The Twelfth Jack Tizard Memorial Lecture

The Twelfth Jack Tizard Memorial Lecture*

*Delivered at the ACPP 2nd European Conference, Winchester, U.K., 2 September 1994.

The Development of Offending and

Antisocial Behaviour from Childhood: Key

Findings from the Cambridge Study in

Delinquent Development

David P. Farrington

Abstract--In the Cambridge Study in Delinquent Development, 411 South London males have been followed up from age 8 to age 32. The most important childhood (age 8‑10) predictors of delinquency were antisocial child behaviour, impulsivity, low intelligence and attainment, family criminality, poverty and poor parental child‑rearing behaviour. Offending was only one element of a larger syndrome of antisocial behaviour that arose in childhood and persisted into adulthood. Marriage, employment and moving out of London fostered desistance from offending. Early prevention experiments are needed to reduce delinquency, targeting low attainment, poor parenting, impulsivity and poverty.

Keywords: Longitudinal study, delinquent development, antisocial behaviour, preventing delinquency

Introduction

I believe that this is an appropriate Memorial Lecture for Jack Tizard for three main reasons. First, Tizard was very interested in offending and antisocial behaviour. He devoted a substantial part of his Presidential Address to the British Psychological Society (Tizard, 1976) to a critical discussion of the delinquency research of Cyril Burt, acknowledging Burt's pioneering epidemiological work but pointing out the limitations of the retrospective case‑control design in which delinquents were compared with matched nondelinquents. Tizard was also, of course, very interested in residential treatment for delinquents (Tizard, Sinclair & Clarke, 1975).

Second, Jack Tizard played an important role in fostering longitudinal epidemiological research in the United Kingdom by co‑directing the Isle of Wight study with Michael Rutter from its inception in 1964 until his death in 1979 (Rutter, 1981). This study has been extremely important in estimating the prevalence and comorbidity of childhood psychiatric disorders in a complete population of children and also in identifying risk factors and documenting the course of disorders over time (Rutter, Tizard & Whitmore, 1970; Rutter, 1989).

Third, Jack Tizard was a leading advocate of the use of fundamental research findings in the formulation of social policy, as is again seen in his Presidential Address and emphasized by its title: "Psychology and social policy" (Tizard, 1976). He thought that there was too much emphasis on classification, diagnosis and testing, and not enough emphasis on the use of social policy experiments to advance knowledge about causes. As a small tribute to Tizard's interest in social policy, I have tried to address policy implications in some detail in my paper.

The Cambridge Study in Delinquent Development

In reviewing the development of offending and antisocial behaviour, I will particularly describe results obtained in the Cambridge Study in Delinquent Development, which is a prospective longitudinal survey of the development of delinquency and antisocial behaviour in 411 South London boys, mostly born in 1953. These males have been followed up by personal interviews from age 8 to age 32. The Study began in 1961, and was originally directed by Professor Donald West. I joined him to work on it in 1969, and I have directed it since 1982. Results of the Study have been described in four books (West, 1969, 1982; West & Farrington, 1973, 1977), and in more than 60 papers listed by Farrington and West (1990). This paper will focus particularly on the most recently obtained results.

The original aim of the Study was to describe the development of delinquent and criminal behaviour in inner‑city males, to investigate how far it could be predicted in advance, and to explain why juvenile delinquency began, why it did or did not continue into adult crime, and why adult crime usually ended as men reached their twenties. The main focus was on continuity or discontinuity in behavioural development, on the effects of life events on development, and on predicting future behaviour. The Study was not designed to test any one particular theory about delinquency but to test many different hypotheses about the causes and correlates of offending. One reason for casting the net wide at the start and measuring many different variables was the belief that theoretical fashions changed over time and that it was important to try to measure as many variables as possible in which future researchers might be interested. Another reason for measuring a wide range of variables was the fact that long-terrn longitudinal surveys were very uncommon, and that the value of this particular one would be enhanced if it yielded information of use not only to delinquency researchers but also to those interested in alcohol and drug use, educational difficulties, poverty and poor housing, unemployment, sexual behaviour, aggression, other social problems, and human development generally.

were carried out by female psychiatric social workers who visited their homes. These took place about once a year from when the boy was about 8 until when he was aged 14‑15 and was in his last year of compulsory education. The primary informant was the mother, although many fathers were also seen. The parents provided details about such matters as the boy's daring or nervousness, family income, family size, their employment histories, their history of psychiatric treatment, their child‑rearing practices (including attitudes, discipline, and parental disharmony), their closeness of supervision of the boy, and his temporary or permanent separations from them. Obstetric records were obtained for boys born in hospital. Also, when the boy was aged 12, the parents completed questionnaires about their child‑rearing attitudes and about his leisure activities.

The teachers completed questionnaires when the boys were aged about 8, 10, 12 and 14. These furnished data about their troublesome and aggressive school behaviour, their restlessness and poor concentration, their school attainments and their truancy. Delinquency rates of secondary schools were obtained from the local education authority. Ratings were also obtained from the boys' peers when they were in the primary schools, about such topics as their daring, dishonesty, troublesomeness and popularity.

Searches were also carried out in the central Criminal Record Office in London to try to locate findings of guilt of the males, of their biological mothers, fathers, brothers and sisters, of their wives and cohabitees, and of people who offended with them (their co‑offenders). The minimum age of criminal responsibility in England is 10. The Criminal Record Office contains records of all relatively serious offences committed in Great Britain or Ireland. In the case of 18 males who had emigrated outside Great Britain and Ireland by age 32, applications were made to search their criminal records in the eight countries where they had settled, and searches were actually carried out in four countries. Since most males did not emigrate until their twenties, and since the emigrants had rarely been convicted in England, it is likely that the criminal records are quite complete.

Convictions were only counted if they were for offences normally recorded in the Criminal Record Office, thereby excluding minor crimes such as common assault, traffic infractions and drunkenness. The most common offences included were thefts, burglaries and unauthorized taking of vehicles, although there were also quite a few offences of violence, vandalism, fraud and drug abuse. In order not to rely on official records for information about delinquency and crime, self-�reports of offending were obtained from the males at every age from 14 onwards (Farrington, 1989c).

Summarizing, the Cambridge Study in Delinquent Development has a unique combination of features. Eight face‑to‑face interviews have been completed with the males over a period of 24 years, between the ages of 8 and 32. The attrition rate is unusually low for such a long‑term survey. The main focus of interest is on crime and delinquency, but the survey also provides information about alcohol and drug abuse, educational difficulties, poverty and poor housing, sexual behaviour, unemployment and other social problems. The sample size of about 400 is large enough for many statistical analyses, but small enough to permit detailed case histories of the males and their families. Information has been obtained from multiple sources, including the males themselves, their parents, teachers, peers, and official (criminal, hospital and school) records. Generally, the data came from parents, teachers, peers or tests completed by the males between the ages of 8 and 14, but primarily from interviews with the males between the ages of 16 and 32. Information has been collected about a wide variety of theoretical constructs at different ages, including biological (e.g. pulse rate, height, weight), psychological (e.g. intelligence, impulsivity), family (e.g. parental supervision and discipline), and social (e.g. poor housing, socioeconomic status) factors. Hence, the relative importance of these different factors as predictors and correlates of offending can be investigated.

Other Prospective Longitudinal Surveys

There have been very few prospective longitudinal studies of community samples of several hundreds, including delinquency and crime data, and several personal interviews with the subjects over a period of at least 10 years (Farrington, 1988a). The most outstanding project of this type is the Dunedin study in New Zealand, in which nearly 1,000 children born in 1972‑73 have been followed up to the present, with nine personal data collection waves between the ages of 3 and 21 (Moffitt & Silva, 1988). After this, the next most useful project (since it has so far provided less information about delinquency) is the Christchurch Health and Development Study, also in New Zealand, in which over 1,000 children born in 1977 were studied at birth, at four months, and annually up to age 13, with data from children, parents and teachers (Fergusson, Horwood & Lynskey, 1993).

Another important project (and the most comparable British research to the Cambridge Study) is the Newcastle Thousand‑Family Study, in which over 800 children born in 1947 were frequently contacted between birth and age 15 and finally interviewed at ages 22 and 32 (Kolvin et al, 1990). Also noteworthy is the American National Youth Survey, in which 1,700 children originally aged 11‑17 in 1976 have been seen nine times up to the present (Elliott, Huizinga & Menard, 1989). However, unlike the Dunedin, Christchurch, Newcastle and Cambridge surveys, which included several data sources, the National Youth Survey only collected self‑report data from the children (in addition to police records, of course).

Another important longitudinal survey is the Kauai study in Hawaii, in which nearly 700 children born in 1955 were followed up from birth to age 32, with four personal data collection waves (Werner & Smith, 1992). The classic American longitudinal surveys of Glueck and Glueck (1968) in Massachusetts, McCord (1979) in Boston, Robins (1979) in St. Louis and Wolfgang, Thornberry and Figlio (1987) in Philadelphia included only between one and three personal contacts with the subjects. A new generation of American longitudinal studies was mounted in the 1980s, with frequent personal contacts and multiple data sources (Huizinga, Esbensen & Weiher, 1991; Loeber, Stoutharner‑Loeber, Van Kammen & Farrington, 1991; Thornberry et al, 1991). When their samples have been followed up for more than a decade, these projects may yet become as important as the New Zealand studies.

Up to age 32, over one‑third of the Study males (153, or 37%) were convicted of criminal offences. The prevalence of offending increased up to age 17 and then decreased. Many other projects show a similar age‑crime curve (Farrington, 1986a). While the peak age for the number of offenders and the number of offences was 17, roughly equal numbers of offences were committed by the Study males as juveniles (age 10‑16), as young adults (age 17‑20), and as adults (age 21‑32). Nearly three‑quarters of those convicted as juveniles were reconvicted between ages 17 and 24, and nearly half of the juvenile offenders were reconvicted between ages 25 and 32 (Farrington, 1992a). Other studies show a similar continuity in offending (Stattin & Magnusson, 1991). On average, official criminal careers began at age 17, lasted 6 years, ended at age 23, and included 4.5 offences leading to conviction. The men first convicted at the earliest ages tended to become the most persistent offenders, in committing large numbers of offences at high rates over long time periods. Moffitt (1993) suggested that the "life‑course‑persistent" offenders, who began early, were different in kind from the "adolescence‑limited" offenders who began later and had short criminal careers.

The "chronic offenders" at age 32 were defined as the 24 men (6% of the sample) who committed half of all officially recorded offences (Farrington & West, 1993). Many other projects report a similar proportion of chronic offenders (Tracy, Wolfgang & Figlio, 1990). They had each committed at least nine officially recorded offences, and they had especially long criminal careers characterized by high rates of offending. They were versatile rather than specialized offenders; 16 of the 24 committed at least five different types of offences (out of 10 types altogether). They also accounted for substantial proportions of the self‑reported offences. For example, between ages 15 and 18, the chronic offenders committed 53% of all self‑reported burglaries and 48% of all self‑reported thefts of vehicles. Because they are so few and account for so much of the crime problem, the chronics are important targets for prevention and treatment.

Most juvenile and young adult offences were committed with others, but the incidence of co‑offending declined with age (Reiss & Farrington, 1991). Burglary, robbery and theft from vehicles were particularly likely to involve co‑offenders. Co‑offenders tended to be similar in age and sex to Study males and to live close to their addresses and to the locations of the offences. It was rare for Study males to offend with their fathers, mothers, sisters or wives, or with unrelated females. Co‑offending with brothers was most likely when a Study male had brothers who were close in age to him.

