The Development Of Attitude Scales To Predict Accident Repeater And Moving Violator Drivers

PDF

Download

Open document

Flip pages

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content This d issertation has been Mic 61-1707 m icro film ed exactly as receiv ed SCHUSTER, Donald Herbert, THE D E V E L O P - MENT OF ATTITUDE SCALES TO PREDICT ACCIDENT REPEATER AND MOVING VIOLATOR DRIVERS. U niversity of Southern C alifornia, Ph.D., 1061 Psychology, general University Microfilms, Inc., Ann Aibor, Miohiqnn THE DEVELOPMENT OF ATTITUDE SCALES TO PREDICT ACCIDENT REPEATER AND MOVING VIOLATOR DRIVERS by Donald Herbert Schuster A Dissertation Presented to the FACULTY OF THE GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY (Psychology) January 1961 UNIVERSITY O F SOUTHERN CALIFORNIA G R A D U A T E S C H O O L U N IV E R SIT Y PA R K L O S A N G E L E S 7 . C A L IF O R N IA This dissertation, written by ................. Pq t m &c I Herbert Schuatar under the direction of h i J k . . .Dissertation Com mittee, and approved by all its members, has been presented to and accepted by the Dean of the iiradunte School, in partial fulfillmerit of rer/nu ements for the de/free of D O C T O R OF P H I L O S O P H Y Dtan Date k£ £ hK iilCj j Q b / (1 I)IS£ERTATION COMMUTEE * . , , . , ( y . / ( s? ( ^ . Chairman ACKNOWLEDGMENTS Thanks are due many people in many places for their help in making this research effort possible. One wonders where to start enumerating. Two people stand out as having been helpful from the beginning of this research five years ago; they are Armand Bowers and Kenneth Brown in the Los Angeles Office of the California Department of Motor Vehicles, Driver Improvement Section. They deserve special mention for their efforts in going far out of their way to provide drivers and facilities for testing, driver record information, and encouragement. Financial support was given freely by the United States Public Health Service. The research for this dis sertation was done under part of their research grant M-2353, called the California Accident Repeater Driver Study, or more simply, the Traffic Project. The Sheridan Supply Company kindly gave permission to use items from several tests and printed special editions of these items for this research. Many organizations were helpful in supplying drivers for testing, a big job. Thanks are due: Farmers Insurance ii Group, Founders Insurance Company, Western Insurance Information Service, North American Aviation, Standard Oil of California, Wilshire Oil Company, the Automobile Club of Southern California, the El Segundo Boosters, San Bernardino Air Materiel Area, Edwards Air Force Base, Norton Air Force Base, Long Beach Naval Station, China Lake Naval Station, adult education classes in the Los Angeles high schools , the California Department of Motor Vehicles, and the University of Southern California. Special thanks are due Sergeant Frank Crewe, of the Los Angeles Police Department, for the continued testing of his Personal Traffic Safety classes. Appreciation and thanks are due Dr. J. P. Guilford, my academic adviser and friend, not only for his over-all encouragement and guidance, but also for letting me go my own way in this research and make my own mistakes. Several people on the Traffic Project staff merit thanks for doing more than just a job; they are Kay Decker, Karen Friedrich, Elinor Gold, and Barbara Levenson. The last, but continuing, word of thanks goes to my wife, Locky, for her patient and enduring encouragement and many hours of statistical computations. TABLE OF CONTENTS Page ACKNOWLEDGMENTS................................... ii LIST OF TABLES..................................... v Chapter I. INTRODUCTION................................. 1 II. REVIEW OF THE LITERATURE..................... 6 Biographical Studies Aptitudes and Psychophysical Studies Attitudes and Temperament Studies Statistical Studies Faking Studies III. METHOD AND PROCEDURE........................ 41 Item Selection Dependent Variables Testing Subjects Criterion Groups Item Analysis Validity Measurement Cross-Validation IV. RESULTS AND DISCUSSION...................... 77 Discrimination on Criterion Groups Significant Items and Traits Validity Analysis Cross-Validation V. SUMMARY..................................... 100 BIBLIOGRAPHY 104 LIST OF TABLES Table Page 1. Discrimination between Problem Driver Criterion Groups and the Selected Better-than-Average (BTA) Group by Means of Derived Scale Scores........... 79 2. Source Trait or Variable of Significant Items in Derived Scales................. 81 3. Validity D a t a .............................. 88 4. Discrimination between Good and Poor Drivers with Physical or Mental (P and M) Health Problems by Means of Derived Scale Scores ................. 90 3. Discrimination between Good and Poor Non-White Drivers by Means of Derived Scale Scores ...................... 91 6. Discrimination between Good and Poor Truck Drivers by Means of Derived Scale Scores.............................. 92 7. Discrimination between Good and Poor Female Drivers by Means of Derived Scale Scores.............................. 94 8. Discrimination between Good and Poor Non-White, P and M, and Truck Drivers Combined by Means of Derived Scale Scores..................................... 96 v CHAPTER I INTRODUCTION For a long time psychologists and statisticians have been interested in accident research, motivated by the practical consideration of predicting persons who would have accidents. Some efforts have met with limited success (Bureau of Public Roads, 1938), while others have been completely unsuccessful (C. Miller, ca 1932). The acci- dent-research field is rather full of perilous rocks, as we shall presently see, and to date, no research effort has avoided these rocks well enough to do a good job of predicting accident-getters. When attention is focused upon driving and traffic safety, a secondary aspect is seen: moving violations. This is an alternate term for traffic citations ("tickets"), which are issued by traffic officers for potentially dan gerous driving behavior. A small percentage of drivers manage to collect a large enough number of traffic 1 2 citations so that their driving privilege (driver's license) la withdrawn. This study alms at the development of short attitude scales to predict problem drivers. The first type of prob lem driver is the accident-repeater driver , with a history of traffic accidents in his driving record. The second type is the negligent-operator (N.O.) driver, with a driv ing record of many moving violations. There is a low positive correlation (eta = 0.26*) between these two types of problem drivers, showing that essentially there are two types of problem drivers. A secondary aim of this study is to investigate the seriousness of faking tendencies in connection with the two problem driver scales, since the scales are of the self-description or inventory type. For operational definitions within this study, the following definitions of problem drivers are employed. An accident-repeater driver is one who has been involved in three or more traffic accidents within the last three years for two of which he was at least partially responsible. Author’s calculations on data from Driver Record Study, Department of Kotor Vehicles, State of California, 1958. 3 The definition of a negligent-operator driver la baaed upon cumulative and weighted pointa. Any peraon whoae driving record ahow3 a count of four or more polnt8 in 12 montha, six or more pointa in 24 months, or eight or more pointa in 36 months is prima facie presumed to be a negligent operator of a motor vehicle (California Vehicle Code, 1957). Traffic convictions involve the unsafe operation of a vehicle on a highway and usually are given a value of one point. Certain violations, such as hit- and-run accidents, drunk or reckless driving, and driving while license is withdrawn, count two points. Being charged with the responsibility of an accident also gives the driver one point. The three-year period was chosen in both definitions primarily for convenience. The State of California watches its problem drivers' records for three years; if the driver has ceased being a problem driver, his file is cleaned out. If still a problem driver at the end of three years , the records continue to accumulate. From a psychometric point of view, it is hoped and assumed that the driver's person ality and attitude characteristics have stayed relatively constant during this three-year period. The validity of 4 this assumption will be examined later. Finally, if a shorter interval of time had been used, it is felt that too small a number of accidents and violations would have accrued to a large enough number of drivers for any sort of statistical reliability. Several basic assumptions in the study should be noted: attitudes determine driving behavior, they remain constant for several years , and they can be measured psychometrically. The scales developed here could be used to help reduce the upward trend of motor vehicle accidents through the early screening of potential problem drivers. Various agencies concerned with traffic safety could use such psychological scales for predicting problem drivers. For example, when a driver is half-way toward meeting the present standard of being a "negligent operator" driver, one scale could be used to predict whether this driver will probably continue up the scale of number of moving violations. If so, the driver could be asked to attend the remedial driving course sooner than otherwise. Like wise, after one accident, the driver could be tested and the probability of his having another accident noted, with 5 appropriate action forthcoming. These scales also could be used to select safe drivers by commercial organizations. CHAPTER II REVIEW OF THE LITERATURE An immense amount of literature has been written about the subject of accidents. Some approach had to be taken to reduce the task of examining this literature. Research concerning industrial, home, and aircraft acci dents will be ignored here, except in those few cases where the results are of direct interest to the hypotheses in this study. Early studies often presented results that were not cross-validated or sometimes gave contradictory results. Later, these conflicting results were in many cases shown to have been due to the many (uncontrolled) factors involved and to the statistical unreliability of accidents. Earlier studies are thus omitted, but key, critical review articles are included. The major emphasis in this review section is on carefully controlled and cross-validated studies. What is left roughly falls into four categories: biographical studies, statistical 6 7 considerations, psychophysical studies, and psychological (personality) studies. Research on moving violations has been a fairly recent secondary aspect. McFarland and Moore (1957) and McFarland (1957a) have recently reviewed and evaluated the literature on human factors in highway safety. Their conclusions were that the accident-liability conditions in accident research should be strictly controlled, and specifically, the nature and amount of accident exposure should be watched. As side remarks, McFarland observed that motor vehicle driving requires more continuous attention than is demanded in other forms of transportation, and that little information is available on the influence of emotional stress on driving. In an earlier publication, McFarland (1954) detailed a human-factors approach to the development of a psychological-testing program and of a procedure for detecting accident repeaters. McFarland (1957a) summarized well the few major research findings concerning the problem of accident- repeater drivers with the following conclusions: (1) the best significant predictor of accidents is a history of previous accidents; (2) the personality picture of the 8 accident-prone person is eccentric, impulsive, and mildly psychopathic; and (3) statistical studies to date have been largely negative due to lack of adequate experimental and statistical controls , although age and sex have been shown to be major factors. Human factors that contribute to the causes of accidents can be subdivided into three areas: faulty attitudes, lack of skills and psychological aptitudes , and lack of information and intelligence (Stack, 1947). A Highway Research Bulletin (Stack, 195 7) lists two or more factors as being of importance in most "near accidents." Drivers' opinions on the causes or factors were poor visibility, "blurry" misjudgment as to speed, misjudgment as to the other driver's intentions, inatten tion, speed, no signal on the part of the other driver, and an unexpected maneuver by the other driver. Biographical Studies Let us consider several biographical or individual characteristics that have been reported at various times in the literature, as predictive of accidents by drivers. One of the first is previous accidents. Wilson (1939) reported that drivers with only one accident have little, 9 if any, greater accident expectancy than the average driver. Of course, the average driver's expectancy of having an accident is that he will have no accidents in the near future. The other side of the picture is reported by Farmer (1945) , who stated that drivers who had a large number of accidents during the first year of their driving had a greater tendency to have accidents in their later years. This is the contagion factor. A Bureau of Public Roads document (1938) has this experience to relate: A company operating a large fleet of motor vehicles during three years eliminated about 1/8 of its drivers with the worst accident records and replaced these drivers with new drivers selected at random. As a result, the number of accidents per year was reduced to 1/5 of the previous accident record, and at the same time, the number of miles between accidents increased about four times. Both improvements were in a favorable direction. A second variable used in predicting accidents , quite often, has been the number of violations that the driver has incurred over a period of years. For example, Comrey (1958), reported a phi coefficie t between moving violations and accidents of .16 and .24 for two separate 10 groups of students going through a high-school driver- training course. This indicates that the common variance between violations and accidents while driving was only 6 per cent. This correlation is quite similar to the correlations found for two random samples of drivers in the State of California where the eta correlation was .25 and .26 for two separate samples. Again, the common variance was 6 per cent. Apparently, except in extreme cases , accidents are not very predictable from knowledge of the number of moving violations that a driver has had over a three-year period. A close but undemonstrated relationship between accidents and violations was reported by the ENO Founda tion for Highway Traffic Control (1948). This report also stated that the repeater-violator drivers were arrested more often on non-traffic charges. A Department of Motor Vehicles, State of California (1958) study reports that most drivers over a three-year period do not have any traffic violations or reports of accident involvements in their record. If the driver has one or more abstracts (violations or accidents) in his record, it places him in a definite minority category and his chances of having an 11 accident eventually are far greater than if he did not have any abstracts in his record. The sex of the driver is an interesting variable that several studies have shown to be related to the violations and accidents records of drivers. The Depart ment of Motor Vehicles study just mentioned shows that men drivers have three times as many violations and twice as many accidents in a three-year period as do women drivers on the average. However, in defense of the male ego, it should be pointed out that women drivers do not drive as many miles per year as the average