Acta Psychologica 42 (1978) 263-276 0 North-Holland Publishing Company

EFFECTS OF AMPHETAMINE AND BARBITURATE SERIAL REACTION TASK UNDER PACED AND SELF-PACED

IN A

CONDITIONS*

H. W. FROWEIN and A. F. SANDERS Institute for Perception

TNO, Soesterberg,

The Netherlands

Received July 1977

Effects of drug conditions (barbiturate, amphetamine, placebo) were investigated in a serial choice RT-task under three pacing conditions: self-paced, paced-fast (approx. 3 signals/set) and paced-slow (approx. 1 signal/set). Drug treatment and pacing condition were varied in a factorial design with 9 male Ss. Drugs affected the reaction time but not the error scores. Barbiturate slowed down RT and this effect was most prominent in pacedfast. Amphetamine, on the other hand, speeded up the RT but it had Bo effect in pacedfast. To account for these differential effects of the two drugs, it was postulated that barbiturate counteracts the arousing effect of time pressure whereas amphetamine speeds up reaction when the task is not arousing. This is consistent with Trumbo and Gaillard (1975) but it contradicts predictions derived from a theory by Broadbent (1971).

Introduction

This study is part of a program to investigate the effects of a barbiturate and an amphetamine derivative on performance in relatively simple tasks such as a contmuing reaction task (Trumbo and Gaillard 1975) or a tracking task (Truijens et al. 1976). The research strategy, as outlined by Sanders and Bunt (197 l), is to carry out experiments which investigate the relationship between these drug effects and the effects of certain task variables. The underlying idea is that a description of the effects of a drug in terms of task variables will lead to conclusions which are more precise and also more generally applicable than can be arrived at by employing a battery of relatively unrelated tasks.

* The investigations were supported by the Foundation for Medical Research FUNGO which is subsidized by the Netherlands Organization of Pure Research (ZWO).

264

H. W. Frowein, A. F. Sanders/Effects

of drug conditions

The present experiment investigates the effects of these drugs in a serial reaction task under conditions of self-paced stimulus presentation (where the stimulus is triggered by the subject’s previous response) and paced stimulus presentation (where the stimulus presentation rate is preprogrammed and independent of the subject’s response rate). This is of methodological as well as theoretical interest. Because all studies of the effects of drugs on performance use either some sort of paced or self-paced mode, it is useful to investigate these drug effects under different pacing conditions. In addition, the experiment is of theoretical interest because two contradictory sets of predictions can be derived from the literature. Firstly, it is consistent with the arousal theory of Broadbent (197 1) that a barbiturate will affect performance in self-paced rather than paced tasks, while the reverse will be true for amphetamine. The theory proposes that arousal involves two mechanisms: a lower mechanism which constitutes the traditionally assumed arousal system (presumably showing an inverted U-shaped relation with performance), and an upper mechanism which monitors the parameters of the lower mechanism to keep it at an optimal level. On the basis of this arousal theory as well as some clinical and physiological studies and some performance experiments reported by Mirsky and Rosvold (1960), Broadbent suggested that barbiturates affect the upper mechanism whereas amphetamines affect the lower mechanism. Thus, he reasoned that a barbiturate would show its effect more readily in a self-paced task: momentary lapses of the lower mechanism would result in extra slow responses and the barbiturate would prevent the upper mechanism, from effectively compensating for this. In a moderately fast paced task, however, each long RT will imply a late start for the next response, and there will be no opportunity for compensation by the upper mechanism. An alternative set of predictions can be derived from Trumbo and Gaillard (1975) who found that a barbiturate slowed down RT to a 70 dB tone but had no effect on RT to a visual stimulus, whereas an amphetamine speeded up visual RT but had no effect on auditory RT. To interpret these data Trumbo and Gaillard started from a proposition by Sanders and Wertheim (1973), later confirmed by Sanders (1975, 1977), that a loud auditory tone might exert an immediate arousing effect, while a visual signal might not. It was suggested that a barbiturate may act to reduce the effect of immediate arousal. Amphetamines, on the other hand, would be beneficial when a task is such that

H. W. Frowein, A. F. SandersjEffects

of drug conditions

265

it is necessary to maintain a continuing preparatory state, while the stimuli themselves do not invoke immediate arousal. Now it seems a reasonable proposition that immediate arousal is not only evoked by imperative stimuli but also by task properties like time-pressure (Kahneman 1973). Hence, a barbiturate might reduce the arousing effect of a paced task and, consequently, degrade performance. Similarly, amphetamines might be mainly beneficial in a visual self-paced task, because response preparation needs to be continuously maintained while neither the stimuli nor the task context are arousing. In summary, when a fastly-paced task is compared with a self-paced task, Broadbent’s theory predicts a barbiturate effect in self-paced rather than paced. Conversely, an explanation along the lines of Trumbo and Gaillard would predict the opposite. When a slow rather’ than a fast pacing tempo is used, the two explanations also make opposite predictions. It follows from Broadbent’s theory that at a slow pacing rate without time pressure, the upper mechanism could play a role again so that the barbiturate effect should be less and the amphetamine effect should be greater. Contrarily, an explanation along the lines of Trumbo and Gaillard predicts that the slow pacing tempo will be less arousing so that the barbiturate effect should be smaller and the amphetamine effect should be greater than in the case of a fast pacing tempo.

