Journal of Applied Psychology1976, Vol. 61, No. 3, 359-367

Decision Processes and Risk Taking in Traffic: Driver Responseto the Onset of Yellow Light

Vladimir J. Konccni, Ebbe B. Ebbesen, and Daiva K, KonecniUniversity of California, San Diego (La Jolla)

. A field study examined the relationship between the drivers' distance from anintersection (0-100 yards [0-91.4 m]) when the light changed from green toyellow, and the probability that the drivers would proceed through the inter-section. The function relating the two variables approximated a normal ogive,but there was additional evidence that drivers took both distance and speedinto account in deciding whether to proceed or stop. Among the drivers whowere at intermediate distances (40-60 yards [36.6-S4.8 m]) when the lightchanged, the younger males were more likely both to proceed and to violatethe red light than were other drivers. The latter finding could be attributedto the younger males' faster driving and the related tendency to ignore theconsequences of the decision conflict induced by the yellow light.

The automobile occupies an importantplace in Americans' daily habits and generallife-style. Unfortunately, it is also a promi-nent factor in death and injury statistics andthe cause of considerable human suffering,medical expenses, and loss of wages (Haddon,Suchman, & Klein, 196S; Roberts, 1971).Nevertheless, most of the research to date hasinvolved the post hoc analysis of accidentrecords, in an attempt to identify accident-prone categories of drivers (Haddon et al.,1965; Roberts, 1971), and the use of mathe-matical models and computer simulation inthe analysis of traffic flow (Gazis, 1972;Heimstra, 1970). In contrast, the knowledgeabout the behavior and decision making ofaverage drivers or categories of drivers intypical traffic situations is quite limited.There is a need for field research on varia-bles related to particular kinds of driver be-havior in specific driving situations.

One possible research strategy is to selectcommon traffic environments and conduct anin-depth analysis of the variables governing

We would like to thank Susan Heim who helpedcollect the data, and Bruce F. Herms, of the TrafficEngineering Section, Department of Transportation,San Diego, California, for his generous cooperationin providing accident statistics.

Requests for reprints should be sent to VladimirJ. Konecni, Department of Psychology, C-009, Uni-versity of California, San Diego, La Jolla, California92093.

driver behavior in such settings (cf. Ebbesen& Haney, 1973). By isolating a single trafficsetting and concentrating on the behavioralalternatives open to drivers in this setting, thecontrolling variables may be reduced to amanageable number, and an understandingof the basic components of driver behaviormight be gained.

Such a strategy was used in the presentresearch. The study focused on the behaviorof drivers who approached an intersection ona green light, which then changed to yellow.In this situation, each driver's alternativeswere (a) to drive through the intersectionor (b) to stop, Given that both risk to lifeand legal repercussions may be involved inproceeding through an intersection after theonset of the yellow or red light, the courseof action that a particular person took clearlyreflected a nontrivial decision. Yet, all drivershave to make such decisions many times aday. The present study attempted to identifyfactors in the individual (e.g., sex and age)and in the situation (e.g., the distance fromthe intersection when the light changed toyellow, speed, lane positioning, and presence/absence of passengers), which might affectthe decision to stop or proceed, and the prob-ability of being "caught" by the red light.In addition, we were interested in the drivers'immediate reaction to the onset of the yellowlight and in the manner in which the de-cision to stop or proceed was executed (i.e.,

359

360 V. J. KONECNI, E. B. EBBESEN, AND D. K. KONECNI

the acceleration and deceleration patterns).Finally, an issue of interest was the way inwhich drivers combined speed and distanceinformation.

The interaction of organismic and situa-tional variables examined by the presentstudy is of considerable applied interest. Maledrivers tend to be involved in a significantlygreater number of fatal and nonfatal acci-dents than female drivers, even when thetotal number of miles driven is taken intoaccount (Roberts, 1971). Young male drivershave to pay considerably higher automobileinsurance premiums than older or femaledrivers. Two different explanations have beenoffered for these facts (Roberts, 1971). Oneemphasizes the sex and age differences in gen-eral driving style, presumably reflecting dif-ferences in psychological characteristics suchas caution and competitiveness. The otherinvokes the hypothesis that male and youngerdrivers, due to various social contingencies,tend to drive a relatively greater proportionof the total miles under unfavorable drivingconditions (notably in rush-hour traffic andat night) in comparison to older or femaledrivers. By concentrating on driver behaviorunder favorable driving conditions in a near-optimal setting, the present research couldhelp elucidate whether or not there are basicdifferences in the driving habits of differentcategories of drivers.

