-go targets and their reaction time to go targets (e.g., Fillmore & Weafer, 2004). The task required finger presses on a keyboard and measured the ability to inhibit prepotent behavioral response of executing a key press. Cues provided preliminary information regarding the type of target stimulus (i.e., go or no-go) that was likely to follow, and the cues had a high probability of signaling the correct target. Participants were instructed to press the forward slash (/) key on the keyboard as soon as a go (green) target appeared and to suppress the response when a no-go (blue) target was presented. The go cue conditions were of particular interest. Go cues generate response prepotency which speeds response time to go targets. However, subjects must overcome this response prepotency to inhibit the response if a no-go target is subsequently displayed. Response inhibition was measured by the proportion of no-go targets in which subjects failed to inhibit a response (p-inhibition failures) during the test. Poor inhibitory control was indicated by a higher proportion of inhibition failures (i.e., greater p-inhibition failure score). A test required approximately 15 minutes to complete. The task has been used in other research, has strong psychometrics, including reliability, and is highly sensitive to dose-dependent impairing effects of alcohol on drinkers’ inhibitory control (Fillmore & Weafer, 2013, Weafer & Fillmore, 2016).

Driver Risk-Taking.

A computerized driving simulator measured driving performance (STISIM Drive, Systems Technology Inc., Hawthorne, CA). In a small room, participants sat in front of a 19-inch computer display which presented the driving simulation at a 60 degree horizontal field of view. The simulation placed the participant in the driver seat of the vehicle which was controlled by steering wheel movements and manipulations of the accelerator and brake pedals. At all times, the participant had full view of the road (lane width = 12 ft), surroundings, and instrument panel, which included an analog speedometer. Crashes, either into another vehicle or off the road, resulted in the presentation and sound of a shattered windshield. The program then reset the driver in the center of the right lane at the point of the crash.

This simulated driving scenario was designed to test risky driving behavior and required participants to drive 21,100 feet on a busy, 4-lane street within a metropolitan setting. There was no posted speed limit. Each direction of traffic was comprised of two lanes. The driver was free to navigate among other vehicles within the driver’s two lanes of traffic. Other vehicles were presented at various speeds in both lanes such that the driver had to change lanes to overtake vehicles in order to maintain speed. To instigate the potential for risk-taking, drivers could earn monetary reinforcement for quickly completing the drive: $5 for completion in 3–4 min, $4 for 4–5 min, $3 for 5–6 min, $2 for 6–7 min, $1 for 7–8 min, and $0.50 for over 8 min. Drivers were penalized $0.50 for each crash. This response conflict scenario is designed to mimic everyday driving behaviors in which drivers are rewarded by arriving at their destination on time at the cost of potential traffic citations, and has been successfully used in other research in our laboratory (e.g., Van Dyke & Fillmore, 2015; Fillmore et al., 2008).

The primary measure of driver risk was time-to-collision (TTC). This is a time-related safety margin measure (Taieb-Maimon & Shinar, 2001) determined by the bumper-to-bumper distance between two vehicles, divided by the closing speed of the vehicles (Zhang & Kaber, 2013). TTC is operationally defined as the time that remains until collision occurs if both the lead and the driven vehicle continue on the same course (Zhang et al., 2006). It is obtained by taking the minimum value of the riskiest instance in which the driven car approaches a lead car throughout the drive, sampled at each foot of the driving test. Riskier driving is indicated by smaller TTC values (seconds). Average drive speed (mph) and accident frequency were also measured.

Perceived Driver Fitness.

Participants self-evaluated their driving fitness during the study on three indicators: perceived level of intoxication, willingness to drive a motor vehicle, and estimated blood alcohol concentration (BAC). The first two indicators were rated on 100 visual-analogue scales ranging from 0 “not at all” to 100 “very much.” Participants estimated their BAC on a scale ranging from 0 to 160 mg/100 ml with a provided midpoint of the current legal driving limit (i.e., 80 mg/100 ml). These scales have been used in other alcohol studies of driving and are sensitive to the effects of the drug (e.g., Harrison & Fillmore, 2005; Harrison, Marczinski & Fillmore, 2007; Van Dyke & Fillmore, 2015).

Driving History and Experience (Harrison & Fillmore, 2005).