The experiment Method Subjects

The Ss were 9 male students from the University of Utrecht with an age ranging from 20 to 26 years. Ss were paid Dfl45,a day, and an additional bonus of about Dfl lo,to Dfl 12,-, depending on the performance during that day. Before the experiment, each S was given a medical examination including urine and blood tests. He also signed an agreement to avoid alcoholic beverages during the 24 hours before each test day. Apparatus

The S was seated at a sloping desk in a soundproof room with dim ceiling illumination. Three pairs of lights and response keys were mounted on the desk in a highly compatible arrangement with each key about 5 cm below its corresponding light. The keys were conveniently operated with the three middle fingers of S’s preferred hand.

266

H. W.Frowein, A. F. Sanders/Effects of drug conditions

The three stimuli were presented in random sequence, with the restriction that no repetitions occurred. In the self-paced condition a stimulus presentation was stopped by the subject’s response, and the next stimulus appeared 100 msec later. Thus, the stimulus duration for the self-paced condition always equalled the RT and there was an interval of 100 msec between offset and onset. For the pacedfast and paced-slow conditions, the stimulus duration was predetermined by the setting of the apparatus, and there was no interval between offset and onset. The programming and timing of the signals and the measurement of the RT’s was performed by the PSARP system (Van Doorne and Sanders 1968). The response data were first registered on tape for subsequent analysis on a PDP- 11 computer. In addition the data were printed out on line to allow the experimenter to have immediate feedback about the subject’s performance. Treatment Treatments consisted of an amphetamine derivative (20 mg. Phentermine HCL), a barbiturate (100 mg. Pentobarbital Sodium) and a placebo. The preparation and administration of the drugs was based on pharmaco-kinetic research on barbiturate by Breimer (1974) and on amphetamines by Vree (1973). To achieve a relatively constant plasma concentration during the post-treatment experimental sessions, treatments were administered by means of a suppository, and testing was started 100 minutes after administration of treatment. Allocation of the drug treatment was ‘double blind’ in the sense that both the Ss and the experimenter knew that an amphetamine, a barbiturate and a placebo would be administered on three different test days, but they did not know which drug would be administered on which day. Design and procedure A within-subjects design was used. For each S the program consisted of one training day and three experimental days at weekly intervals. Treatment conditions were varied between experimental days, while the three task conditions (self-paced, paced-fast and paced-slow) were varied within experimental days. To counterbalance order effects, drug treatments were arranged in a 3 X 3 Latin square, and each of its cells consisted again of another 3 X 3 Latin square in which the order of the task conditions was counterbalanced. During each experimental day two Ss were alternatively run, one S being in the test booth while the other S was in a rest-period. This meant that one S started the program at about 9.00 and finished at about 15.15, whereas the other S started at about 9.30 and finished at about 15.45. The time-line of the program for one S on one experimental day is schematically represented in fig. 1. The pre-treatment phase consisted of two 2-min, self-paced tasks: one with ‘accuracy’ and one with ‘speed’ instructions. For both sets of instructions, the Ss received a 0.5 cent bonus for each correct response, but during ‘accuracy’ there was also a 1 cent fine for incorrect responses, while there were no fines during the ‘speed’ condition. The data obtained during the pre-treatment test served two purposes. Firstly, speed-accuracy trade-off functions were computed for each S for each experimental day to serve as a basis for computing ‘difference scores’ which

H. W. Frowein, A. F. Sanders/Effects

0 00

02

100

610mtn

of drug conditions

2 00

0

10

267

3 00

20

30

LOO

LO

50

500

6 00 h

60m1"

Fig. 1. Time-line for each experimental day (top) and for session (bottom).