METHODSetting

The selling was an intersection controlled bytraffic lights on California Highway 1, designatedas a "prime artery" by the San Diego Departmentof Transporlation. The study focused on a 140-yard(127.9-m) stretch of road in the two northboundlanes, from a 100-yard (91.4-m) point before theintersection to the end of the intersection. This sec-tion of Highway 1 is adjacent lo the University ofCalifornia, San Diego campus, and the pedestriantraffic on the sidewalks was moderate. The speedlimit was 45 miles per hour (72 .4 km) and thetraffic flow was about 11 vehicles per minute, tak-ing into account both northbound lanes.

The side street with which Highway 1 forms theintersection is one of the Ihrce major campus ar-teries. The traffic was about 10 vehicles per minuteeach hour at the ful l hour when classes ended, butwas otherwise much lighter. The lights in the inter-section were fully traffic and pedestrian controlled:The green light for Highway 1 changed only if

(here were waiting pedestrians (about 10% of thetime) or waiting cross traffic (about 90% of thetime). Pedestrians could change the light by push-ing a button, but the change was not instantaneous(the lag ranging from 1 sec to about 30 sec depend-ing on traffic conditions). The yellow light lastedfor an average of 2.99 sec (SD — .08, based on 77observations taken on d i f fe ren t days). This is truefor all intersections in San Diego (information fromthe department of transportation), which means thatthe drivers were generally familiar with the yellowlight duration. The length of the red light dependedon the amount of cross traffic.

The waiting cross traffic was entirely obscuredfrom the view of the drivers approaching the inter-section by trees in the median strip and a continu-ous stretch of elevated ground (3-yards [2.7-mlhigh) east of the northbound Highway 1 lanes.However, the traffic lights were visible from at least300 yards (274.2 m) before the experimental sec-tion and remained visible through the entire ap-proach to the intersection, as well as through 90%of the intersection itself.

SubjectsThe subjects in the experimental sample were 100

men and 44 women traveling in the northboundlanes who encountered a change of light from greento yellow while in the 100-yard stretch before theintersection. Given that the observer was ready totake an observation, the only other qualificationfor a driver to be a subject was that no car traveledahead of him or her in the same lane at any pointin the 100-yard stretch. With these restrictions, allapproaching cars were processed. The subjects were,in essence, randomly assigned to the physical dis-tance from the intersection at which they foundthemselves when the yellow light occurred.

In addition, the speed and characteristics of mem-bers of a control sample of 106 drivers (77 men and29 women) were observed. These people traveledthrough the entire 140-yard stretch without encoun-tering a yellow or red light. Otherwise, they wereobserved under conditions similar to those for theexperimental sample.

ProcedureThe data were collected during a period of over

2 months, on weekdays, during daylight hours (rushhours were avoided), in excellent weather conditions,by two carefully trained observers working one ata time. An observer was located near the pedestriantraffic light control button at the intersection, nextto the southbound lanes. From this point, the ob-server had a clear view of the entire 140-yard stretchof road in the northbound lanes, but was completelyunobtrusive. For the purpose of the study, the 140-yard stretch in the northbound lanes was dividedinto six parts, starting at the 100-yard point beforethe intersection: 100-80 yards, 80-60 yards, 60-40yards, 40-20 yards, 20-0 yards, and 0 to —40 yards,the last 40 yards being the intersection itself. White

