Correctional Service Canada
Symbol of the Government of Canada

Common menu bar links

Research Reports

Warning This Web page has been archived on the Web.

Dynamic and Behavioral Antecedents to Recidivism: A Retrospective Analysis

1991, no. R-17

Prepared by:

Edward Zamble and Vernon L. Quinsey
Department of Psychology
Queen's University
Kingston, Ontario K7L 3N6

A final report to the
Research Branch
Correctional Service of Canada

April 1991

The authors wish to thank Beckey Hooey, Maria Hooey, Sherry Bradford, and Bill Frid for their assistance in conducting this research.


The prediction of both general and violent criminal recidivism of persons released from correctional institutions has received extensive study (for reviews see Gabor, 1986; Monahan, 1981; Quinsey, 1984; Waller, 1974). Enough work has been completed to establish a general consensus within the correctional research community about the classes of variables that are valid predictors of recidivism and the degree to which they are related to the criterion behaviors of interest.

We are fortunate to have a large number of Canadian follow-up studies of released inmates: Andrews and Friesen (1987), Carlson (1973), Gendreau, Grant, and Leipciger (1979), Gendreau, Madden,and Leipciger (1979; 1980), Hart, Kropp, and Hare (1988), Mandelzys (1979), Nuffield (1982), Porporino, Zamble, & Higginbottom (in process), Waller (1974), Wormith (1984), and Zarb (1978). Additional Canadian data come from follow-up studies of released mentally disordered offenders (e.g., Rice, Harris, & Quinsey, 1990; Rice, Quinsey, & Harris, in press; Quinsey & Maguire, 1986; Rice, Quinsey, & Houghton, 1990). American studies include: Adams (1983), Barton and Turnbull (1979), Beck and Shipley (1987), Brown, D'Agostino, and Craddick (1978), Gottfredson, Wilkins, and Hoffman (1978), Gottfredson, Mitchell-Herzfeld, and Flanagan (1982), Heilbrun, Heilbrun, and Heilbrun (1978), Holland, Holt, and Brewer (1978), and Rhodes (1986).

These studies agree that youthfulness and number of previous convictions are positively related to the probability of criminal recidivism. Other predictors, including age at first arrest, criminal versatility (variety of offending), alcohol abuse, and low educational attainment, are usually found to be positively but more weakly related to recidivism rates. Escape and escape attempts have also consistently been found to be related to higher recidivism rates.

A number of investigators have combined these and related predictors in various ways to obtain summary scores that are more closely related to recidivism than any predictor taken singly. Those developed in Canada include the general recidivism and violent recidivism scales of Nuffield (1982), the Level of Supervision Inventory (Andrews, Kiessling, Mickus, & Robinson, 1986; Bonta & Motiuk, 1985), and the Psychopathy Checklist Scale (Hare, 1985; Hare & Jutai, 1983; Hare & McPherson, 1984; Hart, Kropp, & Hare, 1988). Generally, each of these scales predicts recidivism far better than chance within its domain but there is still a great deal of room for improvement.

Limitations Of Previous Work

Regrettably, however, there is a great gap between the empirical predictive literature and practical needs of correctional management, such as aid in designing interventions, or selecting inmates for programs, or assistance in offender release policies (cf. Quinsey & Walker, in press). The empirical literature deals almost exclusively with static or "tombstone" predictors, such as age, offence history, length of institutionalization, etc. Because correctional program managers require information about predictors that they can modify in order to effectively plan interventions, the bulk of this empirical follow-up literature is essentially irrelevant to them, particularly because very few offenders are under indeterminate sentences.

This gap between the needs of program managers and the dominant focus of the empirical literature is most readily apparent in an area where one might least expect it: The prediction of violent re-offending among mentally disordered offenders. Mentally disordered offenders are typically dealt with by mental health professionals working in a psychiatric hospital system that explicitly espouses a treatment-rehabilitation model. Nevertheless, of 28 follow-up studies of released mentally disordered offenders identified in a review of this literature (Quinsey, 1988), 25 employed only static predictors and only three (of which two were essentially pilot investigations) attempted to predict recidivism from measures of therapeutic change.

Actuarial scales for predicting recidivism suffer from other limitations as well. Because they are empirically rather than theoretically derived, the level of prediction can only be improved by repeating the entire process of scale derivation. Moreover, many of the static factors commonly included in these scales are correlated with socioeconomic factors (Porporino, Zamble, & Higginbottom, in process); using such variables to determine release may be compounding inequity with injustice. For the vast majority of offenders, however, the question is not whether they will be released, but when. If the time of release is essentially determined at the beginning of a sentence, then imprisonment may provide little incentive for an inmate to change his current behavior patterns.

Thus, the bulk of the follow-up literature can provide very little information to guide correctional administrators in choosing appropriate programs for offenders or in making decisions based upon offender change. The paucity of well designed intervention evaluation studies carried out on prison populations further contributes to this problem. There is a need to redirect attention from the general determinants of recidivism (except to identify high risk groups for concentrated attention) to questions of how to reduce or prevent it in the community. All of the above leads to a consideration of the specific contemporaneous (dynamic) determinants of recidivism.

What is needed is a better understanding of the role of current factors in the causation of new offences. It is usually both empirically and conceptually unclear, as Mandelzys (1979) has pointed out, how much criminal recidivism is a result of unresolved problems within a released offender that could have been addressed during a period of imprisonment, and what proportion of new offences are caused by new environmental or offender problems. However, there have been some attempts to specify how the way in which offenders interact with their environment will determine future offenses (Zamble & Porporino, 1988) and it has been shown that one can predict recidivism using measures of current behavior as well as from static variables (Zamble and Porporino, in press; Porporino, Zamble, & Higginbottom, in process).

One factor that should obviously be considered is the effectiveness of supervision under early release. Unfortunately, the literature on the effectiveness of parole and mandatory supervision is quite small and it is replete with methodological problems (Nietzel & Himelein, 1987). In summarizing the best executed research on the issue of supervision versus no supervision, Gottfredson, Mitchell-Herzfeld, and Flanagan (1982) concluded that: "First, none of the studies indicates a lasting effect of parole supervision beyond the period of supervision itself. Second, the research seems to indicate an effect of parole supervision on recidivism during the course of the supervision, particularly in the initial period of release. Third, the effect indicated by the research does not appear to be very large."

The limited effectiveness of supervision follows from our arguments above, for it is difficult to know what sorts of parole programs to develop in the absence of knowing the antecedents of parole failure. By "antecedents" in this context, we mean specifiable dynamic conditions of the offender or identifiable environmental events that precede recidivism. Antecedent conditions are, therefore, variables that parole authorities or offenders themselves could potentially do something about in order to prevent the commission of a criminal act. Static personal characteristics of offenders are useful in this context as variables that define the risk group to which an offender belongs and as moderator variables, i.e., variables that determine the manner in which antecedents affect behavior. These static variables, however, must be supplemented by information about the dynamic antecedents of recidivism.

What do we know about the antecedents of parole failure? Not a great deal. With respect to general recidivism, Waller (1974) found that lack of employment, undesirable associates, fighting, not seeing one's children, and frequent drinking predicted re-offending. Hart, Kropp, and Hare (1988) similarly observed that instability in both employment and relationships during the follow-up period predicted re-offending.