The most common reasons given for delinquency were utilitarian, rational or economic: offences were committed for material gain (Farrington, 1993c). The next most common category of reasons were for excitement, for enjoyment, or to relieve boredom. Similar results were reported in Montreal by LeBlanc and Frechette (1989). Other reasons were designed to minimize the offender's responsibility (e.g. I was young; I was drunk) or to blame peers. Reasons for fighting depended on whether the boy was alone or in a group (Farrington, Berkowitz & West, 1982). In individual fights, the boy was usually provoked, became angry, and hit out to hurt his opponent. In group fights, the boy often said that he became involved to help a friend or because he was attacked.

Self‑Reportcd and Official Delinquency

To a considerable extent, the self‑reports and official records identified the same people as the worst offenders, as other researchers have also found (Huizinga & Elliott, 1986). The 80 boys who admitted the highest number of delinquent acts when seen at ages 14 and 16 overlapped significantly with the 85 convicted juvenile delinquents, since 41 boys were in both groups (West & Farrington, 1973). Consequently, conclusions about characteristics of offenders based on convictions were generally similar to conclusions based on self‑reported delinquency. For example, of 39 key predictors and correlates of offending measured between ages 8 and 18, 35 were significantly related to both official and self‑reported delinquency (Farrington, 1992c, Table 6. 1). The only exceptions were that official (but not self-�reported) delinquents were attending high delinquency rate schools and were relatively small at ages 8‑10 and 18, while self‑reported (but not official) delinquents tended to come from low social class families at age 8‑10.

Self‑reported offending and official convictions showed that the most common crimes of burglary, shoplifting, theft of and from vehicles, and vandalism declined in prevalence from the teenage years to the twenties and thirties (Farrington, 1989c). However, theft from work increased with age. The cumulative prevalence of self‑reported offending was very high; 96% of males admitted committing at least one crime that could, in theory, have led to a conviction. Hence, at least in this sample of urban working‑class males, offending was not statistically deviant, although it was less common to commit relatively serious offences such as burglary. Only 22% of males admitted burglary, and only 14% were convicted of burglary, up to age 32.

The relationship between self‑reported offending and official convictions was strongest for burglary and for theft of and from vehicles, but it was also significant for shoplifting, theft from machines, assault, and drug use (Farrington, 1989c). The two measures were not significantly related for theft from work, vandalism, and fraud, because of the low probability of an offender being convicted for these offences. When data were cumulated over the whole period between ages 10 and 32, the probability of an offender being convicted (sooner or later) was quite high for several types of offences: over 50% for burglary and theft of vehicles, and 25% for theft from vehicles. The probability of an offender being convicted increased with age. For burglary, theft of and from vehicles, and drug use, self‑reports of offending among unconvicted males significantly predicted future convictions for these offences, showing that self‑reports provided valid information about offending.

Delinquency and Antisocial Behaviour

The Cambridge Study shows that delinquency is only one element of a much larger syndrome of antisocial behaviour that tends to persist over time, as Robins (1986) has persuasively argued. For example, the boys who were convicted up to age 18 (most commonly for offences of dishonesty, such as burglary and theft) were significantly more deviant than the nonoffenders on almost every factor that we investigated at that age (West & Farrington, 1977). The convicted delinquents drank more beer, they got drunk more often, and they were more likely to say that drink made them violent. They smoked more cigarettes, they had started smoking at an earlier age, and they were more likely to be heavy gamblers. They were more likely to have been convicted for minor motoring offences, to have driven after drinking at least 10 units of alcohol (e.g. 5 pints of beer), and to have been injured in road accidents. The delinquents were more likely to have taken prohibited drugs such as marijuana or LSD, although few of them had convictions for drug offences. Also, they were more likely to have had sexual intercourse, especially with a variety of different girls, and especially beginning at an early age, but they were less likely to use contraceptives.

The convicted delinquents at age 18 tended to hold relatively well paid but low status jobs, and they were more likely to have erratic work histories including periods of unemployment. They were more likely to be living away from home, and they tended not to get on well with their parents. They were more likely to be tattooed, and they had significantly low pulse rates. The delinquents were more likely to go out in the evenings, and were especially likely to spend time hanging about on the street. They tended to go around in groups of four or more, and were more likely to be involved in group violence or vandalism. They were much more likely to have been involved in fights, to have started fights, to have carried weapons, and to have used weapons in fights. They were also more likely to express aggressive and anti‑establishment attitudes on a questionnaire (negative to police, school, rich people and civil servants).

It was interesting that the peak age of offending, at 17‑18, coincided with the peak age of affluence for many convicted males. Convicted males tended to come from low income families at age 8 and later tended to have low incomes themselves at age 32. However, at age 18, they were relatively well paid in comparison with nondelinquents; whereas convicted delinquents might be working as unskilled labourers on building sites and getting the full adult wage for this job, nondelinquents might be in poorly‑paid jobs with prospects, such as bank clerks, or might still be students. These results show that the link between income and offending is quite complex.

The Follow‑Up at Age 32

The main aims of the latest follow‑up were to extend knowledge about criminal careers into the thirties, to investigate the social adjustment and adult life styles of current and past offenders and nonoffenders, and to study how far persistence and desistance of offending could be predicted.

Most efforts in the 2 years between December 1984 and November 1986 were

directed towards locating and interviewing as many of the men in the sample as

possible. Tremendous efforts were made to secure interviews, because of our belief

(based in part on previous results obtained in this survey) that the most interesting

subjects in any research on offending tended to be the hardest to locate and the

most uncooperative. Surveys in which only about 75% of the target sample (or

even less) are inter‑viewed are likely to produce results which seriously underestimate

the true level of criminal behaviour. Generally, an increase in the percentage

interviewed from 75% to 95% leads to a disproportionate increase in the validity

of the results; for example, at age 18, 36% of the one‑sixth of the sample who were

the most difficult to interview were convicted, compared with only 22% of the

majority who were interviewed more easily, a statistically significant difference

(West & Farrington, 1977).

Eventually, after a great deal of detective work, every one of our men was located.

Up to age 32, eight of the men had died, and 20 had emigrated permanently. Of

the remaining 383 who were alive and in the United Kingdom, 360 were interviewed

personally (94%). Seven of the 20 emigrated men were also interviewed, either

abroad or during a temporary return visit that they made to the United Kingdom,

giving a total number inter‑viewed of 367. In addition, nine emigrated men filled

in questionnaires, and two co‑operative wives of refusers filled in questionnaires

on behalf of their husbands, in at least one case with the husband's collaboration and assistance. Therefore, interviews or questionnaires were obtained for 378 of

the 403 men still alive (94%). The median age at interview was 32 years 3 months.

In general, success in tracing the men was achieved by persistence and by using

a wide variety of different methods (Farrington et al, 1990). Searching in electoral

registers and telephone directories, and visits to a man's presumed address, were

the most successful tracing methods for the men who were not particularly elusive.

Searches in the Criminal Record Office, National Health Service records, and

leads from other men were the most useful for the more elusive men. The key

factor in obtaining the men's co‑operation was probably the pleasantness of the

interviewer in the first face‑to‑face meeting.

Later Life Outcomes of Offenders

Generally, the Study males were less antisocial at age 32 than at age 18. Most types of offending declined with age, although binge‑drinking, drunk driving and the use of hard drugs (heroin and cocaine) increased (Farrington, 1990a). While the Study males became less deviant in absolute terms, those who were relatively more deviant at age 18 still tended to be relatively more deviant at age 32. Therefore, there was relative stability but absolute change. Between ages 18 and 32, there was a decrease in aggressive and anti‑establishment attitudes and in self‑reported impulsivity. Most of the men (76%) were living with a wife or female cohabitee at age 32, and their job records had become much more stable. These two factors of settling down with female partners and in stable jobs seemed most plausible in explaining the decrease in offending.

Convicted men differed significantly from unconvicted ones at age 32 in most aspects of their lives (Farrington, 1989b). Convicted men were less likely to be home owners and more likely to be renting (usually from the local council), more likely to have moved home frequently, more likely to be divorced or separated, to be in conflict with their wife or cohabitee and to have assaulted her, and more likely to be separated from a child. Convicted men were more likely to be unemployed, had low take‑home pay, spent more evenings out per week, were more likely to be involved in fights, heavy smokers, heavy drinkers, drunk drivers, drug users and self‑reported offenders. Also, convicted men were more likely to be identified as probable psychiatric cases on the General Health Questionnaire (Goldberg, 1978), which detects anxiety‑depressive types of mental illness, but this was a relatively weak relationship.

Generally, the convicted men who were persisters (those convicted both before and after their 21st birthdays) were more deviant than desisters (those convicted only before their 21st birthdays) or late‑comers to crime (those convicted only after their 21st birthdays). For example, adult social dysfunction at age 32 was assessed on a 9‑point scale including quality of accommodation, cohabitation history, success with children, employment history, fighting, substance abuse, psychiatric disorder, and self‑reported and official offending in the previous 5 years (Farrington et al, 1988b). A similar composite measure was developed by Zoccolillo, Pickles, Quinton and Rutter (1992). While 87% of unconvicted men were living quite successful lives according to this composite measure, this was true of 67% of desisters, 59% of late‑comers and only 44% of persisters (Farrington, 1989b). Even at age 32, the persisters tended to be significantly deviant in a variety of ways.

The Antisocial Personality Syndrome

In order to investigate the syndrome of antisocial behaviour, composite measures of �antisocial personality� were devised at ages 10, 14, 18 and 32, based on indicators of deviant behaviour at each age (Farrington, 1991a). For example, the measure of antisocial personality at age 18 included conviction, self-reported delinquency, self-reported violence, antisocial group behaviour, taking a prohibited drug, heavy smoking, heavy drinking, drunk driving, irresponsible sex (having intercourse without using contraceptives), heavy gambling, an unstable job record, anti-establishment attitudes, being tattooed, and self-reported impulsivity (all referring to the 15-18 age range).

The antisociality scales at the four ages were all significantly intercorrelated, showing the continuity in antisocial behaviour over time. For example, the scales at ages 18 and 32 correlated .55, despite the dramatic environmental changes in people's lives between these ages, as they left the parental home, went through a period of residential instability and then typically settled down with a wife or female cohabitee. Hence, the high correlation probably reflects individual rather than environmental stability. Over half (60%) of the most antisocial males at age 18 were still antisocial at age 32, compared with only 14% of the remainder at age 18 who became antisocial at age 32.

Characteristics of Delinquents

Before anyone was convicted, at age 8‑10, the future convicted juvenile delinquents differed significantly from the nondelinquents in many respects (and similar conclusions were obtained with self‑reported delinquency). For example, the future convicted juvenile delinquents were more likely than nondelinquents to have been rated troublesome and dishonest in their primary schools; 45% of troublesome boys at age 8‑10 were convicted as juveniles, compared with 14% of the remainder, giving an odds ratio of 5.0. Convicted delinquents tended to be from poorer families, from larger‑sized families, living in poor houses with neglected interiors, supported by social agencies, and physically neglected by their parents. However, they did not significantly tend to come from low socio‑economic status families (as measured by the occupational prestige of the family breadwinner) or to have working mothers. The delinquents were more likely to have convicted parents and delinquent older siblings; less than 5% of the families accounted for half of all the convictions of all family members. They tended to be receiving poor parental child-�rearing behaviour, characterized by harsh or erratic parental discipline, cruel, passive, or neglecting parental attitude, and parental conflict. Their parents tended to supervise them poorly, being lax in enforcing rules or under‑vigilant (West & Farrington, 1973; Farrington, 1992c).