Participants completed a Drive History and Experience Questionnaire (DHEQ) that included measures of driving experience such as length of time holding a driver’s license and number of days and miles driven per week. The questionnaire also gathered information about participants’ driving behaviors,such as license revocations, presence and number of DUI citations and punishments, traffic accidents, traffic tickets, typical driving environment (rural, urban, and interstate), and the type of vehicle transmission (manual, automatic, or both).

Recent Drinking Habits.

Recent patterns of alcohol use were measured by the Timeline Follow-back (TLFB, Sobell & Sobell, 1992). The TLFB assessed daily patterns of alcohol consumption over the past 3 months. The measure is structured with prompts to facilitate participants’ recall of past drinking episodes to provide a more accurate retrospective account of alcohol use during that time period. Multiple aspects of alcohol consumption over the past 3 months are measured including the total number of drinking days, the total number of drinks consumed, drinking days that they felt drunk (drunk days), and binge drinking episodes. A binge was defined as a drinking episode in which the individual drank to achieve a resultant BAC that was equal to or greater than 80 mg/100 ml (legal limit for operating a motor vehicle in the United States). The resultant BAC was estimated for each drinking episode based on the participant’s reported number of drinks, the duration of the episode, and the participant’s sex and body weight. Estimated BACs were calculated using well-established, valid anthropometric-based BAC estimation formulae which assume an average clearance rate of 15 mg/dl per hour of the drinking episode (McKim, 2007; Watson, Watson, & Batt, 1981).

Alcohol Use Disorder Identification Test – AUDIT (Babor, De La Fuente, Saunders, & Grant, 1989).

This 10-item self-report questionnaire was used to assess consequences of harmful drinking. Higher total scores indicate greater problems with alcohol.

Drug Abuse Screening Test – DAST (Skinner, 1982).

This 28-item self-report questionnaire screened for drug abuse problems. A score of six or more has been suggested as indicative of a drug use disorder (Skinner, 1982).

Barratt Impulsiveness Scale – BIS-11 (Patton, Stanford, & Barratt, 1995).

This 30-item self-report questionnaire was designed to measure the personality dimension of impulsivity. Impulsivity is thought to contribute to the risk of behavioral disinhibition under alcohol (Fillmore, 2007; Finn, Kessler, & Hussong, 1994). Participants rated 30 different statements (e.g., “I do things without thinking”) in terms of how typical each statement is for them on a 4-point Likert-type scale ranging from Rarely/Never to Almost Always/Always. Higher total scores indicate higher levels of self-reported impulsiveness (score range 30–120).

Procedure

The study was conducted in the Behavioral Pharmacology Laboratory of the Department of Psychology at the University of Kentucky and all volunteers provided informed consent. Volunteers were informed that the purpose of the study was to examine the effects of alcohol on driving performance and other cognitive and behavioral tasks. Participants were tested individually and completed an initial familiarization session to become acquainted with laboratory procedures, practice the tasks, and gather background information.

Participants were tested under 0.65 g/kg alcohol and a placebo on separate days and the dose order was counterbalanced across volunteers and groups. Sessions were separated by a minimum of one day and a maximum of one week. All participants were required to abstain from alcohol for 24 hours and food for 4 hours prior to each session. The alcohol dose was calculated based on body weight and administered as absolute alcohol mixed with three parts carbonated soda. Participants consumed the dose in six minutes. The dose produces an average peak BAC of 80 mg/100 ml approximately 60–70 minutes after consumption. The placebo dose (0.0 g/kg) consisted of a volume of carbonated mix that matched the total volume of the 0.65 g/kg alcohol drink. A small amount (i.e., 3 ml) of alcohol was floated on the top of the beverage and each glass was sprayed with an alcohol mist that provided a strong alcoholic scent as the beverage was consumed.

Testing began 40 minutes post-beverage consumption and each task was separated by a small (i.e., 5 min) rest interval. Test order was identical across each dose session beginning with the cued go/no-go task at 40 min, the drive test at 60 min, and TCIP test at 70 min. The perceived driver fitness scale was administered six times at 70, 85, 130, 175, 220, and 265 min from the onset of drinking. BAC samples were obtained nine times at 20, 40, 60, 70, 85, 130, 175, 220, and 265 min post-beverage. Participants remained in the lab until their BAC fell below 20 mg/100 ml. At 265 min the BAC of most participants was below 20 mg/100 ml. Upon completion of the final session, participants were paid and debriefed. Transportation home by taxi was provided after the sessions.

