we believe that the results of this research should be of interest to a variety of correctional practicioners and administrators, as well as to other researchers or specialized students. Therefore, in writing (and rewriting) this volume we have tried very hard to keep it at a level that should be understandable to an intelligent reader, regardless of background. It has proven to be a difficult task. If we omit or oversimplify too much, we may violate the standards of rigorous evidence and lose the confidence of our academic peers. If we complicate things too much with understatements and caveats, we will convince other readers that they are caught in an unrewarding reprocessed graduate thesis (and our theses were done long ago).

We have not succeeded everywhere as well as we would have liked, but we have tried. We vowed to avoid footnotes in the text and managed to break the habit after only one relapse. The single exception, on the first page, kept us comfortable in getting past the initial stages and should do the same for other academics. We have also minimized the listing of inferential statistics, to the point where the average undergraduate ought to be able to work his or her way through them. Some multivariate analyses are included, but they are offset into discrete sections which the less statistically sophisticated reader can skip with a clear conscience; these sections do strengthen our case overall, but they give very much the same message as those with simpler statistics.

Setting

The core set of subjects in this study were male prisoners in the Ontario region of the Correctional Service of Canada. From this population, we selected only recent recidivists, defined as those who had previously been imprisoned in Canada and had been returned to prison for a new offence committed within a year of their previous release.

Given the constitutional arrangement that mandates responsibility for prisoners in Canada, assignment to the federal system normally means that offenders have sentences of at least two years, with those assigned shorter sentences going to provincial institutions. However, if an offender commits a new offence while under supervision after release from a federal institution, he is returned to federal custody to serve both the remainder of the original term and the new term.

The Ontario region is the largest in the Canadian federal system, drawing from a population of almost 11,000,000. At the time of the study, it included just over 3,000 federal inmates. All prisoners entering the region were sent initially to a Reception Unit in Millhaven Penitentiary, where they were held until sent to their assigned receiving institution. If they had been in the community for only a short time and under supervision when they reoffended, they were sometimes returned within about two weeks to the institution from which they had been previously released, but usually they were held in the Reception Unit for one to two months.

in each of the preceding chapters we have considered implications of our data in context, and it would be redundant to repeat or even review that discussion here. However, there are some general issues raised by the results as a whole. This chapter is largely comprised of examination of some of the wider implications of the research reported here.

In particular, we will survey the consequences of our data in three general substantive areas. First, the results affect our theoretical understanding of the causes and maintenance of criminal behavior. Second, they also have significant implications for policies and practices in the supervision of released offenders. Finally, they hold promise for new directions in the development of instruments and procedures for predicting recidivism.

However, before discussion of those general issues, a brief reconsideration of methodological issues seems appropriate because any conclusions depend on how well one is convinced of the validity of the data. This study has demonstrated what appear to be strong links among poor coping skills, dysphoric emotional states, certain perceptions and cognitions, and criminal recidivism. Before we can consider causation (or its practical consequences) we must reexamine the nature of our information on the period before reoffending.

Any reservations we have about according causal status to the variables measuring preoffence behavior involve the methodological limitations of the study itself.

[this is the version used for recidivists; some questions differ for the control sample. A separate consent sheet contains the subject's name (with signature), FPS (RCMP identification number), and a subject number unique to the study; this is kept separately and securely, and subjects are identified in data files only by their study number.]

(Notes: Bridging dialogue is in italics, but questions are in normal text. Instructions are in parentheses. [Comments added for publication are in brackets.] Additional questions to clarify answers are always permissible. In general, answers should be recorded verbatim as much as possible, even if categories are supplied in the text, to allow later (re) categorization. However, the interviewer should keep possible categories in mind in seeking clarification.)

Introduction

As we explained before, what we are trying to do is to find out what is happening in men's lives while they are out on the street after being in prison. We are trying to find out what sorts of things happen before a new offence, so that in the future we might be able to predict recidivism or maybe even prevent it.