Up to the 10th birthday, the future juvenile delinquents were more likely to have experienced broken homes or separations from their parents for reasons other than death or hospitalization. Their parents tended to be uncooperative towards the research, endorsed authoritarian child‑rearing attitudes on questionnaires, and were uninterested in the boy's education. Their mothers tended to be nervous or in poor physical health, while their fathers tended to have erratic job histories, including periods of unemployment. The boys who became delinquents were more likely to have low intelligence and poor school attainment and to be rated as daring by parents and peers. (Because intelligence and attainment were highly related, they were difficult to disentangle.) The boys' teachers said that they were hyperactive and had poor concentration. Interestingly, hyperactivity at age 8‑10 predicted juvenile convictions independently of conduct problems at age 8‑10 (Farrington, Loeber & Van Kammen, 1990). Delinquents tended to be impulsive on psychomotor tests and personality questionnaires and unpopular with their peers, but they were not nervous. They were likely to have below‑average height and average weight (West & Farrington, 1973; Farrington, 1992c).

Similar results have been reported by many other researchers (Farrington, 1995c). For example, Spivack, Marcus and Swift (1986) in Philadelphia found that troublesome behaviour in kindergarten (age 3‑4) predicted later delinquency. Lynam, Moffitt and Stouthamer‑Loeber (1993) in Pittsburgh showed that low intelligence was correlated with delinquency even after controlling for other variables such as social class and race. Klinteberg, Andersson, Magnusson and Stattin (1993) in Sweden reported that hyperactivity predicted later violent offending. Wilson (1980) in Birmingham concluded that the most important correlate of convictions, cautions and self‑reported delinquency was lax parental supervision at age 10, while parental physical punishment at age 7 predicted later convictions in the Nottingham study of Newson and Newson (1989).

Wadsworth (1979) in the National Survey found that boys from homes broken by divorce or separation had an increased likelihood of being convicted or cautioned up to the 21st birthday. In the Dunedin study in New Zealand, Henry et al. (1993) discovered that children who were exposed to parental discord and many changes of the primary caretaker tended to become antisocial and delinquent. McCord (1977) in Boston, and Robins, West and Herjanic (1975) in St. Louis, reported that criminal fathers tended to have delinquent sons. Numerous studies show that offenders tend to come from large‑sized families (Wadsworth, 1979; Ouston, 1984; Kolvin, Miller, Fleeting & Kolvin 1988; Newson et al., 1993). However, there is less agreement about the importance of socio‑economic status (Hindelang, Hirschi & Weis, 1981).

At age 14, when the boys were in their last year of compulsory schooling, the differences between juvenile delinquents and nondelinquents were similar in many respects to those found at age 8‑10. For example, the convicted delinquents still tended to have cruel, passive, or neglecting parents who were in conflict with each other, and they were still significantly low on measures of intelligence and attainment. They were described by their teachers as frequent liars, truants, daring, lacking in concentration or restless, and they left school at the earliest possible age of 15. Special efforts were made at age 14 to measure aggressiveness, and the delinquents proved to be significantly aggressive according to self‑reports, teacher ratings, and a semantic differential test. Also, the delinquents were likely to have relatively many delinquent friends (West & Farrington, 1973; Farrington, 1992c), as other researchers have found (Elliott, Huizinga & Ageton, 1985).

Predictors of Delinquency

The most important predictors, at age 8‑10, of later delinquency (whether

measured by convictions or by self‑reports) fell into six categories of theoretical

constructs:

(1) Antisocial child behaviour, including troublesomeness in school, dishonesty and aggressiveness.

(2) Hyperactivity‑impulsivity‑attention deficit, including poor concentration,

restlessness, daring and psychomotor impulsivity.

(3) Low intelligence and poor school attainment.

(4) Family criminality, including convicted parents, delinquent older siblings, and siblings with behaviour problem.

(5) Family poverty, including low family income, large family size, and poor housing.

(6) Poor parental child‑rearing behaviour, including harsh and authoritarian

discipline, poor supervision, parental conflict and separation from parents.

In regression analyses, it was generally true that at least one factor from each of these categories predicted offending independently of all the other categories. Far more is known about main effects than about interaction effects (Farrington, 1994c).

At age 8‑10, the best independent predictors of official juvenile delinquency (convictions between ages 10 and 16) were troublesomeness, daring, dishonesty, a behaviour problem sibling, a convicted parent, and poor parental child‑rearing behaviour (Farrington, 1995a). The best independent predictors of convictions up to age 32 were troublesomeness, a convicted parent, daring, low junior school attainment, poor housing and separation from a parent (Farrington, 1990b). When measures of antisocial child behaviour were excluded from the regression analyses in order to investigate explanatory factors only, the best independent predictors of convictions up to age 32 were large family size, a convicted parent, daring, poor housing, separation from a parent, low junior school attainment and not having few friends (Farrington, 1993a); social isolation seemed to act as a protective factor against delinquency.

It was surprising how accurately offending could be predicted. For example, a prediction scale was developed at age 8‑10 based on troublesome child behaviour, economic deprivation, low intelligence, a convicted parent, and poor parental child‑rearing techniques (Farrington, 1985). Of the 55 boys with the highest scores on this scale, 15 became chronic offenders up to the 25th birthday (out of only 23 in the whole sample), 22 others were convicted, and only 18 were unconvicted at that time. Hence, there were few "false positives".

Persistence in crime after the 21st birthday, as opposed to desistance, was predicted especially by heavy drinking at age 18 and by having a father who rarely joined in the boy's leisure activities at age 12, as well as by unemployment at age 16, low verbal IQ at age 8‑10 and not trying to do well at school (Farrington & Hawkins, 1991). Convicted teenagers who were both unemployed and heavy drinkers had an exceptionally high probability of persistence (nearly 90%). The best independent predictors at age 8‑10 of the chronic offenders up to age 32 were troublesomeness, a delinquent sibling, daring and a convicted parent (Farrington & West, 1993).

The late‑comers to crime after the 21st birthday were especially characterized by unemployment at age 18, but they were also predicted by economic deprivation, low attainment and impulsivity at age 8‑10 and by high self‑reported delinquency at age 18. Numerous factors predicted adult social dysfunction at age 32, but the best independent predictors were a poor relationship with the parents at age 18, an unskilled manual job at age 18, no examinations taken by age 18, being nervous‑withdrawn at age 8, small in height at age 14, hospital treatment for illness at age 16‑18, poor concentration or restlessness at age 12‑14 and high neuroticism at age 14 (Farrington, 1993a). Nervousness seemed to be negatively related to offending but positively related to other types of social dysfunction.

The most important predictors of antisocial personality at each age were also investigated (Farrington, 1995b). For example, the best independent predictors of antisociality at age 14 were antisociality at age 10, separation from a parent up to age 10, low nonverbal IQ at age 8‑10, not having few friends at age 8, and low junior school attainment. The best independent predictors of antisociality at age 18 were antisociality at age 14, a convicted parent at age 10, the father not joining in the boy's leisure activities at age 12, an unemployed father at age 14, a poorly educated father, leaving school at the earliest possible age of 15 and not having few friends at age 8.

Effects of Life Events on Development

An advantage of a longitudinal survey is that it is possible to investigate the effects of specific life events on the course of development of delinquency, by comparing before and after measures of offending and carrying out quasi�experimental analyses using each subject as his own control (Farrington, 1988b). For example, the effects on delinquent behaviour of being found guilty in court were studied. If convictions have a deterrent or reformative effect, a person's delinquent behaviour should decline after he is convicted. On the other hand, if convictions have stigmatizing or contaminating effects, a person's delinquent behaviour should increase after he is convicted.

These hypotheses were tested by studying self‑reports of delinquency before and after a boy was first convicted. It was found that boys who were first convicted between ages 14 and 18 increased their self‑reported delinquency afterwards, both in comparison with the level before and in comparison with the offending of a carefully matched group of unconvicted boys (Farrington, 1977). The same result was obtained in studying the effect of first convictions occurring between ages 18 and 21 (Farrington, Osborn & West, 1978). These results are concordant with Gold's (1970) findings in Michigan. There was some indication that the increase in self‑reported delinquency was related to the penalties given in court, although the numbers were small. The most common penalties following a first conviction were fines and discharges. There was no increase in self‑reported delinquency after a fine but a big increase after a discharge, suggesting that the increased delinquency may have been caused by a decrease in the boy's fear of a court appearance after he was discharged.

The effect on delinquency of going to different secondary schools was also investigated (Farrington, 1972). At age 11, most of the boys went to one of 13 secondary schools. These schools differed dramatically in their official delinquency rates, from one which had 20 court appearances per 100 boys per year to another where the corresponding figure was only 0.3. The key issue was whether the boys who went to high delinquency rate secondary schools became more likely to offend as a result, or whether the differing delinquency rates of the different secondary schools merely reflected differences in their intakes of boys at age 11.

As already mentioned, the best predictor of official juvenile delinquency in this survey was the rating of child troublesomeness at age 8‑10 by teachers and peers. Generally, the continuity between troublesomeness and delinquency was not greatly affected by the kind of school to which a boy went. There was a marked tendency

for the more troublesome boys to go to the high delinquency rate secondary schools, and the delinquency rates of different secondary schools largely reflected their different intakes. In contrast to Rutter, Maughan, Mortimore & Ouston (1979), who found somewhat greater school effects, we concluded that the secondary schools themselves had little effect on delinquency.

Another investigation of the effect of a specific event on delinquency focused on unemployment (Farrington et al, 1986). The complete job history of each boy between leaving school at an average age of 15 and the interview at age 18 was obtained, including all periods of unemployment. The key question was whether the boys committed more offences (according to official records) during their periods of unemployment than during their periods of employment.

The results showed that the boys did indeed commit more offences while unemployed than while employed. Furthermore, the difference was restricted to offences involving financial gain, such as theft, burglary, robbery and fraud. There was no effect of unemployment on other offences, such as violence, vandalism, and drug use, suggesting that the boys committed more offences while they were unemployed because they lacked money at these times. Furthermore, the effect of unemployment only applied to those with the highest prediction scores for crime, suggesting that unemployment had a criminogenic effect especially on those boys with the greatest prior potential for offending.

Another event that was investigated was the effect of moving out of London (West, 1982). Most families who moved out were upwardly mobile families who were moving to prosperous suburban areas in the Home Counties, often buying their own houses rather than renting in London. It was clear that both official and self‑reported delinquency of the men decreased after they or their families moved out of London, possibly because of the effect of the move in breaking up delinquent groups.

It is often believed that marriage to a good woman is one of the best treatments for male offending. When we asked our males in their twenties why they had stopped offending, they often mentioned marriage and the influence of women, as well as the fact that they did not hang around so much with delinquent friends. The Study males were growing up in a time period when men and women who wanted to live together usually got married. The effects of marriage were initially studied by following both convictions and self‑reported offending before and after early marriages (up to age 22). While the numbers were small, there was some suggestion that marriage led to a decrease in offending during the following two years, but only for offenders who married unconvicted women. Those who married convicted women continued to offend at the same rate after marriage as matched unmarried offenders (West, 1982). Similarly, the men's fathers who married convicted women incurred more convictions after marriage than those who married unconvicted women, irrespective of their conviction records before marriage.

A more detailed study of marriage up to age 32 (Farrington & West, 1995) showed that, while convicted offenders were no more or less likely than nonoffenders to get married, offenders were more likely to separate from their wife and to conceive a child while unmarried. Regression analyses showed that separation from a wife and conceiving a child outside marriage both predicted later offending (between ages 27 and 32) independently of all other variables, while an enduring marriage was negatively related to offending and seemed to act as a protective factor. Before‑and‑after matching analyses showed that getting married led to a decrease in offending compared with remaining single (irrespective of whether a man married a convicted or unconvicted woman), whereas separation from a wife led to an increase in offending compared with staying married. These results are concordant with those reported by Sampson and Laub (1993) in Massachusetts.