Data Analyses

The performance measures on each task were analyzed individually by a 2 Group (DUI vs. control) X 2 Dose (0.0 g/kg vs. 0.65 g/kg) mixed-design analysis of variance (ANOVA). Measures of perceived driving fitness following the active dose (0.65 g/kg) were analyzed individually by 2 Group (DUI vs. control) X 2 Dose (0.0 g/kg vs. 0.65 g/kg) X 6 Time (70, 85, 130, 175, 220, and 265 min) mixed ANOVAs. BACs under alcohol were analyzed by a 2 Group (DUI vs. control) X 9 Time (20 minutes – 260 minutes) mixed ANOVA. The study was not designed to test for sex differences, nor powered to do so. Therefore the sex factor was not included in the analyses. There was no sex difference in BAC.

Results

Demographics, drinking and driving history, and other drug use

Table 1 lists the demographic and other background characteristics of drivers in the DUI and control groups. The racial makeup of the DUI group was 90% Caucasian and 10% African-American. Two participants in the DUI group were recidivist offenders having 2 previous offenses. In the control group, 85% of the participants self-reported Caucasian and 15% were African-American. Driving experience was determined based on years of licensed driving, number of driving days per week, total weekly miles driven, number of traffic tickets, and number of accidents in which the participant was the driver of the vehicle. Comparisons between DUI and control drivers using post-hoc, two-sample t tests showed no group differences on any measure of driving experience (ps > .24).

Table 1.

Demographics and drinking and driving history

Ratio (n): Men/Women	Controls 15/5		DUI Offenders 15/5
	M	(SD)	M	(SD)	t	p
Age	24.10	(3.59)	25.75	(4.28)	1.25	0.22
Months since DUI	0	0	21.00	(11.56)	-	-
Years driving	8.30	(3.64)	9.04	(4.55)	0.57	0.57
Drive frequency	5.29	(1.77)	5.81	(2.07)	0.86	0.40
Drive distance	26.57	(41.71)	17.71	(13.30)	0.90	0.37
Total traffic tickets	1.25	(1.73)	1.71	(1.58)	0.87	0.39
Total accidents	1.13	(2.02)	1.84	(1.72)	1.20	0.24
TLFB total drinks	92.99	(48.34)	178.26	(118.08)	2.99	0.005
TLFB drinking days	24.48	(12.96)	27.95	(14.31)	0.81	0.43
TLFB binge episode	6.42	(7.13)	13.37	(10.25)	2.49	0.02
TLFB drunk days	7.99	(6.78)	14.56	(8.06)	2.78	0.008
AUDIT total	7.10	(3.11)	10.55	(4.36)	2.88	0.006
DAST total	1.60	(1.64)	3.70	(4.59)	1.93	0.06
BIS total	62.65	(9.10)	62.85	(6.83)	0.08	0.94

Comparison of DUI offenders to controls on background characteristics. Age = years of age; Years driving = total years of licensed driving; Drive frequency = total number of driving days per week; Drive distance = miles driven per week; Total traffic tickets = total number of traffic citations; Total accidents = total number of accidents in which the participant was the driver; TLFB total drinks = TLFB total drinks consumed in the past 3 months; TLFB drinking days = TLFB total drinking days in the past 3 months; TLFB binge episodes = number of binge drinking episodes defined as drinking to or in excess of 80 mg/100 ml; TLFB drunk days = total number of days in which the participant drank to a level that they felt drunk; AUDIT total = total score; DAST total = total score; BIS total = Barratt Impulsiveness Scale (BIS-11) total score.

Compared with controls, DUI offenders reported a greater number of drinks consumed in the past 3 months, t(28) = 2.99, p = .005, as well as number of binge episodes, t(28) = 2.49, p = .02, and drunk days, t(28) = 2.78, p = .008, as well as high AUDIT scores, t(28) = 2.88, p=.006. In terms of other drug use, 10 participants in the DUI group (M = 12.1 days, SD = 3.77) and four control participants (M = 5.25 days, SD = 8.45) reported using cannabis in the past month. Eight participants in the DUI group and three participants in the control group tested positive for THC at testing. However, all participants self-reported not using cannabis for at least 24 hours prior to the study sessions. No other drug use was reported in the past month. With regard to DAST scores, although the groups were not statistically different, DUI offenders were trending toward higher DAST scores, t(38) = 1.93, p = .06, d = 0.61. The groups did not differ in self-reported impulsivity on the BIS (p=.940).