DECISION PROCESSES IN TRAFFIC 361

sticks were placed unobtrusively in the median stripbetween the northbound and southbound lanes atthe 100-, 80-, 60-, 40-, 20-, 0-, and —40-yard(91.4-m, 73.1-m, 54.8-m, 36.6-m, 18.3-m, 0-m, and—36.6-m) points, controlling for parallax. An ob-server could clearly see these sticks, and thus esti-mate when a car entered a particular stretch of roadin the northbound lanes. The observer was equippedwith a Datamyte portable recorder (Model DAK-8,Electro/General Corporation). The recording deviceconsisted of an electronic "clipboard" on which en-tries were made by pressing appropriate keys, abuilt-in clock, and a cassette tape recorder (Norelco1420) that registered entries on magnetic tape inthe form of sound signals. Information from thetapes subsequently served as input to a Datamyte608 Coupler, which was connected to a PDP-12computer. The type, time of occurrence (with refer-ence to an arbitrary zero point), and relative dura-tion of observed events could thus be retrieved andanalyzed, the smallest possible time unit being 1 sec.

When a car approached the 100-yard point in oneof the northbound lanes facing a green light, theobserver pressed the pedestrian traffic light controlbutton. This usually caused the light to changefrom green to yellow when the subject's car wassomewhere between the 0- and lOO-j'ard points. Assoon as the car reached the 100-yard point, the ob-server made an entry. Such entries were made forthe same car at each of the six points if the carcame to a stop at the intersection, or at seven points(including the —40-yard point) if the car wentthrough the intersection. An entry was also mademarking the time when the yellow light came on.The observer estimated each driver's sex, age (be-low 30 years old or over 30 years old), andwhether or not the vehicle contained passengers. Thelane in which the subject's car traveled, whether ornot there were any other cars in the other lane atany point in the 100-yard stretch, and whether ornot an>' of these cars passed the subjects were alsonoted. Finally, the observer recorded whether therewas any car within 20 yards behind the subject inthe same lane at the time when observation startedat the 100-yard point.1

RESULTSControl Sample

A control sample of 106 drivers who pro-ceeded through the entire 140-yard stretch onan uninterrupted green light was studied toobtain information about speed and lane posi-tioning, The sample consisted of 34 youngermen and 10 younger women and 43 men and19 women of older appearance. In comparisonto older males, the younger were far morelikely to travel in the left (vs. right) lane,X 2 ( l ) = 23,66, p < .001, corrected for con-tinuity. They were also somewhat less likely

to have passengers, x 2 U ) — 3 . 6 6 , p < .06,also corrected. No such patterns were evidentfor the women. In all, there were 24 youngermen without passengers traveling in the pass-ing lane (23% of the whole sample). Thesepeople differed significantly from the remain-ing 82 people in terms of average speed andthe severity of speed limit (45 miles perhour) violations. They took an average of3.29 sec to travel the 100-yard stretch beforethe intersection (about 62 miles per hour[99.8 km]), as opposed to 4.2S sec (48 milesper hour [77.2 km]) for the rest of the sam-ple (z - 6.08, p < .001, two-tailed test, bythe Mann-Whitney U test for large samples).Similarly, they took an average of 1.13 sec(72 miles per hour [115,8 km]) to cross the40 yards of the intersection itself, in com-parison to 1,56 sec (52 miles per hour [83.7km] ) for the remaining members of the sam-ple (2 = 6.53, / > < .001). Thus, the entiresample further increased the speed in the in-tersection. The speed increment (from 62 to72 miles per hour, calculated on the timedata) was quite drastic for the younger pas-sengerless males in the left lane (Wilcoxon'sT = 7, p < .001), but also significant (from48 to 52 miles per hour) for the remainder ofthe sample (z = 4.89, p < .001, two-tailedtest, by the matched-pairs signed-ranks Wil-coxon's T test for large samples).

Experimental SampleSpeed, The first issue to be considered is the

effect of the onset of the yellow light on thespeed of drivers who encountered it and never-theless proceeded through the intersection. Ofthe 144 people (100 men and 44 women) whoencountered a change of the traffic light toyellow within 100 yards of the intersection, atotal of 48 (33%, i.e., 38% of the men and23% of the women) decided to go through

1 Observers' task was facilitated by the features ofthe equipment and the practice they had had. Thebreadth of the median strip, and the fact that ob-servers were stationed by the southbound lanes, com-bined to make the observation unobtrusive. Reliabil-ity checks were carried out at random intervalsthroughout the study. These checks revealed sub-stantial interobservcr agreement (.82-.9S) on allmeasures.