New Work On Dynamic Factors

In terms of information about dynamic behavioral processes in re-offending, recent work from two sources offers the possibility of helping to fulfill some of the needs described above. The first comes from a relatively large scale Canadian study that attempted to specify how offenders interact with their environment, and especially how they cope with their problems (Zamble and Porporino, 1988). Although it was primarily designed to study behavior in prison, the study also included data on problems experienced by inmates prior to imprisonment, and the resulting coping attempts. Not only was the general level of coping disastrously poor, but there was also evidence of an association between poor coping and criminal behavior. For example, there were significant negative correlations between measures of the efficacy of coping and previous criminal history (retrospective); measures of coping and associated behavior from the original study were also useful in predicting recidivism (prospective) with accuracy in the same range as that for commonly used actuarial scales (Porporino, Zamble, & Higginbottom, in process).

These and other similar results led to the formulation of a "coping-criminality" hypothesis linking the repetition of criminal behavior to inadequate coping resources, along with a particular set of generalized behaviors, e.g., a large amount of time spent socializing in a diffuse network of casual acquaintances. Whether or not it is correct in detail, this hypothesis is supported by data, and some portion of it is very likely veridical. It represents a useful step in understanding the causes of recidivism because it emphasizes the role of deficiencies in ordinary behavioral interactions between the individual and his environment as a determinant of new criminal actions.

The second line of research of interest here is very similar, although it was developed as an explanation of relapses in addictive behaviors. Marlatt and Gordon (1980) have proposed that relapse is largely triggered by negative emotional states, interpersonal conflict, and social pressure.

Some work has already been done in applying relapse theory to one type of criminal behavior. Determinants of sexual re-offending among 136 child molesters and 64 rapists have been examined in some detail by Pithers, Kashima, Cumming, Beal, and Buell (1988). Nearly 90 percent of the sex offenders reported experiencing strong emotional states before relapse (the commission of a new sex offence): 94 percent of the rapists reported feeling anger, usually occasioned by interpersonal conflict; 46 percent of the child molesters reported experiencing anxiety and 38 percent reported depression (these emotional states appeared to be related to social disaffiliation). The chain leading to relapse seemed to begin with negative affect leading to paraphiliac sexual fantasies, then cognitive distortions, and, finally, passive planning just prior to the offence.

Based on 550 interviews of 311 child molesters under supervision, Frisbie (1969) concluded that, in addition to alcohol abuse, factors predicting recidivism were "...the desire for and selection of physically immature children as sexual objects, unorthodox ethical values, and grave difficulties in establishing meaningful relationships with adult females on a mature basis." (p. 223). The similarities between Frisbie's observations and those of Pithers et al., (fantasies, disaffiliation, and cognitive distortion) are striking. Planning and behavioral rehearsal as antecedents to serious sexual offences have also been noted by MacCulloch, Snowden, Wood, and Mills (1983).

Although they have been formulated quite independently and in different contexts, the coping-criminality hypothesis and relapse theory are clearly compatible, and probably even synergistic. The link between coping difficulties and criminal actions is likely emotional distress which triggers violent or irrational behavior, or which causes the individual to reduce his efforts at self-monitoring and self-control. Conversely for relapse theory, the precursors of the emotional responses preceding criminal recidivism are probably inappropriate or inadequate coping behavior.

The purpose of this research was to investigate the specific dynamic behavioral antecedents of criminal recidivism by interviewing offenders who had been returned to prison as release failures. This method raises questions concerning the informational value and reliability of information that can be obtained in this fashion. We would certainly not argue that offenders will tell openly or even know why it is they committed their offences. However, they can usually remember what their thoughts and feelings were when they committed a crime, and they will be willing to tell these to a researcher. Although offenders cannot be expected to have a theory that can explain their own criminal behaviors, they can report details of events preceding re-offending that will allow an investigator to construct a theory of re-offending. It is of interest in the present context that Frisbie (1969) was surprised at how much her interviewees would disclose to a project interviewer; indeed, the research team were often aware of impending relapse before the parole authorities.

The quality of the information appears to be dependent upon asking the right questions in the first place: for example, Quinsey & MacCulloch (in preparation) obtained detailed descriptions of antecedent conditions by asking minute questions in a lengthy clinical interview, whereas Quinsey (in preparation) frequently received logically unintelligible and extremely vague answers from offenders to the question "Why did this offence occur?" Similarly, Zamble and Porporino (1988) received detailed answers to questions about responses to specific situations but often only vague and self-contradictory answers to questions about the purposes of those responses.

Because most of the work described above has not dealt with serious offenders against persons (that is, those of most concern), this study concentrated on previously imprisoned men who had been returned to prison for violent offences.

Previous studies have also failed to deal with the problem of base rates. What are we to make, for example, of the finding (Pithers et al., 1988) that most sex offenders report negative affect before they committed their offence? What is the base rate of negative affect among released offenders of any type and is it related to the probability or type of re-offending? Perhaps more to the point, we are not even sure of the base rate of negative affect among demographically similar members of the general population. The observation that negative affect precedes sexual re-offending, therefore, may be correct but uninformative. Two types of comparisons are relevant here. The first is between re-offenders and those who successfully adapt to the free world after release; the second is among offenders who commit various different types of crimes, e.g., sex offenders vs. robbers.


Subject Recruitment

Potential subjects in the recidivist group were selected from among men who had committed violent (Schedule 1) offences while on Parole or Mandatory Supervision for a previous offence. The previous term must have been imprisonment of at least 6 months and the new offence must have occurred within one year of the previous release (although a few exceptions were made to include longer periods when suitable subjects were not available).

Subjects were recruited from among men returned to federal institutions in the Ontario region after Parole or Mandatory Supervision revocation for either conviction or being charged with the new offence. It was stipulated in advance that subjects subsequently found not guilty of the new offense for which they were charged would be dropped from the study. No such cases were encountered.

Potential subjects were identified from lists of new admissions for convictions or charges for Schedule 1 offences obtained from Regional Headquarters (Ontario) at weekly intervals. Subjects identified as having been released from federal institutions were targeted first. In addition, names of other possible subjects were screened with the aid of institutional staff in order to identify those who had served provincial terms from RCMP FPS records. Most of the inmates were located in the Reception Unit of Millhaven Institution, although the sample included a few subjects from Collin's Bay and Joyceville Institutions.

Lists of potential subjects were submitted to contact persons in each institution for clearance according to the criteria for eligibility. Inmates were excluded if they showed evidence of active or florid psychosis, subnormal intelligence (I.Q. more than one S.D. below the mean), literacy level so low that they could not complete questionnaires even with assistance, insufficient competence in English or French to allow an interview, or if they were judged by security officials to present a likely danger to an interviewer. These criteria excluded fewer than 10% of potential subjects.

The proposed comparison group of subjects will be comprised of men who have successfully completed at least one post-release year under supervision. Initially, they will be recruited from a random sample of cases under supervision in the Kingston office of the Parole Service. Subjects will be recruited from other districts as necessary to fill the required number. This part of the research is funded by the Social Sciences and Research Council of Canada and is currently underway.

Interview Measures

Data were obtained from a variety of sources. However, the most important was a structured interview with subjects. The interview form is presented in the appendix.

Following from the work cited above on relapse prevention, many of the questions focussed on the events and the sequences of behavior that ultimately led to release failure, concentrating on the period immediately prior to the subject's revocation. Information was elicited on the level of psychosocial functioning, the nature and severity of life problems, and the relationship of these problems to re-offending. Similarly, following from previous work on the relationship between coping and criminal behavior, questions were asked for detailed information about how the inmate dealt with his problems.

All questions did not apply to every subject, as the details of offences and their precursors varied. For example, for robbers we explored expectations about the amount of money to be gained, while for sex offenders, precursors to sexual aggression such as deviant sexual fantasies were covered. For subjects in the comparison group there are no new offences to discuss, but a parallel form is used to deal with events in the month preceding the interview.