Blood Alcohol Concentrations (BACs)

BACs following the alcohol administration were examined by a 2 (Group) X 9 (Time) ANOVA. A main effect of time was found owing to the rise and fall of BAC during the course of testing, F(9, 342) = 177.76, p < .001, η_p² = 0.82. No main effects (p = .12; η_p² = .06) or interactions involving group or time were found (p = .26; η_p² = .03). Figure 1 plots the mean BACs for easch group.. The figure shows that mean peak BAC was achieved between 40 and 60 min after drinking and began to decline 85 min after drinking. No detectable BACs were observed in the placebo condition.

Figure 1.

BACs following 0.65 g/kg alcohol for DUI and control participants. Error bars indicate standard error of the mean.

Choice Impulsivity

A 2 Group X 2 Dose mixed-design ANOVA examined participants’ impulsive choices as the proportion of trials in which they chose the short-delayed, smaller rewards. The analysis revealed a significant group X dose interaction, F(1, 38) = 4.53, p = .040, η_p² = .11. No significant main effects of group, F(1, 38) = 1.77, p = .191, η_p² = .04, or dose, F(1, 38) = 1.30, p =.261, η_p² = .03, were found. The interaction is illustrated in Figure 2. The figure shows that impulsive choices of controls were unaffected by alcohol as their choice of short-delayed rewards under alcohol showed little difference from placebo. However, for DUI offenders, impulsive choices increased under alcohol, relative to placebo. Planned comparisons between dose conditions confirmed that alcohol increased impulsive choices compared with placebo for DUI offenders t(19) = 1.9, p = .035, d = 0.43, but not for controls, t(19) = 0.9, p =.180, d = 0.20.

Figure 2.

Impulsive responding under 0.65 g/kg alcohol and placebo for DUI and control participants. Impulsive choice scores calculated by the proportion of impulsive trials over the total number of trials. Higher values indicate a greater preference for impulsive choices. Error bars indicate standard error of the mean.

Response Inhibition

A 2 Group X 2 Dose mixed-design ANOVA of drivers’ proportion of inhibitory failures on the cued go/no-go task revealed a significant main effect of dose, F(1, 38) = 8.35, p = .006, η_p² = .18. No main effect of group (p = .33; η_p² = .02) or interaction was found (p = .95; η_p² = .00). Table 2 presents the mean proportion of inhibition failures for each group following placebo and alcohol. The means show that inhibition failures increased under alcohol compared with placebo, and this increase was similar for DUI participants and controls. A 2 Group X 2 Dose ANOVA of reaction time to go cues found no significant main effects of dose or group, or an interaction (all ps > .24; η_p²: .01 – .04).

Table 2.

Mean (SD) measures under placebo and 0.65 g/kg alcohol for controls and DUI offenders

Men/Women (Ratio)	Controls 15/5		DUI Offenders 15/5
	0.0 g/kg M (SD)	0.65 g/kg M (SD)	0.0 g/kg M (SD)	0.65 g/kg M (SD)
Cued Go\No-Go Task
Inhibition Failures	0.12 (0.12)	0.18 (0.16)	0.08 (0.11)	0.14 (0.14)
Reaction Time	303.70 (17.92)	308.51 (17.13)	311.08 (26.26)	312.71 (26.05)
Simulated Driving
Risky Driving (TTC)	0.07 (0.07)	0.05 (0.05)	0.08 (0.06)	0.06 (0.06)
Driver Speed	51.68 (8.02)	53.12 (7.25)	52.50 (7.52)	56.01 (7.62)
Collisions	0.60 (0.75)	0.80 (0.95)	0.50 (0.76)	0.70 (1.08)

Inhibition Failures = proportion of no-go targets following go cues in which there was a failure to inhibit a response in the cued go/no-go task; Reaction Time (msec) = RT to go-targets following go cues in the cued go/no-go task; Risky Driving (TTC) = minimum time to collision (secs) during simulated drive; Driver Speed = average miles per hour across a drive; Collisions = average number of vehicle collisions during a drive