male driver (Lauer, 1959). Comrey (1958) also mentioned a slight correlation between sex and accidents , the variables had specifically 4 per cent common variance. This is surprisingly small in view of the relative insurance risk for young drivers , women being much better risks in the eyes of most insurance companies. On the other side of the picture, women drivers were found to be very significantly worse in an unusual but relatively safe situation, as reported by Uhr (1959). The Department of Motor Vehicles (State of Califor nia, 1958) study mentioned before pointed out that age 12 seems to be one o£ the most Important differentiating factors on the driving records in terms of violations and accidents. For example, male drivers in the 22-year-old age bracket showed only 22.9 per cent of their number in the conviction-free category, whereas 56-year-old men showed 59.2 per cent of their number in the conviction- free or no-abstract category. The previous Comrey and ENO Foundation studies also report that age or youthfulness is a significant factor in accidents and violations. Biesheuvel and Barnes (1959) report that the South African youthful motorcycle drivers are the accident-repeater drivers. Case and Stewart (195 7) report that negligent-operator drivers (drivers with primarily many moving violations) differ primarily in age from the better drivers. Others who reported age to be involved with problem drivers are McFarland (1957) , Heath (1957), and Billion (1956). The relationship of accidents to the speed at which the motorist drives is not a simple one. Lefeve (1956) states that faster drivers have more accidents than slower drivers , especially when judged by their speeds in the afternoon. However, in the morning the speeds are slightly 13 higher for drivers without accident records. Apparently the time of day reverses the relationship of accidents to speeds driven. Barnes (1958) reports that the tendency to speed is responsible for only one in five of the accidents caused, so that traveling at excessive speed is not the all- embracing cause of motorcycle accidents that the public presumes it to be. Concerning reported driving speeds and previous accidents, Stewart (1957) reports that students with consistently higher-than-average driving speeds have traffic records just as free of accidents as drivers who drive at lower-than-average driving speeds. Peculiar to accident, or to potential accident situations, is another interesting variable. The Highway Research Board (1939) found that the annual miles driven per year correlated slightly (0.20) with accidents. Boek (1958) found that persons driving lower annual mileage had fewer accidents per year, and that this finding was sig nificant at the 1 per cent level. However, the correlation involved was still quite low, approximately 0.09. Boek also showed that low mileage men with 10 to 14 years of driving experience had fewer accidents than did those with 14 either more or fewer years of driving experience. However, this variable is probably related to age. Bletzacker and Brittenham (1958) also report that driving experience appears to be a very significant factor in accident causation. Specifically, the Inexperienced driver has difficulty on highways with restrictive fea tures. Accordingly, if the process of gaining driving experience can be accelerated, this factor could be mini mized. Driver-training courses and simulators or trainers seem to be approaches to accelerating such driving experience. Aptitudes and Psychophysical Studies Fletcher (1939) conducted a survey of driving tests , including many psychological and road tests. Specifically, he gave tests for brake reaction time, steering, vigilance, estimation of relative speeds, night-time glare resistance, color vision, field of vision, visual acuity, visual depth, fusion, and astigmatism. The road tests were written rules of the road, driving limitations and deficiencies, and a driving road test. Whether unmatched, or matched on age, mileage per year, or number of years driving, the "good" group (professional drivers) was superior in all the tests 15 to the negligent-operator and accident-repeater drivers. Soma (1953) reported that nearly 100 per cent of the over-accident makers on the Japanese National Railroad could be picked out by means of a card-sorting test scored In terms of a ratio of motor to perceptual speeds. How ever, his study was not cross -validated, nor was his prediction formula specified. In a review of studies , Webb (1956) reported no relationship of ability or aptitude to accidents for pilots of aircraft. Instead, he suggested the study of transitory states of the pilot in contrast to his persistent states, aptitudes or psychological skills. In a study of the prediction of accident proneness of motorcycle operators , Buchanan (1950) reported that a vlsual-attention test and a perceptual-speed test (Ruch- Wilson tests, Nos. 1 and 6) were found to be the best pre dictors of the accidents of police motorcycle operators. However, only 1.7 per cent of the variance in the accident scores for the officers was predicted by these two tests. Fletcher (1945) found that there was no significant difference in the number of traffic accidents or violations between drivers with normal color vision and those who were 16 red-green blind. The lack of significance for the impor tance of color vision here may be partially offset by the fact that in California the traffic signal lights are not true colors , which partially compensates for red-green blindness. Further, for visual acuity, no significant differences were found between good drivers and drivers involved in fatal accidents or drivers who were habitual violators. However, for drivers with one poor eye, a marked relationship was found between the side with defec tive vision and the accident side for intersectional accidents. Drivers with violations for improper passing or crowding the car ahead ("following too closely") also showed defective visual acuity. Drivers involved in acci dents where they were "blinded by glare" revealed that nearly all of them were either oversensitive to glare or glare-blind. However, over half of these same cases had normal visual acuity, showing that the visual-acuity test alone is insufficient in determining night driving privi leges or predicting violations and accidents at night. Knraishl, Masahide, and Tsujioka (1947) reported that the Uchida-Kraepelin Psychodiagnostic Test in Japan predicted 99 per cent of the accident makers on the 17 Japanese National Railroad operation. This test, along with the card-sorting test, and the ratio of motor speed to perceptual speed, predicted 99 per cent of the accident makers. There are three factors in the Uchida-Kraepelin test: (1) a general volitional energy level, (2) fatigue, and (3) volitional tension. A fourth and insignificant factor was probably temperamental. King and Sutro (1957) estimated that an obstruction to vision contributed to one out of every eight motor vehicle accidents. In addition, there are other cases where through inattention, distraction or some other cause, the visual stimulus was visible to the eye but failed to register. The authors subdivided this failure of percep tion into five categories: free and controlled attention, shifting and fluctuating attention, distraction, divided attention, and span of attention. Uhlaner (1956) discarded most of a psychological test battery, but retained the following valid items as predictors of driving in the Army: visual attention to detail, visual judgment, emergency judgment, two-hand coordination, along with driver know-how, driver self description and his age. 18 Conger, et al. (1959) reported no Intelligence or psychophysical differences between non-accident subjects and the accident subjects tested. Finally, we have a preliminary statement by the Drivers Safety Service (1959) that reported the following areas of psychophysical weaknesses to be involved in drivers with accidents: visual acuity, field of vision, depth perception, night depth perception, glare recovery, and complex reaction time. These results surprisingly survived a cross-validation and in particular, a previously puzzling finding may be resolved by this study: the better drivers (no accidents) had a faster complex reaction time. However, the accident drivers had a faster simple reaction time. Attitudes and Temperament Studies Schulzinger (1956) reported on 35 ,000 industrial accident injuries as studied by a physician over an 18-year period. He states that the accident syndrome is a dynamic variable: a constellation of signs, symptoms and circumstances which together determine or influence the occurrence of an accident. The syndrome is a synthesis of "environmental, psychological, physiological, 19 characterological and temporal factors which forms a unified series of accident causation." First, there is a universal risk factor plus a transitory or additional risk Incident to a particular injury. Then there is the abnormal physical side of the environment, along with the maladjustment and irresponsibility of the individual which can be triggered by a specific episode. The corresponding behavior of the person then results in an accident with or without injuries. Causative emotional stages of the individual may be due to faulty perception, faulty thinking (reasoning or interpretation) , or faulty mechanisms for discharging tensions. Thus, there may be a temporary breakdown in the adequate control of the person's behavior. Schulzinger reaches two very interesting conclu sions: Most accidents (74 per cent) are due to relatively Infrequent solitary experiences of large numbers of people (this applies to Industrial and nonindustrial accidents alike, and presumably to driving accidents as well). Second, those people who suffer injuries year after year, the accident repeaters , account for a very small percentage of all accidents (one-half of one per cent). The irrespon sible and the maladjusted individuals are significantly 20 more liable to accidents. Further, almost anyone can become accident prone under emotional stress. Most people find solutions to their problems, develop defenses against their emotional conflicts, and drop out of the "accident prone" group. Accidents are also related to the degree of adjustment in life and of the person's ability to get along with others. In a symposium of the psychological factors contrib uting to traffic accidents, Forbes (1957) reported that drivers involved in accidents were among the socially and economically maladjusted groups of people. Emotional dis organization has also been reported as a factor in acci dents. Some studies have involved periods of frustrations and unconscious motives. Another study reported an amazingly high incidence of apparently emotional disorgani zation as a cause of fatal accidents. This finding has not been verified and needs cross-validation. Forbes particularly stressed the idea that a combination of these factors is involved in driving and that this large variety of possible factors makes the problem quite difficult. A different approach to the over-all aspects of traffic safety was the one taken by Haber, Brenner, and Hulbert (1954). These authors analyzed highway-accident focal points in terms of possible factors related to driver behavior. They evolved a "psychology of trip geography." This concept demonstrates that known psychological behavior patterns combined with the geography of the area on which a particular trip is planned tend to produce driving con ditions that may lead to accidents. Some of the recognized psychological behavior patterns considered are the setting of aspiration levels in driving, the rigid adherence to previously made plans, general fatigue, and performance letdown as the end of the driving task nears. Trip geog raphy refers not only to terrain and weather, but also to the spatial layout of routes as well as stopping places enroute. Tillman and Hobbs (1949) state that accidents are not always chance happenings, and that sometimes they reflect the basic personality of the individuals. On the basis of interviewing taxi drivers and observing them at work, it was demonstrated that the high-accident driver most frequently comes from a home marked by parental divorce and instability. During childhood his life is marked by evidence of instability and disrespect for 22 organized authority. In adulthood his occupation record is marked by frequent short-term employment and involuntary terminations; he has a police record aside from traffic violations, more frequently so than drivers with low- accident records. His personal life is marked by some evidence of social disregard as noted in other aspects of his life. His driving Is marked by some tendency of aggressiveness, impulsiveness, lack of thought for others, and a disrespect for authority. It may be said that "a man drives as he lives." Just because we know that emotional adjustment is involved in the problem of traffic safety does not mean that we can thus settle the matter. We cannot properly separate safety attitudes from emotional maladjustment; they are too interrelated. We must deal with the whole man, a person of many facets, and one dominant personality subject to many possible influences. This thought comes from Brody (1959). Stewart (1958) raised the interesting question of whether psychologists can measure driving attitudes , and gave a survey of the previous attitude research. Conover developed the driver attitude inventory in 1947. Siebrecht 23 developed an attitude scale in 1941. Both of these inven tories had high reliabilities but undemonstrated validities. In 1948, the ENO Foundation and Lauer found very low cor relations between the Iowa State Multi-Attitude Scale and accident records of drivers. Here, however, there was a question of the range of accidents and the reliability of scales. Selling (1940) did a clinical study but did not show how to measure driving attitudes and how to determine their relationship to driving behavior. He cited evidence to show that the accident-repeater driver is primarily a medical (psychiatric) problem. This leads to some problem areas. What is attitude? What is the correlation between action tendency and the real action behavior? What is the criterion for good, safe driving? There is also the bias in reporting viola tions and accidents , as only an unknown fraction of all traffic violations and accidents is observed and reported to local enforcement authorities. What is the extent to which the problem drivers have near accidents , which are unreported? Are driving habits specific or general? If driving habits prove highly specific to the individual and 24 to the traffic situation, then it is doubtful that more than a small correspondence between measures of attitude and driving behavior will ever be obtained. Furthermore, there may not be a correlation between attitude and driving behavior, since Case and Stewart (1956) showed that while test scores changed as a result of safety slogans and emphasis on safe driving, no corresponding change in driv ing behavior was shown. Miller (1952) provides a good review of the litera ture and an excellent bibliography of the personality and attitude characteristics of accident drivers. She found that originally significant Rohrshach differences between streetcar drivers Involved in accidents and those not involved in accidents washed out upon cross-validation. However, she lumped chargeable and non-chargeable accidents together, and adopted a different criterion in the cross- validation group for the measure of accidents. She also conducted the cross-validation in a different section of the country. Matching in the control group for the cross- validation was significantly off in terms of experience. These differences could have contributed to the lack of cros8-validated results, although the Rohrshach may not 25 have been a valid instrument. However, her original findings were that meticulousness and normal