time of day variations in driving performance

Accid. Anal. and Prm.. Vol. 29, No. 4, 431-437. 1997 pp. 8 1997 Elsevier Science Ltd

All rights reserved. Printed in Great Britain OOOI-4575.‘97 s17.00 + 0.00

PII: SOOO1-4575(97)00022-5

TIME OF DAY VARIATIONS IN DRIVING

PERFORMANCE

MICHAEL G. LENNY*, THOMAS J. TRIGGS and JENNIFER R. REDMAN

Department of Psychology, Monash University, Wellington Road, Clayton 3168, Australia

Abstract-Numerous factors may contribute to the 24-hour pattern of automobile accidents. One factor may be a time of day variation in driving ability. In the present study, I I male subjects operated a driving simulator for 30 minutes at six times of day. Subjects were instructed to maintain a stable position in the left-hand lane and to drive at a constant speed of 80 km/hour. In addition subjects performed a secondary reaction time task. Subjective mood was measured at the beginning and end of each session. Driving performance was measured in terms of the mean and standard deviation of lateral position and speed. The mean and standard deviation of speed varied significantly across the day for both curved and straight segments. Reaction time was also affected by time of day. Performance was more impaired at 0600 and 0200 hours, with improvements in driving performance between 1000 and 2200 hours and an early afternoon dip. These results suggest that driving performance is subject to diurnal variations. Of particular importance is the result that impairments in driving performance in the early afternoon are of a similar magnitude to those occurring in the late evening and early morning. 0 1997 Elsevier Science Ltd.

Keywords-Driving, Time of day, Secondary task

INTRODUCTION

Levels of performance and psychological measures, such as subjective mood, have been found to exhibit reliable variations across the day (Colquhoun, 1981; Folkard, 1983). It has yet to be established whether driving performance also varies across the day.

Simple rural driving entails a range of tasks, including vigilance, reaction time (RT) and tracking. Previous studies have found that performance on these components, when executed individually, also varies with time of day (Blake, 1967; Buck, 1977; Craig et al., 1981). Performance on these individual tasks was shown to improve across the waking day

and decline in the late evening and early morning

hours. In light of these findings it is possible that when

these tasks are combined, as in a simple rural driving scenario, that a time of day difference in driving performance may be evident. Further support for this prediction is gained from examining the rhythm in driving accident data. Even when two of the major

factors, traffic volume and alcohol consumption, are taken into account, reviews of driving accident data report a prominent time of day effect, with highest accident rates between ca 0300 and 0500 hours, and

*Corresponding author.

a much smaller secondary peak between 1400 and 1500 hours (Lisper et al., 1979; Langlois et al., 1985; Schwing, 1990; Summala and Mikkola, 1994). The secondary peak in the number of accidents corresponds with the already established post-lunch dip in performance (Smith, 1989).

To ascertain the source of this accident rhythm, it is critical to measure the demand of the driving task across the day. To this end, a secondary task can be administered concurrently with the driving task. It is possible that driving may be more demand-

ing at particular times of day, and therefore requiring more attentional resources. A secondary task will

help to ascertain this because secondary task performance is dependent upon the amount of residual resources, and therefore should reflect the resource demands and task difficulty associated with the driving task (Wickens, 1984; Eggemeier, 1988). This claim, of course, assumes that performance on a task is dependent upon the amount of resources allocated to it, that the total amount of resources available is fixed, and that resources are derived from a single source (Wickens, 1984; Gopher and Donchin, 1986; Meshkati and Loewenthal, 1988 ).

Secondary tasks have been used as measures of driving performance. For example, Lisper et al. ( 1986) reported that changes in RT mirrored changes in wakefulness, such that the length of time spent

431

432 M. G. LENNY et al.

driving before falling asleep at the wheel could be predicted adequately by secondary RT. Riemersma et al. (1977) reported that between the fifth and eighth hours of continuous driving at night, errors on a secondary vigilance task increased from ca 0 up to 40% errors. Brown (1962) tested auditory RT while driving in heavy and light traffic conditions. It was expected that driving in heavier traffic would involve an increased processing load, and therefore result in lower residual capacity. As expected, more errors were found on the secondary RT task while driving in heavier traffic.