Risky Driving

Table 2 presents the mean time-to-collision (TTC) values under each dose. Mean TTC values indicate that alcohol increased risky driving by reducing drivers’ TTC with both groups, showing similar reductions in their TTC under alcohol compared with placebo. A 2 Group X 2 Dose mixed-design ANOVA confirmed a significant main effect of dose, F(1, 38) = 8.85, p = .005, η_p² = .18. No main effect of group (p = .437; η_p² = .02) or interaction (p = .861; η_p² = .00) was found. Secondary measures of risky driving (mean driver speed and number of collisions) were also analyzed. A 2 Group X 2 Dose ANOVA found a significant main effect of dose on drive speed, F(1, 38) = 6.90, p = .012, η_p² = .15. The mean speeds in Table 2 show that drivers drove faster under alcohol compared with placebo. No main effect of group or interaction was obtained (ps > .28). Table 2 also presents the number of collisions incurred during the drive. On average, collisions were infrequent, with less than a single collision per drive. No significant main effect or interactions were found (ps > .19).

Perceived Driver Fitness

Table 3 presents the group mean ratings following alcohol and placebo for each indicator of perceived driver fitness (subjective intoxication, willingness to drive, and estimated BAC) from 70 min to 265 min post-administration.

Table 3.

Subjective effect measures under placebo and 0.65 g/kg alcohol

Placebo (0.0 g/kg) M (SD)								Alcohol (0.65 g/kg) M (SD)
Minutes Post-Beverage								Minutes Post-Beverage
		70	90	135	180	225	270	70	90	135	180	225	270
Subjective Intoxication	Control	11.15 (11.8)	10.95 (18.8)	4.55 (6.9)	2.90 (5.2)	1.50 (1.9)	2.16 (3.4 )	61.25 (19.8)	44.95 (23.0)	36.20 (21.4)	26.65 (20.7)	13.60 (15.4)	5.15 (4.5)
	DUI	12.05 (16.8)	8.60 (11.8)	5.40 (8.1)	3.35 (5.0)	2.95 (4.0)	1.45 (1.7)	48.85 (25.7)	37.55 (22.3)	21.50 (14.3)	15.60 (12.6)	16.80 (19.8)	6.90 (10.0)
Willingness to Drive	Control	60.20 (27.7)	69.15 (27.5)	78.35 (24.1)	82.30 (21.8)	85.15 (18.9)	90.65 (15.1)	20.05 (16.2)	27.70 (24.0)	42.35 (30.8)	57.85 (29.3)	66.45 (27.4)	76.75 (18.1)
	DUI	55.45 (34.9)	67.45 (31.2)	80.85 (25.0)	82.25 (29.1)	90.15 (23.1)	95.15 (8.6)	33.30 (29.1)	40.05 (26.7)	48.10 (26.5)	59.79 (32.5)	66.75 (32.3)	73.05 (29.1)
BAC Estimation	Control	46.75 (23.0)	36.00 (22.5)	19.75 (18.7)	14.00 (16.9)	6.75 (11.7)	3.25 (6.1)	95.25 (28.5)	87.00 (35.7)	72.00 (31.8)	56.25 (32.9)	42.50 (30.1)	26.50 (23.4)
	DUI	39.25 (21.9)	28.75 (23.3)	18.25 (17.8)	10.25 (13.5)	6.25 (8.7)	3.50 (5.9)	86.88 (35.1)	74.75 (33.1)	58.00 (32.8)	41.75 (28.9)	32.25 (26.8)	21.00 (24.4)

Subjective Intoxication = Participant’s subjective level of intoxication; Willingness to Drive = Participant’s willingness to drive a vehicle; BAC Estimation = Participant’s estimation of current BAC

Subjective Intoxication.

A 2 (Group) X 2 (Dose) X 6 (Time) ANOVA of subjective intoxication ratings revealed a significant three-way interaction, F(5, 190) = 2.74, p = .021, η_p² = .07. Table 3 shows that subjective intoxication differed as a function of dose, with higher ratings under alcohol versus placebo, that diminished over time as BAC fell. The table also shows group differences. DUI offenders and controls reported similar levels of intoxication following placebo, but following alcohol, DUI offenders reported less intoxication compared with controls and this difference was evident throughout most of the time course.

Willingness to Drive.