Protective Factors

Many of the findings of this survey are probably not surprising to people who work with offenders. Often, people are less interested in the continuity in offending and antisocial behaviour than in discontinuity. For example, why do some boys from criminogenic backgrounds nevertheless become successful nonoffenders, and why do some boys from favourable backgrounds nevertheless become antisocial offenders? We investigated good boys from bad backgrounds, and searched for protective factors against delinquency.

About one‑sixth of the boys (63) were identified as vulnerable at age 8‑10, because they possessed three out of five adverse background features (low family income, large family size, convicted parents, poor parental child‑rearing behaviour, and low nonverbal intelligence). Three‑quarters of these vulnerable males were convicted of criminal offences up to age 32, and the vulnerable males were also likely to be identified as social failures at age 32 on the combined measure of living circumstances and behaviour. We investigated whether the unconvicted quarter of these boys were affected by any protective factors that might have helped them to achieve successful life outcomes (Farrington et al, 1988b).

Our earlier research suggested that being nervous and withdrawn might act as a protective factor in insulating vulnerable boys against juvenile delinquency (West & Farrington, 1973). Similar tendencies were apparent in the latest analyses up to age 32, but the effects were relatively weak. The most important results were that boys with few or no friends at age 8, and those without convicted parents or behaviour problem siblings at age 10, tended to remain unconvicted; and that boys who were rated favourably by their mothers at age 10 tended to be leading relatively successful lives at age 32. Of course, it may be that the mothers accurately perceived who were the good boys at age 10, rather than that an approving mother had a positive effect on a boy's self‑concept. There was some evidence that shyness acted as a protective factor for nonaggressive boys and as an aggravating factor for aggressive boys.

One problem with these analyses was that the unconvicted boys at age 32 were not necessarily leading the most successful lives. We studied the characteristics of unconvicted vulnerable men, convicted vulnerable men, unconvicted nonvulnerable men and convicted nonvulnerable men at age 32 (Farrington et al, 1988a). Surprisingly, the unconvicted vulnerable men were often the most unsuccessful, for example in not being home owners, in living in dirty home conditions, in having large debts and in having low status, lowly paid jobs. They were also the most likely of these four groups to have never married, to have no wife or cohabitee, and to be living alone. Also, they were the most likely to be in conflict with their parents. However, they were generally well‑behaved, for example in not taking drugs other than marijuana and in being least likely to commit offences. Their good behaviour may be connected with the fact they were also the most likely to stay in every night.

For boys from vulnerable backgrounds, the most protective factors against offending in our research were the avoidance of contact with other boys in the neighbourhood, and the absence of convicted fathers and delinquent siblings, who all may have exerted undesirable influences. However, the unconvicted vulnerable males at age 32 were not necessarily leading the most successful lives. Their social isolation at age 8 seemed to lead to social isolation at age 32.

Recent Research on Aggression, Bullying and Truancy

There was significant continuity between childhood aggression and adult violence (Farrington, 1989a, 1991b), as other researchers have also reported (Eron & Huesmann, 1990). Boys who were aggressive in childhood or adolescence tended to be more deviant in adulthood: living in worse home circumstances, more in conflict with and violent towards their wife or cohabitee, more likely to be unemployed, heavier smokers and drinkers, more likely to be drunk drivers and drug takers, and committing more offences (including violence). This continuity, however, was probably not specific to aggression and violence but was part of the general continuity in antisocial and deviant behaviour from childhood to adulthood. This was why aggressive children had deviant life styles 20 years later as adults.

The most important predictors of adolescent aggression and adult violence tended to fall into the categories listed earlier. For example, the best independent predictors at age 8‑10 of convictions for violence up to age 32 were a low parental interest in education, daring, authoritarian parents, a convicted parent, low verbal intelligence and harsh parental discipline. The best independent predictors up to age 18 of self‑reported violence (involvement in fights) at age 32 were antisociality at age 18, the father rarely joining in the boy's leisure activities at age 12, no money saved at age 18, antisocial group membership at age 18, daring at age 8‑10, and a hostile attitude to the police at age 16 (Farrington, 1989a).

Conclusions drawn from these predictive analyses of aggression and violence were similar to those drawn about the prediction of delinquency and frequent offending. This further confirmed the argument put forward by West and Farrington (1977) that aggression was merely one element of a more general antisocial tendency, which arose in childhood and continued through the teenage and adult years. Violent offenders were very similar to nonviolent frequent offenders in childhood, adolescent, and adult features (Farrington, 1991b), suggesting that the causes of aggression and violence were essentially the same as the causes of persistent and extreme antisocial, delinquent and criminal behaviour.

The predictors of football violence at age 18 and violence against spouses and cohabitees at age 32 were also investigated (Farrington, 1994a). The best predictors of football violence were being relatively small, having a father who was not interested in the boy, not attending church, not being nervous, leaving school early and having authoritarian parents. Speculatively, relatively small boys whose fathers pay little attention to them might try to compensate for these handicaps by showing off and being aggressive in football crowds, especially if they are not inhibited by nervousness or religious beliefs. The best predictors of spouse assault were having a convicted parent, unpopularity, daring, separation from a parent and low intelligence. There seemed to be a link between experiencing parental disharmony and early separation from a parent in childhood, difficulties in relationships with peen and parents, and later difficulties in relationships with spouses and cohabitees. Teenage violence tended to develop into later spouse assault, but particularly for those aggressive males who had long‑lasting difficulties in their relationships with other people.

Research on bullying (Farrington, 1993d) shows continuity both within and between generations. There was a significant tendency for the males who reported that they were bullies at age 14 also to report that they were bullies at age 32 and that their children were bullies when the males were age 32. In addition to being bullies themselves, the men who had children who were bullies tended to be poor readers, heavy gamblers and unpopular in their teenage years, and they tended to have authoritarian parents. The men were also asked at age 32 about whether their children were victims of bullying. Those who had children who were victims tended to be those who had been unpopular and had few friends at age 8‑10, and those who were nervous and regular smokers at age 14. Knowing that these factors are associated with being bullied, it seemed likely that there was intergenerational continuity in being bullied as well as in bullying (see also Olweus, 1994).

Research on truancy (Farrington, 1980, 1995a) shows that primary school truants at age 8‑10 tended to become secondary school truants at age 12‑14. The best independent predictors, at age 8‑10, of secondary school truancy were troublesomeness, a behaviour problem sibling, a nervous or psychiatrically treated father, low nonverbal intelligence, separation from a parent, low parental interest in the boy's education, low school attainment and daring. Generally, truants and delinquents were similar in childhood, adolescent and adult features, but the most important difference was that nervousness was positively related to truancy but negatively related to delinquency. This suggested that, for some children, truancy was a behavioural symptom of a nervous‑withdrawn temperament rather than of an antisocial personality, thus confirming the distinction between truancy and school refusal (Boots, Foster, Brown & Berg, 1990).

Explaining the Development of Delinquency

There is far more agreement about risk factors than about their theoretical interpretation. The major risk factors for delinquency include poverty, poor housing, and living in public housing in inner city, socially disorganized communities (Farrington, 1995c). They also include poor parental child‑rearing techniques, such as poor supervision, harsh or erratic discipline, parental conflict, and separation from a biological parent. They also include impulsivity and low intelligence or attainment (which may reflect a poor ability to manipulate abstract concepts and deficits in the "executive functions" of the brain). It seems likely that communities influence parenting, and that parenting influences the development of impulsivity and low intelligence, which in turn are conducive to delinquency (Farrington, 1993b).

Other risk factors may be linked to poverty, poor parenting, impulsivity or intelligence. For example, teenage mothers tend to live in poverty, tend to use poor child‑rearing techniques, and tend to have impulsive children with low intelligence (Morash & Rucker, 1989). Perinatal complications, in combination with other risk factors, may cause neurological dysfunction, which in turn causes impulsivity or low intelligence (Kandel & Mednick, 1991). Large family size may lead to poor parenting, because of the problem of dividing attention between several children at once. Convicted parents may be poor supervisors of children and disproportionally separated from their spouses, or alternatively there may be genetic transmission of a biological factor linked to offending (Di Lalla & Gottesman, 1991). The links between delinquent friends, delinquent schools and offending are less clear, but may involve learning from deviant models. It is also likely that the occurrence of offences depends on situational factors such as perceived costs, benefits and opportunities (Clarke & Cornish, 1985).

It is important to establish which factors predict delinquency independently of other factors. Based on the independent predictors in the Cambridge Study, it might be suggested that impulsivity, low intelligence or attainment, poor parenting, an antisocial family and poverty, despite their inter‑relations, all contribute in some way to the development of offending. In addition, of course, there is significant continuity in offending and antisocial behaviour from childhood to adulthood, even though the prevalence of offending peaks in the teenage years. Any theory needs to give priority to explaining these results.

My theory of delinquency and antisocial behaviour (Farrington, 1986b, 1992b, 1993c) was designed to explain offending by working‑class males. It distinguishes explicitly between the long‑term development of antisocial tendency and the immediate occurrence of offences and other antisocial acts. The level of antisocial tendency depends on energizing, directing and inhibiting processes. The occurrence of offences and other antisocial acts depends on the interaction between the individual (with a certain degree of antisocial tendency) and the social environment, and on a decision‑making process in criminal opportunities.

It is proposed that the main energizing factors that ultimately lead to long‑term, between‑individual variations in antisocial tendency are desires for material goods, status among intimates, and excitement. The main energizing factors that lead to short‑term, within‑individual variations in antisocial tendency are boredom, frustration, anger and alcohol consumption. The desire for excitement may be greater among children from poorer families, perhaps because excitement is more highly valued by lower‑class people than by middle‑class ones, because poorer children think they lead more boring lives, or because poorer children are less able to postpone immediate gratification in favour of long‑term goals (which could be linked to the emphasis in lower‑class culture on the concrete and present as opposed to the abstract and future).

In the directing stage, these motivations lead to an increase in antisocial tendency if socially disapproved methods of satisfying them are habitually chosen. The methods chosen depend on maturation and behavioural skills; for example, a 5‑year‑old would have difficulty stealing a car. Some people (e.g. children from poorer families) are less able to satisfy their desires for material goods, excitement and social status by legal or socially approved methods, and so tend to choose illegal or socially disapproved methods. The relative inability of poorer children to achieve their goals by legitimate methods could be because they tend to fail in school and tend to have erratic, low status employment histories. School failure in turn may often be a consequence of the unstimulating intellectual environment that lower‑class parents tend to provide for their children, and their lack of emphasis on abstract concepts.

In the inhibiting stage, antisocial tendencies can be reduced by internalized beliefs and attitudes that have been built up in a social learning process as a result of a history of rewards and punishments. The belief that offending is wrong, or a strong conscience, tends to be built up if parents are in favour of legal norms, if they exercise close supervision over their children, and if they punish socially disapproved behaviour using love‑oriented discipline. Antisocial tendency can also be inhibited by empathy, which may develop as a result of parental warmth and loving relationships. There are individual differences in the development of these internal inhibitions. Perhaps because of associated neurological dysfunctions, children with high impulsivity and low intelligence are less able to build up internal inhibitions against offending, and therefore, tend to have a high level of antisocial tendency.

The level of antisocial tendency can also be increased in a social learning process if children are surrounded by antisocial models (criminal parents and siblings, delinquent peers, in delinquent schools and criminal areas). The belief that offending is legitimate, and anti‑establishment attitudes generally, tend to be built up if children have been exposed to attitudes and behaviour favouring offending (e.g. in a modelling process), especially by members of their family, by their friends, and in their communities.