362 V. J. KONEOJI, E. B. EBBESEN, AND D. K. KONECNI

t3OJ

I

.7 -

.6

.5

-I .4

.3

_80-

0-

Men /

Women

100-80 80-60 60-40 40-20 20-0

Distance in yards from the intersectionwhen the yellow light occurred

FIGURE 1. Probability of going through the inter-section as a function of the sex of driver and thephysical distance from the intersection when the yel-low light occurred. (In the 100-80-, 80-60-, 60-40-,40-20-, and 20-0-yard conditions, there were, respec-tively, 18 men, 10 women; 29 men, 7 women; 26men, 12 women; 13 men, 9 w o m e n ; and 14 men, 6women.)

the intersection rather than stop. The 48 sub-jects who went through the intersectioncrossed the 100-yard stretch before the inter-section in 5.50 sec on the average (about 37miles per hour [59,5 km] ) , and the 40 yardsof the intersection in 1.88 sec (about 44 milesper hour [70.8 k m ] ) . Like the controls, thesesubjects accelerated in the intersection (z =5.15, p < .001, by the Wilcoxon T test), butgenerally traveled significantly slower thanthe controls (z - 5.84, p < .001, by theMann-Whitney U test), possibly because theoccurrence of the yellow light induced a con-flict as to whether or not to proceed, resultingin a decrease in speed.

Probability of proceeding through the In-tersection. The physical distance from the in-tersection at which the 144 experimental sub-

jects found themselves when the yellow lightoccurred had a considerable effect on theprobability of their proceeding through theintersection. This relationship is shown inFigure 1 for men and women separately. It isevident that the probability of going throughincreased in a lawful manner as the distancefrom the intersection decreased, Also, therewas an interaction between the distance andsex-of-driver factors, such that only in the60-40-yard and 40-20-yard groups were menmore likely than women to proceed throughthe intersection, x 2 ( l ) = 4 . 4 7 , p < .05 andX~(l) = 2.57, p < .15, respectively, correctedfor continuity. Of the 60 subjects in these twoconditions, 27 went through; of these, 23 weremale (only 4 of the 21 women did so).

The 60-40-yard and 40-20-yard distanceconditions may be considered critical in thesense that the associated probability of pro-ceeding through the intersection was close to.50. Therefore, these groups were brokendown further in terms of age of driver, travel-ing in the right versus left lane, and presenceversus absence of passengers, with stoppingversus not stopping as the criterion. Thischoice of variables was dictated partly by thefact that they differentiated the control sam-ple drivers in terms of speed. A significantlygreater proportion of younger, as opposed toolder, males in the 60-40-yard group pro-ceeded through the intersection, ,\2(1) = 7.96,p < .01, corrected for continuity. In contrast,age had no differential effects for women inthe 60-40-yard group and for either men orwomen in the 40-20-yard group. In addition,younger males in the 60-40-yard group weresomewhat more likely to drive in the leftrather than the right lane, x s ( l ) = 3.43, p <.08, corrected for continuity. Of the 13 menin the 60-40-yard condition who proceededthrough the intersection, 9 appeared under 30years of age, and of these, 7 drove in the lef tlane. Among the men in the 40-20-yard group,there was a nonsignificant predominance ofthe younger ones among those who drove inthe left lane (p < .20). However, driving inthe lef t versus right lane had no effect on theproportion of men proceeding through versusstopping (p < ,40). Other variables, such asthe presence versus absence of passengers and

DECISION PROCESSES IN TRAFFIC 363

the presence versus absence of vehicles in theadjoining lane or behind the subject, had noeffect on the probability of driving throughthe intersection.