A 2 (Group) X 2 (Dose) X 6 (Time) ANOVA of willingness ratings revealed a significant three-way interaction, F(5, 190) = 2.32, p =.045, η_p² = .64. Table 3 shows that willingness to drive was generally less under alcohol versus placebo, and increased over time, as BAC declined. DUI offenders and controls were similar in their willingness to drive following placebo, but under alcohol, offenders reported greater willingness to drive compared with controls, particularly early in the time course of the dose when BAC was at peak and just beginning to decline (70 min – 135 min).

Estimated BAC.

A 2 (Group) X 2 (Dose) X 6 (Time) ANOVA of drivers’ BAC estimations revealed a significant dose X time interaction, F(5, 190) = 22.58, p < .001, η_p² = .37. Table 3 shows that both groups estimated higher BACs under alcohol compared with placebo and that estimations decreased over time. No main effect or interactions involving Group were observed, ps > .05.

Discussion

The present study tested the degree to which DUI offenders would display greater impulsive choice and driver risk-taking in response to alcohol compared with non-offender control drivers. The results showed that alcohol impaired drivers’ inhibitory control and increased their impulsive choice behavior and risky driving behavior. Moreover, the results showed that it was DUI offenders whose impulsive choice behavior was specifically affected by alcohol, as control drivers showed no alcohol-induced increase in their impulsive choices. With regard to self-perceptions of intoxication and driver fitness, the results showed that, compared with controls, offenders reported feeling less intoxicated and were more willing to drive after drinking.

The finding that impulsive choice behavior is increased by alcohol in DUI offenders, but not in controls, is a new finding and could highlight a specific vulnerability that is relevant to DUI behavior. When intoxicated, these individuals might demonstrate strong bias towards choices that yield immediate gratification in favor of delayed, but often safer behaviors. One such choice is the decision to drive after drinking with the immediate reward of convenience versus leaving one’s vehicle and making alternative travel arrangements. This finding is also consistent with other laboratory evidence that found college students with self-reported histories of driving after drinking were more impulsive in response to alcohol versus those with no such history (McCarthy et al., 2012).

Although findings of increased impulsive choice behavior in DUI offenders are intuitively appealing, the evidence is based on a single task, and other factors could account for the results. Increased choice for sooner rewards in the TCIP could also indicate an intolerance for waiting or delay aversion, possibly due to boredom, regardless of reward value (Johnson, Sweeney, Herrmann, and Johnson, 2016; Smits, Stein, Johnson, Odum, & Madden, 2013). Choosing smaller rewards has the effect of shortening the duration of the entire test, which could be a desired outcome for a participant regardless of monetary reward, especially in those with boredom proness or aversion to delay. Other discounting models of choice impulsivity do not involve delay, such as probabilistic discounting models. In these models, participants choose between larger rewards with low probability of payout versus smaller rewards with more likely payout. The application of such models to studies of DUI offenders would provide a more complete account of the specific situational factors that contribute to their impulsive choice behavior. The present study found no evidence for group differences in the degree to which alcohol decreased inhibitory control. This finding is consistent with a prior study in our laboratory that also failed to observe a difference between DUI offenders and controls in the degree to which alcohol impaired their inhibitory control (Van Dyke & Fillmore, 2014a). Although considered an important aspect of impulsivity, measures of response inhibition differ considerably from measures of impulsive choice. The cued go/no-go task in the current study measures inhibitory control over the execution of pre-potent response. The inhibitory response is a discrete and extremely brief act of control, operating in the order of milliseconds. The measure of impulsive choice (TCIP) on the other hand, requires explicit decision-making for rewards and tolerating delays in their delivery. Factor analyses of these and other impulsivity tasks find no interdependence between tasks that measure response inhibition and those measuring impulsive choice (MacKillop et al., 2016). Thus, our evidence that offenders can differ from controls in response to alcohol in one facet of impulsivity but not another, is not surprising. It might be that DUI offenders are more sensitive to the impairing effects of alcohol on the ability to delay reward, but not necessarily on the ability to inhibit pre-potent action.