counon think ing differentiated high and low accident streetcar operators. Case, et al. (1956) used trained interviewers to determine the factors surrounding the violations and the psychological characteristics of habitual violators. There was a tendency for the violators to view both the law and police favorably. A tendency appeared not to commit the traffic violation which a violator classified as being the most serious. However, there were no significant attitudes or attitude tendencies to differentiate habitual violators from other drivers. Brody (195 7) in the analysis of personality adjust ment and accidents found the following areas of difference between control drivers and accident-repeater drivers: attitudes toward parents, guilt feelings, fears, and reality level. He concluded that the problem of safe, lawful and courteous driving is primarily a problem of emotional make-up and social adequacy. The so-called psychophysical functions (reaction time, glare recovery time, and so forth) do not by themselves differentiate 26 between good and bad drivers. The bad drivers may actually excel in these functions in many instances, while the good drivers may occasionally be inferior without apparent jeopardy to their driving records. Finally, the chronic violators and accident repeaters, the extreme problem drivers, are apt to be aggressive, intolerant of others, resentful of authority, lacking in responsibility, and impulsive. The origin of such characteristics is likely to be obscure. McFarland (1957) lists a history of previous accidents as the best predictor of accidents , and follows with the personality structure of the accident-prone drivers , which he regards as eccentric , impulsive and mildly psychopathic. Furthermore, he states that it is possible to identify 85 per cent of the accident-repeater drivers using the number of contacts with social agencies , such as the police, car violations, and so forth. However, the number of cases was low, only twenty. Suhr (1953) developed a prediction equation using selected scales from the Cattel Sixteen Personality Factor Test. The factors with the highest weights from the 16 PF inventory were super-ego strength (the responsibility and 27 the emotional maturity of the individual), and second, autla (or the unconventlonality) of the driver. While these two factors had the highest loadings, other factors that were significantly involved were the following: G or general intelligence, emotional stability or ego strength of the individual, dominance or ascendance, surgency or relative stability of mood, the guilt proneness or worry ingness of the individual, control or exactingness of self-sentiment and, finally, erglc tension or how excitable or tense the person is. Conger (1956) reported on research that investigated many personal and interpersonal factors in accidents. He had a control for exposure and made a special effort to ensure the accuracy of the accident and violation data. Using airmen it was found that the only major attitude or temperament scales to survive the cross-validation were three of the scales from the Allport-Vernon Study of Values. The problem drivers were higher than good drivers on theoretical values and aesthetic values, and lower on religious values. Other personality and temperament scales were significant originally, but were not signifi cant in the cross-validation. Conger's conclusions were 28 that the accident subjects are more complex, they are less conventional, less in harmony with their environment, and have complex defense mechanisms. McGuire (1956b) administered the Rozenzweig Picture Frustration Study to a group of persons who had been involved in a recent accident and second t to an accident- free group. He found the Ego-Defensive higher and the Need-Persistence scores significantly lower for the acci dent groups and that he could predict these drivers 65 per cent correctly in a cross-validation. The drivers were matched on number of years of driving experience, estimated annual mileage, age, parental and marital status, educa tion, number of years in the military service, as well as on GOT scores. McGuire (1956a) also did an item analysis of the MMPI, the Bell Adjustment Inventory, and the Kuder Personal Preference Record. He found that the Pd (Psycho pathic Deviate) and the Sc (Schizoid tendency) scales on the MMPI were significantly different for these two driver groups. He concluded that the accident- and violation-free driver is more mature, conservative, intellectual in his interests and tastes, has a higher aspiration level, and is the product of a happier family background than is the 29 accident- and violation-incurring driver. Brown and Berdie (1960) followed up on drivers for whom they had obtained MMPI data when the drivers had been college freshmen four to six years earlier. The driving records were investigated and very small but statistically significant relationships were obtained between scores on the Psychopathic deviate and the Hypomanla scores on the MMPI and the number of accidents and violations for these drivers. The small size of the correlation might be due to the existence of more than one kind of personality pattern related to poor driving, but also it very well could be related to the fact that there may have been per sonality changes in the subjects as a result of the lapse of four to six years. That such changes do occur has been demonstrated. Part of the Status Report about driver rehabilita tion by the Drivers Safety Service (1959) concerns the cross-validated results using the Thurstone Temperament Schedule for drivers processed by the Accident Prevention Clinics in New Jersey. The drivers with two or more accidents in the previous 18 months were less active, more stable, and more sociable than the control or better 30 drivers. In a second comparison, drivers with two or more accidents were also more dominant and less reflective than the better drivers. Age and mileage were controlled. Heath (1957) studied the relationship between driv ing records, selected personality characteristics, and biographical data of traffic offenders and non-offenders. A non-offender was a driver with a clear record over approximately a five- or six-year period, whereas a traffic offender was defined as a driver with three or more traffic violations and/or three or more accidents for the same five- or six-year period. Using the Thurstone Temperament Schedule, he found that the offender drivers were higher on the impulsive and sociable scales , and lower on the reflective scales than the non-offender drivers. He also reported this biographical information: The offender drivers typically were young (between 18 and 33), unmarried, had not graduated from high school or college, were in non-official or non-professional jobs , earned less than $5,000 a year, had had more than one job in the previous five years , and had terminated jobs for other reasons than self-betterment. He had not had time to do a cross- validation study. He had matched the groups only on 31 driving experience as a rough control for exposure. He found, surprisingly, that the biographical data were better than temperament data in predicting the offending drivers. Nathanson (1953) also used the Thurstone Temperament Schedule in an attempt to differentiate drivers with one or more citations versus those with none. He could predict the number of citations of the driver using the driver's score on the Vigorous scale. Other scales were not pre dictive and accidents were not predictable at all. Age, exposure, and length of driving period investigated were not controlled. This study is interesting further because the fakeability of each item was investigated. All items were fakeable at the 1 per cent level, but some were more so than others. A special scale was constructed of hard- to-fake items; also, a ratio was made of scores on hard- to-fake items compared to easy-to-fake items. These scales apparently were not useful in predicting citations or accidents. Morman (1955) tried to develop a selection system for motor-coach operators. However, he found that he needed more reliable criteria and relevant predictors. He 32 used the Otis , the MMPI, the Johnson Temperament Analysis, the Kuder Personal Preference Record and biographical items. It is not known why he could not predict accidents or moving vehicle violations. It could be that motor- coach operators are a different breed of drivers, or that different factors are involved in the safe operation of motor buses. Statistical Studies Jacobs (1955) points out some of the problems in applying mathematical models to accident processes. The driving situation is not transferable to the laboratory for study, observations are limited to accident records for some particular risk, there is little chance of delib erately exposing a selected sample in true experimental fashion to some risk, and the data are usually limited since accidents are rare. For these reasons, statisticians usually turn to mathematical models in an attempt to describe a given set of accident observations. Jacobs goes on to list three factors involved in accidents as (1) the accident liability is randomly distributed in population; (2) liability to accidents changes linearly with accident experience (linear contagion); and (3) there 33 is an effect due to tine alone, primarily the reliability, or lack of reliability, of the accident criterion itself. Statistically, it is quite difficult to tell the different causes of accidents as judged by accident distributions for a given sample of people. Marltz (1950) in a discussion of the Polsson and negative binomial distributions , observed that fitting of the observed distributions to these theoretical distribu tions is not an adequate criterion for detecting accident proneness. The indispensable technique is the correlation of consecutive periods of exposure to accidents. In discussing repeater pilot accidents , Webb and Jones (1952) point out that there is little relationship between the number of accidents that a pilot has in one period and those he has in another time period. This is supported by several methods of analysis. The maximum reliability found was 0.45. Kraft and Forbes (1944) point out that the truly accident-prone driver is relatively rare, accounting for less than 4 per cent of the traffic accident problem. Accidents for streetcar operators are governed by the laws of chance. The authors developed an accident index for 34 each driver representing the number of accidents he would have been expected to have under various conditions of time, weather, mileage, route and so forth. They found a low degree of reliability between the accident index and personal items. Hughes (1950) discusses the discrimination of the accident-prone individual. Even if all persons in a given population had the same degree of accident proneness, a large proportion of the accidents would occur to a small number of Individuals. This fact in itself would account for the Polsson distribution of accidents as usually found. However, since the Poisson distribution does not quite fit the observed distribution of accidents, one can conclude that the accident rates within a group of people are not homogeneous; that is, the accident rate varies from person to person. However, before one could conclude that the people with the greater number of accidents in a given time are accident-prone individuals , one would have to correlate their accident histories over consecutive time periods. Wolbers (1955) used an index of flying experience as a measure of exposure to hazard, and then used the dif ference between the number of accidents expected for an 35 Individual on the basis of this index of flying experience and the number of accidents that actually occurred to a person as an index of accident proneness for the individ ual. He also used the traditional number of accidents as an index of accident proneness. The accidents above those accounted for on the basis of chance and exposure were Investigated. A reliability measure proposed by Newbold and Cobb was applied to the distribution of accidents for aircraft. This index of reliability was found to be quite small (0.34). Hakkinen (1958) discusses the underlying logical and statistical concepts of traffic accidents and driver char acteristics. He evaluates the strength and weakness of each of the various measures of accident proneness , such as comparisons of the distributions, mean differences, and correlations. In a study of 1,000 bus and tram drivers in Finland, Hakkinen divided the causes of accidents into two groups: those due to individual factors , and those due to the environment or exposure factors. He equalized the environmental risk for each subject, to get at their indi vidual factors of accident proneness, if any. Even though he found a high reliability (0.60) for the individual 36 factors in the accident frequencies of the drivers, the individual factor was very small compared to the environ mental risk factors. That is, exposure accounted for most of the accidents. The author concluded by stating that accident research needs more concepts and theoretical systems to understand accident proneness, rather than just more correlations and other statistical work. The exposure index for measuring the risk element of different types of behavior is in question when it comes down to what is actually used to measure exposure. For instance, Battey (1959) is in favor of evaluating exposure to motor-vehicle accidents as varying as the square of the number of vehicles. As a logical extension of this, he recommends retaining the usual method of calculating exposure by squaring the vehicle mileage per year, regard less of the amount of travel on different types of high ways . However, exposure should be calculated separately for urban streets and rural highways. Mathewson and Brenner (1956) find limited usefulness in expressing risk in terms of the annual mileage driven. Instead, they favor a risk or an exposure index that is related to the number of accidents occurring at a given 37 point on a given highway during some period of time that is related to the volume of vehicles passing that point. This is a volume exposure index instead of a mileage exposure index. Computing such an index poses practical difficul ties in obtaining the necessary data per driver. Obviously, research will be needed to evaluate whether this, or annual mileage, would be an adequate index to control for exposure. Many other non-personal factors could account for accidents, as well as exposure to accidents; these could be volume of vehicles past a given point, or the annual mileage given by the driver, or the relative annual mileage on urban-versus-rural highways. Since the frequency of accidents by itself is not a reliable index, Fine (1958) proposed to investigate driving behavior in order to discover valid criteria for evaluating driver behavior. Specifically, he proposed the use of the critical-incident technique, to be used as follows: a professional jury would pass on collected data for validity and list the critical requirements for the near-accident situations. This listing of critical requirements then would be used to construct a test. 38 Another approach to accident proneness may apply to the traffic-accident situation and is reported by Keehn (1959). He reported a factor analysis on one hundred university students and utilized statements of all acci dents and minor mishaps. He found a general factor to run through all types of accidents and minor mishaps, indi cating that individuals who admit to having accidents in one situation, also indicate they have been involved in accidents in other situations. Such a finding does not contradict the notion of accident proneness and suggests the possibility that some minor accidents and mishaps might be predictive of subsequent major accidents. This, of course, calls for further research. Faking Studies In attempting to develop an attitude scale or scales predictive of problem-driving behavior, consideration must be given to faking. Mayo and Guttman (1959) reported that significantly higher MMPI lie scores were made under directions to fake vocational inventories and second, that significantly higher fake scores were found in one out of two groups under the condition involving knowledge that a falsification score would be computed. The faking scores 39 were in the favorable direction, but the authors conclude that faking on the two self-report inventories in a mil itary, vocational-classification situation is minimal, even though the experiment provided a favorable opportunity for faking to manifest itself, as it was not observed to any appreciable extent. The authors' conclusion here is that the significant differences reported made no practical differences; that is, the shifts were significant but slight. Sheldon (1959) experimented with conditions affect ing the fakeability of teacher-selectlon inventories. For all but one of the scales , the order in which the tests were given (standard conditions versus faking Instruction conditions) was a significant condition in determining how well the scale could be faked. In each case, the groups of students who first responded honestly changed the scores more than those who faked on the first administration of the test. Sheldon concludes that fakeability should be con sidered a characteristic of psychological inventories just as reliability and validity are characteristics. Fake ability varies in the same way from one group of subjects 40 to another> just as reliability and validity vary. Steinmetz (1932) measured the ability of testees to fake interest in any occupation. The subjects were able to fake interest in any occupation inversely in proportion to their initial score on that occupation. Second, faking distorts at least half of the remaining occupational scores. CHAPTER III METHOD AMD PROCEDURE After the literature was studied, these procedural decisions were made: 1. This research would concentrate on the attitude and temperamental variables involved in problem-driving behavior. 