In the decision‑making stage, which specifies the interaction between the individual and the environment, whether a person with a certain degree of antisocial tendency commits an antisocial act in a given situation depends on opportunities, costs and benefits, and on the subjective probabilities of the different outcomes. The costs and benefits include immediate situational factors such as the material goods that can be stolen and the likelihood and consequences of being caught by the police, as perceived by the individual. They also include social factors such as likely disapproval by parents or spouses, and encouragement or reinforcement from peers. In general, people tend to make rational decisions. However, more impulsive people are less likely to consider the possible consequences of their actions, especially consequences that are likely to be long delayed.

The consequences of offending may, as a result of a learning process, lead to changes in antisocial tendency or in the cost‑benefit calculation. This is especially likely if the consequences are reinforcing (e.g. gaining material goods or peer approval) or punishing (e.g. legal sanctions or parental disapproval). Also, if the consequences involve labelling or stigmatizing the offender, this may make it more difficult for him to achieve his aims legally, and hence there may be an increase in his antisocial tendency. In other words, events that occur after offending may lead to changes in energizing, directing, inhibiting or decision‑making processes in a dynamic system.

Applying the theory to explain some of the results reviewed here, children from poorer families may be likely to offend because they are less able to achieve their goals legally and because they value some goals (e.g. excitement) especially highly. Children with low intelligence may be more likely to offend because they tend to fail in school and hence cannot achieve their goals legally. Impulsive children, and those with a poor ability to manipulate abstract concepts, may be more likely to offend because they do not give sufficient consideration and weight to the possible consequences of offending. Also, children with low intelligence and high impulsivity are less able to build up internal inhibitions against offending.

Children who are exposed to poor parental child rearing behaviour, disharmony or separation may be more likely to offend because they do not build up internal inhibitions against socially disapproved behaviour, while children from criminal families and those with delinquent friends tend to build up anti‑establishment attitudes and the belief that offending is justifiable. The whole process is self �perpetuating, in that poverty, low intelligence, and early school failure, lead to truancy and a lack of educational qualifications, which in turn lead to low status jobs and periods of unemployment, both of which make it harder to achieve goals legitimately.

Limitations of the Study

The Cambridge Study provides information about the development of delinquency and antisocial behaviour in an inner‑city, working‑class British white male sample born about 1953. How far the same results would be obtained with females, Afro-�Caribbean or Asian children, suburban or rural children, middle or upper‑class children, children born more recently, or children born in other countries are interesting empirical questions. Generally, results obtained with the Cambridge Study are similar to those obtained with comparable male samples from Sweden (Farrington & Wikstrom, 1994), Finland (Pulkkinen, 1988) and from other Western industrialized countries, but more replications would be desirable. In particular, new longitudinal studies should be mounted in England, to investigate the effect of social changes such as the increased number of one‑parent families, and delinquent development in different ethnic groups. No new English studies on the development of delinquency have been mounted for many years.

The Cambridge Study inevitably has the usual methodological problems of longitudinal surveys. While the problem of attrition was largely overcome, the problem of testing effects (the effects on the subjects of repeated interviews) was not investigated. In retrospect, other boys from the same schools, who were not interviewed, should have been followed up in records to control for testing effects, but this was not a feature of the initial design (Farrington & West, 1981, p. 139).

Inevitably, some of the initial measures now appear rather old‑fashioned. For example, the method of measuring parenting variables (relying on psychiatric social workers interviewing parents) caused problems, and great efforts had to be made to try to achieve consistency and objectivity. While the main aim was to measure all relevant variables, the measurements of biological and neighbourhood variables and of psychiatric conditions were clearly inadequate, and the measurement of individual factors also left a lot to be desired in retrospect. Because we did not include a behaviour genetic design (e.g. studying twins or adopted children), we cannot investigate how much of the link between poor parenting and behaviour was mediated by genetic factors. Because of the focus on English personality questionnaires, key constructs such as empathy, egocentricity, guilt and depression were not measured. The attempts to study protective factors may have been inadequate, because far more was known about risk factors, and hence the measurement and analyses focussed on risk factors.

Validity checks were made where possible to compare interview data with external information from records. For example, admissions of convictions were compared with criminal records of convictions, and the mother's report of the boy's birth weight was compared with hospital records. Reliability checks were also made. For example, information about the same topic (e.g. school leaving age) from different interviews was compared, as was information about the same topic from different parts of the same interview. Generally, the men were randomly allocated between our two or three interviewers in order to study interviewer effects, which were rare. Sometimes, information could be tested for predictive validity. For example, more than twice as many of those who said that they had sexual intercourse without using contraceptives at age 18 subsequently conceived a child outside marriage as of the remainder (Farrington & West, 1995). All of these checks suggested that our subjects were genuinely trying to tell the truth. Of course, more validity checks would have been desirable, but these would have required more access to external records than we could achieve.

The sample was too small to study rare events, such as sex offenders or perinatal complications, effectively. The interviews were too infrequent to establish the exact or relative timing of many life events, and hence to establish developmental sequences between presumed causes and observed effects. The lack of experimental interventions made it difficult to test causal models effectively. The single cohort design made it difficult to distinguish between aging and period effects; for example, between ages 14 and 18 the percentage who had taken drugs increased from less than 1 % to 31 %, but this was probably a function of the time period (from 1967 to 1971).

In order to overcome these kinds of problems, it would be highly desirable in future longitudinal studies of delinquency to include multiple age cohorts and experimental interventions, to begin prenatally with one cohort, and to have more frequent interviews and larger samples (Farrington, 1991c; Farrington, Ohlin & Wilson, 1986; Tonry, Ohlin & Farrington, 1991). However, because ideal designs cost a great deal of money, more limited compromise designs like the Cambridge Study are likely to predominate in the foreseeable future.

Policy Implications

The major policy implication of the Cambridge Study is that, in order to reduce delinquency and antisocial behaviour, early prevention experiments are needed that target four important risk factors: low intelligence/attainment, poor parental child‑rearing behaviour, impulsivity and poverty (Farrington, 1994b). To the extent that these experiments lead to reductions in offending, this will increase our confidence that these risk factors have causal effects or are part of causal chains leading to offending.

Prevention experiments on offending and antisocial behaviour have been reviewed by Gordon and Arbuthnot (1987), Kazdin (1985,1987) and McCord and Tremblay (1992). I will focus on randomized experiments with reasonably large samples and with outcome measures of offending, since the effect of any intervention on offending can be demonstrated most convincingly in such experiments (Farrington, 1983; Farrington, Ohlin & Wilson, 1986). Many interesting experiments are not randomized (Jones & Offord, 1989), or do not have outcome measures of offending (Kazdin, Esveldt‑Dawson, French & Unis, 1987; Kazdin, Bass, Siegel & Thomas, 1989), or are based on very small samples (Shore & Massimo, 1979). It must be admitted that the experiments reviewed here provide glimmers of hope rather than a convincing basis for large‑scale social policy.

It is difficult to know how and when it is best to intervene, because of the lack of knowledge about developmental sequences, ages at which causal factors are most salient, and influences on onset, persistence and desistance. For example, if truancy leads to delinquency in a developmental sequence, intervening successfully to decrease truancy should lead to a decrease in delinquency. On the other hand, if truancy and delinquency are merely different behavioural manifestations of the same underlying construct, tackling one symptom would not necessarily change the underlying construct. Experiments are useful in distinguishing between developmental sequences and symptoms, and indeed, Berg, Hullin and McGuire (1979) found experimentally that decreases in truancy were followed by decreases in delinquency.

The ideas of early intervention and preventive treatment raise numerous theoretical, practical, ethical and legal issues. For example, should prevention techniques be targeted narrowly on children identified as potential delinquents or more widely on all children living in a certain high‑risk area (e.g. a deprived housing estate)? It would be most efficient to target the children who are most in need of the treatment. Also, some treatments may be ineffective if they are targeted widely, if they depend on raising the level of those at the "bottom of the heap" relative to everyone else. However, targeting potential delinquents may lead to damaging publicity and a refusal to cooperate. The most extreme group may also be the most resistant to treatment or difficult to engage, so there may be a greater pay‑off from targeting those who are not quite the most in need. Also, it might be argued that early identification could have undesirable labelling or stigmatizing effects, although the most extreme cases are likely to be stigmatized anyway and there is no evidence that identification for preventive treatment in itself is damaging. The degree of stigmatization, if any, is likely to depend on the nature of the treatment. In order to gain political acceptance, it may be best to target areas rather than individuals.

The ethical issues raised by early intervention depend on the level of predictive accuracy and might perhaps be resolved by weighing the social costs against the social benefits. In the Cambridge Study, as already mentioned, three‑quarters of vulnerable boys identified at age 10 were convicted up to age 32. It might be argued that, if preventive treatment had been applied to these boys, the one‑quarter who were "false positives" would have been treated unnecessarily. However, if the treatment consisted of extra welfare benefits to families, and if it was effective in reducing the offending of the other three‑quarters, the benefits might outweigh the costs and early identification might be justifiable. In any case, since the vulnerable boys who were not convicted had other types of problems, such as social isolation, even these boys might have needed and benefited from some kind of preventive treatment designed to alleviate their problems. Blumstein, Farrington and Moitra (1985) developed an explicit method of taking social costs and benefits into account in prediction exercises.

Improving School Achievement

If low intelligence and school failure are causes of offending, then any programme that leads to an increase in school success should lead to a decrease in offending. One of the most successful delinquency prevention programmes has been the Perry pre‑school project carried out in Ypsilanti (Michigan) by Schweinhart and Weikart (1980). This was essentially a "Head Start" programme targeted on disadvantaged African‑American children, who were allocated (approximately at random) to experimental and control groups. The experimental children attended a daily pre‑school programme, backed up by weekly home visits, usually lasting two years (covering ages 3‑4). The aim of the programme was to provide intellectual stimulation, to increase cognitive abilities, and to increase later school achievement.

About 120 children in the two groups were followed up to age 15, using teacher ratings, parent and youth interviews, and school records. As demonstrated in several other Head Start projects, the experimental group showed gains in intelligence that were rather short‑lived. However, they were significantly better in elementary school motivation, school achievement at 14, teacher ratings of classroom behaviour at 6‑9, self‑reports of classroom behaviour at 15 and self�-reports of offending at 15. Furthermore, a later follow‑up of this sample by Berrueta-�Clement et al (1984) showed that, at age 19, the experimental group was more likely to be employed, more likely to have graduated from high school, more likely to have received college or vocational training, and less likely to have been arrested.

By age 27, the experimental group had accumulated only half as many arrests on average as the controls (Schweinhart, Barnes & Weikart, 1993). Also, they had significantly higher earnings and were more likely to be home‑owners. More of the experimental females were married, and fewer of their children had been born out of wedlock. Hence, this pre‑school intellectual enrichment programme led to decreases in school failure, to decreases in offending, and to decreases in other undesirable outcome.

The Perry project is admittedly only one study based on relatively small numbers. However, its results become more compelling when viewed in the context of 10 other similar American Head Start projects followed up by the Consortium for Longitudinal Studies (1983) and other pre‑school programmes such as the Carolina Abercedarian Project, which began at age 3 months (Horacek et al, 1987). With quite impressive consistency, all studies show that pre‑school intellectual enrichment programmes have long‑term beneficial effects on school success, especially in increasing the rate of high school graduation and decreasing the rate of special education placements. The Perry project was the only one to study offending, but the consistency of the school success results in all projects suggests that the effects on offending might be replicable.