However, secondary task performance while

driving across the day has not been defined clearly. Lisper et al. ( 1979) examined secondary auditory RT during 3-hour driving episodes at four times of day: 0300; 0900; 1500; and 2100 hours. Although there was no significant difference in RT between the four

times, there was a trend suggesting that RT was slowest at 0300 hours.

There has been no study which has systematically examined driving performance under constant condi-

tions using a simulator across the entire circadian cycle, and has examined secondary task performance at the same times. The present study examined driving performance and RT on a secondary task at six

4-hourly intervals across the 24-hour day.

METHODS

Suhject,F Eleven male subjects participated in this study

(age range 21-26 years, mean age of 23.6 years). These subjects were required to have possessed a

driver’s licence for between 3 and 8 years, and were required to be neither morning- nor evening-type people according to the Horne and ijstberg (1976)

Morningness-Eveningness Questionnaire. High

caffeine users, defined as those who consumed an

equivalent of more than six cups of coffee per day, were excluded.

The drilling tusk The task utilized a fixed base Systems

Technology Incorporated (ST1 ) model driving simulator. Computer generated graphics were displayed on a 14” VGA multisynching monitor (Rosenthal et al., 1991). Normal operation of the simulator required the use of the steering wheel, and the acceler- ator and brake pedals. The driving scenario appeared as a rural two-lane highway, with flat green surrounds and background mountains. Each session involved 30 minutes of continuous driving, comprising 3 minutes practice, and then six 4.5 minute tracks. Each track contained four left curves, four right

curves, two S-bends and two oncoming vehicles. Curves were either 200 or 400 m in length, and were separated by straight sections of road between 200 and 600 m in length. Within each track the order of the left curves, right curves and the S-bends, was varied. The presentation of tracks was counterbalanced across each session.

The ST1 simulator also contained an additional task which could be used to measure probe RT while driving. Two small grey squares on the upper left and right corners of the simulator screen contained a red diamond. At pre-specified points along the

track, the diamond was momentarily replaced by a shape resembling a horn. This event occurred six times during each track, at either the beginning, midpoint or end of a curve or straight section of road. Subjects responded to this event by depressing a pedal with their left foot. This response produced a horn-like sound. Acoustic signals simulated skid- ding on curves, approaching vehicles and the sound

of the engine.

Procrduw Subjects were recruited via pamphlets placed

around Monash University. Prior to the commence- ment of the experiment, subjects completed a 30-minute practice session. Subjects then attended six experimental sessions, one per week, at each of the following times: 0600; 1000; 1400; 1800; 2200; and 0200 hours. The allocation of subjects to experimental sessions was counterbalanced. Subjects were asked to abstain from eating for 2 hours prior to each session, because the presence of food in the stomach may alter alertness levels and performance (Smith and Kendrick, 1992). In addition subjects abstained from caffeine and nicotine for 2 hours prior, and from

alcohol and other psychotrophic drugs for 24 hours prior to each session. Subjects were also asked to retire between 2300 and 2330 hours the night before, and rise between 0730 and 0800 hours on the morning of each experimental session. For the 0600 hours session, subjects were required to wake up at 0500 hours. Subjects were instructed not to sleep or nap before the 0200 hours session.

Given that vehicles in Australia drive in the left-

hand lane, subjects were asked to drive in the left- hand lane, but in addition were asked to drive as close as possible to 80 km/hour at all times, and to respond to the secondary task when required. Subjects were never prompted to maintain the desired speed. The lighting and display presentation was the same in all conditions, both night and day. Curvature angles were set such that safe negotiation at 80 km/hour was possible, and therefore no slowing was necessary. For each individual curve and straight

Time of day and driving performance 433

segment, the simulator recorded the mean and standard deviation of lateral position (m), the mean and

standard deviation of speed (km/hour), and the RT (seconds) to the secondary task. In the event of a crash, a short tone was presented, and then the simulation continued from the point of the crash.

At the beginning and end of each session, subjects were asked to complete three visual analogue

scales (VAS) to assess subjective alertness, sleepiness and motivation. Subjects received Aus$21 for their participation.

RESULTS

For the purposes of the analyses, each 4.5-minute track was defined as a block. Each session therefore contained six blocks, which were analysed separately. Driving performance measures were consequently analysed in six blocks per session, whereas subjective moods were analysed as two blocks per session (either before or after the driving test). Performance and subjective measures were analysed using two-way repeated measures ANOVAs; the factors being time of day, and block number within each session. Performance was analysed separately on straight and

curved segments, as driving around curves is associated with higher mental workloads than negotiating straight sections (Hancock et al., 1990). Post-hoc analyses were performed using the Tukey test (Keppel, 1991).