2. Biographic or demographic variables would be measured for matching or screening purposes , as well as for checking their relevancy to problem-driving behavior. 3. Aptitude and psychophysical variables, while important, would largely be ignored due to their relatively long administration time. 4. Statistically, an attempt would be made to detect unfavorable test-taking attitudes, or biases. Emphasis would be placed on careful checking of driver records to obtain accurate histories of accidents and violations for each driver. 41 42 Traits or factors included in the surveys either had been reported by more than one investigator or by the same investigator in a cross-validation. The traits utilized were the variables or scales closest to these twice-reported variables and that could be found from local tests. Temperament and attitude scales will be discussed first. Item Selection Nine traits were selected from the Guilford- Zinmerman Temperament Schedule (Guilford and Zimmerman, 1949). The definition of these traits and the number of Items selected and used per trait are as follows: General Activity (15 items) High score: energecic, active; has strong drive; likes speed; is enthusiastic; quick in action. Low score: likes slow pace; pauses for rest; takes time, Inactive due to physical causes. Restraint (15 items) High score: over-restrained; over-serious; delib erate and persistent; self-controlled. 43 Low score: happy-go-lucky; carefree; loves excitement; is impulsive. Ascendance (12 items) High score: self-assertive; socially bold; likes to take the role of a leader. Low score: submissive; socially timid; avoids conspicuousness. Sociability (10 items) High score: likes social contact and activities; interested in making conversation; has many friends. Low score: dislikes social activities; avoids social contact; has few friends. Emotional Stability (12 items) High score: optimistic, cheerful; has evenness of moods; is composed. Low score: pessimistic, gloomy; daydreams; excit able; has fluctuation of moods. Objectivity (15 items) High score: less egocentric, less sensitive; "thick skinned." 44 Low score: hypersensitive; self-centered; suspicious. Friendliness (18 items) High score: compliant; agreeable; respects other people. Low score: hostile; resists domination and desires to dominate; is ready to fight; has contempt for others. Thoughtfulness (15 items) High score: reflective, meditative; is observant of self and of behavior of others; interested in thinking; has mental poise. Low score: is interested in overt activity; poor observer of people and of himself. Personal Relations (15 Items) High score: is tolerant and understanding with respect to other people. Low score: finds fault; is critical of other people; has difficulty in getting along with others. 45 Eight traits and their items were selected from the DF (Dynamic Factors) Opinion Survey (Guilford, Christensen and Bond, 1954). These were as follows: Need for Attention (12 iteau) High score: craves recognition; enjoys status; is exhibitIonistic. Low score: recognition means little; status is unimportant. Adventure vs. Security (15 items) High score: likes to explore, to take personal risks; is generally bold. Low score: seeks security; avoids danger. Self-Reliance vs. Dependence (15 items) High score: self-reliant; responsible; dependable. Low score: dependent; seeks support; subservient. Aesthetic Appreciation (11 items) High score: enjoys art in ell forms--drama, music, literature. Low score: not inclined to enjoy art. 46 Cultural Conformity (14 items) High score: fully accepts social customs; highly developed conscience and ethical awareness. Low score: rejects social customs; little bothered by conscience. Need for Freedom (15 items) High score: likes freedom; nonconformist; dislikes system and order. Low score: likes order, systems, and organized life; accepts control. Realistic Thinking (15 items) High score: takes realistic view of self; matter- of-fact attitude; forthright and direct. Low score: prone to wishful thinking; appreciates humor; expresses hostility indirectly. Need for Precision (10 items) High score: likes exactness, precision, and detail. Low score: dislikes exactness, precision, and detail. 47 One trait was selected from the (unpublished) Guilford-Holley Inventory, an inventory by J, P. Guilford and J. W. Holley. This was: Aabitiousness (15 items) High score: places high value on success in life, likes to possess material things. Low score: not interested in success. An additional scale used was the Driving Attitude scale which had been empirically derived several years earlier by the author from reactions of problem drivers compared with better-than-average drivers matched on age and sex. This was: Driving Attitude (31 items) High score: typical of drivers with many moving violation tickets and/or accidents. Low score: typical of better-than-average drivers (few moving violation tickets and/or accidents). Still another experimental attitude trait was called Spiritual Values. This had items similar to those of the 48 Religious Values scale from the Allport-Vernon Study of Values. Instead of a forced-choice format, the Items were to be answered by simply "yes" or "no." The hypothetical definition was as follows: Spiritual Values (15 items) High score: likes thinking about ethics, values religious activities. Low score: dislikes thinking about ethics, not interested in religious activities. The published scales mentioned from the various standard inventories typically had 30 items in each scale. It was desired to cover as many traits or variables as possible; accordingly, the length of the scales was reduced from 30 items to about 15 items. The length of a scale was reduced in proportion to its reliability. Each scale was given enough length so that the reliability did not fall below about 0.65. The items were selected so as to be most appropriate to drivers. The items were put in a rotated order in the inventory to facilitate scoring. A few other traits were hypothesized by the investi gator to be involved in differentiating the accident and 49 violator drivers from the average or better-than-average driver. These hypothesised characteristics were as follows: 1. A tendency while driving to let attention wander occasionally to other personal problems, Instead of concentrating on driving. Attentiveness vs. preoccupation. (8 items) 2. Emotional immaturity in solving life problems such that an emotional conflict perseverates for a period of time. Emotional perseveration and frustration. (20 items) 3. A type of solution to personal problems which involves exposure to accident situations and to potential or actual injury. Risk, domination and lawlessness. (26 items) Special items were written in an attempt to measure the attitudes underlying the above three hypothesized personality characteristics. Typical items were as follows: 1. Attentiveness: You usually have the radio playing while driving. 2. Frustration: The slow driver in traffic is 50 Che reel menece. 3. Risk: You like to drive £est for Che Chrlll of 1C. In Che mldsc of laCer testing, It was suddenly realised Chat no items had been Included Co measure atti tude Coward consumption of alcohol and driving. That this is important can be seen from the Department of California Highway Patrol, Annual Statistical Report, 1959. Drivers had been drinking (HBD) in 11 per cent of all accidents involving injury an> in 21 per cent of all fatal accidents. Checking the traits measured in the Manson Evalu ation (1948) for the prediction of alcoholic addiction was consoling; most of the traits were covered by items used in this research. However, in the revised Driver Attitude Survey, 10 items relating directly to alcoholism were included for research purposes. Most of the theoretical items and the Driver Attitude items developed previously were placed with bio data items and items related to accident involvement in the first survey called the Driver Attitude Survey. This survey consisted of 65 attitude items (the hypothesized and some driving-attitude items), the 20 bio-data items, 51 and 10 driving emergency reaction iterne to make up a total of 95 Items. Finally, 9 more questions were asked concern ing each accident in which the driver might have been Involved while driving within the last three years. The items for the 20 personality traits were placed in rotated order in the second survey, called the General Attitude Survey. They totaled three hundred. Ten of the scales were placed on one side of the answer sheet and the remain ing 10 scales were placed on the reverse side. The biographic, demographic or "bio-data" items were also selected on the basis of research by others. Three of these items were recorded only on the Driver Attitude Survey answer sheet: age, sex and race. Due to the current controversy over desegregation, race was not answered directly as an item by subjects. Instead, the test administrator, at the time of checking over the answer sheet for completeness, discretely recorded race in one of five answer spaces in the lower right-hand corner of the answer sheet. Actually, "race" should be modified to race or ethnic group, since the five classifications were as follows: White, Negro, Mexican or Spanish, Oriental, and other. The bio-data items included in the Driver Attitude 52 Survey were insurance coverage for Che subject's car, accident history and Insurance claim history, physical and mental health items, years of school completed, years of operating a motor vehicle, average annual (or daily) miles of driving, marital status, number of dependent children, number of non-driving accidents in the last five years , amount of military service, number of employers in the last two years, occupation, mechanical condition of the subject's car, amount of rural or country driving, amount of rush hour traffic driving, amount of night driving, whether the subject had taken a course in school to learn how to drive or to improve his driving (i.e. , the Driving Improvement Course), the type of vehicle driven most (such as car, bus, motorcycle or other), the number of traffic accidents or collisions in which the subject had been involved while driving within the last five years , and the nuaiber of traffic violations in the last five years. As part of the hypothetical items relating person ality characteristics to driving behavior, it was decided to construct several items that had to do with the reaction of the driver in emergency driving situations. There were ten of these five-choice items. A typical item was as 53 follows: You *rs signaling to turn left in an Intersection; there Is e car coning rather fast toward you in the opposite lane. As you start your left turn, a pedestrian starts across in front of you. What do you do? a) Check for traffic behind you and then back up in order to clear the lane for the oncoming car. b) Warn the pedestrian by blowing your horn and then continue slowly ahead. c) Come to a complete stop even though you block the oncoming car. d) Turn your wheel more to miss the pedestrian and keep going. e) Drive up on the curb to avoid both the pedestrian and the other car. If the person had been Involved in an accident while driving, within the last five years, he was requested to fill out the following information for each such accident Involvement: year, quarter, side (direction) of the collision or accident, what was wrong (such as vehicle failure, poor visibility, poor road condition, condition 54 of the driver or pedestrian), the extent of damage to the vehicle, the extent of injury to person(s), what could have been done to prevent the accident, whether the driver was thinking about something else, distracted, in a hurry, or finally, whether the driver was in an emotional frame of mind. Dependent Variables The two dependent variables to be predicted were the number of accident involvements while driving in the last three years and the number of moving violations that the driver had had while driving within the last three years. These variables were treated as independent. In an analysis of data published by the State of California Department of Motor Vehicles, it was determined that the eta correlation between violations and accidents is nearly 0.26. This is sufficiently low so that for all practical purposes the variables could be treated as inde pendent variables. Accordingly, this was done. The accident variable was treated by assigning drivers to one of several categories of accident involve ment, depending on a number of such accident involvements. The mildest accident group was called the Recent Accident 55 (RA) group. For this a driver had to have been Involved In an accident within the laat 45 daye previous to testing. The driver need not have been responsible for this accident as determined by the police records. The second accident group was the Two Accident (TA) group. For this a driver must have been involved In two accidents within the pre vious three years. Actually, if the person could not have been put in the following category, he might have been put in this group with more than two accidents because respon sibility for them could not have been determined. The third accident group was the Accident Repeater (AR) group. For this the driver must have been involved in three accidents within the last three years, of which he was responsible, or partially responsible, for at least two of the accidents; or the driver may have been involved in four accidents with responsibility shown for at least one of the accidents; or, finally, the driver might have been involved in five or more accidents in the last three years without any responsibility demonstrated for the accidents. This graduated definition of an accident repeater was adopted because responsibility could not be determined in all cases from the police or Department of Motor Vehicle 56 records concerning the accident. It was felt that if a driver had been involved in a greater number of accidents , it was more likely that he was responsible for a greater number of his accidents. The second dependent variable was that of moving violations. There were four groups varying in severity in the violations. The first group was the Minimum Violator (MV) group. To be eligible for this group, the driver must have met the Department of Motor Vehicles definition of a negligent-operator (N.O.) driver. This definition is as follows: If the driver has had a total of 4-point counts in the preceding 12 months, or 6-point counts in the preceding 24 months, or 8-point counts in the preceding 36 months, he is classified as a negligent-operator driver. In the point-count system, any ordinary non-serious moving violations (such as going through a red light or failing to stop at a stop sign) are given a point count value of one. More serious or dangerous types of violations are given a point count of two. These are five more serious violations: falling to stop after an accident (hit and run); striking an unattended vehicle and leaving the scene of the accident (a second type of hit and run); driving 57 while intoxicated; reckless driving; and finally, driving while the subject's driver's license had been revoked, cancelled, suspended, or not issued. Finally, in the point-count system, an accident for which the driver was responsible or partially responsible was counted as one point. The second violator group was the Extended Violator (EV) group. To be classified here, the driver must have had a minimum violator record within the preceding three years (a current record) and a record extending back beyond the preceding three years. That is , some time within the previous three years the driver must have had at least a point count of four, as well as currently he must have had at least 2-point counts within the previous year; and further, he must have had at least 2-point counts beyond the previous three years. The third violator group was the Serious Violator (SV) group. The qualifications for this group were as follows: The driver must have had at least a minimum negligent-operator count; and second, at least one of his violations must have been one of the two-count serious violations, such as drunk driving. 