Improving Child‑Rearing Methods

If poor parental supervision and erratic child rearing behaviour are causes of delinquency, it seems likely that parent training might succeed in reducing offending. Many different types of family therapy have been used (Kazdin, 1987; Tolan, Cromwell & Brasswell, 1986), but the behavioural parent management training developed by Patterson (1982) in Oregon is one of the most hopeful approaches. His careful observations of parent‑child interaction showed that parents of antisocial children were deficient in their methods of child rearing. These parents failed to tell their children how they were expected to behave, failed to monitor the behaviour to ensure that it was desirable, and failed to enforce rules promptly and unambiguously with appropriate rewards and penalties. The parents of antisocial children used more punishment (such as scolding, shouting or threatening), but failed to make it contingent on the child's behaviour.

Patterson attempted to train these parents in effective child rearing methods, namely noticing what the child is doing, monitoring behaviour over long periods, clearly stating house rules, making rewards and punishments contingent on behaviour, and negotiating disagreements so that conflicts and crises did not escalate. His treatment was shown to be effective in reducing child stealing and antisocial behaviour over short periods in small‑scale studies (Dishion, Patterson & Kavanagh, 1992; Patterson, Chamberlain & Reid, 1982; Patterson, Reid & Dishion, 1992).

Another parenting intervention, termed functional family therapy, was evaluated by Alexander and his colleagues (Alexander & Parsons, 1973; Alexander, Barton, Schiavo & Parsons, 1976; Klein, Alexander & Parsons, 1977). This aimed to modify patterns of family interaction by modelling, prompting and reinforcement, to encourage clear communication between family members of requests and solutions, and to minimize conflict. Essentially, all family members were trained to negotiate effectively, to set clear rules about privileges and responsibilities, and to use techniques of reciprocal reinforcement with each other. This technique halved the recidivism rate of minor offenders in comparison with other approaches (client-�centred or psychodynamic therapy). Its effectiveness with more serious offenders was confirmed in a replication study using matched groups (Barton et al, 1985).

Reducing Impulsivity

Impulsivity might be decreased using the set of techniques variously termed cognitive‑behavioural interpersonal skills training, which have proved to be quite successful (e.g. Michelson, 1987). For example, the methods used by Ross to treat juvenile delinquents (Ross, Fabiano & Ewles, 1988; Ross & Ross, 1988) were solidly based on some of the known individual characteristics of offenders: impulsivity, concrete rather than abstract thinking, low empathy and egocentricity.

Ross believed that delinquents could be taught the cognitive skills in which they were deficient, and that this could lead to a decrease in their offending. His reviews of delinquency rehabilitation programmes (Gendreau & Ross, 1979,1987) showed that those which were successful in reducing offending generally tried to change the offender's thinking. Ross carried out his own "Reasoning and Rehabilitation" programme in Ottawa, Canada, and found (in a randomized experiment) that it led to a significant (74%) decrease in reoffending for a small sample in a short 9‑month follow‑up period. His training was carried out by probation officers, but he believed that it could be carried out by parents or teachers.

Ross' programme aimed to modify the impulsive, egocentric thinking of delinquents, to teach them to stop and think before acting, to consider the consequences of their behaviour, to conceptualize alternative ways of solving interpersonal problems, and to consider the impact of their behaviour on other people, especially their victims. It included social skills training, lateral thinking (to teach creative problem solving), critical thinking (to teach logical reasoning), values education (to teach values and concern for others), assertiveness training (to teach nonaggressive, socially appropriate ways to obtain desired outcomes), negotiation skills training, interpersonal cognitive problem‑solving (to teach thinking skills for solving interpersonal problems), social perspective training (to teach how to recognize and understand other people's feelings), role‑playing and modelling (demonstration and practice of effective and acceptable interpersonal behaviour).

Reducing Poverty

If poverty causes offending, then providing increased economic resources for the more deprived families should lead to a decrease in offending by their children. The major problem is to identify the active ingredient to be targeted. Low income and poor housing are inter‑related, but the causal chain linking these factors with offending is unclear. Experiments are needed which target each of these factors separately. It seems likely that relative rather than absolute income and housing quality are important, since there have been great improvements in recent years in average living standards.

The most relevant studies may be the income maintenance experiments carried out in the United States, which provided extra income for poor families. However, the only evaluation of the effect of income maintenance on children's delinquency (Groeneveld, Short & Thoits, 1979) did not yield positive results. There were no significant differences between the experimental and control groups in the later official offending records of children who were aged 9‑12 at the time of the treatment. Nevertheless, there is some evidence that extra welfare benefits given to ex‑prisoners can in some instances lead to a decrease in their offending (Rossi, Berk & Lenihan, 1980). In light of the clear link between poverty and antisocial behaviour, it is surprising that more prevention experiments targeting this factor have not been conducted.

Combined Interventions to Prevent Offending

A combination of interventions may be more effective than a single method, although combining interventions makes it harder to identify the active ingredient. As an example, Kazdin, Siegel and Bass (1992) found that a combination of parent management training and problem solving skills training was more effective in reducing self‑reported delinquency than either method alone.

An especially important combined prevention experiment was carried out in Seattle by Hawkins and his colleagues (Hawkins, Von Cleve & Catalano, 1991; Hawkins et al., 1992). This combined parent training, teacher training and skills training. About 500 first grade children (aged 6) in 21 classes in eight schools were randomly assigned to be in experimental or control classes. The children in the experimental classes received special treatment at home and school which was designed to increase their attachment to their parents and their bonding to the school, on the assumption that offending was inhibited by the strength of social bonds. Their parents were trained to notice and reinforce socially desirable behaviour in a programme called "Catch them being good". Their teachers were trained in classroom management, for example to provide clear instructions and expectations to children, to reward children for participation in desired behaviour, and to teach children prosocial methods of solving problems (Hawkins, Doueck & Lishner, 1988).

In an evaluation of this programme 18 months later, when the children were in different classes, Hawkins et al. (1991) found that the boys who received the experimental programme were significantly less aggressive than the control boys, according to teacher ratings. This difference was particularly marked for Caucasian boys rather than African‑American bur. The experimental girls were not significantly less aggressive, but they were less self‑destructive, anxious and depressed. Importantly, by the fifth grade, the experimental children were less likely to have initiated delinquency and alcohol use.

Another important combined experiment was carried out in Montreal by Tremblay and his colleagues (1991, 1992). They identified about 250 disruptive (aggressive/hyperactive) boys at age 6. Between ages 7 and 9, the experimental group received training to foster social skills and self‑control. Coaching, peer modelling, role playing and reinforcement contingencies were used in small group sessions on such topics as "how to help", "what to do when you are angry" and "how to react to teasing". Also, their parents were trained using the parent management training techniques developed by Patterson (1982). Generally, this prevention programme was quite successful. By age 12, the experimental boys committed less burglary and theft, were less likely to get drunk, and were less likely to be involved in fights than the controls. Also, the experimental boys were higher in school achievement. Interestingly, the differences in antisocial behaviour between experimental and control boys increased as the follow‑up progressed.

Conclusions

The Cambridge Study shows that the types of acts that lead to convictions (principally, crimes of dishonesty) are components of a larger syndrome of antisocial or deviant behaviour. Generally, males are versatile rather than specialized in their delinquency and antisocial behaviour. The high degree of continuity between ages 18 and 32, during a period of enormous environmental change, suggests that stability lies in the individual rather than in the environment. Our conclusion is that there are individual differences between people in some general underlying theoretical construct which might be termed "antisocial tendency", which is relatively stable from childhood to adulthood.

Importantly, there is relative stability but absolute change in antisocial behaviour. Offending may increase or decrease over time, but the worst offenders at one age still tend to be the worst at another age. Changes in the social environment inevitably lead to age‑appropriate changes in the behavioural manifestations of the underlying construct. For example, adults cannot truant from school, just as juveniles cannot hit their spouses. While the relative position of individuals on this underlying dimension is sufficiently stable to allow significant prediction from age 8 to age 32, the stability should not be exaggerated. Significant predictability does not mean that outcomes are inevitable or that people cannot and do not change. The good news is that most juvenile delinquents are leading quite successful lives by the age of 32.

The Cambridge Study also shows how far delinquency can be predicted in advance, in childhood. Previous projects did not measure such a wide range of theoretical constructs in advance of offending, and so they were not able to show so effectively which variables predicted delinquency independently of others or the relative importance of different variables. The most important independent childhood predictors of delinquency in this research could be grouped under the headings of antisocial child behaviour, impulsivity, low intelligence or attainment, family criminality, poverty and poor parental child‑rearing behaviour.

The Study also shows how far convicted offenders and unconvicted males were significantly different in numerous respects before, during and after their criminal careers. Furthermore, differences between convicted offenders and unconvicted males were similar to differences between high and low self‑reported delinquents, where boys in the high category were more frequent, serious and versatile offenders. Previously, it was argued that everyone committed delinquent acts and that differences between convicted and unconvicted males largely reflected police or court biases, but this view can be firmly rejected. The concordance between official records and self‑reports, and the ability of self‑reports to predict future convictions, shows that both measures are validly detecting the worst offenders.

The Study provides detailed information about criminal careers and co‑offending, and shows how far offending is concentrated in certain persons and certain families. A small number of chronic offenders, usually coming from multi‑problem families, accounted for substantial proportions of all official and self‑reported offences, and they were to a considerable extent predictable in advance. The fact that a large proportion of the crime problem is attributable to a small number of persons and families who are identifiable is potentially good news for prevention and treatment.

The Study also shows the importance of within‑individual analyses, in documenting the effects of life events on the course of development of offending. In particular, the Study provides perhaps the most detailed quantitative information about factors influencing desistance: being employed as opposed to unemployed, getting married and staying married, and moving out of London. It also shows that desistance from offending generally coincides with an improvement in other life circumstances.

The main policy implication of the Cambridge Study is that, in order to reduce offending and antisocial behaviour, early prevention experiments are needed targeting four important predictors that may be both causal and modifiable: low attainment, poor parental child‑rearing behaviour, impulsivity and poverty. There are hopeful results from existing small‑scale prevention experiments suggesting that improving the first three of these factors, at least, leads to reductions in offending. If larger‑scale experiments are similarly successful, the prevention measures should be implemented more widely by governmental agencies.

The Cambridge Study has influenced the thinking of the British government about crime policy. Certainly, between 1990 and 1992, it influenced Home Secretary Kenneth Baker and his junior minister John Patten, who drafted out a Green Paper on early prevention in 1992 that drew on the Study's conclusions. In 1989, John Patten stated that:

This important study has been influential both in this country and in the United States. It bears out much of the Home Office's current thinking about juvenile crime. It is an excellent example of how academic work, funded by government, can help in policy‑making ... I will be examining Dr. Farrington's conclusions, and their pointers toward future action, very fully indeed (Rock, 1994, p. 153).

In documenting the development of early prevention ideas in the Home Office in 1990‑92, Rock (1994) reported that:

There was one article [Farrington & West, 1990] that circulated about the [Home] Office at just that time, an article that had again made out the case for using longitudinal studies to predict and control delinquency, and for social prevention experiments to prevent the development of crime and antisocial behaviour ... An official remarked of Farrington's influence: "There is a theory for a particular time and maybe what he is saying is just one of the things that particularly suit at the moment". (Rock, 1994, p. 150).

Unfortunately, because of the 1992 general election and the change of Home Secretary from Kenneth Baker to Kenneth Clarke, the Green Paper was not published, but the policy ideas contained in it may yet reappear in another Home Office document.

Elliott, D. S., Huizinga, D. & Menard, S. (1989). Multiple problem youth. New York: Springer‑Verlag.

Eron, L. D. & Huesmann, L. R. (1990). The stability of aggressive behaviour‑even unto the third generation. In M. Lewis & S. M. Miller (Eds), Handbook of developmental psychopathology (pp. 147‑156). New York: Plenum Press.