Questionnaire Measures

Subjects were also given a set of questionnaires to gather supplementary information on specific topics. Some of the measures that we have chosen were developed for use with other populations or in very different circumstances, and required adaptation for the present purpose. Other measures were developed and tested for this project. For these reasons, the first few months of this project were devoted to developing, adapting and testing the measures.

The questionnaire measures are presented in the appendix.

  1. Alcohol and drug abuse is frequently found among offender populations and has also been found to predict re-offending (Waller, 1974). The Alcohol Dependence Scale (Skinner & Horn, 1984) as well as the Drug Abuse Screening Test (Skinner, 1982) were employed. These are brief self-report scales that measure the dimension of dependence on alcohol or drugs and the problems associated with it. Subjects were also been asked to answer in terms of their alcohol/drug use around the time of their release failure.
  2. Emotional state prior to and at the time of release failure was measured in several ways. Subjects were asked in the interview about their mood and emotional state in the period preceding the events which resulted in revocation. In addition, several questionnaires were included to assess subjects' emotional reactions. The Beck Depression Inventory (Beck, 1967) was employed as a standard measure of depression with good psychometric properties. Anger was measured with the Multidimensional Anger Inventory (Siegel, 1986). This inventory yields scores on anger arousability, range of anger eliciting situations, hostile outlook, as well as precipitants and expression of anger. It is internally consistent and has good test-retest stability. Anxiety was measured with the Spielberger Trait Anxiety Inventory (Spielberger, Gorsuch and Lushene, 1970). This scale's content, concurrent and construct validity compare favourably with other published tests of anxiety.
  3. Ability to cope in various situations was measured with the Coping Situations Questionnaire (Zamble, unpublished). A variety of common life situations that individuals must cope with are described. The respondent was instructed to imagine that each situation is currently happening to him, and to describe how he would deal with it or work it out. He was also asked questions about his thoughts and feelings in each case. While this instrument has not been standardized, it is derived from those used successfully in the past (Porporino, 1983; Zamble, Porporino, and Kalotay, 1984; Zamble and Porporino (1988).
  4. Other scales measured some specific behavioral patterns. A Socialization Questionnaire (Zamble, unpublished) asked subjects about relationships, leisure time and accommodation. Zamble and Porporino (1988) have found that offenders often depend on fluid and generally superficial relationships and that instability is found in most aspects of their lives, e.g., residences, activities and personal relationships. This questionnaire attempts to measure such a pattern. A Time Use Questionnaire (Zamble, unpublished) was also included. Developed from Zamble and Porporino's (1988) findings that many offenders often live in the present, without much concern for the past or future, this scale seeks to determine how people organize their time, and how they conceptualize time generally. The Social Desirability Scale from the PRF was used to assess the degree to which subjects attempted to "fake good".

File Data

Information on previous institutional behavior, primarily disciplinary offenses, and on problems seen during supervision by parole officers, were obtained from files. File data were also used for comprehensive information on previous criminal history.

Measures Of Risk

As discussed above, one purpose of this research was to examine and interpret dynamic risk factors in comparison with static antecedents of parole failure. The resultant information may be useful in improving the level of prediction of release failure possible from current instruments. Therefore, two of the best measures of risk that use static variables, each of them developed in Canada, will be included: The Nuffield General Recidivism Scale (Nuffield, 1982) and the Level of Supervision Inventory (LSI; Andrews, Kiessling, Mickus, and Robinson, 1986). Most of the items necessary for these scales were included in the interview, the others were taken from files.


Because the quality of much of the information of interest is dependent on the accuracy of recall, candidates were contacted as soon as possible after they were identified, usually within the first two weeks after their return to the penitentiary, usually in the Reception Unit at Millhaven, but some of them had already been sent to their receiving institutions. All interviews took place within 60 days of the offence.

Over 80% of those asked agreed to participate. For those who agreed, the interview started immediately after consent was given. The interview took between 60 and 90 minutes. The interviews were held in the common room or in interview rooms at the Millhaven reception units. The interviews were conducted by a single interviewer under semiprivate conditions.

Questionnaires were given to the subjects after the interview, although in some cases, the questionnaires had to be read to the subjects. Completion of the questionnaires took about 45 minutes. File data were taken from a systematic search done after the interview.

Confidentiality Of Information

In accordance with the standards of professional research, and the requirements for clearance by our Ethics Review Committee, we guaranteed confidentiality to subjects (see the consent form in the appendix). They were asked to sign a consent sheet in which this condition was clearly stated, and the researcher also signed to establish the contract. Each subject was assigned a unique number and all records were identified only with this number. The list of assignments of names to numbers were stored under lock and were accessible only to the principal investigators and the research assistants.

Data Analyses

The purpose of this study was to identify dynamic antecedents of recidivism that can be measured in the post-release period. In order to accomplish this aim, a large number of measures were examined simultaneously for their predictive value. These measures can be divided into several classes: historical (e.g., offence history), individual differences (e.g., anger proneness, intelligence, coping ability), general post-release (e.g., self ratings of mood in the time before the re-offense or during a specified period of successful supervision), and specific post-release (i.e., the particular events that preceded a specific arrest). The initial data analysis task was to reduce the number of variables within a given predictor domain. Measures that had intractable problems (e.g., missing data, extremely skewed distributions, no real variance) were eliminated. The remainder of the variables were examined for redundancy; highly intercorrelated variables were combined or eliminated.

The analyses in the present report are basically descriptive. In addition, we have compared offenders who were convicted of robbery to offenders who were convicted of serious assaultive offences on key variables.

When data from successful supervisees become available, hierarchical setwise analyses on the reduced data set using the measurement domains specified above as units will be employed to determine the best combination of variables for differentiating offender groups from each other and from the successful releasees.

Since the specific post-release variable set provided a unique set of events for each subject, it will not be included in the above analyses. This variable set is nonetheless of great practical and theoretical interest; it allows us to examine what particular antecedent events recidivists reported (for example, whether an offence was planned and how far back this planning began). Such information is directly relevant to programming for recidivism prevention and for parole supervision. This information will also be useful in other comparisons and analyses, e.g., close inspection of subjects without previous history of violence.


Data were collected from 100 inmates; 39 of them had been convicted of robbery and 52 were convicted of violent or sex offences. The principal comparative analyses in this report contrast these two groups of inmates.

Personal And Criminal History

The inmates' background characteristics and life histories are what one would expect in a sample of core offenders. Their average age was 29.4 years, ranging from 19 to 60 years. On average, they had successfully completed 9.5 years of school.

There was clear evidence of a long history of personal instability. The longest time they had held any single job was only 30.6 months, and one-third had never held a job longer than 6 months. The longest relationship with a woman was only a little longer, with a mean of 38.1 months, and 26% of subjects had never had a relationship longer than 6 months. Forty-eight percent said that they had been previously treated for an alcohol or drug problem, and 40% claimed to have been treated for a psychological or psychiatric problem, other than alcohol abuse, while living in the community. As expected in light of these statistics, 32% said that they had seriously considered suicide and 22% reported actually attempting it.

These indicators of poor mental health are matched by long and extensive criminal histories. Subjects reported that they were, on average; 14.6 years of age when they first had trouble with the law. A majority (57%) had other family members who had criminal records. They averaged 24.4 previous convictions recorded on their FPS forms, with a range from 3 to over 100. Most had violent offences among those listed, even before the current offence: The mean total of violent offences was 5.2. Of course, all had previously served time in prison, because this was one of our selection criteria. Most had served their preceding term in Ontario, and for 75% their last release was from a medium-security institution, most commonly a federal penitentiary. Most (89%) had been under supervision and revoked for the current offence. The average time on the outside between the previous release and the new arrest was only 4.7 months, with a range of a few days to almost 19 months, with 96% reimprisoned within a year.