Probability of being caught by the red light.In terms of risk taking, we considered theprobability of drivers being in the intersectionat the onset of the red light as a function ofthe distance from the intersection at whichthe yellow light occurred. Since the 20-yardstretch in which a driver had been located atthe time when the yellow light occurred wasknown for each subject, his or her positionwhen the red light came on could be cal-culated quite accurately from the individualcumulative-time data, given that the durationof the yellow light was also known.2 For thepurpose of the estimate, it was assumed thateach subject was at the midpoint of a 20-yardstretch when the yellow light came on (i.e.,at the 10-, 30-, SO-, 70-, or 90-yard [9.1-m,27.4-m, 45.7-m, 63.9-m, or 82.3-m] points, de-pending on the experimental condition). Aconservative estimate was also made assumingthat each subject was at the point closest tothe intersection in a 20-yard stretch when theyellow light occurred (e.g., the 40-yard pointfor a subject in the 60-40-yard condition).

The probability of being in the intersectionwhen the red light occurred is shown in Figure2 as a function of the distance from the inter-section. The data for men and women arepresented separately, and both the non-conservative and conservative estimates aregiven.

None of the subjects in the 100-80-, 80-60-and 20-0-yard blocks were in the intersectionwhen the red light occurred (except for oneman in each of the two latter conditions).Drivers in the further two blocks generallydid not even attempt to enter the intersection,while those in the closest block had ampletime to get through before the onset of thered light. In contrast, a considerable numberof the 60-40-yard and 40-2 0-yard subjectswere in the intersection when the red lightcame on, particularly male subjects in the60-40-yard condition (among whom thoseappearing under 30 years of age were over-represented). Even by the conservative esti-mate, 8 such people overlapped the red light(31% of all male subjects in the 60-40-yard

i.o

.9

§> .6

.5

.3

nonconservative estimateconservative estimateMENWOMEN

100-80 80-60 60-40 40-20 20-0

Distance in yards from the intersectionwhen the yellow light occurred

FIGURE 2. Probability of going through the redlight as a function of the sex of driver and thephysical distance from the intersection when the yel-low light occurred.

group and 62% of those males in the groupwho crossed the intersection). Across all con-ditions in the study, 9 men and no womenwere in the intersection when the red lightoccurred, even by the conservative estimate.

Acceleration and deceleration patterns. Inorder to gain an understanding of the decisionprocesses involved, it is important to examinethe behavior of various groups of subjectsover the entire stretch of road for which thedata are available. The mean cumulative timetaken by groups of subjects to reach various20-yard points before and in the intersectionis presented in Figure 3. The data for sub-groups of people who did or did not gothrough the intersection within each conditionare presented separately.

- This procedure was necessary because it had beenexpected that only a negligible number of peoplewould drive through the red light. For this reason,observers did not directly record the onset of thered light.

364 V. J. KONECNI, E. B. EBBESEN, AND D. K. KONECNI

FIGURE 3. Mean cumulative time in seconds takento reach di f ferent distances before and in the inter-section as a func t ion of the physical distance f romthe intersection when the yellow light occurred, andof whether or not subjects proceeded through theintersection. (In the two cases where several con-ditions are represented by a single line, as shown inthe legend, the standard deviations of the conditionmeans are so small that they cannot be adequatelyindicated by the customary vertical bars.)

Several points should be made about thesedata. First, people who eventually came to astop seem to have reached the various 20-yard points at about the same time, irrespec-tive of the distance condition they were in. Asingle line portrays quite accurately the de-celerating behavior of subjects in the 100-80-,80-60-, 60-40-, and 40-20-yard conditionswho came to a stop.

Second, it seems that the division of sub-jects into those who eventually stopped andthose who eventually went through was com-pleted by the 80-yard point. Prior to thispoint, and again after the 40-yard point, thesubjects who eventually stopped had traveledmore slowly than those who eventually wentthrough. Thus, it seems that the driving speedas far as 80 yards from the intersection hadmuch to do with whether or not a personwent through, and with the likelihood that atleast a part of the crossing would overlap thered light. Data on the behavior of young maledrivers from both the control and experimen-tal samples illustrate this point.