Most of the significant results reported did not violate the sphericity assumption, as indicated by the Mauchly Sphericity Test (Keppel, 1991). If the assumption was violated, the data were subjected to

the Huynh-Feldt correction (Huynh and Feldt, 1976). Only significant results which passed these tests are reported here.

Drbing task perjbrmance measuws Time of duy &XY. Significant time of day varia-

tions were found for mean speed on both curved and straight segments [F( 550) = 2.83, p ~0.05; F( 5,50) = 3.52, p <0.05]. Figure 1 illustrates curve mean speed, with the line at 80 km/hour representing the speed that subjects were asked to maintain. Poorest performance, being the furthest from 80 km/hour, occurred

at 1400 hours (~~0.05). Similarly to mean speed for curved segments, mean speed for straight segments was also poorest at 1400 hours (p < 0.05 > (see Fig. 1).

Standard deviation of speed (km/hour) also varied significantly across the day for both curved and straight segments [F( 5,50) = 2.93, p <0.05; F( 5,50) = 2.89, p < 0.051. Standard deviation of speed on curves was highest at 0600, 1400 and 0200 hours (see Fig. 2). Performance was better at 2200 hours than at any

81.5

81 I

80 5

E- 80

z _ 795

z % 79

CT) 6 78 5

’ 78

77 5

77 ,

O&O Id00 1400 IsbO 22.00 oioo

Time of Day

Fig. 1. Mean speed on curved and straight segments across time of day, collapsed across block (*SE).

2

Fig. 2.

0600 1000 1400 1800 2200 0200

Time of Day

Standard deviation of speed on curved and straight segments across time of day, collapsed across block (i SE).

other time, better at 1000 hours than at 0600, 1400 and 0200 hours, and better at 1800 hours than at 1400

and 0600 hours (p < 0.05). In addition, performance was improved at 2200 hours than at 0600, 1400 and 0200 hours for straight segments (p ~0.05).

The secondary RT task also yielded a significant time of day variation [F(5,50) =2.84, p <0.05]. Figure 3 shows that the pattern of RT across the 24-hour cycle is similar to that observed for standard deviation of mean speed for both curved and straight segments of road. Reaction times were significantly

434 M. G. LENNY et al

lateral position measures indicate a position that is closer to the left edge of the road.

3 --.__ *

,’

0600 1000 1400 1800 2200 0200

Time of Day

Fig. 3. Probe RT across time of day, collapsed across block (+ SE).

higher at 0600, 1400 and 0200 hours than at 1000,

1800 and 2200 hours (p < 0.05).

Block effects Some of the performance measures did yield

significant block effects, although the patterns across block were inconsistent. No time by block interactions were found for any of the performance measures, indicating that variations across block occurred at all times of day.

The only results of note were for lateral position on curved and straight segments, where a steady shift in lateral position towards the left edge of the road was observed across blocks [F( 5,50) = 8.01, p < 0.05; F(5,50)=2.71, p<O.O5], as seen in Fig.4. Higher

1 I-

1 05.-

f :

1 2 3 4 5 6

Block Number

Fig. 4. Lateral position for curved and straight across block, collapsed across time of day (&SE).

Subjective measures

Subjective alertness, motivation and sleepiness all showed significant variations across time of day

[F(5,50)=6.45, p<O.OOl; F(5,50)= 19.32, p<O.OOl; F( 5,50) = 7.02, p < 0.0011. Figure 5 illustrates that alertness was low at 0600 hours, rose to a peak at 1400 hours, and then steadily declined to a trough at 0200 hours. Subjective motivation displayed a similar trend, and subjective sleepiness showed a reversed trend. Furthermore, alertness was significantly lower after the completion of the session [F( l,lO)= 16.62, p<O.O5]. Similarly, motivation was lower at the end of each session [F( 1,10)=23.25, p<O.OOl], whilst subjective sleepiness was higher at this time [F(1,10)=10.19, p<O.O5]. There were no time by block interactions.