Thus, it can be concluded that we were successful in defining and recruiting a population of recidivists with serious criminal histories. In addition, because we had specified a new violent offence as a selection criterion, all (except one) had such offences. The most serious current offence is shown in Table 1.

Table 1 - Principal Current Offence
Offence Number
Robbery 39
Assault (general) 34
Homicide, or aggravated assualt 9
Sexual Assualt (rape) 8
Sexual Assualt (pedophilia) 1
Other 9

The mean length of the aggregate current term was 52 months, with a range up to 14 years.

Behavior While On Release

The information we gathered was centered on the month before the occurrence of the current offence. The largest proportion of subjects (32%) lived in a nuclear family with wives and children, and another 8% lived with their parents; in contrast, 29% lived on their own, and another 18% lived with friends, situations likely to be somewhat less settled than living with family.

Looking at their living accommodations provides some further detail. Only 12% owned the accommodations in which they lived, while 59% resided in rented houses or apartments. Others lived in less stable places, 17% in rooming houses or institutions and 5% having no fixed address at all. At the time of their arrest, about as many (43%) were married or living in common-law relationships as were single (42%); the others were mostly divorced or newly separated from relationships.

In terms of employment, the greatest number (42%) were unemployed, an unemployment rate somewhat higher than found in previous samples of offenders. Only 39% were occupied in full-time jobs or school programs, with the others working part-time. The main source of income was from employment for 39%, while 33% were supported principally by government benefits including disability and welfare. In response to the same questions, 16% admitted that their principal income had come from illegal activities.

Much of the information on lifestyles is encapsulated in Table 2, which shows the amount of non-working time spent in a variety of categories. As can be seen, most subjects spent the greatest amount of time in unstructured activities, with the highest value in "Nonspecific activities with friends (hanging around)". In contrast, family duties and organized activities occupied on the average much less time.

Table 2 - Time In Various Activities
Category Mean Hours/Week Percent with None (0)
Family 8.2 38
Hobbies 2.8 73
Listening to Music 19.9 28
TV 11.0 27
Physical Activities 6.0 54
Specific Activity with friends 5.6 53
Nonspecific activity with friends 20.6 30
Other 10.3 65

We asked subjects about a variety of problems that they may have encountered. The highest percentage (59%) said that they had had some problem in their relationships with their wives or girlfriends, while other kinds of interpersonal problems, such as conflicts with work supervisors or co-workers, were far less frequently reported (7 and 6%, respectively). In response to specific questions aimed at assessing financial needs and motivations, many said that they had had money problems, but only 18% thought they had not had enough money to survive, while another 20% described their money problem as the (almost universal) gap between means and desires.

Given the offenders' perceptions of these problems, one might have expected subjects to look for assistance. However, only a minority said that they sought help anywhere. The greatest number (36%) asked for help from friends, most frequently loans when money was short. Only about half as many (19%) ever sought help from their wives. We had also asked whether they could have gone to any source for help, and about as many said that they could have gone to persons such as wives, parents, or friends (but did not) as actually did seek help. This appears to be confirmation of some of the Macho or "independent" descriptions of offender socialization patterns, although it-may also be the result of prison experience.

Simply asking subjects to name areas in which they experienced difficulties can be misleading because problems vary in seriousness from trivial to overwhelming. Near the end of the interview, therefore, subjects were asked to rate the seriousness of any problems they had mentioned as occurring during the month before the current offence on a 10-point scale. The results can be seen in Table 3. The three highest ranked problem areas, substance abuse, physical and (mostly) emotional problems, and money, were also the most frequently mentioned, each cited in some form by about two-thirds of subjects, so it is evident that they had felt seriously troubled in some respects.

Table 3 - Mean Rated Seriousness Of Selected Problem Categories
Problem Mean Seriousness
Alcohol/drug use 5.4
Money 5.0
Physical/emotional health 3.9
Wife/family 3.6
Housing, physical environment 3.5
Work/school (inc. unemployment) 3.4
Leisure time (inc. boredom) 2.8
Friends 2.7
Release supervision/conditions 2.7

Substance Abuse

On the basis of our theoretical expectations and prior knowledge, substance abuse and emotional dysphoria had been particularly targeted for detailed investigation. The resultant data are of some interest, and they also lead us into discussion of the precursors of the new criminal offences; they will be considered below.

It has been previously shown that offenders have substantial problems with alcohol and drug use, and that intoxication often accompanies the commission of criminal acts. For these reasons, prohibitions on drinking are a common condition of release supervision. Indeed, 78% of our subjects reported that they were under such restrictions during the last month before their new offence; this included most of those who were under supervision.

To determine how effective such restrictions were, we asked subjects when they first had violated any of their release conditions. The first conditions violated were almost always those prohibiting alcohol and drug use. To our surprise, 56% of those under supervision admitted to having violated their terms in the first week of release. When we asked for more specific information on how soon after release they had begun drinking, 44% said that they had at least one drink on their first day, and another 18% had drunk within the first week. Similarly, 31% had taken at least one illegal drug on the first day, and another 16% within the first week.

These figures are mutually supportive and they are strong evidence that the simple imposition of restrictions on alcohol and drug use had little effect. There was almost immediate violation of release conditions by a large majority of offenders. Given this, it is interesting to consider the reasons for the violation. In general, the motivations provided by subjects are not very profound: "Social circumstances or pressure" was the reason most commonly cited (in 22% of the sample).

Additional detail is provided in Table 4, which summarizes answers to specific questions about the reasons for the resumption of alcohol and drug use. In general, the resumption of alcohol or drug use does not seem to have been in response to any particular environmental circumstances. Rather, it appears to be just a resumption of previous behavior, after the interruption of a term of imprisonment.

Table 4 - Reasons For (Re-)Starting Alcohol And Drug Use After Release
Description Percent (Alcohol) Percent (Drug)
Pleasent Consequences 29 22
Social Pressure 25 22
Response to Problem 16 6
Habitual use 8 11
Don’t Know 4 3

In addition to information on first use, we also gathered a variety of data on the amount of drinking and drug use. Most of the sample drank some amount of alcohol, with only 15% claiming to abstain. The mean frequency of drinking occasions was 12.6 per month, i.e., almost half of all days. The mean quantity of drinks per occasion was 12.7. Combined, these figures yield a Frequency-Quantity Index (arguably the best single measure of consumption) with an average of 8.1 standard drinks daily for all subjects, including abstainers (or 9.6 excluding nondrinkers). This figure is very similar to the value of 8.0 found before (Zamble and Porporino, 1988).

Use of other drugs is harder to quantify, because the dosage and potency of street drugs varies so much, but we can at least show the frequency of use. Table 5 lists the percentages of subjects who said that they commonly used substances in each of a variety of categories. Overall, 65% of subjects used one of these substances. In general, the results are as expected, and as previously shown. However, the amount of use of cocaine seems much higher than in previous comparable surveys (Zamble and Porporino, 1988), providing confirmation for recent observations that cocaine use has increased considerably over the last decade. There appears to be compensatory decreases in some other categories, e.g., amphetamines and stimulants.