Given what we're interested in, you should be able to see why we're asking most of the questions that follow. If not, you can ask for an explanation, andI'll try to explain, although it may not be until we've finished the interview because it's very important that we finish in the time we have available.

although the information on behavior of reoffenders in the community allows us to construct a picture of recidivists' lifestyles, moods, and outlooks that has some use in understanding criminal behavior generally, the data presented so far do not allow us to address questions of proximal causation very effectively. On the basis of the data, we cannot say which events actually influence the commission of new offences, and which are epiphenomenal. For example, high rates of unemployment or of depression may be determinative of recidivism, or they may be characteristic of marginal populations, or of people who have recently undergone major transitions in their lives. Moreover, some of the specifics of the results may have been produced by features of the method of inquiry. In short, conclusions are impossible without information on comparative or marginal rates.

Our group of men who had been released and who had not reoffended provides comparative data. Any area in which this group does not differ from the set of recidivists is unlikely to be a precursor of renewed offending. The use of this comparison group does not control for all possible confounding variables, but it does match recent past experience of imprisonment and release, and it will allow us to narrow the broad picture shown in the preceding chapter. Although the comparison group is relatively small in comparison to the recidivist group, it is large enough to allow statistical comparisons, certainly where differences are sizeable or consistent.

Static and Dynamic Variables

The prediction of both general and violent criminal recidivism of persons released from correctional institutions has received extensive study (for reviews see Andrews and Bonta, 1994; Blackburn, 1993). This literature indicates that a variety of measures are positively and reliably related to the probability of criminal recidivism. Enough work has been completed to establish a 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.

Among the best commonly available predictors are youthfulness and number of previous arrests. 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 less strongly related to recidivism rates. Although there are conflicting findings on the use of institutional behavior in predicting postrelease recidivism, escape and escape attempts have always been found to be related to higher recidivism rates. Measures of antisociality, such as psychopathy (Hare, 1991), yield higher correlations with recidivism than single predictors commonly available in institutional files, although they are more expensive to collect.

It is of interest that the types of predictors found useful in predicting recidivism among convicted offenders are very similar to those that have been found to predict the initiation of criminal behavior in longitudinal studies of relatively unselected samples of children and adolescents.

while the design of the study and the selection of subjects were intended to elucidate the precursors of offending according to our trinary division of offence types, many other comparisons of interest may be done. Some of these deal with possible differences within groups, such as comparisons of subtypes of the current offence within groups. These within-group analyses can provide some additional detail on the principal direction of the study and will be presented first in this chapter. Because of the lesser numbers, our sample may not be so statistically sensitive in showing these effects, but the sample sizes are still large enough to test adequately some ideas of theoretical interest.

A great many other analyses are possible with the present dataset, including those that assess the effects of factors other than current psychological functioning. For example, we could evaluate the influence of any included historical measure on offence precursors and process. We have performed a limited set of the possible analyses, chosen because they may help to connect this study to other parts of the literature. The results form the basis of the second half of this chapter.

Among Thieves: Violent versus Nonviolent

In the preceding chapter, we considered evidence for a variety of differences associated with the type of new offence. Given that the comparisons generally neglect the specifics of previous offences, the emergence of a consistent pattern of differences in offence precursors may seem in some ways surprising.

the preceding chapters have elucidated some aspects of the transition to recidivism, but the picture is still fragmentary. In several areas of inquiry we received information about a variety of specific events in subjects' lives and thoughts, such as the changes in specific emotional states occurring just before the new offence or the motives for a return to criminal behavior. It is possible that this shows only the variability inherent in the offence process but more likely that at least some of the variance in the overall sample comes from grouping together many sorts of offenders, with diverse current experiences and varying behavior patterns. We will argue that different sets of specific psychological events are associated with different types of current offences.

It has often been noted (for example, Gottfredson and Hirschi, 1990) that a large proportion of repeat offenders commit a variety of types of crimes. Some of the men in our sample had at some time committed all of the types of crimes according to which subjects were grouped in this study. However, the observation that these men are not specialists says nothing about whether there are unique dynamic antecedents of particular types of crimes. Rather, the latter question is best addressed by comparing the antecedents across their most recent offences, as is done here.

In terms of our understanding of the recidivism process, probably the most important question here is that of the specificity of the dynamic antecedents we have seen in the preceding chapters.