DISCUSSION

This study demonstrated that some aspects of driving performance and subjective mood varied across the day. In particular, performance was more impaired at 0600 and 0200 hours, with improvements in driving performance between 1000 and 2200 hours and an early afternoon dip. Furthermore, deteriorations in driving performance and subjective mood within sessions were observed.

Time of duy eflects

Mean speed on both curved and straight segments of track was found to vary significantly across the day, with poorest performance (being furthest

01 : 0600 1000 1400 1800 2200 0200

Time of Day

Fig. 5. Subjective alertness before and after each session. across time of day (100 corresponds to maximum alertness) (i SE).


from 80 km/hour) occurring at 1400 hours, and better

performance at 0600 and 0200 hours. However, this result should be considered with the results for the standard deviation of speed, showing poor performance at 0600 and 0200 hours. The results suggesting better maintenance of mean speed at these times, therefore, do not appear very meaningful, given that speed was actually significantly more variable, and less predictable, at these times. For this reason, it is perhaps more meaningful to focus upon the standard deviation of speed.

RT was also influenced by time of day. The pattern of performance across the 24-hour cycle is similar to that observed for standard deviation of speed. Reaction times were prolonged at 0600, 1400 and 0200 hours. Again, these are the times at which poorer performance would be expected. Arousal is lowest at 0600 and 0200 hours, and the post-lunch dip, which will be discussed shortly, may account for

the increased RT at 1400 hours. It is, however, difficult to speculate about how

resources are being utilized across the 24-hour cycle. If we assume that the amount of resources available varies as a function of basal arousal level (Kahneman, 1973), then the similar patterns in performance observed for RT and standard deviation of speed could easily be accounted for, provided that subjects were adopting a similar resource allocation strategy at all times of day. Perhaps the time of day variations were not related to resource availability. Instead,

arousal may have influenced the strategy adopted by subjects, as evidenced with other tasks (Monk and Leng, 1982).

There was a trend (p =O.OS) suggesting that standard deviation of lateral position on curves was influenced by time of day. This trend was in the same direction as the results for the standard deviation of speed and RT, with poorer performance at 0600, 1400 and 0200 hours. It was surprising that standard deviation of lateral position on curves was not found to be the most sensitive measure of time of day variations in driving performance. Deteriorations in lateral position and increasing deviations in lateral

position have been found to be characteristics associated with driver fatigue, resulting from prolonged periods of driving (Dureman and Boden, 1972; Riemersma et al., 1977; Haworth et al., 1988). Lateral deviation has also been found to be a sensitive measure of driving experience in simulators, with inexperienced drivers producing the greatest deviation (Blaauw, 1982; Drummond et al., 1992). It could, therefore, have been expected that standard deviation of lateral position would be a sensitive measure of time of day variations in driving performance. Increasing deviation in lateral position is associated

with the fatigued driver. Perhaps if the duration of driving in the present experiment was increased from

27 minutes, and the curved segments made more challenging, then this result may have reached significance.

One of the most interesting aspects of the results of this study is the predominance of performance dips at 1400 hours. Such dips were observed for all of the performance measures. Similar ‘post-lunch dips’ in performance have been identified for a range of performance measures (Blake, 1971; Craig and Phil, 1986; Smith and Miles, 1986). In light of findings suggesting that the post-lunch dip in performance is more prevalent in tasks requiring longer periods of sustained attention (Folkard, 1983; Craig and Phil, 1986), such as that used in the present study, the post-lunch dip seems a likely explanation for these results. Interestingly, the performance dips found at 1400 hours correspond with the early afternoon

increase in automobile accidents (Lisper et al., 1979; Langlois et al., 1985; Schwing, 1990; Summala and Mikkola, 1994). It is also interesting to note that despite an increase in sleep propensity in the early afternoon (Clodore et al., 1986), it is common for subjective alertness to peak at 1400 hours, as in this study, despite dips in performance at this time (Folkard, 1983; Monk et al., 1983).

Block effects Significant block effects were observed for mean

lateral position for both curved and straight segments. In both cases, mean lateral position shifted more towards the left-hand edge of the left lane as the session progressed. Similar trends were obtained by Riemersma et al. (1977) in their on-the-road study. In particular, they observed shifts in lateral position during an S-hour driving session. Riemersma et al. (1977) found that the fatigued driver tended to adopt a lateral position that became closer to the right- hand edge of the road. It should be noted that this study was conducted in The Netherlands where vehicles travel on the right-hand side of the road, whereas in Australia vehicles travel on the left-hand side of the road. The similar trends found by Riemersma

et al. (1977) on-the-road, and those of the present study on the simulator, provide further validation for using the simulator.