Third, the curve for subjects who eventuallycame to a stop begins to deviate radically only

at the 40-yard point from the curves for sub-jects who eventually went through (betweenthe 80- and 40-yard points the curves are es-sentially parallel). However, the deviation isprimarily due to the former subjects' decelera-tion. The 60-40-, 40-20-, and 20-0-yard sub-jects who eventually went through had begunto accelerate only as late as the 20-yard point.In view of this, it is not surprising that somany of the 60-40-yard subjects who crossedoverlapped the red light. The curve for the60-40-yard subjects who went through is in-distinguishable from that for the 40-20-yardsubjects who also went through (see Figure3 ) ; yet, the former subjects were further fromthe intersection when the yellow light oc-curred. Apparently, a judgmental error hadbeen committed by many of the 60-40-yardsubjects.

Immediate reaction to yellow light. In-formation about the subjects' immediate reac-tion to the onset of the yellow light could beobtained by comparing the time taken bysubjects in a given condition to travel the 20-yard stretch where the 3'ellow light occurredto the time these subjects took to cover thenext 20 yards. These data, presented for indi-vidual subjects, are shown in Figure 4. PanelsA to D contain the results for subjects in the100-80-, 80-60-, 60-40-, and 40-20-yardconditions who came to a stop. The data forsubjects in the 80-60-, 60-40-, 40-20-, and20-0-yard conditions who went through theintersection are shown in Panels E to H.

From Panels A to D, it is clear that sub-jects who eventually came to a stop reacted tothe onset of the yellow light by slowing downconsiderably, irrespective of how far from theintersection they found themselves when theyellow light occurred. A high proportion ofthe 100-80-yard subjects slowed clown in the80-60-yard stretch, in comparison to the100-80-yard stretch (p = .064 by the two-tailed sign test; see Panel A in Figure 4). Asimilar, immediate slowdown response wasalso exhibited by subjects in the 80-60-,60-40-, and 40-20-yard conditions who even-tually came to a stop (p — .038, .001, and.016, respectively, by the two-tailed signtests; see Panels B, C, and D).

In contrast, the speed with which the60-40-, 40-20-, and 20-0-yard subjects who

DECISION PROCESSES IN TRAFFIC 365

65432]

0 -

a.(Cond. 100-80) b. (Cond. 80-60) c.(Cond. 60-40) d. (Cond. 40-20)

Do

100 80 60e. (Cond. 80-60) MCond. 60-40) g.(Cond. 40-20) h (Cond. 20-0)

/

"..I

80 60 40 60 40 20 40 20 0Distance in yards from the intersection

20 0 -40

FIGURE 4. Cumulative time in seconds that it took individual subjects in the various distanceconditions to reach the end of the 20-yard stretch in which the yellow light occurred, and toreach the end of the subsequent 20-yard stretch. (Panels A to D provide the data for subjectsin various conditions who came to a stop; Panels E to H are for subjects who eventually wentthrough the intersection. The moment at which a subject entered the 20-yard stretch in whichhe or she encountered the yellow light was treated as time zero. Numbers on the lines indicatethe number of subjects who took a particular length of time to travel a 20-yard stretch.)

eventually crossed the intersection coveredthe 20 yards in which they, respectively, en-countered the yellow light, did not differ fromtheir speed in the next 20 yards. As is clearfrom Panels G and H (see Figure 4), boththe 40-20- and 20-0-yard subjects sped upsomewhat in the 20-yard stretch after en-countering the yellow light, but this trendwas not significant (p — .22, by the two-tailed sign test, for the 40-20-yard subjects;p ̂ .SO for the 20-0-yard subjects). Thespeed with which the 60-40-yard subjectstraveled the 40-20-yard stretch was also notsignificantly different from that in the 60-40-yard stretch (p — ,45); in fact, these subjectstended to slow down (see Panel F in Figure4), even though they were initially furtherfrom the intersection than the 40-20-yardand 20-0-yard subjects.