With respect to simulated driving, the study showed that alcohol reliably increased risky driving behavior as indicated by less TTC and increased speed under alcohol compared with placebo. Compared with controls, offenders showed no difference in their sensitivity to the alcohol-induced increase in risk-taking. Given driving record analyses that find DUI offenders generally display risk-taking behaviors while driving (e.g., greater speeding violation and motor vehicle crashes) (e.g., Dahlen & White, 2006; Lajunen & Parker, 2001; Matthews, Dorn, & Glendon, 1991), we expected offenders might display heightened sensitivity to alcohol-induced risk-taking compared with controls. The specific simulated drive scenario used in the study could be a reason for the failure to observe heightened risk-taking in DUI offenders. A limitation commonly reported in simulated driving studies concerns the degree to which driving simulators model driving behavior outside the laboratory. Although measures of simulated driving performance attempt to model more complex, “real-life” activities, ironically they often come under greater scrutiny with regard to their ecological validity than do simple laboratory tasks. A common criticism is that simulated driving might overestimate poor or reckless driver behavior because it does not engender the same degree of driver motivation as actual driving, since there is no actual risk to personal injury. However, despite the lack of injury risk in the laboratory, drivers display little tendency for risk-taking unless there is some explicit incentive to do so. In this study, risk-taking was instigated by providing monetary incentive to complete the drive quickly. Such incentives are necessary to prompt risky driving maneuvers in order to examine how this behavioral tendency can be exacerbated by alcohol. We chose monetary incentives because they are potent reinforcers of risky driving behavior in the laboratory as demonstrated in previous studies (e.g., Fillmore et al., 2008; Laude and Fillmore, 2015). Outside the laboratory, instigation to engage in risky driving is commonplace in many driving situations. Being late and in a hurry to get somewhere is a familiar example of an instigation for a driver to risk-take and there is a strong incentive to speed in order to arrive on time and avoid possible punishment for being late for work or some other important engagement. The proxemics measure of driver risk-taking (TTC) is also well recognized in studies of driver risk behavior (e.g., Taieb-Maimon & Shinar, 2001). However, the study represents only one method of assessing driver risk and other research groups have implemented alternative methods to assess risk-taking behaviors in driving simulations (Burian et al., 2002, 2003; Cohen et al., 1958; Leung & Starmer, 2005). It should also be noted that in a prior study using the same TTC measure reported here, we found drivers overall tended to be more risky as evident by their lower TTC scores compared with the control and offender samples in the current study (Van Dyke & Fillmore, 2017). It is not clear why the samples in the current study tended to be less risky in the drive simulator compared with those in previous studies. Nonetheless, future studies of driver risk-taking in DUI offenders would benefit from a broader assessment of these and other aspects of risk-taking.

It is also important to note a limitation with respect to experimental control over the target BACs in the study. Although the mean BACs were similar in each group, unavoidable variation in rates of absorption and metabolism following oral alcohol administration results in variability between individual participant’s BACs. Although oral dosing has ecological validity, future studies could limit individual variation in BAC by employing intravenous administration of alcohol that uses clamping techniques to maintain steady-state BACs during assessment of risk-taking over longer periods of time. Such clamping techniques would be of particular utility in studies examining acute tolerance to the subjective intoxication in DUI offenders. Given that decisions to drive after drinking often occur as BACs are declining when acute tolerance might accrue, it is important for future research to employ greater control over BAC variation between participants and over time.

In conclusion, there is growing interest in the application of new cognitive behavioral (CBT) approaches that emphasize behavioral planning and evaluation of potential negative consequences, as well as mindfulness techniques to recognize and refrain from impulsive or rash behaviors in DUI offenders (DiStefano & Hohman, 2010; Moore, Harrison, Young, & Ochshorn, 2008). However, without clear identification of the specific behavioral dysfunctions involved in DUI behavior, treatments lack guidance on the particular behavioral dysfunctions that need to be targeted. Little is known about how DUI drivers actually respond to alcohol once drinking has begun. Yet, it is during the intoxicated state that the decision to drive or not to drive is made. Laboratory studies are beginning to show that DUI offenders differ from non-offenders in their acute responses to alcohol. This study identified two alcohol response characteristics of DUI offenders that are key indicators of their lack of risk awareness during intoxication: heightened impulsivity and reduced subjective intoxication. Strategies and treatments to alter these response characteristics in DUI offenders could enhance their risk awareness during the intoxicated state and possibly reduce risk of DUI recidivism.

Public Significance Statement.

This study identified two alcohol response characteristics of DUI offenders that are key indicators of their lack of risk awareness during intoxication: heightened impulsivity and reduced subjective intoxication. Strategies and treatments to alter these response characteristics in DUI offenders could enhance their risk awareness during the intoxicated state and possibly reduce risk of DUI recidivism.