Although subjects were only required to drive for 27 minutes in the present study, it may be possible to attribute these results to ‘time on task’ fatigue. It has been found that, from the onset of driving, fatigue effects are observed much sooner on a driving simulator than when driving an actual vehicle (Haworth et al., 1988). This is believed to occur as a result of the monotony and consequent boredom associated

436 M. G. LENNY et al

with simulator driving. Given that subjects were significantly more sleepy and less alert at the comple-

tion of each session, it appears reasonable to suggest that fatigue was a pivotal factor in the degradation of performance throughout the sessions. As stated

earlier, lateral position measures have been found to be the most sensitive measures of fatigue. Hence it is not surprising that these measures produced significant block effects in the present study.

Conclusions As with many performance studies involving

collection of data during the night, there is the option of keeping subjects awake until the testing session,

or letting them sleep and then waking them for the session. Either way, the performance measures are masked by sleep deprivation or sleep interruption (Colquhoun, 1981). There is therefore a possibility

that the results in the present study were confounded by partial sleep deprivation. The circadian variations in driving performance will be investigated further in

a subsequent study using sleep deprivation. This study has found that driving performance

is affected by time of day. Further confirmation is

provided for the strong relationship between driving at night and impaired driving performance, and con-

sequently accident risk. The significance of the early afternoon period and associated dips in driving performance is also highlighted. Although many drivers consider the effects of drugs and environmental factors on driving ability, the effects of time of day are rarely considered. These results are of particular relevance and importance to those involved in the transportation industry. Where possible, strategies should be developed to minimise driving during the late night and early morning hours, and to encourage extra care in the early afternoon.

Ackno,2/edgement-We would like to thank the Monash University Accident Research Centre (MUARC) for the use of the driving simulator.

REFERENCES

Blaauw, G.J. (1982) Driving experience and task demands in simulator and instrumented car: a validation study. Hum. Fuctors 24, 473-486.

Blake, M.J.F. (1967) Time of day effects on performance in a range of tasks. Psychon. Sci. 9, 349-350.

Blake, M.J.F. (1971) Temperament and time of day. In Biological Rhythms and Human Perjtirmance, ed. W.P. Colquhoun, pp. 109-149. Academic Press, London.

Brown, I.D. (1962) Measuring the ‘spare mental capacity’ of car drivers by a subsidiary auditory task. Ergonotnics 5, 247-250.

Buck, L. (1977) Circadian rhythms in step-input pursuit tracking. Ergonomics 20, 19-3 1.

Clodor&, M., Foret, J. and Benoit, 0. (1986) Diurnal varia-

tion in subjective and objective measures of sleepiness: the effects of sleep reduction and circadian type. Cllro- nobiol. Int. 3, 255-263.

Colquhoun, P. ( 1981) Rhythms in performance. In Hand- book of Beharioural Neurobiology: Biological Rhythms, ed. J. Aschoff, pp. 333-349. Plenum Press, New York.

Craig, A. and Phil, D. 1986) Acute effects of meals on perceptual and cognitive efficiency. Nutrition Rev. 44, Suppl., 163-171.

Craig, A., Wilkinson, R.T. and Colquhoun, W.P. ( 1981) Diurnal variation in vigilance efficiency. Ergonomics 24, 641-651.

Drummond, A.E., Triggs, T.J. and Schulze, M.T. ( 1992) Basic driving performance as a function of driving experience. Paper presented at IMAGE IV Cottftirence, Scottsdale, AZ, 1992.

Dureman, E.I. and Boden, C. (1972) Fatigue in simulated car driving. Ergonomics 15, 299-308.

Eggemeier, F.T. (1988) Properties of workload assessment techniques. In Human Mental Workload.. Adwnces in Psychology, eds P.A. Hancock and N. Meshkati. Vol. 52, pp. 41-62. Elsevier Science, North-Holland.

Folkard, S. (1983) Diurnal variation. In Stress and Fatigue in Human Pe~f+mance, ed. G.R.J. Hockey, pp. 245-272. Wiley, New York.