Table 5 - Categories Of Drug Use
Drug Class Percent Used
Cannabis 57
Amphetamines 1
Stimulants or hallucinogens 12
Sedatives or barbiturates 10
Opiates 5
Cocaine 28

It is important to consider not only the amount of substance use, but also the effects of those substances. The interview included questions asking about how frequently the respondent was verbally or physically aggressive, or, in contrast, socially withdrawn, after consuming alcohol or other drugs. About half of the subjects (from 47 to 53% for the respective effects) said that alcohol increased each of these behaviors. In the case of other drugs, about 30% reported aggression, while 73% said that they became withdrawn. Thus, subjects perceived a link between aggression and alcohol, but a smaller effect for other drugs, which they saw as more frequently linked with social withdrawal and isolation.

Some further information on alcohol and drug use was gathered using instruments developed by the Addiction Research Foundation. The mean score on the Alcohol Dependency Scale (ADS) was 9.6, with only 28% having scores of 0 (no evidence of dependency). On the Drug Abuse Screening Test (DAST) the mean was 5.7, with only 31% having scores of 0. These scores are useful as summary measures of substance abuse for comparison with other populations and for multivariate statistical analyses. In addition, information about the relationship between substance use and offending will be educed in a later section of this report.

Emotional States

Relapse theory argues that one of the primary antecedents of failure is the emergence of dysphoric emotional states, just as would be predicted from the failure of coping efforts, so the interview included items to measure emotional states. Overall, it was evident that subjects were far from enthusiastic about their lives in the period preceding their new offence.

Some summary evidence of this is available in answers to questions asking subjects to evaluate their lives in general for the relevant period. For example, 81% said that they had worried that their lives were not going the way they wanted. When we asked them to rate their lives in the community on a scale from 1 ("unbearable") to 100 ("all you'd ever want from life"), the mean was only 44.9, with about half of subjects giving ratings of below 50.

Much more detail can be seen in Table 6. A fairly comprehensive list of emotional states was presented to subjects and they were asked to specify which they had experienced in the final month before the offence. The emotions of principal clinical concern, i.e., depression, anxiety, and anger, had each been experienced by the majority of subjects. However, one might expect that a variety of states would be experienced by fairly high proportions of people over a period as long as a month, so the absolute numbers can be neglected, although the ordering of frequencies is probably meaningful.

Table 6 - Emotions Experienced In The Month Preceding The Offence
State Percent Experencing Percent Strongest
Hopelessness 38 4
Depression 58 23
Moody/brooding 41 5
Anger 52 17
Frustration 57 9
General stress 41 5
Anxiety 54 15
Guilt 31 4
Loneliness 37 4
Boredom 31 1
Sexual frustration 14 1
Numbness 23 0
Positive emotions 51 8
Other 11 3

In addition, and more revealing for descriptive purposes, is the choice of which emotion predominated for each subject in the same period. As can be seen from the table, depression (23%) was the most frequent choice, followed by anger (17%) and anxiety (15%). Although a majority of subjects had reported feeling positive emotions during this period, such emotions were the strongest for only 8%. Other states, such as boredom and sexual frustration, although frequent, seem to be only relatively minor constituents of most subjects' palette of emotions.

Thus, it would appear that subjects' lives on the outside were often characterized by dysphoric emotions. This may have been reactive and specific to what they were experiencing at that time, the result of generalized personality dispositions, or, more likely, a mix of the two. Some additional information on emotional experiences closer to the time of the offence will be considered in the following section.

However, we have additional evidence that the dysphoria persisted, and that in many cases it may have been of clinical magnitude. Our set of questionnaires included the Beck Depression Inventory, the most widely used measure of clinical depression. The mean score was 16.9, substantially higher than even that for a random sample of inmates at the beginning of their terms (the mean in Zamble and Porporino (1988) was 13.0) and far higher than any non-clinical population norm. Almost half of the sample had scores higher than 15, a level used to indicate symptoms at a clinical level.

Similarly, scores on the Spielberger State Anxiety Index averaged 48.1. On this scale, a standardization group of patients diagnosed as suffering from anxiety reactions had a mean of 46.0, so the obtained scores indicate a very substantial amount of disturbance.

Thus, although concurrent information on the degree of disturbance experienced by subjects on the outside is not available, there is strong, if retrospective, evidence of major disturbances during their community tenure.

It is very likely that these emotional disturbances are linked to difficulties in coping with problems in the community. One of our questionnaires assessed the effectiveness of coping efforts, using a set of five common problem situations; the results ought to be comparable to those found previously when coping was assessed with each individual's unique set of problem situations. On this scale, the mean Coping Efficacy was 9.9, on a scale with a practical working range of 4 to 20. In contrast, the comparable value from a sample of correctional workers (Hughes, 1990) was about 16. (Moreover, the current values are very likely overestimates, since questionnaire administration requires that problems which the subject omits be counted as missing, while in face-to-face administration these are likely the cases that would have the lowest scores.) Only 12% were above the level of 12, considered the point of mediocre coping, and none were as high as the mean for the noncriminal sample. Although coping was not the focus of this study, it can be said that subjects had major deficiencies in the ability to cope with everyday problems.

Offence Precursors

Although the preceding sections have established the occurrence of certain behaviors and emotional patterns while subjects were on the outside, it remains to be seen whether such events were more closely involved with the commission of their new violent crimes. This was one of the primary purposes of the study, so much of the interview focused on experiences and events preceding the offence.

For many subjects, the beginning of the offence process seems to have been related to some exacerbation of their negative mood states. We surveyed their emotional responses in the last 48 hours preceding the offence, just as we had done for the longer period of a month reported in Table 6. The comparisons between these results, shown in Table 7, and those for the longer period, are instructive.

Table 7 - Emotions Experienced Within 48 Hours Prior To The Offence
State Percent Experencing Percent Strongest
Hopelessness 24 5
Depression 37 17
Moody/brooding 24 2
Anger 43 22
Frustration 37 4
General stress 28 2
Anxiety 38 14
Guilt 16 3
Loneliness 25 4
Boredom 15 1
Sexual frustration 9 1
Numbness 14 3
Positive emotions 32 14
Other 15 7

As one would expect, the numbers of subjects who reported experiencing various emotions was generally lower in the shorter period, with the numbers decreasing overall by about one-third. However, the percentage of those reporting some anger fell by less than the others, from 52 to 43. At the same time, the frequency of feelings of guilt and boredom dropped more than the others. From this, one might infer that the approach of the offence was characterized by a relative shift from general dysphoria to anger. The occurrence of events and thoughts that might lead to an offence would likely drive out boredom and suppress guilt feelings.

The above reasoning is quite speculative, and the figures in the tables make it clear that no dramatic shift was visible for most inmates. However, the argument is strengthened somewhat when one looks at the mood states named by subjects as the strongest they had experienced in the two periods. The percentages increased in only two categories: Anger, and positive mood. Again there seems to be a shift to anger, although the significance of the corresponding increase in positive mood states is unclear.

Regardless of the validity of the above interpretation, the evidence does show that dysphoric emotional states predominated in the period immediately preceding the offence. Anger (22%) was clearly the most common state at that time, followed by depression and anxiety. This supports the argument that such states are linked with the occurrence of criminal acts, although the strength of the argument must be limited by the present lack of baseline or comparative data.

The argument for links between mood state and offending is also strengthened by other results. We asked subjects to tell us whether they thought that something in particular had happened just before their first thoughts of the offence. Most of them (76%) responded affirmatively. Such a recognition is of course retrospective and may contain some elements of self-justification, but it is interesting nevertheless.

Much more direct was a question asking subjects to specify what they thought had led them to the offence. For this, we had a set of categories established in advance, with the results shown in Table 8. Although financial gain is the single most frequently chosen motive, it was cited by only one-quarter of subjects, and the two anger categories together contain the same proportion. Over a third of the sample gave reasons that were definitely linked to emotional or mood states. Interestingly, the numbers citing social or peer pressure, or boredom and the need for excitement, were low. One might conclude that there were several subpopulations of offenders here, one acting for financial gain, and the other precipitated by poor coping and emotional states. Together, these categories account for about three-quarters of the known reasons.