58 The fourth violator group was the Chronic Violator (CV) group. To be eligible for membership in this group, the driver must have had an extreme violations record. Most of the drivers in this group had had such a poor record that their driver's license had been revoked within the previous nine months. A few of the drivers were put in this category because their driver's license had been suspended for a short period of time, but suspended for the second time as punitive action by the Department of Motor Vehicles. This second (most recent) suspension must have occurred within nine months previous to testing. The three accident-driver categories were mutually exclusive, as were the four-violator categories. However, if a driver qualified for an accident category and a moving violations category, the driver could have been placed in both categories and his test records shifted accordingly in the item analysis. However, this shifting was reduced so that only the more extreme drivers appeared in two categories, one accident group and one violator group. No driver appeared in two accident groups or in two violator groups. 59 Since there were many more drivers in the MV (minimum violator) group than in the more extreme groups, none of the drivers with minimum records were transferred. There were 44 drivers who shared one accident category and also one violation category. An initial analysis was made with a small number of chronic violators, accident repeaters, two-accident drivers, and extended-violator drivers. This initial analysis was done with the 20 personality variables and the bio-data items thought to be most important. The results of this analysis were reported at the National Safety Congress (Schuster and Guilford, 1959). A group of better-than-average drivers was selected at random and compared with the accident repeaters and chronic violators. The better-than-average drivers had had no moving violations and no accident involvements while driving within the previous three years. Initially, the accident repeaters and chronic violators were very significantly different in age and in the annual mileage driven as compared to the better-than-average drivers. Therefore, it was decided to match in the subsequent item analysis on age and annual mileage. Several of the other 60 bio-data items ware significant, but not for two or more groups. Accordingly, it was felt that the influence of these other variables was not as important as age and mileage. These items were occupation, amount of rural driving, years driven, years of school completed, amount of night driving, and amount of rush-hour driving. Testing Subjects Subjects were tested under the direction or super- vision of one of the Traffic Project staff members. Subjects were obtained in many places and from many organizations. Problem drivers were obtained primarily through the Department of Motor Vehicles. The Department of Motor Vehicles, as part of its driver improvement pro gram, recommended to drivers who were at least at the minimum violator level, that they should attend a Personal Traffic Safety (PTS) course given in the high schools in the adult education program. Many of the problem drivers (those with accident involvements or a record of many moving violations) were so tested at high schools in the *The term "Traffic Project" was the unofficial title given to the small group conducting research on problem drivers at the University of Southern California. 61 Los Angeles area, at Long Beach State College, or In Fullerton. Problem drivers were also tested by the sta££ as they came in for interviews at the Department of Motor Vehicles at 3500 South Hope Street, in Los Angeles. Problem drivers were also obtained incidentally through testing various groups throughout the area. At one point a random sample of better-than-average drivers was desired. Accordingly, a list of approximately 100 such drivers was obtained through the courtesy of the Department of Motor Vehicles in Sacramento. Letters were sent out to these drivers offering them $10, as well as a brief interpretation of their test results, for taking the two experimental attitude surveys. Of these, approximately 25 were finally tested; about 50 per cent of the drivers had moved and left no forwarding address, or for some other reason were ineligible for testing. Subjects were also tested through various other adult education classes. The students volunteered to come in for the attitude surveys in response to announcements posted on bulletin boards for the adult education classes. One civic group, the El Segundo Boosters, was tested one evening and provided approximately 50 subjects. A 62 number of insurance companies similarly participated: Farmers Insurance Group tested numbers of its staff, as did Founders Insurance Group and the Western Insurance Information Group. North American Aviation, Standard Oil, Wllshire Oil, Auto Club of Southern California, and other industrial companies in the area supplied subjects as part of their safety programs. There were also several military organi zations which provided subjects for testing: San Bernardino Air Materiel Area, Long Beach Naval Station, China Lake Naval Station, Edwards Air Force Base, and Norton Air Force Base. No attention was paid to the source from which the subjects were obtained for testing. The subjects were classified according to their accident or violations history within the previous three years and assigned to criterion groups. Subjects were informed that they would learn of the results if they were in the better-than-average category. This was the motivation or inducement for the subjects to take the two attitude surveys. Group testing wa9 done from the safety motivation of the company or 63 organization concerned and the company may have got the reaults back, but not the individual. A few of the drivers were paid $4 for taking the test. The number of such sub jects was quite small compared to the approximately 2,000 drivers tested. Finally, on the random better-than-average drivers, the drivers were paid $10 because it was quite desirable to test as many of these drivers eligible as possible. These drivers also had the results reported to them. The results were given with a brief description of the 20 personality traits and the driver's results within four levels (categories). These levels were arranged by centile scores: zero to tenth centile (or below average) , the eleventh to the fiftieth centile (or average), the fifty- first to 90th centile (above average), and the 91st to 100th centile (or outstanding). For problem drivers, a more detailed profile was sent to form a part of the Department of Motor Vehicles (Los Angeles Special file) records. Forty-two of the drivers were retested under faking conditions and, accordingly, were paid $4 each as soon as they completed the retesting. Only problem drivers were 64 requested to reteke the test under feking instructions. The £eking instructions were as follows: "We want you to pretend that you are an excellent driver. Look at each item or question to see if it would be answered differently by an excellent driver as compared to a driver with acci dents or violations. Answer each question in the 'good1 direction. Pretend that keeping your driver's license depends on making a good score on this test." Criterion Groups Depending upon their driving record for the past three years, drivers were placed in one of the criterion groups. First, however, drivers were selected who met these criteria: they were male Caucasians in good health, who had been driving only a car for at least three years in California. The driving records of the drivers were checked very carefully because it was felt that the previously reported lack of reliability for accidents and moving violations has been in part due to insufficient checking on the part of the investigators. Accordingly, the driving records of the subjects were checked three ways. Within the Driver Attitude Survey itself, several items related 65 to accident involvement and moving violations. The answer concerning motor vehicle insurance was checked, as well as the driver's answers about accidents while driving. The driver was later asked, within the Driver Attitude Survey, for the number of accident involvements while he had been driving within the preceding five years. Five years was specified to account for overlap in memory, in spite of the fact that the score or record was kept only for the preceding three years. Finally, if the driver had had any accidents , he was asked to describe them in detail on the nine special items at the end of the Driver Attitude Survey. Then the driver's record was checked according to his driver's license number, at the State of California Department of Motor Vehicles in Sacramento. If the driver was a problem driver and had been involved locally with the Department of Motor Vehicles, he had a file called the Los Angeles Special (LAS) file, and accidents and moving violations were cross-checked there. Of course, there was a duplication between the Sacramento and local driver records. However, in spite of a considerable overlap, the records did not coincide completely. The Sacramento 66 records had up-to-date Information about accidents and moving violations from all parts of the state of Califor nia, while the local (LAS) records might have more detailed information about the accidents of the subjects, and they would also have information about the driver that extended back in time for more than just the last three years. The Sacramento files are purged of any record of accidents or violations that happened more than three years ago unless some legal action, such as suspension of license, had been taken against the driver. To summarize, the two dependent variables very carefully cross-checked were the number of accident involvements in the previous three years and the number of moving violations within the previous three years. Accord ing to his number of accidents and moving violations, a driver was assigned to one of the three accident groups, to one of the four violator groups , to one of the three better-than-average groups, or was excluded. A few drivers may have been screened out because they had had just one accident that had occurred more than 45 days previous to the date of testing, or drivers may have been excluded because they had had between one and 67 three moving violations in the previous three years. Thus, the drivers would have fitted neither into the better-than-average driver category, nor into any of the accident or moving violator driver categories. In justi fication of the term "better-than-average," the average number of accidents per driver over a three-year period was 0.16 in California; the figure for violations was 0.85. The criterion groups that were submitted for item analysis were as follows: the better-than-average groups were drivers with no moving violations and no accident involvements within the preceding three years. The first such group was composed of better-than-average drivers (BTA-no restrictions) with no restrictions on either age or annual mileage driven, N being 59. The second better- than-average driver group (BTA-younger) was composed of drivers who were younger than the average better-than- average driver, N being 20. The third better-than-average group (BTA-selected) was one consisting of drivers who were younger than average, but also who drove considerably more miles than the average better-than-average driver, N being 39. 68 The eccident-driver groups were as follows: The recent-accident (RA) group consisted of drivers who had been Involved in an accident while driving within the 45 days previous to testing, M being 52. The two-accident- driver group (TA) consisted of those drivers who had been involved in two accidents within the three years prior to testing, or three or wore accidents, but did not fit into the next group, N being 65. The third accident group was the accident-repeater (AR) group, people who had been involved in three or more accidents within the preceding three years, with responsibility for at least two of the three accidents, or involved in four accidents and respon sibility demonstrated for one, or the driver had been involved in five accidents, N being 71. The four violator groups included the mlnimum- vlolator group (MV) who had met the Department of Motor Vehicles legal definition of a negligent-operator (N.O.) driver, N being 100. The extended violators (EV) , in addition to having met the legal definition of a negligent- operator driver (with a minimum violator record) had a current record as well as a record extending beyond the last three years, N being 64. The serious violators 69 (SV), in addition to having at least a minimum number of violations, had as part of their record, at least one serious violation, II being 51. The chronic violators (CV) had an extreme record such that most of then had had their drivers' licenses revoked (or suspended for the second time) within the previous nine months, N being 27. Finally, to check on item fakeability, the item percentages for all items under standard administration instructions were compared to those under faking-excellent- driver (FED) instructions. The number of problem drivers in this test-faking retest group was 42. Item Analysis The test records for each group were analyzed using the IBM test scoring machines of the Testing Bureau at the University of Southern California. Since three categories of response (yes, 7, and no) were permitted on the answer sheet for most items, the tallies for both the yes and no answers were computed for each group. From the item tally sheets, the totals of the yes and no response were con verted to percentage responses for each item for each group. The percentage of 7 ^ responses for each item was included with either the yes or no response, whichever 70 maximized the difference between the problem-driver groups and the BTA group. Approximately 5,000 percentages were computed. An abac by Guilford (1941) was used to convert the percentages on each item per group to Chi Square values. A significance level of 5 per cent was adopted. Since the abac was constructed only for groups where the number of people per group was the same, several prelim inary calculations were done to compute the equivalent N's so that the abac could be used in spite of s different number of subjects per experimental group. In case of doubt, where the Chi Square value was very close to the 5 per cent level, the actual value was computed by hand. In all cases, the accident- and violator-group percentages (p-values) were compared separately with the combined better-than-average (BTA) group of 118 people. If the item were significant, the age and mileage groups were checked for differing p's within the better-than-average groups. If the younger and higher mileage groups had significantly different probabilities for that item, even though it was significant between the problem-driver group and the combined-BTA group, then the item was discarded. However, there were only a few such items. 