Farrington, D. P. (1972). Delinquency begins at home. New Society, 21, 495‑497.

Farrington, D. P. (1977). The effects of public labelling. British Journal of Criminology, 17, 122‑135.

Farrington, D. P. (1980). Truancy, delinquency, the home and the school. In L. Hersov & I. Berg (Eds), Out of school. modern perspectives in truancy and school refusal (pp. 49‑63). Chichester: Wiley.

Farrington, D. P. (1983). Randomized experiments on crime and justice. In M. Tonry & N. Morris (Eds), Crime and justice, vol. 4 (pp. 257‑308). Chicago: University of Chicago Press.

Farrington, D. P. (1985). Predicting self‑reported and official delinquency. In D. P. Farrington & R. Tarling (Eds), Prediction in criminology (pp. 150‑173). Albany, NY: State University of New York Press.

Farrington, D. P. (1986a). Age and crime. In M. Tonry & N. Morris (Eds) Crime and justice, vol. 7 (pp. 189‑250). Chicago: University of Chicago Press.

Farrington, D. P. (1986b). Stepping stones to adult criminal careers. In D. Olweus, J. Block & M. R. Yarrow (Eds), Development of antisocial and prosocial behaviour (pp. 359‑384). New York: Academic Press.

Farrington, D. P. (1988a). Advancing knowledge about delinquency and crime: the need for a coordinated programme of longitudinal research. Behavioural Sciences and the Law, 6, 307‑331.

Farrington, D. P. (1988b). Studying changes within individuals: the causes of offending. In M. Rutter (Ed.), Studies of Psychosocial risk (pp. 158‑183). Cambridge: Cambridge University Press.

Farrington, D. P. (1989a). Early predictors of adolescent aggression and adult violence. Violence and victims, 4, 79‑100.

Farrington, D. P. (1989b). Later adult life outcomes of offenders and non‑offenders. In M. Brambring, F. Losel & H. Skowronek (Eds), Children at risk: assessment, longitudinal research, and intervention (pp. 220‑244). Berlin: De Gruyter.

Farrington, D. P. (1989c). Self‑reported and official offending from adolescence to adulthood, In M. W. Klein (Ed.), Cross‑national research in self-reported crime and delinquency (pp. 399‑423). Dordrecht, Netherlands: Kluwer.

Farrington, D. P. (1990a). Age, period, cohort, and offending. In D. M. Gottfredson & R. V. Clarke (Eds), Polity and theory in criminal justice: contributions in honour of Leslie T. Wilkins (pp. 51‑75). Aldershot: Gower.

Farrington, D. P. (1990b). Implications of criminal career research for the prevention of offending. Journal of Adolescence, 13, 93‑113.

Farrington, D. P. (1991a). Antisocial personality from childhood to adulthood. The Psychologist, 4, 389‑394.

Farrington, D. P. (1991 b). Childhood aggression and adult violence: early precursors and later life outcomes. In D.J. Pepler & K. H. Rubin (Eds), The development and treatment of childhood aggression (pp. 5‑29). Hillsdale, NJ: Erlbaum.

Farrington, D. P. (1991c). Longitudinal research strategies; advantages, problems and prospects. Journal of the American Academy of Child and Adolescent Psychiatry, 30, 369‑374.

Farrington, D. P. (1992a). Criminal career research in the United Kingdom. British Journal of Criminology, 32,521‑536.

Farrington, D. P. (1992b). Explaining the beginning, progress and ending of antisocial behaviour from birth to adulthood. In J. McCord (Ed.), Facts, frameworks and forecasts. Advances in criminological theory, vol. 3 (pp. 253‑286). New Brunswick, NJ: Transaction.

Farrington, D. P. (1992c). Juvenile delinquency. In J. C. Coleman (Ed.), The school years, 2nd edn (pp. 123‑163). London: Routledge.

Farrington, D. P. (1993a). Childhood origins of teenage antisocial behaviour and adult social dysfunction. Journal of the Royal Society of Medicine, 86, 13‑17.

Farrington, D. P. (1993b), Have any individual, family or neighbourhood influences on offending been demonstrated conclusively? In D. P. Farrington, R.J. Sampson & P.‑O. Wikstrom (Eds), Integrating individual and ecological aspects of crime (pp. 3‑37). Stockholm: National Council for Crime Prevention.

Farrington, D. P. (1993c). Motivations for conduct disorder and delinquency. Development and Psychopathology, 5, 225‑241.

Farrington, D. P. (1993d). Under‑standing and preventing bullying. In M. Tonry (Ed.), Crime and justice, vol. 17 (pp. 381‑458). Chicago: University of Chicago Press.

Farrington, D. P. (1994a). Childhood, adolescent and adult features of violent males. In L. R. Huesmann (Ed.) Aggressive behaviour current perspectives (pp. 215‑240). NewYork: Plenum Press.

Farrington, D. P. (1994b). Early developmental prevention of juvenile delinquency. Criminal Behaviour and Mental Health, 4, 209‑227.

Farrington, D. P. (1994c). Interactions between individual and contextual factors in the development of offending. In R. K. Silbereisen & E. Todt (Eds), Adolescence in context (pp. 366‑389). New York: Springer‑Verlag.

Farrington, D. P. (1995a). Later life outcomes of truants in the Cambridge Study. In I. Berg & J. Nursten (Eds) Unwillingly to school (4th edn). London: Gaskell, in press.

Farrington, D. P. (1995b). Psychosocial influences on the development of antisocial personality. In G. Davies, M. McMurran, C. Wilson & S. Lloyd‑Bostock (Eds) Psychology, law and criminal justice. Berlin: De Gruyter, in press.

Farrington, D. P. (1995c). The explanation and prevention of youthful offending. In J. D. Hawkins (Ed.), Current theories of delinquency and crime. New York: Cambridge University Press, in press.

Farrington, D. P., Berkowitz, L. & West, D. J. (1982). Differences between individual and group fights. British Journal of Social Psychology, 21, 323‑333.

Farrington, D. P., Gallagher, B., Morley, L., St. Ledger, R. J. & West, D. J. (1986). Unemployment, school leaving, and crime. British Journal of Criminology, 26, 335‑356.

Farrington, D. P., Gallagher, B., Morley, L., St. Ledger, R. J. & West, D. J. (1988a). A 24‑year follow-�up of men from vulnerable backgrounds. In R. L. Jenkins & W. K Brown (Eds), The abandonment of delinquent behaviour promoting the turnaround (pp. 155‑173). New York: Praeger.

Farrington, D. P., Gallagher, B., Morley, L., St. Ledger, R. J. & West, D. J. (1988b). Are there any successful men from criminogenic backgrounds? Psychiatry, 51, 116‑130.

Farrington, D. P., Gallagher, B., Morley, L., St. Ledger, R. J. & West, D.J. (1990). Minimizing attrition in longitudinal research: methods of tracing and securing cooperation in a 24‑year follow‑up study. In D. Magnusson & L. Bergman (Eds), Data quality in longitudinal research (pp. 122‑147). Cambridge: Cambridge University Press.

Farrington, D. P. & Hawkins, J.D. (1991). Predicting participation, early onset, and later persistence in officially recorded offending. Criminal Behaviour and Mental Health, 1, 1‑33.

Farrington, D. P., Loeber, R. & Van Kammen, W. B. (1990). Long‑term criminal outcomes of hyperactivity‑impulsivity‑attention deficit and conduct problems in childhood. In L. N. Robins & M. Rutter (Eds), Straight and devious pathways from childhood to adulthood (pp. 62‑81). Cambridge: Cambridge University Press.

Farrington, D. P., Ohlin, L. E. & Wilson, J.Q. (1986). Understanding and controlling crime. New York: Springer‑Verlag.

Farrington, D. P., Osborn, S. G. & West, D. J. (1978). The persistence of labelling effects. British Journal of Criminology, 18, 277‑284.

Farrington, D. P. & West, D. J. (1981). The Cambridge Study in Delinquent Development (United Kingdom). In S. A. Mednick & A. E. Baert (Eds), Prospective longitudinal research (pp. 137‑145). Oxford: Oxford University Press.

Farrington, D. P. & West, D. J. (1990). The Cambridge Study in Delinquent Development: A long‑term follow‑up of 411 London males. In H-J. Kerner & G. Kaiser (Eds), Kriminalitat: Personlichkeit, lebensgeschichte und verhalten (Criminality: Personality, behaviour and life history (pp. 115‑138). Berlin: Springer‑Verlag.

Farrington, D. P. & West, D. J. (1993). Criminal, penal and life histories of chronic offenders: risk and protective factors and early identification. Criminal Behaviour and Mental Health, 3,492‑523.

Farrington, D. P. & West, D.J. (1995). Effects of marriage, separation and children on offending by adult males. In Z. S. Blau & J. Hagan (Eds), Current perspectives on long and the life cycle. Delinquency and disrepute in the life course (vol. 4, pp. 249‑281). Greenwich, CT: JAI Press.

Farrington, D. P. & Wikstrom, P‑0.H. (1994). Criminal careers in London and Stockholm: a cross-�national comparative study. In E. G. M. Weitekamp & H-J. Kerner (Eds), Cross‑national longitudinal research on human development and criminal behaviour (pp. 65‑89). Dordrecht, Netherlands: Kluwer.

Fergusson, D. M., Horwood, L. J. & Lynskey, M. T. (1993). The effects of conduct disorder and attention deficit in middle childhood on offending and scholastic ability at age 13. Journal of Child Psychology and Psychiatry, 34, 899‑916.

Gendreau, P. & Ross, R. R. (1979). Effective correctional treatment: bibliotherapy for cynics. Crime and Delinquency, 25, 463‑489.

Gendreau, P. & Ross, R. R. (1987). Revivification of rehabilitation: Evidence from the 1980s. Justice Quarterly, 4, 349‑407.

Glueck, S. & Glueck, E. T. (1968). Delinquents and nondelinquents in perspective. Cambridge, MA: Harvard University Press.

Gold, M. (1970). Delinquent behaviour in an American city. Belmont, CA‑ Brooks/Cole.

Goldberg, D. (1978). Manual of the General Health Questionnaire. Windsor, Berks: NFER‑Nelson.

Gordon, D. A‑ & Arbuthnot, J. (1987). Individual, group and family interventions. In H. C. Quay (Ed.), Handbook of juvenile delinquency (pp. 290‑324). New York: Wiley.

Groeneveld, L P., Short, J. F. & Thoits, P. (1979). Design of a study to assess the impact of income maintenance on delinquency. Final Report to the National Institute of Juvenile Justice and Delinquency Prevention, Washington, DC.

Hawkins, J. D., Catalano, R. F., Morrison, D. M., O'Donnell, J., Abbott, R. D. & Day, L. E. (1992). The Seattle social development project: Effects of the first four years on protective factors and problem behaviours. In J. McCord & R. Tremblay (Eds), Preventing antisocial behaviour (pp. 139‑161). New York: Guilford.

Hawkins,J. D., Doueck, H.J. & Lishner, D. M. (1988). Changing teaching practices in mainstream classrooms to improve bonding and behaviour of low achievers. American Educational Research Journal, 25, 31‑50.

Hawkins,J. D., Von Cleve, E. & Catalano, R. F. (1991). Reducing early childhood aggression: Results of a primary prevention programme. Journal of the American Academy of Child and Adolescent Psychiatry, 30, 208‑217.

Henry, B., Moffitt, T., Robins, L., Earls, F. & Silva, P. (1993). Early family predictors of child and adolescent antisocial behaviour who are the mothers of delinquents? Criminal Behaviour and Mental Health, 3, 97‑118.