Gopher, D. and Donchin, E. (1986) Workload-an exami- nation of the concept. In Handbook of Perception and Human Perfbrmance: Cognitive Processes und Perfor- mance, eds K.R. Boff, L. Kaufman and J.P. Thomas, pp. l-49. Wiley, New York.

Hancock, P.A., Wulf, G., Thorn, D. and Fassnacht. P. ( 1990) Driver workload during differing driving maneuvers. Accid. Anal. Prev. 22. 28lL290.

Haworth, N.L., Triggs, T.J. and Grey, E.M. ( 1988) Dritet Futigue: Concepts, Measurement and Crash Countermea- sures. Federal Office of Road Safety, Canberra, Australia.

Horne, J.A. and &tberg, 0. ( 1976) A self-assessment questionnaire to determine morningness-eveningness in human circadian rhythms. Int. J. Chronobiol. 4, 97-l 10.

Huynh, H. and Feldt, L.S. (1976) Estimation of the box correction for degrees of freedom from sample data in randomised block and split-plot designs. J. Educ. Stat. 1, 69-82.

Kahneman, D. (1973) Attention and Eflorr. Prentice Hall, Englewood Cliffs, NJ.

Keppel, G. (1991 ) Desigtl and Analysis: A Researcher’s Handbook, 3rd edn. Prentice Hall, Englewood Cliffs, NJ.

Langlois, P.H., Smolensky, M.H., Hsi, B.P. and Weir, F.W. (1985) Temporal patterns of reported single-vehicle car and truck accidents in Texas, U.S.A., during 1980-1983. Chronobiol. Int. 2, 131-146.

Lisper, H.O., Eriksson, B., Fagerstrom, K.-O. and Lind- helm. J. (1979) Diurnal variation in subsidiary reaction time in a long-term driving task. Arcid. Anal. Ptw. 11, l-5.

Lisper, H.O., Laurell. H. and Van Loon, J. (1986) Relation between time to falling asleep behind the wheel on a closed track and changes in subsidiary reaction time during prolonged driving on a motorway. Ergonomics 29, 445-453.

Meshkati. N. and Loewenthal, A. ( 1988) The effects of individual differences in information processing behavior on experiencing mental workload and perceived task difficulty: a preliminary experimental investigation. In


Human Mental Workload: Advances in Psychology, eds P.A. Hancock and N. Meshkati, Vol. 52, pp. 2699288. Elsevier Science, North-Holland.

Monk, T.H. and Leng, V.C. (1982) Time of day effects in simple repetitive tasks: some possible mechanisms. Acta Psychol. 51, 207-221.

Monk, T.H., Leng, V.C., Folkard, S. and Weitzman, E.D. (1983) Circadian rhythms in subjective alertness and core body temperature. Chronobiologia 10, 49-55.

Riemersma, J.B.J., Sanders, A.F., Wildervanck, C. and Gaillard, A.W. (1977) Performance decrement during prolonged night driving. In Vigilance: Theory, Opera- tional Performance and Physiological Correlates. ed. R.R. Mackie, pp. 41-58. Plenum Press, New York.

Rosenthal, T.J., Parseghian, Z., Stein, A.C. and Wade Allen, R. (1991) A Users Guide to the Systems Tech- nology, Inc. Driving Simulator. Systems Technology, Hawthorne, CA.

Schwing, R.C. ( 1990) Exposure-controlled highway fatality rates: temporal patterns compared to some explanatory variables. Alcohol, Drugs Driving 5, 275-285.

Smith, A.P. (1989) Diurnal variations in performance. In Acquisition and Performunce qf Cognitive Skills, eds A.M. Colley and J.R. Beech, pp. 301-325. Wiley, Chichester.

Smith, A.P. and Kendrick, C. (1992) Meals and performance. In Handbook of Human Performance, Vol. 2: Health and Performance, eds A.P. Smith and D.M. Jones, pp. l-23. Academic Press, London.

Smith, A.P. and Miles, C. (1986) The effects of lunch on cognitive vigilance tasks. Ergonomics 10, 1251-1261.

Summala, H. and Mikkola, T. (1994) Fatal accidents among car and truck drivers: effects of fatigue, age, and alcohol consumption. Hum. Factors 36, 315-326.

Wickens, C.D. (1984) Engineering Psychology and Human Pecformance. Charles F. Merrill, Columbus, OH.

time of day variations in driving performance

Documents