Table 8 - Thoughts About What Led To Offence
Category Percent
General anger or frustration 12
Specific anger 14
(Emotionally) out of control 11
Boredom 3
Money 25
Sexual frustration 1
Peer pressure 6
Don’t know 19
Other 7

This perceived motivation is strongly related to the type of offence. Financial gain was cited for 88% of robbers, while it was rarely listed for other types of offenders. Statistically, the association is significant at the most stringent level of confidence (X2 = 40.68, p < .001). This provides some evidence that offenders' cognitions are differentially associated with the nature of their offences.

In addition to emotional difficulties, the preceding sections had indicated the possible contribution of substance abuse to the occurrence of the new offence. Information on drinking and drug-taking in the 24 hours before the offence is presented in Table 9.

Table 9 - Drugs In The 24 Hours Preceding Offence
Drug Class Percent Using
Alcohol 67
Cannabis 22
Amphetamines 0
Stimulants or hallucinogens 1
Sedatives or barbiturates 10
Opiates 3
Cocaine 17

For those who were drinking, the mean number of drinks in the 24-hour period was 11.2, somewhat higher even than the overall daily average for drinkers. It is interesting that only a minority of those who commonly used cannabis or hallucinogens reported use just before the offence, while all of those using "downers", two-thirds of the cocaine users, and most of the alcohol drinkers had used their respective substances in that period. This may reflect differences in addictiveness between these types of drugs, or it may show some relationship to the occurrence of violent behavior.

There are also some relationships between substance abuse and the type of offence committed. For purposes of comparisons, all offences clearly involving personal attacks, i.e., homicide, assault, rape, etc., were combined into a single category, and contrasted with robbery and with a third category (other) where the nature of the interaction between offender and victim was unclear, e.g., paedophilia. Measures of substance abuse were also reduced to two or three categories to increase the power of nonparametric statistical analyses.

These analyses showed that robbers used (nonalcohol) drugs more frequently than others (X2 = 9.81, p < .05) and also that they used more types of drugs (X2 = 10.88, p < .05). In contrast, robbers seemed to use alcohol less than the other subjects, and, although the results did not quite reach statistical significance for any single measure (e.g., for the Frequency-Quantity Index, (X2 = 5.40,p < .07) the various measures all show the same tendency.

In summary, there is evidence that some of the features evident in offenders' lives on the outside continued until the time of their most recent criminal behavior. There is also (weaker) evidence of shifts in emotional state just before the offence, and some indication of heavy substance abuse at the same time. The type of substances used may be differential across offences. Many subjects themselves thought that emotional difficulties were the cause of their offences, and that their substance abuse was linked to aggressive and violent behavior. If these are truly precursors of a relapse into recidivism, then we ought to be able to link them to events in the offence process, so we are led to consider some evidence on just what happens during that process.

Offence Process

In addition to looking for the situational and internal precursors of offences, we wanted to be able to describe the progression of events actually involved in the offence itself. This is important for a number of reasons: For example, it can tell us about the realistic chances of being able to intervene just before an offence, or at least how to recognize that a critical series may have begun. Therefore, much of this study was concerned with offenders' actions and thoughts that were directly involved in the offence.

One can think of a criminal offence as the endpoint of a decision process that progresses from casual thoughts to a definite commitment to action. Therefore, we plotted a series of six landmarks: The first passing or fleeting thought of committing an offence; the first longer thought, defined as lasting a minute or more; the first time the person considered that he might actually carry out an offence; the first casual plans; the first specific or detailed plans; and, finally, the point of no return, where carrying out the offence was inevitable.

One would expect that not every offence would include each of these stages. For example, the first two stages might occur together, or the planning stages might also be combined; in very impulsive or spontaneous cases, the planning stages might be abbreviated or even eliminated. Therefore, during the interview, we asked subjects whether - and, if so, when - they had experienced each stage.

To help maintain consistency and to make the task simpler, subjects were given a visual image of a time line (p. 19 of the Interview Protocol, see Appendix). The time represented spans the time from several months before the offence up to its occurrence, in "psychological logarithm" units that telescope the entire period backward from the offence, so that short periods close to the offence are represented by the same lengths as longer periods further away. Subjects were asked to place each of the critical sequence events on this line.

Table 10 - Thinking Landmarks In Offence Process (Percentages Of Subjects)
Time Interval First Longer Consider
One month or more before offence 26 14 7
One week or more (but less than a month) 11 11 9
One day or more (but less than a week) 10 12 13
Hours (but less than a day) 8 11 13
Within an hour of the offence 6 10 8
At the time of the offence 38 44 50

First = First passing though of offence Longer = First longer thought ( > 1 min.) of offence Consider = First time considered might actually do it

The results can be seen in Tables 10 and 11, which show how long before the offence subjects located each of the landmarks. Overall, the lack of anticipation is impressive, whichever measure one considers. For close to half of the subjects, the entire process, from first passing thought to commission of the offence, was collapsed into an hour or so. While one can understand that the point of no return would be close to the event, it is somewhat surprising that so few did any real planning, e.g., only 26% say that they considered definite plans or rehearsed for more than an hour before the offence.

Table 11: Planning Landmarks In Offence Process (Percentages Of Subjects)
Time Interval Plan 1 Plan 2 No Return
One month or more before offence 4 3 0
One week or more (but less than a month) 5 3 0
One day or more (but less than a week) 10 11 6
Hours (but less than a day) 11 9 11
Within an hour of the offence 12
At the time of the offence 59 62 76

Plan 1 = First thoughts of planning offence Plan 2 = First definite plans No Return = When offence was inevitable

type of offence and perceived motivation, subjects' cognitions about the cause of their offence were also significantly related to the length of anticipation, i.e., those who acted for financial gain had thought of the offence earlier and planned longer than others. For example, 52% of subjects motivated by financial gain planned more than one hour, as compared to only 16% of the others (X2 = 10.92, p < .01).

In addition, antecedent factors captured in some of our measures also affected the time line. Means for each of the landmarks (again, with the exception of the point of no return) were significantly shorter for subjects who consumed more than 3 standard drinks of alcohol in the 24 hours preceding the offence. Using again the time of first planning as representative, 11% of those consuming more than 3 drinks began planning more than an hour before the offence, as opposed to 38% of those who drank less or not at all (X2 = 7.43, p < .01). Similarly, there was evidence of effects from the amount of drug use, although it was weaker.

Other analyses showed relationships between the offence time line and emotional events. Subjects who had felt angry during the final month were more likely to think of an offence earlier than others (X2 = 5.19, p < .05) although they did not plan any earlier. However, those who felt generally stressed or anxious not only thought of offences earlier (X2 = 4.42 and X2 = 7.31 for the two respectively, p < .05) but they also decided significantly sooner that they might actually commit the offence (X2 = 3.99 and X2 = 5.97 respectively, p < .05) and differences for the other landmarks approach significance. In contrast, those who reported feeling positive emotions in the month before the offence were likely to enter into the critical sequence later, with significant differences visible up to the time of definite planning. Interestingly, feelings of depression or hopelessness show no suggestion of a relationship to the onset of the offence pattern.