Hindelang, M. J., Hirschi, T. & Weis, J. G. (1981). Measuring delinquency. Beverly Hills, CA: Sage Publications.

Horacek, H.J., Ramey, C. T., Campbell, F. A., Hoffmann, K P. & Fletcher, R. H. (1987). Predicting school failure and assessing early intervention with high‑risk children. Journal of the American Academy of Child and Adolescent Psychiatry, 26, 758‑763.

Huizinga, D. & Elliott, D. S. (1986). Reassessing the reliability and validity of self‑report measures. Journal of Quantitative Criminology, 2, 293‑327.

Huizinga, D., Esbensen, F‑A. & Weiher, A. W. (1991). Are there multiple paths to delinquency? Journal of Criminal Law and Criminology, 82, 83‑118.

Janson, C‑G. (1984). Project Metropolitan: A presentation and progress report. Stockholm: Department of Sociology, Stockholm University.

Jones, M. B. & Offord, D. R. (1989). Reduction of antisocial behaviour in poor children by non�school skill‑development. Journal of Child Psychology and Psychiatry, 30, 737‑750.

Kandel, E. & Mednick, S. A. (1991). Perinatal complications predict violent offending. Criminology, 29,519‑529.

Kazdin, A. E. (1985). Treatment of antisocial behaviour in children and adolescents. Homewood, IL: Dorsey Press.

Kazdin, A. E. (1987). Treatment of antisocial behaviour in children: current status and future directions. Psychological Bulletin, 102, 187‑203.

Kazdin, A. E., Bass, D., Siegel, T. & Thomas, C. (1989). Cognitive‑behavioural therapy and relationship therapy in the treatment of children referred for antisocial behaviour. Journal of Consulting and Clinical Psychology, 57, 522‑535.

Kazdin, A. E., Esveldt‑Dawson, K, French, N. H. & Unis, A. S. (1987). Effects of parent management training and problem‑solving skills training combined in the treatment of antisocial child behaviour. Journal of the American Academy of Child and Adolescent Psychiatry, 26,416‑424.

Kazdin, A. E., Siegel, T. C. & Bass, D. (1992). Cognitive problem‑solving skills training and parent management training in the treatment of antisocial behaviour in children. Journal of Consulting and Clinical Psychology, 60, 733‑747.

Klein, N. C,, Alexander, J. F. & Parsons, B. V. (1977). Impact of family systems intervention on recidivism and sibling delinquency: a model of primary prevention and programme evaluation. Journal of Consulting and Clinical Psychology, 45, 469‑474.

Klinteberg, B. A., Andersson, T., Magnusson, D. & Stattin, H. (1993). Hyperactive behaviour in childhood as related to subsequent alcohol problems and violent offending: a longitudinal study of male subjects. Personality and Individual Differences, 15, 381‑388.

Kolvin, I., Miller, F. J. W., Fleeting, M. & Kolvin, P. A. (1988). Social and parenting factors affecting criminal‑offence rates: findings from the Newcastle thousand family study (1974‑1980). British Journal of Psychiatry, 152, 80‑90.

Kolvin, I., Miller, F. J. W., Scott, D. M., Gatzanis, S. R. M., & Fleeting, M. (1990). Continuities of deprivation? Aldershot: Avebury.

LeBlanc, M. & Frechette, M. (1989). Male criminal activity from childhood through youth. New York: Springer‑Verlag.

Loeber, R., Stoutharner‑Loeber, M., Van Karnmen, W. B. & Farrington, D. P. (1991). Initiation, escalation and desistance in juvenile offending and their correlates. Journal of Criminal Law and Criminology, 82, 36‑82.

Lynam, D., Moffitt, T. & Stouthamer‑Loeber, M. (1993). Explaining the relation between IQ and delinquency: class, race, test motivation, school failure or self‑control? Journal of Abnormal Psychology, 102,187‑196.

Magnusson, D. (1988). Individual deveopment from an interactional perspective. Hillsdale, NJ: Lawrence Erlbaum.

McCord, J. (1977). A comparative study of two generations of native Americans. In R. F. Meier (Ed.), Theory in criminology (pp. 83‑92), Beverly Hills, CA: Sage Publications.

McCord, J. (1979). Some child‑rearing antecedents of criminal behaviour in adult men. Journal of Personality and Social Psychology, 37, 1477‑1486.

McCord, J. & Tremblay, R. (1992). (Eds), Preventing antisocial behaviour. New York: Guilford.

Michelson, L. (1987). Cognitive‑behavioural strategies in the prevention and treatment of antisocial disorders in children and adolescents. In J. D. Burchard & S. N. Burchard (Eds), Prevention of delinquent behaviour (pp. 272‑310). Beverly Hills, CA: Sage Publications.

Moffitt, T. E. (1993). Adolescence‑limited and life‑course‑persistent antisocial behaviour: A developmental taxonomy. Psychological Review, 100, 674‑701.

Moffitt, T. E. & Silva, P. A. (1988). IQ and delinquency: a direct test of the differential detection hypothesis. Journal of Abnormal Psychology, 97, 330‑333.

Morash, M. & Rucker, L (1989). An exploratory study of the connection of mother's age at childbearing to her children's delinquency in four data sets. Crime and Delinquency, 35, 45‑93.

Newson, J. & Newson, E. (1989). The extent of parental physical punishment in the UK London: Approach.

Newson, J., Newson, E. & Adarns, M. (1993). The social origins of delinquency. Criminal Behaviour and Mental Health 3,19‑29.

Olweus, D. (1994). Bullying at school: basic facts and effects of a school based intervention programme. Journal of Child Psychology and Psychiatry, 35, 1171‑1190.

Ouston, J. (1984). Delinquency, family background, and educational attainment. British Journal of Criminology, 24, 2‑26.

Patterson, G. R. (1982). Coercive family process. Eugene, OR: Castalia.

Patterson, G. R., Chamberlain, P. & Reid, J. B. (1982). A comparative evaluation of a parent training programme. Behaviour Therapy, 13, 638‑650.

Patterson, G. R., Reid, J. B. & Dishion, T. J. (1992). Antisocial boys. Eugene, OR: Castalia.

Pulkkinen, L. (1988). Delinquent development: theoretical and empirical considerations. In M. Rutter (Ed.), Studies of psychosocial risk (pp. 184‑199). Cambridge: Cambridge University Press.

Reiss, A. J. & Farrington, D. P. (1991). Advancing knowledge about co‑offending: results from a prospective longitudinal survey of London males. Journal of Criminal Law and Criminology, 82, 360‑395.

Robins, L. N. (1979). Sturdy childhood predictors of adult outcomes: Replications from longitudinal studies. In J. E. Barrett, R. M. Rose & G. L. Klerman (Eds), Stress and mental disorder (pp. 219‑235). New York: Raven Press.

Robins, L. N. (1986). Changes in conduct disorder over time. In D. C. Farran & J. D. McKinney (Eds), Risk in intellectual and psychosocial development (pp. 227‑259). New York: Academic Press.

Robins, L. N., West, P. J. & Herjanic, B. L. (1975). Arrests and delinquency in two generations: a study of black urban families and their children. Journal of Child Psychology and Psychiatry, 16, 125‑140.

Rock, P. (1994). The social organization of a Home Office initiative. European Journal of Crime, Criminal Law and CriminaI Justice, 2, 141‑167.

Ross, R. R., Fabiano, E. A. & Ewles, C. D. (1988). Reasoning and rehabilitation. International Journal of Offender Therapy and Comparative Criminology, 32, 29‑35.

Ross, R. R. & Ross, B. D. (1988). Delinquency prevention through cognitive training. New Education, 10,70‑75.

Rossi, P. H., Berk, R. A. & Lenihan, K.J. (1980). Money, work and crime. New York: Academic Press.

Rutter, M. (1981). Isle of Wight and Inner London Studies. In S. A. Mednick & A. E. Baert (Eds), Prospective longitudinal research (pp. 122‑131). Oxford: Oxford University Press.

Rutter, M. (1989). Isle of Wight revisited: twenty‑five years of child psychiatric epidemiology. Journal of the American Academy of Child and Adolescent Psychiatry, 28, 633‑653.

Rutter, M., Maughan, B., Mortimore, P. & Ouston, J. (1979). Fifteen thousand hours. London: Open Books.

Rutter, M., Tizard, J. & Whitmore, K. (1970). Education, health and behaviour. London: Longman.

Sampson, R.J. & Laub, J. H. (1993). Crime in the making. Cambridge, MA: Harvard University Press.

Schweinhart, L. J., Barnes, H. V. & Weikart, D. P. (1993). Significant benefits, Ypsilanti, MI: High/Scope.

Schweinhart, L. J. & Weikart, D. P. (1980). Young children grow up. Ypsilanti, MI: High/Scope.

Shore, M. F. & Massimo, J. L. (1979). Fifteen years after treatment a follow‑up study of comprehensive vocationally‑oriented psychotherapy. American Journal of Orthopsychiatry, 49, 240‑245.

Spivack, G., Marcus, J. & Swift, M. (1986). Early classroom behaviours and later misconduct. Developmental Psychology, 22,124‑131.

Stattin, H. & Magnusson, D. (1991). Stability and change in criminal behaviour up to age 30. British Journal of Criminology, 31, 327‑346.

Thornberry, T. P., Lizotte, A.J., Krohn, M. D., Farnworth, M. & Jang, S.I. (1991). Testing interactional theory: an examination of reciprocal causal relationships among family, school and delinquency. Journal of Criminal Law and Criminology, 82, 3‑35.

Tizard, J. (1976). Psychology and social policy. Bulletin of the British Psychological Society, 29, 225‑234.

Tizard, J., Sinclair, I. & Clarke, R. V. G. (1975). (Eds), Varieties of residential experience. London: Routledge and Kegan Paul.

Tolan, P. H., Cromwell, R. E. & Brasswell, M. (1986). Family therapy with delinquents: a review of the literature. Family Process, 25, 619‑649.

Tonry, M., Ohlin, L E. & Farrington, D. P. (1991). Human development and criminal behaviour. New York: Springer‑Verlag.

Tracy, P. E., Wolfgang, M. E. & Figlio, R. M. (1990). Delinquency carom in two birth cohorts. New York: Plenum Press.

Tremblay, R. E., McCord, J., Boileau, H., Charlebois, P., Gagnon, C., LeBlanc, M. & Larivee, S. (1991). Can disruptive boys be helped to become competent? Psychiatry, 54, 148‑161.

Tremblay, R. E., Vitaro, F., Bertrand, L., LeBlanc, M., Beauchesne, H., Boileau, H. & David, L. (1992). Parent and child training to prevent early onset of delinquency: The Montreal longitudinal-�experimental study. In J. McCord & R. Tremblay (Eds), Preventing antisocial behaviour (pp. 117‑138). New York: Guilford Press.

Werner, E. E. & Smith, R. S. (1992). Overcoming the odds. Ithaca, NY: Cornell University Press.

West, D.J. (1982). Delinquency: Its roots, careers and prospects. London: Heinemann.

West, D. J. & Farrington, D. P. (1973). Who becomes delinquent? London: Heinemann.

West, D. J. & Farrington, D. P. (1977). The delinquent way of life. London: Heinemann.

Wilson, H. (1980). Parental supervision: a neglected aspect of delinquency. British Journal of Criminology, 20,203‑235.

Wolfgang, M. E., Thornberry, T. P. & Figlio, R. M. (1987). From boy to man, from delinquency to crime. Chicago: University of Chicago Press.

Zoccolillo, M., Pickles, A., Quinton, D. & Rutter, M. (1992). The outcome of childhood conduct disorder. implications for defining adult personality disorder and conduct disorder. Psychological Medicine, 22,971‑986.