71 Additionally, If the item ware answered differently by the drivers who retook the same tests under faking- aatcellent-driver instructions , then the itesi was put in the F or faking scale, in spite of the fact that it was significant for the problem drivers versus the BTA drivers. Again, there were just a few such items. Most of the items that were significantly discriminating between the problem- driver groups and the BTA drivers were not fakeable. A scale was made of all the iteais significant at the 5 per cent level for each criterion group. The appro priate criterion group and the selected (matched) better- than-average driver group were scored using this derived scale. Then the discrimination value of the derived scale was checked for the original criterion group and the better-than-average driver group. Instead of having so many scoring keys for each of the accident and violator groups, it was decided to combine the accident groups to derive one scale, and the items from the violator groups to derive a second scale. This pro cedure rests upon some cross-validation. Items for the A (Accident Attitude) scale were significant at the 5 per cent level for two out of three accident groups or were 72 significant for one of the three groups, and for the other two groups combined at the 5 per cent level. The V (Violations Attitude) scale was composed of items that were significant at the 5 per cent level for any two of the four violator groups. Mote that for some purposes, it might be desirable to retain the more extrsme driver scales for further research purposes, notably the accident-repeater and the chronic-violator scales. The discrimination pro vided by the A and V scales was checked on the original accident and violator groups versus the better-than-average groups. Validity Measurement Three validation scales were derived. The items that were answered significantly differently under faking- excellent-drlving instructions comprised an F or Faking scale. For these items the percentage of response shifted significantly under the faking Instructions. Upon inspect ing these items, it was found that they are desirable Items in terms of driving behavior, and as such, the items would be similar to the L or Lie scale items of the MMPI, but in a driving context. 73 A second validity seals was constructed using items with an extreme percentage of response in one direction. The items had to be answered almost always In one direction for all of the driver groups. The unusual direction (P <■ .10) was scored for the Deviance (D) scale. This scale Is similar to the JMPI "F" scale, but is somewhat different from the F scale because the probabilities or percentages were extreme for all the driver groups, not just one group versus the other groups. As such, the scale would measure such things as mistakes in answering the items, or trouble in reading the items, or finally, an unusual attitude and the items were correctly measuring this unusual attitude. It is Interesting to note in pass ing, that many of the hypothesised items (e.g., attentive ness) that were specially constructed, were significant for either the D or the F scales, but most were not significantly discriminating between the problem-driver groups and the better-than-average groups. Problem drivers from the original criterion groups who were not correctly predicted in the checking of the derived A and V scales were used in a second item analysis. Misses were defined as either of two classes. If a driver 74 had a low V or A score and the driver actually belonged in one of the accident or violator groups (problem-driver groups), then the man was counted as one type of miss. If the driver were actually a better-than-average driver, but yet had a high V or A score such that he would have been counted with the problem drivers, then this was a second type of miss. Accordingly, 42 problem drivers who were type-I misses on the A scale, and a second 42 problem drivers who were type-I misses on the V scale, were used In a second item analysis. Only the 300 items in the General Attitude Survey were analyzed by comparing answers with those from the better-than-average driver group. The itesm that were significantly different between the two miss groups and the better-than average group were put in the X or ''misses" scale. About half of the X items were close to being significantly different for one group and were significant when both groups' data were combined. Cros s-Validation In the Interest of saving time and generalizabl11ty, it was decided to conduct the major part of the cross- validation using records for drivers already tested, but who had previously been omitted from the item-analysis 75 studies. Drivers could harve been screened out due to physical or mental health probleats, being finale, usually driving a vehicle other than a car (such as a truck, bus or motorcycle), or being non-White. Obviously, if the driver had not been in the state for at least three years, he was screened out. Finally, a driver could have been screened out because he did not complete both of the tests (395+ items). The drivers who were screened out for bio data reasons (the four first variables mentioned) comprised the bulk of the cross-validation. These four driver groups were each subdivided into good (BTA) and problem drivers, once for violations and again for accidents. Subjects were added or deleted appropriately to the subgroups to match on age and annual mileage. Since very little prediction was possible for women's violations and accidents, only the first three groups' data were combined for a larger N. Work with a new format of 100 items was also done for further cross-validation. The new test format was called the Driver Attitude Survey (revised). The new format comprised 100 selected items in rotated order for the D, F, X, V and A scales. Finally, in the new format, 76 10 Items were Included as research Items to get at alcoholism as a factor previously omitted In the attitude Items. CHAPTER IV RESULTS AND DISCUSSION Discrimination on Criterion Groups Each derived scale was checked for its discrimina tion between the criterion groups upon which the scale was based. Thus, for example, the AR (Accident Repeater) scale was checked for its power to predict membership in the Accident Repeater group or the Selected BTA group. Obvi ously, if the derived scale proved unable to discriminate between the groups upon which it was based, there would be no reason to attempt a cross-validation. The method was as follows: Cumulative distributions were made of the derived scale scores both for the respec tive criterion group and for the Selected BTA group. (This group was used for the comparisons because the average age and annual mileage were closer to the problem- driver group averages than were the Combined BTA group's data.) A cutoff score was selected empirically which 77 78 maximized the percentage of drivers correctly classified, or "hits." Since the umber of drivers in the criterion groups varied considerably, this cutoff score was restricted to lie between the median scores of the two criterion groups; in an extreme case, the cutoff score may have been the same as one median score. The results of the discrimination checks are pre sented in Table 1. Note that the A (Accidents attitude) and V (Violations attitude) scales were checked along with the specific driver group scales. The discrimination in these internal consistency checks was considered satisfactory. In Table 1, note that the various accident attitude scales resulted in 75 to 84 per cent correct classification of the criterion group drivers. The various Violation attitude scales correspond ingly gave figures from 74 to 88 per cent correct classifi cations . In general, the more severe criterion groups were classified more accurately. Significant Items and Traits A correlational analysis between the dependent variables of moving violations and driving accidents, and the various predictors is very interesting. One can also 79 TABLE X.— Discrimination between problem driver criterion groups end the Selected Better-than-Average (BTA) group by means of derived scale scores Problem Driver Group Scale Cutoff Score Per Cent Correct* Accident Repeater (AR) AR 17 84 A 7 80 Two Accident (TA) TA 6 78 A 7 75 Recent Accident (RA) RA 7 79 A 7 77 Chronic Violator (CV) CV 9 85 V 12 88 Serious Violator (SV) SV 8 81 V 12 77 Extended Violator (EV) EV 14 84 V 12 85 Minimum Violator (MV) MV (Not Computed) • * V 12 74 *This refers to the total number of good and problem drivers correctly classified. Drivers with a raw score lower than the cutoff score specified were predicted to be in the Selected BTA group. Drivers with a raw score equal to, or greater than, the cutoff score were pre dicted to be in the corresponding problem driver group. 80 look at the traits or variables from which come the items that were significant in the item analysis. The way in which the items were selected for the final Violations Attitude (V) and the Accidents Attitude (A) scales Is a procedure that resembles a cross-validation. The items to be put into the Accidents Attitude scale were already sig nificant for at least two accident groups, and those selected for the Violations Attitude scale were significant for at least two of the four violator groups. This pro cedure is a partial cross-validation helping to ensure against the reliance on chance differences. It would have been a complete cross-checking if the item? had also been compared with a separate better-than-average group instead of simply the combined better-than-average group. The item content will be discussed by type of item. Refer to Table 2. First, let us consider the bio-data or demographic information. By type, this category had the most signifi cant items. Items significant for the V and A scales were as follows: fewer years of driving, higher annual mileage driven, marital status (single), no dependent children, no military service, more than one employer in the last 81 TABLE 2 --Sourca trait or variable of significant items in darivad aoalaa Darivad Scales D r X V A RA TA AR MV EV SV CV Sourca Trait or Variabla 1 3 Attantivanaes 2 2 Em. Peraev. and Praoccup 2 2 1 Pruatration U 1 1 2 2 2 2 3 2 2 1 2 Lawlessness Attitude 2 1 2 Intelligence 1 2 Doainatloo 1 1 1 Suppraaaion 1 7 7 6 7 6 8 9 7 6 Bio-Data 3 1 I 1 1 1 4 1 1 3 1 Risk 1 1 2 2 2 2 I 3 3 2 1 I G , CanaraL Activity 1 2 R , Rhathymia 2 2 3 1 1 3 1 1 3 1 A, Ascendance 1 1 1 I 1 1 1 S, Sociability 1 1 E, Estttlonal Stability 1 1 1 1 1 0, Objectivity 2 2 1 F, Friendllnaaa I 1 1 T , Thoughtfulness 2 1 1 L 1 3 P, Peraonal Ralationa 3 j 2 1 1 2 2 2 2 2 1 DA, Driver Attitude 1 1 NA, Need for Attention 1 2 3 1 3 3 3 2 2 AS, Advent, va. Security 1 2 2 3 2 2 SR , Self-Reliance 2 2 2 1 1 1 1 1 AA, Aesthetic Apprec 2 2 1 1 3 1 2 CC, Cultural Conformity I i 1 2 i• 1 NF, Need for Freedom 1 1 1 I RT, Realistic Thinking 1 2 1 NP, Need for Precision 1 3 1 1 SV, Spiritual Values 1 1 1 2 1 2 1 Am, Ambitlousneas 82 two years, and having had no driving course. In addition, significant for the V scale alone, were the mechanical condition of the driver's car, and age. In addition, for the A scale by itself, the amount of night driving (more by the problem drivers) was important. Many of these variables are age dependent, such as marital status, number of dependent children, years of driving, and military service. It appears that the match ing should have been expanded from age and annual mileage alone to include marital status and military service; these two items were the farthest mis-matched between problem drivers and the better-than-average drivers. It is a good question, of course, when to stop matching and use the variable as a predictor! These responses, as were the trait items, were assigned a weight of one in the Viola tions Attitude or Accidents Attitude keys. The hypothesized variables, interestingly enough, turned up primarily in the validity (Deviance and Faking) keys. This was true for the variables of attentiveness, intelligence, frustration, risk, emotional perseveration and preoccupation items. Only one of the hypothesized variables had several items significant in the V and A 83 keys. This was Che lawlessness-attitude type of item. Apparently, if an item were obvious, it wound up in one of the validity keys. If an item were more subtle (apparently some were quite subtle) , it wound up in the key for the V or A scales. Not Involved at all in the violations or accidents keys were hypothesized items for domination or suppression. The fact that these face-validity items wound up primarily in the validity scales instead of actually being discriminating in terms of attitude is not too surprising. Apparently, most of them are too obviously related to safe driving. However, the fact that these were not signifi cantly discriminating does not mean that they are not involved; it merely means that they are too obvious. Even the problem drivers can see through them, so that the items are not discriminating between the problem and the better-than-average driver. One needs but to look at accident records and to hear the reports of drivers involved in accidents for one to conclude that these qual ities are important in traffic safety. Next, we come to temperament and attitude items-- the 20 traits from the Guilford-Zimmerman Temperament 84 Survey, the DF Opinion Survey, the Guilford-Holley Inventory, and a previous Driver Attitude scale. With two or more items apiece in the V scale are these traits: General activity, Ascendance, Driver Attitude, Adventure vs. Security, Aesthetic Appreciation, and Cultural Con" fonnity (G, A, DA, AS, AA, and CC). Those traits from which two or more items came in the Accident Attitude scale are General activity, Ascendance, Self-Reliance, and Aesthetic Appreciation (G, A, SR, and AA). Scales from which came just one item for the Violations attitude scale were Sociability, Personal relations, Self-Reliance, Need for Freedom, Need for Precision, and Ambitiousness (S, P, SR, NF, NP, and Am). Scales from which came just one item for the Accidents attitude scale were Sociability, Objec tivity, and Adventure vs. Security (S, 0, and AS). Scales from which came two or more items for the X or "miss" scale were General activity, Ascendance, Adventure vs. Security, and Cultural Conformity (G, A, AS, and CC). Scales from which came just one item for the X scale were Frustration, Lawlessness, Thoughtfulness, Realistic Think ing, Spiritual Values, and Ambitiousness (Fr, L, T, RT, SV, and Am). Scales for which items primarily turned up 85 in the validity scales (Deviance and