Some further data provide significant additional detail about subjects' thought processes preceding their criminal acts. In addition to locating events on the time line, we asked them whether they had ever daydreamed or fantasized about a new offence. Most (86%) said that they had not entertained themselves with such thoughts. However, the more interesting question was the follow-up, which asked why they had not done so. Only 7 men answered that they had been worried about the possible negative consequences, and only 4 said that they had considered possible victims. In answer to a different question, which asked specifically and directly whether they had ever thought of the victim of their most recent offence before committing it, only 1 subject said yes, and the other 99 said that they had not.

When we asked whether they had ever fantasized about pleasant or rewarding consequences of possible offences, a much greater proportion (61%) of those answering admitted to having done so. For these, the most frequently cited anticipated reward was material gains (33%), while self-esteem or peer-esteem were very rarely cited (1% each). This strongly reinforces evidence considered earlier about perceived motivations.

In general, these results document the spontaneous nature of violent offences far better than anything we are aware of in the previous literature. Although the impulsive nature of habitual offenders had been widely reported for some time, this may be the first quantitative description of the process. The implications for models of rational considered decision making, on which most of the criminal law is based, would appear to be devastating.

It would appear that for many offenders the offence process has something of a ballistic nature: Once it is set off, it runs very quickly on its course. At the same time, the onset of the chain is affected by certain behavioral and emotional events. Thus, treatments or interventions that concentrate on criminal cognitions or habits of impulsive thinking are unlikely to succeed in the prevention of violent new offences. Rather, we would argue that intervention must deal instead with the priming events involved, e.g., mood states and substance use.


To date, most of the information we have reported is descriptive in nature. We plan, however, to move forward on two fronts. The addition of data from a group of successful supervisees will permit us to draw sharper conclusions about the likely causal influence of the precursors to criminal offending that we have identified in the recidivist sample. Similarly, the addition of more recividists, including nonviolent offenders, will allow finer grained analyses according to offence type.

After dealing with the task of specifying the dynamic aspects of the model more completely, we need to address some issues in the substantive interpretation of static risk factors and their possible relation to dynamic antecedents of parole failure. For example, how are static predictors to be interpreted in the present context? We know that they are related to recidivism and treatment failure but not by what mechanism. From a theoretical viewpoint, it is unsatisfying to rest with the identification of a group of miscellaneous variables that are empirically linked to outcome. However, it is not at all clear how relative risk is related to the sequence of events in the recidivism process. We will address these and similar questions by exploring the use of static risk factor scores as moderator variables.

In an important but often neglected study, Chaiken and Chaiken (1984) gathered self reported and officially recorded-offence data from 2,200 inmates in American prisons from three states. These investigators succeeded in identifying 15% of their sample as a type of criminal that they labelled "violent predators" who reported committing robbery, assaults, and illegal drug deals. Violent predators turned out to have committed crimes of all types much more frequently than other offenders, most of whom committed crimes at quite low rates. Because of the greatly disproportionate share of serious crimes committed by violent predators, their identification and study is a high priority. Unfortunately, they cannot accurately be identified by officially recorded offence data alone.

Because the group of offenders we have selected for study are already recidivists, a substantial proportion of them are likely to be violent predators whom we can identify with the interview and LSI data we have collected. These offenders will be young, unmarried, have previous robbery convictions, have very poor employment records, demonstrate criminal versatility, have very high LSI scores, drug and alcohol problems, and a juvenile history of frequent violent crimes (Chaiken & Chaiken, 1984). Many, if not all of these inmates, will meet Hare's criteria for psychopathy.

Based upon our knowledge of violent predatory psychopathic criminals, we predict that the conditions leading to their criminal offences will be markedly different from those leading to the criminal acts of other offenders. Based upon the work of Newman and his colleagues (Newman, 1991; Newman, Patterson, Howland, & Nichols, 1990; Wallace & Newman, 1991), we expect that violent predatory criminals will be less likely to consider the possible negative consequences of criminal behavior because they are "reward driven". In addition, their offences will be more impulsive, occur earlier in their supervision, and be less tied to dysphoric emotional states. The personality of these offenders will be relatively more responsible for the criminal act than environmental precipitants. Situational phenomena and dysphoric mood are predicted to be more important precipitants of criminal behavior among other offenders.

The separation of our recidivists into violent predator and other inmate groups should permit the more accurate delineation. of the relationship of emotional and environmental precipitants to criminal acts in both types of offenders. These results will have direct implications for the development of more effective programs of treatment and supervision. Similarly, we expect that we will be able to separate the "other offenders" into several groups that will differ according to precipitants and motivations, as some of the results presented above imply. These differentiations may lead to some very useful guidelines for supervision.

The principal threat to the inferential value of the information acquired in this study is posed by retrospective bias. Newly arrested releasees could very well distort their recall of events through the filter of their current difficulties. To the extent that this is operative, it changes the status of many of the measures from predictive to postdictive, i.e., measures that might appear to have predictive validity could have no validity when used in prediction. Partly, this problem can be circumvented by the comparisons between different types of recidivists because the issue of retrospective bias will be identical for violent predators and other inmates and among inmates who have committed different types of offences. The ultimate method of addressing this methodological problem, is our plan to conduct a follow-up of the comparison group of releasees who have not yet re-offended, because the model generated using data collected from the recidivist group can be used to predict subsequent offending within this group.


Adams, K. (1983). Former mental patients in a prison and parole system: A study of socially disruptive behavior. Criminal Justice and Behavior, 10, 358-384.

Andrews, D.A. & Friesen, W. (1987). Assessments of anti-criminal plans and the prediction of criminal futures: A research note. Criminal Justice and Behavior, 14, 33-37.

Andrews, D.A., Kiessling, J.J., Mickus, S., & Robinson, D. (1986). The construct validity of interview-based risk assessment in corrections. Canadian Journal of Behavioral Science, 18, 460-471.

Barton, R.R. & Turnbull, B.W. (1979). Evaluation of recidivism data: Use of failure rate regression models. Evaluation Quarterly, 3, 629-641.

Beck, A.T. (1967). Depression: Clinical, experimental and theoretical aspects. New York: Harper & Row.

Beck, A.J. & Shipley, B.E. (1987). Recidivism of young parolees. Bureau of Justice Statistics Special Report. Washington, D.C.: U.S. Department of Justice, Bureau of Justice Statistics.

Bonta, J. & Motiuk, L. (1985). Utilization of an interview-based classification instrument: A study of correctional halfway houses. Criminal Justice and Behavior, 12, 333-352.

Brown, R.C., D'Agostino, C.A., & Craddick, R.A. (1978). Prediction of parole outcome based on discriminant function. Corrective and Social Psychiatry and Journal of Behavior Technology. Methods and Therapy, 24, 93-101.

Carlson, K.A. (1973). Some characteristics of recidivists in an Ontario institution for adult male first incarcerates. Canadian Journal of Criminology and Corrections, 15, 1-15.

Chaiken, M.R. & Chaiken, J.M. (1984). Offender types and public policy. Crime and Delinquency, 30, 195-226.

Frisbie, L.V. (1969). Another look at sex offenders in California. California Mental Health Research Monograph, 12, Sacramento: Department of Mental Hygiene.

Gabor, T. (1986). The prediction of criminal behavior: Statistical approaches. Toronto: University of Toronto Press.

Gendreau, P., Grant, B.A., & Leipciger, M. (1979). Self-esteem, incarceration and recidivism. Criminal Justice and Behavior, 6, 67-73.

Gendreau, P., Madden, P.G., & Leipciger, M. (1979). Norms and recidivism for first incarcerates: Implications for programming. Canadian Journal of Criminology, 21, 1-26.

Gendreau, P., Madden, P.G., & Leipciger, M. (1980). Predicting recidivism with social history information and a comparison of their predictive power with psychometric variables. Canadian Journal of Criminology, 22, 328-337.

