Impulsivity and Inhibitory Control

We report an experiment testing the hypothesis that impuistve behavior reflects a deficit in the ability to inhibit prepotent responses. Specifically, we examined whether impulsive people respond more slowly to signals to inhibit ("stop signals) than nonimpulsive people. In this experiment, 136 undergraduate students completed an impulsiviry questionnaire and then participated in a stop-signal experiment, in which they performed a choice reaction time (go) task and were asked to inhibit their responses to the go task when they heard a stop signal. The delay between the go signal and the stop signal was determined by a tracking procedure designed to allow subjects to inhibit on 50% of the stop-signal trials. Reaction time to the go signal did not van' with impulsivity, but estimated stop-signal reaction time was longer in more impulsive subjects, consistent with the hypothesis and consistent with results from populations with pathological problems with impulse control. Impulsivity is an important construct in several domains. In personality theory, it is part of the construct of extraversion (Eysenck & Eysenck. 1969; Revelle, Humphreys, Simon, & Gilliland, 1980). In childhood psychopathology, it is part of the construct of attention deficit hyperactivity disorder (ADHD) and conduct disorder (Quay, 1988), And in adult psychopathology, it is an important characteristic of psychopathic and sociopathic personalities (Gorenstein & Newman, 1980: Patterson & Newman, 1993), We have found it useful to operationalize impulsivity in terms of the ability to inhibit prepotent courses of action: People who are impulsive have trouble inhibiting action, whereas people who are not impulsive find it easier to do so. We found .support for this hypothesis in studies of children with ADHD, They had more trouble inhibiting prepotent actions than control children with other psychiatric diagnoses and control children with no apparent psychopathology (for a review, see Schachar, Tannock, & Logan, 1993), Moreover, the inhibitory difficulties of children with ADHD were ameliorated by administration of stimulant medication (methylphenidate), which also improves behavioral symptoms of ADHD (including impulsivity: Tannock, Schachar, Carr, Chajcyzk, & Logan, 1989; Tannock, Schachar, & Logan, 1995), The purpose of the present article is to examine the relation between impulsivity and inhibitory control in young adults with no psychiatric diagnoses. THE STOP-SIGNAL PARADIGM Our measure of inhibitory' control comes from the stop-signal paradigm (Lappin & Eriksen, 1966; Logan & Cowan, 1984; Logan, Cowan, & Davis, 1984; Oilman, 1973; Osman, Komblum, & Meyer, 1986, 1990; Vince, 1948), The paradigm involves two concurrent tasks, a go task and a stop task. The go task is a choice reaction time Address correspondence to Gordon D. Logan, Department of Psychology, University of Illinois, 6()3 Easi Daniel St,, Champaign, IL 51820; e-mail: [email protected],uiuc.edu. task that requires subjects to discriminate an X from an O, The stop task, which occurs on 25"* of go-task trials, involves presentation of a tone (a stop signal) that tells subjects to inhibit their response to the go task on that trial. Whether or not subjects are able to inhibit depends on a race between the stop task and the go task: If they finish the stop task before the go task, they inhibit their response to the go task. However, if they fini.sh the go task before the stop task, they fail to inhibit their response to the go task, responding much as they would if no stop signal had been presented. Thus, inhibitory control depends on the latency of the response to the go signal (go reaction time) and the latency of the response to the stop signal (stop-signal reaction time). The race model has been developed formally and shown to be able to account quantitatively for all of the data in stop-signal experiments (Logan & Cowan, 1984: O.sman et al., 1986: for a review, see Logan, 1994), We treat the stop-signal paradigm as a model of inhibitory control of an impulse. We interpret the go signal as the impetus for the impulse, and we interpret the response to the go signal as the prepotent response. The stop signal is a control signal that makes the prepotent response inappropriate. In the real world, the stop signal may come from an external source, like a teacher or a stoplight, or from an internal source, like the person's reevaluation of the situation. In the laboratory, it comes from an external source (the computer) that is under the control of the experimenter. From this perspective, inhibiting when given a stop signal is evidence of good impulse control, and failing to inhibit when given a stop signal is evidence of poor impulse control. According to the race model, poor inhibitory control could result from responding too quickly to the go signal or responding too slowly to the stop signal. Fast responses to the go signal would be executed before the person could respond to the stop signal, and slow responses to the stop signal would allow normally speeded responses to the go signal to escape inhibition. Our studies of ADHD children suggest that slow stop-signal reaction time is responsible for poor impulse control. Children with ADHD inhibit less often than control children even though their go-signal reaction times are longer than control children's. Children with ADHD exhibit disproportionately longer .stopsignal reaction times and therefore inhibit less often (Schachar & Logan, 1990; Schachar, Tannock, Marriott, & Logan, 1995), Thus, one might expect that long stop-signal reaction times are responsible for inhibitory control deficits in other populations with impulse control problems (i,e,, young impulsive adults). MEASURING STOP-SIGNAL REACTION TIME Unlike go-signal reaction time, stop-signal reaction time cannot be measured directly. Subjects either itihibit or fail to inhibit when a stop signal is presented. If they fail to inhibit, stop-signa! reaction time must have been slower than the observable latency of the go-signal response, but it is not clear how much slower it was If they success in inhibiting, stop-signal reaction time must have been faster than gosignal reaction time, but neither the stop process nor the go process 60 Copyright © 1997 American Psychological Society VOL. 8, NO. 1, JANUARY 1997 PSYCHOLOGICAL SCIENCE Gordon D. Logan. Russell J. Schachar. and Rosemary Tannock •rovides an observable re.sponse with a measurable latency. Somehing beyond direct observation is required. The race tnodel of the stop-signal paradigm provides at least three different ways to estimate stop-signal reaction time (see Logan, 1994). The methods are relatively complex and require a deep understanding of the race model (at the level of reaction time distributions). The purpose of this article, in part, is to introduce a fourth method for estimating stop-signal reaction time, also derived from the race model, that is easier to compute and much easier to understand than the other methods. In stop-signal experiments, re.searchers vary the delay between the stop signal and the go signal (stop-signal delay) in order to handicap the race in favor of one process or the other. Most often, stop-signal delays are selected at random from a fixed set that is held constant throughout the experiment (e.g., Logan & Cowan, 1984), but many researchers let them vary dynamically, contingent on the subject's behavior (e.g., Osman et al., 1986, 1990; Schachar & Logan, 1990; Schachar et al., 1995). The new method for estimating stop-signal reaction time uses a tracking procedure in which stop-signa! delay changes after ever)' stop-signal trial, increasing by 50 ms if subjects inhibit and decreasing by 50 ms if they respond. This tracking procedure, introduced by Osman et al. (1986, 1990), converges on a stop-signal delay at which signals inhibit 50% of the time. That delay is important because it represents the amount of handicapping necessary to " t i e" the race. At that delay, the stop process and the go process finish at the same time, on average, and the one that happens to win on a particular trial depends on random variation. Thus, that delay is the average point in time at which the stop process finishes, and that information can be used to estimate stop-signal reaction time. The estimation of stop-signal reaction time is illustrated in Figure. 1. The race depends on three quantities—go reaction time, stop-signal reaction time, and stop-signal delay—and the experimenter knows two of them. Moreover, because subjects inhibit 50% of the time at the critical delay, stop-signal reaction time plus stop-signal delay must equal mean go reaction time. Stop-signa) reaction time can be calcuGO-SIGNAL REACTION TIME STOP-SIGNAL DELAY STOP-SIGNAL REACnON TIME