Faking) were Personal relations, Driver Attitude, and Spiritual Values (P, DA, and SV). Scales which did not have even one item in either the V or the A keys were Rhathymia, Emotional stability, Friendliness, Thoughtfulness, Need for Attention, Realistic Thinking, and Spiritual Values (R, E, F, T, NA, RT, and SV) . If one combined the Violations attitudes and Acci dent attitudes scales, one finds that the most common items underlying problem-driving behavior are those coming from General activity, Ascendance, and Aesthetic Appreciation scales (G, A, and AA). The 20 temperament and attitude values were orig inally selected because of their similarity or resemblance to traits and variables involved in safe driving and described by other Investigators at least twice. As such, it is very interesting to note traits that were not involved in discriminating items in the item analysis. This produces just as much information as the variable that did produce discriminating items. In passing, however, it should be noted that the variables that did not provide at least one item for either the Violations attitude or 86 Accident attitude scales did have one or more items sig nificantly discriminating for one problem-driver group. Now, this could have been chance or random significance differences. On the other hand, it could mean that these variables are significant at a marginal level; they are not of primary importance. A correlation analysis would help clear up the relationships involved. Looking at the X, or "misses" scale, the items came mostly from the common traits General activity and Ascendance. However, the other two important scales from which the X-items came, were Adventure vs. Security and Cultural Conformity. Collectively, this means that the people that were missed in the original prediction from the V and A scales are not too different from other prob lem drivers. The same traits are there, but with just a slightly different emphasis. It should be noted that the X or "miss" scale improved the predictions noticeably in the case of truck drivers (see Table 6). Validity Analysis When tested, the drivers were informed that their answers would not affect their driver's license in any way. Problem drivers were as careful as the better-than-average 87 (BTA) drivers in answering the 395+ items; this was shown by the lack of significant differences between D and F scores for the selected BTA group and the Extended Violator (EV) group, a typical problem-driver group. Centiles corresponding to D and F raw scores nearest the median are shown below, along with those of a stratified random sample (Norms) for comparison: Selected BTA Norms Extended Violators Deviance (score of 0) 56 55 41 Faking (score of 10) 59 61 55 Further evidence for the sincerity of the problem drivers under standard instructions comes from data with 42 problem drivers (OF) retested under faking instructions (FED). Only the Faking (F) score changed significantly under faking instructions (see Table 3). Cross-Validation It was decided to use people already tested for the major part of the cross-validation. People who had been previously screened out in the selection of drivers for TABLE 3.--Validity data Means (and Standard Deviations in parentheses) Scale Original Instructions (OF) Faking Instructions (FED) t Ratio Faking (F) Scores 8.35 (3.69) 14.28 (3.47) 9.75* Deviance (D) Scores 1.19 (1.10) 1.31 (2.58) 0.27 NS Violations (V) Attitude Scores 12.59 (3.86) 11.92 (3.51) 0.79 NS Accidents (A) Attitude Scores 10.16 (3.40) 8.93 (3.22) 1.71 NS i t Significant at the 1 per cent level. NS: Not significant at the 5 per cent level. oo 00 89 the original criterion groups provided the bulk of the cross-validation samples. That is, people's records were scored using the Violations and Accident Attitude keys. Then the distributions were computed for the people with no accidents and no violations versus those with either one or more accidents or some violations within each group. The selection for prediction within the health problem (P and M) and non-White groups was rather favorable: 65 per cent correct in predicting accidents, and approximately 70 per cent (or better) correct for violations. Refer to Tables 4 and 5. This means that these two variables, physical and mental health (P and M), and race (non-White) were not too important in the original screening. As such, these two variables could be left uncontrolled in future research. It is possible that separate norms, i.e., means and standard deviations would be necessary for these groups. At least the same items would be useful and discriminating. For truck drivers (including a few bus and motor cycle drivers) the X or "misses" scale was useful in increasing the prediction. Refer to Table 6. Drivers could be classified 70 per cent correctly with respect to 90 TABLE 4.--DlacrlalnatIon b«tw«n good and poor drivers with physical or nental (P and M) haalch probleas by naans of darivad scala acoras Watching Data BTA Drivers Problaa Drivers Average Age 38.46 (S.E Mileage 8,000 >. “ 12.19) 39 8, Driver ^ Scores Violations V - 8 V ’ 9 Total 3 and over 3 0-2 7 21 7 24 14 Total 10 28 38 Correctly predicted: 21 + 7 74X, Jt J - 6 34* I)r*«er A Scores Accidents A *- 7 A 8 Total 1 and over 8 0 11 14 5 22 16 Total 19 19 38 Correctly predicted: 11 ♦ 14 66X I 1 3 88* Driver * Scores Accidents X ' 5 X ' 6 Total 1 and over 4 0 8 Total 12 ----- 4 3 7 8 11 19 Correctly predicted: 6 + 4 64X Accidents correctly predicted by A and X scores cofsblned (A * 7 or less, X 5 or less): 4 + 14 + 8 68V, ) ' 4 20* Hvber and Type of Health Problens Reported 7 fainting spells 6 epilepsy 7 any disability that slight affect driving 8 nental illness 4 paralysis 8 abdication that alght affect driving Significant beyond the 5 per cent level. 91 TABLE 3.— Dlacriadnatlon b i t m n good and poor aoa-tfhlte drivers by dorlvod iul* iconi of Hatchlog Dot* Average Ago Average Annual Miloogo BTA Orlmt 34.00 (S.D. - 11.83) 6,300 Problaa Drivers 33.34 (S.D. - 9.83) 6,300 Scores Total Violations 9 < H i - , O 3 and over 10 38 48 0 6 6 12 Total 16 44 60 Correctly predicted: 6 + 38 - 831; X * « 4 19* Driver A Scores Accidents A^ 7 A Z 8 Total 1 and over 10 30 40 0 9 11 20 Total 19 41 60 Correctly predicted: 9 + 30 * 651; ]tX 2.53 Driver X Scores Accidents X <r 5 X 3” 6 Total 1 and over 5 5 10 0 4 5 9 Total 9 10 19 Correctly predicted: 4 + 5 - = • 42T Accident* correctly predicted by A and X score* conbined (A 7 or leaa, X 5 or lea*): 30 + 5 + 4 - 65t, / J - 0.59 N«ber of Drivers bv Race Type BTA Probli Negro Mexican Other Total 4 5 3 12 30 12 6 48 Significant at the 5 per cent level. 92 TABLE 6.— Discrimination between good and poor truck drivers by Mini of darivad acala scores Matching Data BTA Drlvara Problaw Drlvara Average Aga 35.BO (S.D. - 9.03) 33.SB (S.D. - 10.45) Avaraga Annual Mllaaga 20,000 32,000 V Scores Drlv«r Violations V \ 9 V ' 10 Total I and over 9 24 33 0 6 4 10 Total 15 28 43 Corractly predicted 6 + 24 ~ 70%; t 1 ’ 3 57 Orlvar A Scoraa Accldanta A - 7 A 8 Total 1 and ovtt 12 16 28 0 7 8 15 Total 17 24 4 3 Corractly pradlctad: 7+16 53X, t J 0 07 Driver X Scores Accidents X ' 3 X ' 4 Total 1 and over 2 10 12 0 3 7 Total 6 13 19 Corractly pradlctad: 4+10 74% A c c i d e n t s c o r r a c t l y p r e d i c t e d b y A a n d X s c o r e s c o n b i n e d (A 7 o r l a s s . X 3 o r l e s s ) : 16 + 10+ 4 707., 1 1 67 Huaibar of Drivers by Type of Vehicle Driven Car and Type Truck Truck Bus Cycle BTA 3 4 2 1 j _ Problaa 9 22 0 2 Total 10 33 93 their violations and accidents. It is quite possible that a slightly different personality or temperament is involved here. Screening presumably was necessary for truck drivers. The review of the literature suggests that there is a difference in the environmental/individual factors for professional drivers as compared with non-professional drivers. Environmental (exposure) factors for the profes sional driver account for most of his accident experience. Apparently, professional driving (or flying) tends to eliminate many of the individual factors (such as temper- ament) that contribute to accidents. For non-professionals, apparently the contribution of individual factors is more important relative to exposure factors; the extent is another research question. For women drivers, the Violations Attitude scale predicted to some extent their violations (62 per cent correctly), but the Accident Attitude scale was no help at all in predicting the accident involvement of women. Refer to Table 7. It appears from this that the original screen ing of drivers in terms of sex was very necessary. Possibly different attitudes, variables or items are involved in the case of problem women drivers than in 94 TABLE 7.--Discrimination batwaan good and poor fanala drlvara by naana of darlvad aeala acoraa Hatching Data Avaraga Age Avaraga Annual Mllaaga BTA Drlvara 44.04 (S.D. - 9.71) 3,200 Problan Drlvara 44.29 (S.D. - 13.37) 8 ,000 Drlvar V Scoraa Vlolatlona v l 11 V : 12 Total 3 and ovar 13 0-2 20 10 5 23 25 Total 33 15 48 Correctly pradlctad: 10 + 20 ^ 621; ( - 3.03 Drlvar ^ Scoraa Accldanta A 7 A ' 8 Total 1 and ovar 11 0 15 9 13 20 28 Total 26 22 48 Correctly pradlctad: 9 + 15 » 501. y J * NS X Scoraa Accldanta X * 6 X ' 7 Total 1 and ovar 7 0 9 4 6 11 15 Total 16 10 26 Co tractly pradlctad: 4 + 9 - 501 Accldanta correctly pradlctad by A and X acoraa coaiblnad (A - 7 or laaa, X - 6 or laaa): 9 + 4 + 9 - 541, NS Nuaber of Drlvara by Hlacallaneoua Varlablaa Typa Nagroaa Mexlcana P and M Non-Car BTA 0 0 .... " 2 0 Problan 2 2 1 1 NS » not algnlflcant at tha 5 par cant lavel. 95 the case of problem women drivers than in the case of problem men drivers. Possibly Involved is the fewer number (restricted range) of accidents for women drivers. Also note that mileage was poorly controlled; BTA women drivers that drive 10f000 miles or more a year are rare. Of course, it could also be that no prediction would be pos sible for problem women drivers. It was gratifying that the discrimination or pre dictive efficiency of the derived scales held up in the cross-validation for three driver groups on the original 395+ item format. This was probably due to the care originally used in selecting the items for the Violations Attitude and Accidents Attitudes keys. The data from the P and M, non-White and truck drivers were pooled for a larger N. Using the V and X scores combined, drivers could be classified 75 per cent correctly with respect to their violation records. Using the A and X scores combined, these drivers could be categorized 64 per cent correctly with respect to their accident record. See Table 8. These are fair predictive efficiency figures in spite of the range of driver charac teristics in this combined group. TABLE 8.--Dlacrlalnatlon between good and poor non-White, P and M, and truck driver* combined 'ey a«ani of derived acale acorea Hatching Data Average Age Average Annual Mileage BTA Drlvera 26 17 (S.O. 1 0 ,0 0 0 11 41) Probiea Drlvera 34 89 (S.D. 9.84) 11 ,000 Driver Violation* 1 and over 0 V Score* V ^ 9 25 21 10 80 15 Total 46 95 Correctly predicted: 21 ♦ 80 7 2 7, , Total 105 36 141 16 26* A Score* Accident* A - 7 A ' 8 Tota 1 1 and over 10 60 90 0 27 29 5 1 Tota 1 57 84 141 Correctly predlcted: 27 + 60 62L; , " 5 22 V and X Score* - --- Driver V • 9 V ' 10 Violet Iona X - 4* X 5 Total I and over 1 98 105 8 28 16 Total 15 126 1 4 1 Correctly predlcted: 8 + 98 75Z, . 6 95 A and X Score* Driver A ' 7 A ' 8 Accident* X - 5** X • 6 Total I and over 17 71 90 0 17 34 51 Total 34 107 141 Correctly predicted: 17 +■ 71 64T, 3.70 * - X acorea for driver* with V ‘ 9. « X acorea for drlvara with A - 7. ^Significant at the 5 per cent level. Significant at the 1 per cent level. 97 A major consideration is that the items once taken out of their 3954* item format, might no longer be discrimi nating. Actually, a more cogent reason for selecting just the discriminating items is the matter of testing time. If one is interested primarily in picking out problem drivers from the better-than-average drivers, then one wants a short attitude scale comprised primarily of dis criminating items with the irrelevant items excluded. Accordingly, a shorter scale called the Driver Attitude Scale (revised) was constructed of 100 items from several of the previous keys. The keys included the Violations and Accidents attitude items , the X items, the Faking and Deviance items. Originally, items measuring the person’s tendency toward the use of alcohol, or an alcoholism attitude had been omitted. In this revised Driver Attitude Survey, accordingly, ten research items designed to measure an alcoholism attitude were included for future research. In a limited trial with the 100-item format, it was found that discrimination for accidents and violations was approximately the same as with the old format. The pre dictive efficiency was approximately 65 per cent for accidents, and approximately 70 per cent or somewhat 98 better la the case of violations. Further research to develop norms and check the predictive efficiency more exactly Is definitely Indicated with the revised Driver Attitude Survey. In commenting on the change from the old to the new format, it appears that the items by themselves in new contexts are still discriminating. The cross-validation using groups previously screened out is particularly Interesting in a sense that it means driving attitudes are generalized considerably beyond the Caucasian male driver only, good health only, drivers on whom the original item analysis was performed. The results certainly are generalizable: to drivers with physical and mental health problems, to non-White drivers, and to truck drivers. However, the results are not generalizable to women drivers. Further research is definitely indicated here. The predictive efficiency of the derived scales was not as high as originally hoped for. It is true that the discrimination decreased only slightly from the criterion groups to the cross-validation groups. But why was not even greater success found in predicting drivers with violations and accidents? The author feels that this study 99 encompassed most of the attitude and temperament variables likely to be important in safe driving. Obviously, some important attitude trait could have been skipped, but the author believes that, more likely, some aptitudinal or psychophysical trait would help the prediction. Or, due to chance, some accidents are "accidents." In conclusion, it can be stated that the original objective of this dissertation was accomplished. The development of separate scales to predict accidents and violations of drivers was accomplished with a reasonable predictive efficiency. Second, validation scales were developed which do their job fairly satisfactorily. CHAPTER V SUMMARY The goal of this dissertation was to develop and check attitude scales to predict drivers with moving violations and accidents. A secondary scale was to be developed to detect faking tendencies. An extensive review of the literature was made to determine the biographical and psychological characteris tics of drivers with many moving violations and accidents. Those characteristics reported at least twice were selected for further investigation. Traits from the Guilford- Zimmerman Temperament Survey and the DF Opinion Survey most similar to these selected characteristics were included in the experimental tests. Biographical items were included, but aptitude scales were not. The temperament and opinion scales were limited in number of items, but reliability (split-half, corrected) was kept above 0.65. One test booklet, called the General 100 101 Attitude Survey, contained 300 items from 20 temperament and motivational scales. The second test booklet, called the Driver Attitude Survey, contained 104 items. In this booklet were 65 experimental items (for traits of atten tiveness , lawlessness, risk-taking, frustration, and emotional perseveration), 20 biographical items , 10 driving-reaction items, and 9 items related to accident involvements. Drivers of many types took the attitude surveys. Over 2 ,000 subjects were tested at the local Department of Motor Vehicles, Personal Traffic Safety classes, several insurance companies, several military bases in the area, several local industries, and at the University of Southern California. The drivers' violations and accidents were checked both by the subjects' reports and through the Department of Motor Vehicle records. The drivers were classified as better-than-average or as problem drivers. Those whose driving records showed an Intermediate number of violations were omitted. Prob lem drivers were classified into one of three accident types and/or one of four violator types , depending upon severity. 