Speed and distance processing. The way inwhich the 60-40-yard male subjects combinedinformation about the speed at which theywere traveling when the yellow light came onwith information about their distance fromthe intersection at that time should also beexamined. (Such analyses could not be carriedout for women, due to the small number ofsubjects.) For this purpose, males in the60-40-yard and 40-20-yard groups were com-pared. The median cumulative time subjectstook to cover the respective 40 yards preced-ing the 20-yard stretch where the yellow lightoccurred was 2 sec. Of the 10 males in the60-40-yard group who traveled slower thanthe median prior to their 20-yard yellow lightstretch, only 2 (20%) went through the inter-section. Both of these accelerated in the 60-40-yard stretch in comparison to the average

366 V. J. KONECNI, E, B. EBBESEN, AND D. K. KONECNI

for the 100-80-yard and 80-60-yard stretches,but they still overlapped the red light even bythe conservative estimate. However, 11 of the16 60-40-yard subjects (69%) who traveledfaster than the median prior to the 20-yardstretch in which the yellow light occurredproceeded through the intersection. Of these,4 traveled the 60-40-yard stretch only as fastas the average of the 100-80-yard and 80-60-yard stretches, and 4 traveled slower than theaverage; all 8 went through the red light evenby the conservative estimate. The remaining3 subjects accelerated in the 60-40-yardstretch and overlapped the red light only bythe nonconservative estimate. A similar break-down of the 40-20-yard male subjects, interms of traveling slower or faster than themedian in the 80-.60-yard and 60-40-yardstretches, yielded 61% and 80%, respectively,going through the intersection. The pattern ofresults for the 60-40-yard subjects differedsignificantly from that for the 40-20-yardsubjects, X

2(3) = 8.81, p < .05. The drivingspeed prior to the occurrence of the yellowlight had far less of an effect on subjects whowere closer to the intersection (the 40-20-yard group), in comparison to those who were20 yards further on the average (the 60-40-yard group).

Thus, the 60-40-yard males took into ac-count both speed and distance in making thedecision whether or not to proceed. The fewslow-moving subjects who eventually wentthrough accelerated immediately, but over-lapped the red light despite this. However, itwas predominantly the fast-moving subjectswho eventually drove through the intersec-tion. These subjects slowed down or failed toaccelerate, the consequence being passagethrough the intersection in the presence ofthe red light.

DISCUSSIONLike other complex behaviors, driving in-

volves many related decisions that are un-doubtedly affected both by situational andorganismic factors. An example of such a de-cision is the speed at which a person drives ina given situation relative to other drivers andthe speed limit. Data from the control sampleshowed that, in comparison with other drivers,young male drivers drove faster and increasedtheir speed in the intersection more. Young

males also chose to drive predominantly inthe passing lane and were often without pas-sengers who could potentially exert a restrain-ing influence (cf. Ebbesen & Haney, 1973).

Such a driving-speed decision increased therisk involved in driving in general,3 Furtherincreasing speed in the intersection may alsobe hazardous, because of various kinds of in-terference typical of intersections (e.g., pro-truding automobiles waiting to make a leftturn; right-hand turns on a red light, whichare legal in California). Moreover, the de-cision to drive fast probably contributed tothe young males' overrepresentation amongdrivers in the experimental sample who pro-ceeded through the intersection. Namely,drivers seemed to take both speed and dis-tance into account, that is, to "compute"temporal distance. It follows from this state-ment, and is substantiated by the data inFigure 3, that across different physical dis-tances the fast-moving drivers (mostly youngmales) would be more likely to proceedthrough the intersection.

A passage through an intersection after theonset of the yellow light may increase theprobability of accidents because it is accom-panied by an increase in speed, a slowing-down pattern of other drivers who have de-cided to stop, and the mentioned interferencewithin the intersection. The risk is pre-sumably further augmented if the passage iscarried out in the presence of the red light.In 1974, about 65% of automobile accidentsin San Diego (excluding freeways, ramps, al-leys, and parking accidents) occurred in in-tersections, as opposed to midblocks (datafrom the San Diego Department of Trans-portation). In our stud}', the Highway 1driver's vision of the waiting and approachingcross traffic was obscured by the structuralaspects of the intersection. In addition, thelights were traffic controlled. To most of theHighway 1 drivers, the yellow light shouldhave indicated the presence of cross traffic orpedestrians (lights in the newer areas of SanDiego are usually traffic controlled).4

3 Roughly 7% of all automobile accidents in SanDiego are directly related to speeding (unpublisheddata by the San Diego Department of Transporta-tion) .