Gottfredson, D.M., Wilkins, L.T., & Hoffman, P.B. (1978). Guidelines for parole and sentencing. Toronto: Lexington.

Gottfredson, M.R., Mitchell-Herzfeld, S.D., & Flanagan, T.J. (1982). Another look at the effectiveness of parole supervision. Journal of Research in Crime and Delinquency, 19, 277-298.

Hare, R.D. & McPherson, L.M. (1984). Violent and aggressive behavior by criminal psychopaths. International Journal of Law and Psychiatry, 7, 35-50.

Hart, S.D., Kropp, P.R., & Hare, R.D. (1988). Performance of male psychopaths following conditional release from prison. Journal of Consulting and Clinical Psychology, 56, 227-232.

Heilbrun, A.B., Heilbrun, L.C., & Heilbrun, K.L. (1978). Impulsive and premeditated homicide: An analysis of subsequent parole risk of the murderer. Journal of Criminal Law and Criminology, 69, 108-114.

Holland, T.R., Holt, N., & Brewer, D.L. (1978). Social roles and information utilization in parole decision-making. Journal of Social Psychology, 106, 111-120.

Lutey, C. (1977). Individual intelligence testing: A manual and sourcebook, (2nd edition.). Greeley, Colo.: Carol L. Lutey Publishing.

MacCulloch, M.J., Snowden, P.R., Wood, P.J.W., & Mills, H.E. (1983). Sadistic fantasy, sadistic behavior and offending. British Journal of Psychiatry, 143, 20-29.

Mandelzys, N. (1979). Correlates of offense severity and recidivism probability in a Canadian Sample. Journal of Clinical Psychology, 35, 897-907.

Marlatt, G.A., and Gordon, J.R. (1980). Determinants of relapse: Implications for the maintenance of behavior change. In P.O. Davidson and S.M. Davidson (Eds.) Behavioral Medicine: Changing Health Lifestyles, 5. New York: Bruner/Mazel.

Milan, M.A. (1987a). Basic behavioral procedures in closed institutions. In E.K. Morris & C.J. Braukmann (Eds.), Behavioral approaches to crime and delinquency: A handbook of application, research and concepts. (pp. 161-193). New York: Plenum.

Monahan, J. (1981). Predicting violent behavior: An assessment of clinical techniques. Beverly Hills, Cal.: Sage.

Newman, J.P. (1987). Reaction to punishment in extraverts and psychopaths: Implications for the impulsive behavior of disinhibited individuals. Journal of Research in Personality, 21, 464-480.

Newman, J.P., Patterson, C.M., Howland, E.W., & Nichols, S.L. (1990). Passive avoidance in psychopaths: The effects of reward. Personality and Individual Differences, 11, 1101-1114.

Nietzel, M.T. & Himelein, M.J. (1987). Probation and parole. In E.K. Morris & C.J. Braukmann (Eds.), Behavioral approaches to crime and delinquency: A handbook of application, research and concepts. (pp. 109-133). New York: Plenum.

Nuffield, J. (1982). Parole decision-making in Canada: Research towards decision guidelines. Ottawa: Supply and Services Canada.

Pithers, W.D., Kashima, K.M., Cumming, G.F., Beal, L.S., & Buell, M.M. (1988). Relapse prevention of sexual aggression. In R.A. Prentky & V.L. Quinsey (Eds.), Human sexual aggression: Current perspectives, 528, (pp. 244-260). Annals of the New York Academy of Sciences.

Porporino, F. (1983). Coping behavior in prison inmates: Description and correlates. Unpublished Ph.D. thesis, Queen's University at Kingston.

Porporino, F.J., Zamble, E., & Higginbottom, S.F. (in press). Assessing models for predicting risk of criminal recidivism. Journal of Criminal Justice.

Quinsey, V.L. (1980). The baserate problem and the prediction of dangerousness: A reappraisal. Journal of Psychiatry and Law, 8, 329-340.

Quinsey, V.L. (1984). Institutional release policy and the identification of dangerous men: A review of the literature. Criminologie, 17, 53-78.

Quinsey, V.L. (1988). Assessments of the treatability of forensic patients. Behavioral Sciences and the Law, 6, 443-452.

Quinsey, V.L. & Maguire, A. (1986). Maximum security psychiatric patients: Actuarial and clinical prediction of dangerousness. Journal of Interpersonal Violence, 1, 143-171.

Quinsey, V.L. & Walker, W. (in press). Dealing with dangerousness: community risk management strategies with violent offenders. In R. Peters, R.J. McMahon, & V.L. Quinsey (Eds.), Aggression and violence throughout the lifespan. Sage.

Rhodes, W. (1986). A survival model with dependent competing events and right-hand censoring: Probation and parole as an illustration. Journal of Quantitative Criminology, 2, 113-137.

Rice, M.E., Harris, & Quinsey, V.L. (1990). A followup of rapists assessed in a maximum security psychiatric facility. Journal of Interpersonal Violence, 5, 435-448.

Rice, M.E., Quinsey, V.L., & Harris, G.T. (in press). Sexual recidivism among child molesters released from a maximum security psychiatric institution. Journal of Consulting and Clinical Psychology.

Rice, M.E., Quinsey, V.L., & Houghton, R. (1990). Predicting treatment outcome and recidivism among patients in a maximum security token economy. Behavioral Sciences and the Law. 8, 313-326.

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

Siegel, J.M. (1986). The multidimensional anger inventory. Journal of Personality and Social Psychology, 51, 191-200.

Skinner, H.A. (1982). Drug Abuse Screening Test. Addictive Behaviors, 7, 363-371.

Skinner, H.A. & Horn, J.L. (1984). Alcohol Dependence Scale User's Guide. Toronto: Addiction Research Foundation.

Spielberger, C.D., Gorsuch, R.L., & Lushene, R.E. (1970). STAI manual for the State-Trait Anxiety Inventory ("Self-evaluation questionnaire"). Palo Alto, Calif.: Consulting Psychologists Press.

Wallace, J.F. & Newman, J.P. (1991). Failures of response modulation: Impulsive behavior in anxious and impulsive individuals. Journal of Research in Personality, 25, 23-44.

Waller, I. (1974). Men released from prison. Toronto: University of Toronto Press.

Watson, D., Clark, L.A. & Tellegen, A. (1988). Development and validation of brief measures of positive and negative affect: The PANAS scales. Journal of Personality and Social Psychology, 54, 1063-1070.

Wormith, J.S. (1984). Attitude and behavior change of correctional clientele: A three year follow-up. Criminology, 22, 595-618.

Zamble, E. (1989). Coping Situations Questionnaire: Form M. Unpublished. Queen's University.

Zamble, E. (1988a). A questionnaire for assessing patterns of socialization for criminal offenders. Unpublished. Queen's University.

Zamble, E. (1988b). A questionnaire for assessing time framing and time use for criminal offenders. Unpublished. Queen's University.

Zamble, E., and Porporino, F. (in press). Coping, imprisonment, and rehabilitation: Some data and their implications. Criminal Justice and Behavior.

Zamble, E. & Porporino, F.J. (1988). Coping, behavior, and adaptation in prison inmates. N. Y.: Springer-Verlag.

Zamble, E., Porporino, F.J. & Kalotay, J. (1984). An analysis of coping behavior in prison inmates. Programs Branch User Report No. 1984-77. Ottawa: Ministry of the Solicitor General of Canada.

Zarb, J.M. (1978). Correlates of recidivism and social adjustment among training-school delinquents. Canadian Journal of Behavioral Science, 10, 317-328.