102 Only Caucasian, male drivers In good health, who had been driving in California at least three years, and who usually drove only a car (and not a truck or bus) were Included In the survey Item analysis. Since the problem- drlver groups varied In age and annual mileage, a better- than-average subgroup was selected from drivers who were younger and who drove more than the other better-than- average drivers In the comparison group. The 395 Items were analyzed for differences by driver group. Items significantly different beyond the 5 per cent level between the driver groups were screened further for the effects of age and annual mileage. Items were retained only If not significant beyond the 20 per cent level on age and annual mileage within the comparison driver groups. A Violations attitude scale was comprised of items significant for discriminating at least two of the four violator groups. An Accidents attitude scale was comprised of items significant for discriminating at least two of the three accident driver groups. Three validity scales were developed. Four groups not used in the item analysis and pre viously screened out were used for cross-validation 103 purposes. These Included drivers with physical or mental health problems, non-White drivers, truck drivers, and women drivers. Drivers with poor health, non-White or truck drivers could be discriminated with 70-80 per cent accuracy with respect to their violations by the Violations attitude scale. With regard to accidents, these same drivers could be categorized with an accuracy of 65-70 per cent, primarily by their Accidents attitude score, but also with the aid of a suppressor scale. Women drivers could be categorized only 62 per cent correctly on violations and not at all on accidents (50 per cent was chance). In conclusion, two attitude scales were derived and cross-validated to predict reasonably well drivers with violations or accidents. Three auxiliary scales were also derived to help ensure the validity of the test results. BIBL IO GRA PH Y SELECTED BIBLIOGRAPHY Allport, G. W., Vernon, P. E., end Llndzey, G. Study of values. Revised; Boston: Houghton Mifflin Co., 1951. The ATA Foundation, Inc., and The Pure Oil Company. Centering traffic safety communications around driver's motivations, 1953, 32-33. Barnes, P. M. Motorcycling accidents in South Africa--A survey of statistics. J. Natn'l. Inst. Personnel Res. , 7, 1958, 104-108. Battey, A. D. The measurement of exposure to motor-vehicle accidents. Traf. Saf. Res. Rev., 3, March, 1959, 19-22. Biesheuvel, S. and Barnes, P. M. A motorcycle accident study. J. Natn'l. Inst. Personnel Res., South African Counc. Sci. Ind. Res., 1959, 32 pp. Billion, C. E. Community study of characteristics of drivers and driver behavior related to accident experience. High. Res. Bd. Bull., 172, December, 1956. Bletzacker, R. W. and Brittenham, T. G. An analysis of one-car accidents. Unpublished master's thesis, January, 1958, Ohio State U., Eng. Exp. Sta. Boek, J. K. Automobile accidents and driver behavior. Traf. Saf. Res. Rev., 2, December, 1958, 2-12. Brody, L. Personal characteristics of chronic violators and accident repeaters examined at the Accident Preven tion Clinics of the New Jersey Division of Motor Vehicles. Paper presented at the 35th Annual Meeting of the High. Res. Bd., Washington, D. C., January, 1956. 105 106 Brody, L. Get the whole picture. Traf. Saf., February, 1959, 30. Brown, P. L. and Berdie, R. F. A Study In the relation ship between personality traits and driving behavior. A paper. Saf. Dlv. , Minn. High. Dept. Director of Stud. Counc. Bur. , Univ. of Minn. , 1958. ________. Driver behavior and scores on the MMPI. J. appl. Psychol. . 44, No. 1, February, 1960, 18-21. Buchanan, P. C. A study of the prediction of the accident proness of motorcycle operators. Doctor's disserta tion, Univ. of South. Calif., June, 1950. Case, H. W. , et al. The habitual traffic violator. High. Res. Bd. Bull. , 120, Traffic Accidents and Violations, 1956. Case, H. W. and Stewart, R. G. Some personal and social attitudes of habitual traffic violators. J. appl. Psychol. , 41, No. 1, February, 1957, 46-52. Chalfant, M. W. and King, G. F. An evaluation of the effectiveness of driver improvement procedures. Paper presented at the Annual Meeting of the High. Res. Bd. , January, 1959. Comrey, A. L. A factor analysis of variables related to driver training. J. appl. Psychol., 42, August, 1958, 218-221. Conger, J. J. Personal and Interpersonal factors in accidents. Paper presented at Amer. Psychiatric Assoc. Meeting, May, 1956. A report of the Driving Survey, Colorado Sch. Med. and Fitzsimmons Army Hospital. Conger, J. J. , et al. Psychological and psychophysio- logical factors in motor vehicle accidents. J.A.M.A., 169, April, 1959, 1581-87. Conover, D. W. Development of certain techniques for the measurement of driver attitudes. Master's thesis, Iowa State Coll., 1947. 107 Department of California Highway Patrol, State of California. Annual Statistical Report, 1959, 106 pp. Dept. Calif. High. Patrol, B. M. Crittenden, Commis sioner. Sacramento, Calif. Department of Motor Vehicles. Vehicle Code, State of California, 1957, 627 pp. Sacramento, Calif. 6. J. Knight, Gov.; P. Mason, Director, Dept. M. V. _______ . Division of Drivers Licenses, State of Califor nia. F. P. Williams, Chief. Driver Record Study, 1958. Drivers Safety Service, Inc. A status report concerning driver rehabilitation in New Jersey--A preliminary statement. 298 Broadway, N. Y. 7, N. Y. Eno Foundation for Highway Traffic Control. Personal Characteristics of Traffic Accident Repeaters , 1948, 64 pp. , Saugatuck, Conn. Farmer, E. Accident-proneness on the road. Practitioner, 154, 1945, 221-226. Fine, J. L. The use of the critical incident technique in developing an instrument to predict and evaluate driving behavior. Proposed Ph. D. study at N. Y. U., December 8, 1958. Fletcher, E. D. Comprehensive survey of driving tests made. The Calif. High. Patrolman, Part I, August, 1939, 4 et seq. _______. Comprehensive survey of driving tests made. The Calif. High. Patrolman, Part II, 3, No. 7, September, 1939, 14 et seq. _______. An evaluation of the vision testing of motor vehicle operators. Master's thesis, Univ. of South. Calif., September, 1945. Forbes, T. W. Analysis of "near accident" reports. High. Res. Bd. Bull.. 152, 1957. 108 Forbes, T. W. Psychological factors in traffic accidents, in The Mind and the Motorist, 1957. A symposium of psychological factors contributing to traffic accidents, sponsored by the Nebraska Psychiatric Inst. Guilford, J. P. The Phi coefficient and Chi Square as indices of item validity. Psychometika, 6, No. 1, February, 1941, 11-19. Guilford, J. P., Christensen, P. R. , and Bond, N. A. The DF Opinion Survey. Sheridan Supply Co. , Beverly Hills, Calif., 1954. Guilford, J. P. and Holley, J. W. The Guilford-Holley L Inventory (unpublished). Sheridan Supply Co., Beverly Hills, Calif., 1953. Guilford, J. P. and Michael, W. B. The prediction of categories from measurements: with applications to personnel selection and clinical prognosis. Sheridan Supply Co., Beverly Hills, Calif., 1949. Guilford, J. P. and Zimmerman, W. S. The Guilford- Zimnerman Temperament Survey. Sheridan Supply Co. , Beverly Hills, Calif., 1949. Hakkinen, S. Traffic accidents and driver character istics: A statistical and psychological study. Scientific Researches. No i? 19^8, 198. Finland's Inst, of Tech. helsinki: Akateeminen Kirjakkauppa. Heath, E. D. The relationships between driving records, selected personality characteristics, and biographical data of traffic offenders and non-offenders. Doctor's dissertation, N. Y. U., 1957. Heinz, H. , Brenner, R. , and Hulbert, S. Psychology of trip geography. High. Res. Bd. Bull. , 91, January, 1954. Henderson, H. L. A study of pre-clinic and post-clinic accident and violation records of drivers processed by the Accident Prevention Clinics of the State of New 109 Jersey. Drivers Safety Service, Inc., 298 Broadway, N. Y. 7, N. Y. March, 1958. Henderson, H. L. Preliminary findings from psychological and psycho-physical tests administered in the New Jersey Accident Prevention Clinics. Drivers Safety Service, Inc., 298 Broadway, N. Y. 7, N. Y. October, 1959. _______ . Driver rehabilitation through small group dis cussion. Drivers Safety Service, Inc., 298 Broadway, N. Y. 7, N. Y. January, 195:. Highway Research Board. Tests of motor vehicle drivers and the relation of test scores to the officially recorded accidents sustained. Motor Vehicle Conditions in the United States, July 31, 1959. Washington, D. C. (Percy Cobb.) Hughes, H. M. Discriminatory analysis III. Discrimina tion of the accident prone individual. Project No. 21-49-004, Report No. 3, October, 1950. USAF Sch. Avia. Med. Jacobs, H. H. Mathematical models applied to accident processes. Math. Models Hum. Behav., 1955, 25-31. Proc. of a symposium. Dunlap and Assoc., Stamford, Conn. Keehn, J. D. Factor analyst of reported minor personal mishaps. J. appl. Psychol.. 43, No. 5, October, 1959, 311-314. King, 3. G. and Sutro, P. J. Dynamic visual fields. Driver characteristics. High. Res. Bd. Bull., 152, 1957. Kraft, M. A. and Forbes, T. W. Evaluating the influences of personal characteristics on the traffic accident experience of transit operators. Proc. High. Res. Bd., 24, 1944, 278-291. Washington, D. C. 110 Kuraishi, S., Masahide, K., and Tsujioka, B. Development of the "Uchida-Kraepelin Psychodiagnostic Test" in Japan. Psychologia. No. 1, December, 1957, 104-109. Lauer, A. R. Age and sex in relation to accidents. Traf. Saf. Res. Rev., 3, No. 4, December, 1959, 21-25. _. Iowa State Multi-Attitude Scale, Form A. Lefeve, B. A. Speed habits observed on a rural highway. Proc. High. Res. Bd., 22, 1954, 409-429. _. Relation of accidents to speed habits and other driver characteristics. High. Res. Bd. Bull., 120, Traffic Accidents and Violations, 1956, 6-30. Kanson, M. P. The Manson Evaluation. Western Psycho logical Services, Beverly Hills, Calif., 1948. t'aritz, J. S. On the validity of inferences drawn from the fitting of Poisson and negative binomial distribu tions to observed accident data. Psychol. Bull. , 47, No. 5, September, 1950, 434-443. Mathewson, J. H. and Brenner, R. Indices of motor vehicle accident likelihood. High. Res. Bd. Bull.. 161, January, 1957, 1-8. Mayo, G. D. and Guttman, 1. Faking in a vocational classification situation. J. appl. Psych., 43, April, 1959, 117-120. McFarland, R. A. Psychological and psychiatric aspects of highway safety. J.A.M.A., 163, January 26, 1957. _. The role of preventive medicine in highway safety. A. J. Pub. Health. 47, No. 3, March, 195 7. _. Human Factors in Highway Transport Safety, Harvard Sch. Pub. Health, 1954. McFarland, R. A. and Moore, R. C. Human factors in highway safety, a review and evaluation. New Engl. J. Med. , 256, 4 and 5, 1957, pp. 792-799, 837-845, 890-897. Ill McGuire, F. L. Psychological comparison of automobile drivers: Accident-and-violation-free versus accident- violation-incurring drivers. U. S. Armed Forces Med. J. , 7, 1956, 1741-1748. _______ . Rozenzweig P-F study for selecting safe drivers. U. S. Armed Forces Med. J., 7, 1956, 200-207. Miller, Carmen. A comparison of personality character istics of high-accident and low-accident bus and street car operators. Doctor's dissertation, Western Reserve Univ. , 1952. Merman, R. R. A selection study of motor coach operators involving six criteria. Doctor's dissertation, Univ. of South. Calif., June, 1955. Nathanson, S. N. The relationship between the traits measured by Thurstone's Temperament Schedule and convictions for traffic violators among junior college students. Master's thesis, Univ. of South. Calif., August, 1953. National Safety Council. Accident Facts, 97 pp. 425 Michigan Ave. , Chicago 11, Illinois, 1959. Schulzinger, M. S. The Accident Syndrome: The Genesis of Accidental Injury, 234 pp. C. C. Thomas, Springfield, Illinois, 1956. Schuster, D. H. and Guilford, J. P. An analysis of accident repeater and chronic violator drivers. Traffic Project Report No. 1. Dept. Psychol., Univ. of South. Calif., October, 1959. Selling, L. S. Neuropsychiatric study of traffic offenders. Psychosomatic Med.. 2, 1940, 384-397. Sheldon, M. S. Conditions affecting the fakeability of teacher-selection inventories. Educa. psychol. Measmt. , 19, No. 2, Summer, 1959, 207-219. Siebrecht, E. B. Siebrecht Attitude Scale, Manual of Directions , 4 pp. Center for Saf. Educa. , N. Y. U., 1941. 112 Siebrecht, £. B. , Schumacher, C. F., and Lauer, A. R. Accident characteristics of drivers at various age levels. Traf. Saf. Res. Rev., 3, No. 4, December, 1959, 29-32. Soma, M. A psychological study of the over-accldent makers. Railroad Labor Scl. . 4, 1953, 27-30. Stack, H. J. What makes drivers act that way? Traf. Quarterly. January, 1947, 29-39. Eno Foundation, Saugatuck, Conn. Stelnmetz, H. S. Measuring ability to fake occupational Interest. J. appl. Psychol., 16, 1932, 123-30. Stewart, R. G. Reported driving speeds and previous accidents. J. appl. Psychol. , 41, No. 5, October, 1957, 293-296. ________. Can psychologists measure driving attitudes? Educa. psychol. Measmt. , 18, No. 1, Spring, 1958, 63-73. Suhr, V. W. The Cattell <6 ns prognosticator of accident susceptibility, lioe. low* Acad. Sci. , 60, 1953, 553-561. Tillman, W. A. and Hobbs, G. E. The accident prone driver: A study of the psychiatric and social background. A. J. Psychiat.. 106, No. 5, November, 1949, 321-331. Uhlaner, J. E. Tests for selecting drivers. Paper. Personnel Res. Br., A. G. 0., Dept. Army, March, 1956. IThr, L. Sex as a determinant of driving skills: Women drivers! J. appl. Psychol., 43, February, 1959, 35-39. Webb, W. B. The prediction of aircraft accidents from pilot centered measures. USN Sch. Avia. Med., Res. Rep., 1956, Proj. No. NM 001 106 100, Rep. No. 1, 9 pp. 113 Webb, W. B. and Jones, F. R. Repeater pilot accidents in the USAF. H. R. R. L., Bolling AFB, 1952, Washington, D. C. Wilson, 0. W. Police control of the automobile driver. J. crim. Law. Crim. » 30, 1939, 89-95. Wolbers, H. L. A study of the relationship between Air Force classification test scores and aircraft acci dents. Doctor's dissertation, Univ. of South. Calif., June, 1955. The accident prone driver. Bur. Pub. Roads, Part 6 of House Document No. 462. Motor veh. traf. conditions in the U. S. G. P. 0., 1938, X + 52 pp.

Asset Metadata

Creator Schuster, Donald Herbert (author)

Core Title The Development Of Attitude Scales To Predict Accident Repeater And Moving Violator Drivers

Contributor Digitized by ProQuest (provenance)

Degree Doctor of Philosophy

Degree Program Psychology

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag OAI-PMH Harvest,psychology, general

Language English

Advisor Guilford, Joy P. (committee chair), [Harvey, Herman] (committee member), Warren, Neil D. (committee member)

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c18-93208

Unique identifier UC11358112

Identifier 6101707.pdf (filename),usctheses-c18-93208 (legacy record id)

Legacy Identifier 6101707.pdf

Dmrecord 93208

Document Type Dissertation

Rights Schuster, Donald Herbert

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the au...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus, Los Angeles, California 90089, USA