4 Of all the intersection-related accidents in San

DECISION PROCESSES IN TRAFFIC 367

In the experimental sample, the red lightviolations were committed almost exclusivelyby the young male drivers. While this tend-ency seemed to be a consequence of drivingfast prior to the onset of the yellow light, andthe decision to proceed through the intersec-tion after it came on, an important qualifica-tion is in order here. Most of the drivers whooverlapped the red light were at a certaincritical distance (40-60 yards) from the in-tersection when the yellow light came on. Thedata for the 60-40-yard subjects who drovethrough—mostly young males—also indicatedthat these subjects tended to slow down afterthe onset of the yellow light. It can be as-sumed that such behavior was due to thedecision conflict. The decision to drive throughmay have been based on the temporal dis-tance from the intersection, one componentof which was speed prior to the time whenthe decision was made (i.e., just before theonset of the yellow light). The decision mayhave been correct had the speed remainedconstant. However, the judgment to proceedseemed not to take into account the timetaken by the processing of the yellow lightonset, and the resulting decrease in speed, Inshort, drivers failed to revise the decision toproceed following the conflict-induced time(speed) loss.

It is likely that the underestimation of thetime taken" by the decision making is not re-stricted to the young male drivers. However,other drivers may not find themselves in situa-tions where a conflict would arise. They mayrarely be at critical distances from the inter-section while driving at a high speed. More-over, lower speeds and smaller distances fromthe intersection may be processed morequickly. A driver moving slowly, who is 30yards away when the light changes to yellow,may need less time for decision making thana person driving fast, who is SO yards away,although the temporal distance may be equal.To extend this reasoning, young males may beinvolved in a disproportionately large numberof accidents because they often have to en-

Diego in 1974, about 55% were two-vehicle, right-angle collisions (data from the San Diego Depart-ment of Transportation). This implies frequent fail-ures of the traffic lights to control drivers' behaviorand/or drivers' reluctance to yield to cross traffic.

gage, as a consequence of driving fast, in time-consuming decision processes.

Apart from several near misses, no acci-dents were observed in the study, but it seemsjustifiable to conclude that the driving be-havior of young males increased the proba-bility of accidents. Had any accidents in-volving young male drivers occurred, theresponsibility could have been assigned tothese people's driving habits (rather than un-favorable driving conditions). However, theresults showed that the physical aspects of thedriving environment (such as the relativelyshort duration of the yellow light) interactedwith driving habits to increase the probabilityof accidents.

Traffic safety efforts have included drivereducation, punishment of violators, develop-ment of safety standards for automobiles, andtraffic engineering measures. The human fac-tors have generally been either ignored or dis-cussed in terms of personality traits and ac-cident proneness (see Carlson & Klein, 1970,for a critique). Our results suggest thatsubtler variables, such as the decision pro-cesses and conflict, should be taken into ac-count in planning the features of the drivingenvironment. For example, an improved sig-nal system could be introduced, includinglonger "dead" (all red) periods. Alternatively,the onset of the yellow or red light could bedelayed by sensors detecting drivers travelingat relatively high speeds at critical distancesfrom the intersection. Obviously, such changescould be efficient only as components of aplan to suit the physical features of the driv-ing environment to the drivers' behavior andcognitive limitations.

REFERENCESCarlson, W. L., & Klein, D. Familial versus institu-

tional socialization of the young traffic offender.Journal of Safety Research, 1970, 2, 13-25.

Ebbesen, E. B., & Haney, M. Flirting with death:Variables affecting risk taking at intersections.Journal of Applied Social Psychology, 1973, 3, 303-324.

Gazis, D. C. Traffic flow and control: Theory andapplications. American Scientist, 1972, 60, 414-424.

Haddon, W., Jr., Suchman, E, A., & Klein, D. Ac-cident. research. New York: Harper & Row, 1965.

Heimstra, N. W. Injury control in traffic safety.Springfield, 111.: C. S. Thomas, 1970.

Roberts, H. J. The. causes, ecology, and prevention oftraffic accidents. Springfield, 111.: C. S. Thomas,1971.

(Received May 5, 1975)