constitute a measure of performance incorporating both errors and reaction times (see ‘Treatment of data’). Secondly, the pacing tempo for paced-fast and paced-slow was computed on the basis of reaction times obtained in the self-paced ‘accuracy’ pretest, i.e. the interval between stimulus onsets was computed as follows: for paced-fast, 70th percentile of the RT distribution of S.P. accurate +lOO msec; and for paced-slow, 3 X 70th percentile of the RT distribution of S.P. accurate +lOO msec. These pacing tempos were computed separately for each subject for each experimental day. In practice, this amounted to about one presentation per set for paced-slow and about three presentations per set for paced-fast. During the post-treatment part of the experiment, the S worked through three experimental sessions, i.e. one session for each task condition. As fig. 1 illustrates, the program for each task condition consisted of a 25min ‘work’ period which was preceded by a 2-min pre-test and followed, after a 30-min rest period, by a 2-min post-test. This procedure was previously used by Sanders and Hoogeboom (1970) to allow the separate assessment of the effects of skill acquisition and reactive inhibition which might occur during time-on-task during the 25min work-period. The assumption of this procedure is that improvement from the pre-test to the post-test can be interpreted as a skill acquisition effect, and that improvement between performance at the end of the work-period and the post-test is attributable to release from the reactive inhibition which might build up over the 25-min period. In addition, the effects of time-on-task themselves were assessed by comparing performance on the first and the second half of the 25-min work-period. The pay-off schedule during the post-treatment sessions was similar to that of the pre-treatment accuracy test in that there was a bonus of 0.5 cent for correct responses and a fine of 1 cent for incorrect responses. In addition, Ss were also told that 1 cent would be deducted for each response which was ‘too long’.

268

H. W.Frowein, A. F. Sanders/Effects of drug conditions

Treatment of data A particular problem associated with paced stimulus presentation is that, because some RT’s might be longer than the onset-onset interval, one has to adopt a decision rule to distinguish between a late response to a stimulus (S,) and a response to the next stimulus (S,). The following rule was decided upon: if (a) there was r .dsponse to Si during the Si-S2 interval, and (b) the first response following SZ was incorrect to SZ but correct to Sr, that particular response would be counted as a RT to Sr. It must be realized, however, that any decision rule will result in some erroneous classifications. In this experiment, an incorrect late RT to Si might be wrongly classified as a correct RT to Ss, and, vice versa, an incorrect RT to Ss might be classified as a correct RT to Si. There is no reason, however, to presume that these cases might occur frequently or bring about a systematic bias in the data analysis. Several measures of performance were computed, i.e. (a) To assess the effect of drug treatment and task variables upon the RT-distribution, the 25th, 50th and 75th percentile were computed. (b) Effects of accuracy were assessed by taking the percentage of incorrect RT’s (errors). (c) To have a measure of performance in which both speed (RT’s) and accuracy (errors) are represented, ‘difference scores’ were computed on the basis of the tradeoff functions which were obtained for each subject for each day of testing on the basis of their performance on the self-paced accurate and self-paced speed tasks during the pre-treatment phase of the experiment. The method used was essentially the same as used by Wagenaar and Stakenburg (1975) who based their method on the finding of Pew (1969) that for a variety of experimental situations the trade-off function could be described by a linear relation between median RT and log odds (correct responses/errors). Given this linearity, one can determine an individual trade-off function by computing the median RT and the log odds for self-paced accurate and self-paced speed and drawing a straight line connecting the plots of these two conditions. This procedure was followed, and subsequently the difference scores were computed by subtracting the RT-values predicted by the trade-off functions from the RT-values obtained in the post-treatment conditions. Thus, difference scores are positive for longer RT’s in the post-treatment conditions. (d) The subject’s ability to keep up during paced conditions, was assessed by computing the percentage of missed responses. Results The main body of data are those obtained during the 25-min work-period. The median RT’s and the difference scores obtained during this period are pictured in figs. 2 and 3 respectively, while the percentages of errors and missed RT’s are shown in figs. 4 and 5. The effects of skill acquisition which might occur during the 25-min work-period can be assessed by comparing performance on the pre-test and the post-test. Separate analyses of variance were carried out for the difference scores, the median RT’s, the errors and the missed responses, and in none of these cases there was a significant

H. W. Frowein, A. F. Sanders/Effects

of drug conditions

269

FAST

I

lb-----e L

SLOW 1

J

1

I

I

II

I

II

half-periods

Fig. 2. Median reaction times obtained during the 25-min work periods, as a function of timeon-task (first half-period vs. second half-period), pacing condition (self-paced, paced-fast, pacedslow) and treatment condition (barbiturate o-o, amphetamine o-o and placebo +- - -+).

effect of skill acquisition or of an interaction between skill acquisition and the other independent variables. This indicates that subjects were sufficiently well-trained to preclude significant learning effects within sessions. Similarly, the analyses of variance showed no significant order effects between consecutive testing days or between consecutive sessions on the same day, and there were no significant interactions of these order effects with drug treatment, pacing condition or time-on-task.

SELF

FAST

1

,

I

half-

1

periods

Fig. 3. Differences obtained by subtracting the median RT-values predicted by pre-treatment tradeoff functions from the median values obtained in post-treatment conditions. Data were obtained during the 25min work periods, and experimental variables were time-on-task (first haif-period VS. second half-period), pacing condition (self-paced, paced-fast, paced-slow), and treatment condition (barbiturate o-o, amphetamine 0-o and placebo +- - -+).

270

H. W.Frowein, A. F. Sanders/Effects of drug conditions

Time-on-task as measured by the difference between the first and the second half of the 25-min work-period had a significant effect on the median RT at F(1,32) = 152.63, p