Desistance in high-risk prisoners: Pre-release self-reported desistance commitment and perceptions of change predict 12-month survival

Devon L. L. Polaschek & Julia A. Yesberg

Victoria University of Wellington

Author biographies:

Devon Polaschek, PhD DipClinPsyc is a Professor of psychology at Victoria University of Wellington, New Zealand. Her research interests include theory, intervention, and intervention evaluation with serious violent and sexual offenders, psychopathy, desistance, reintegration, parole and experimental approaches to offender assessment.

Julia Yesberg is a PhD student in the School of Psychology at Victoria University of Wellington, New Zealand. Her research interests include the rehabilitation and reintegration of high-risk offenders, theories of offender change and change generalisation/maintenance, desistance, and risk assessment.

Much of what we know about successful desistance is the product of hindsight. Although many people decide while in prison that they want to desist from crime, only some are successful during their next release into the community; the process of getting to long-term desistance typically zigzags (Burnett, 2004), and may be affected significantly by external factors (e.g., job loss, financial debt, relationship conflict), no matter how committed the offender was before these set-backs occurred.

Maruna and Farrall (2004) proposed that desistance can be broken into two types: (a) primary desistance is any gap or pause in the flow of criminal behaviour; (b) secondary desistance, they suggested, is likely to be a longer period free of offending. Much of the desistance literature has focused on how to define desistance based on the length of time between offences (Kazemian, 2007). But crucially, Maruna and Farrall (2004) suggested that secondary desistance is also associated with awareness on the offender’s part that he or she is not offending, and "the movement from the behaviour of non-offending to the assumption of a role or identity of a non-offender or ‘changed person’" (p. 4). Thus there are two dimensions to their definition of desistance: the length of time since the last offence, and the presence or absence of self-conscious efforts to refrain from offending which may be associated with or lead to a change in identity.

We often talk of desistance as a process that starts after an offender’s release from prison; there is little mention of whether or how desistance processes may begin in prison. But being sentenced to a long prison sentence is likely to be a significant turning point for some offenders. Going back to prison may lead to decisive momentum, the first step in a secondary desistance process according to the recently proposed Integrated Theory of Desistance from Sexual Offending (ITDSO; Gobbels, Ward, & Willis, 2012). And even though offenders are in custody, because there are many opportunities to commit offences in prison, abstinence from crime can also be demonstrated: the prisoner can be misconduct-free, and staff can describe their attitudes as prosocial and law abiding. Furthermore, probation officers and psychologists will attest that prisoners being assessed for parole often talk about a current commitment to "going straight", and about their perceptions of positive change during their sentence. Consistent with the ITDSO, we speculate that it is possible for offenders to make several initial and meaningful steps in the process of secondary desistance while they are still in prison. The ITDSO postulates that those who attend rehabilitation will enter the second stage of their desistance model—moving from thinking about change (initial desistance) to beginning to change (promoting desistance; Gobbels et al., 2012).

Scepticism about the validity of offender self-report may lead us to wonder about the value of offenders’ talk about desistance and change while they are still in prison, especially if they are high-risk offenders who have not undergone any form of intensive rehabilitation during their sentence. In this brief report, we examine how high-risk offenders view desistance and change, using interview data from the Parole Project collected just prior to release on parole. Consistent with the ITDSO, we examined whether ratings differed for men who were graduates of high-intensity psychological treatment programmes vs. those who were released without such treatment. We also examined whether prisoners’ perceptions of desistance and change were related to reconviction in the first 12 months after release.

Based on research showing that contrary to popular belief, offender self-report can provide valid indicators of later offending behaviour (Walters, 2006), we hypothesised that offenders’ ratings would be predictive of reconviction. We expected the strongest relationship to be between these ratings and the occurrence of any reconviction (including breaches): with weaker relationships to reconviction outcomes that did not assume the offender was completely free of offending behaviour (e.g., reconvictions for violence).



The sample for this study was 141 completers of one of four high-risk special treatment units (HRSTU)1 and 147 similarly high-risk prisoners who did not complete treatment at one of these units on their current sentence2. Each man was recruited for the study and interviewed just prior to release from prison, then followed up in the community. We examined whether they had been reconvicted for a breach of parole, for any other type of offence, for a new violent offence, or reconvicted for an offence leading to a new prison sentence, all within 12 months of the date of release.

See Box 1 for the sample characteristics. Statistically, the two samples were indistinguishable on characteristics that predated the possible effects of HRSTU treatment, except that treatment-completing men were more likely to be lifers, and had been given longer sentences. Post-treatment, they were also much more likely to be released early onto parole (80 percent vs. 32 percent) than those in the comparison sample, but this difference is likely to be influenced by treatment completion.


This research was approved by the Victoria University of Wellington School of Psychology Human Ethics Committee. Participants were recruited between November 2010 and January 2014. Each month, Corrections and New Zealand Parole Board records were used to identify those male prisoners over 20 years old who had release dates in the next 10 weeks, were serving sentences of at least two years and had a RoC*RoI of at least 0.65. A team of four senior PhD students from Victoria University of Wellington was trained for the project. These research assistants travelled each month to prisons where several eligible prisoners were located, and met with each prisoner who was willing to take part. Prisoners were advised that the study was being conducted independently of the Department of Corrections and were invited to consent to take part. If they provided written consent, a research assistant then interviewed them for between 1.5 and 2.5 hours. The interview covered experiences in prison, and thoughts about their impending release.

The interviews were semi-structured and at various points, the interviewee was asked to make ratings on a 6-point scale that always ranged from very negative to very positive. For example, we asked each prisoner: "With regard to future crime, are you planning at this stage to go straight, or do you think you’ll possibly or probably still get involved in some crime?" (desistance commitment; see below). The rating options for this question varied from "definitely still involved in crime" through to "definitely going straight". During the interview, each time a rating was to be made a written copy of the rating scale for that question was put in front of the respondent for reference.

Box 1
Sample characteristics: means for the two samples

HRSTU completers

Comparison sample

Age at parole






Age first conviction



No. previous convictions



No. previous violent convictions



Sentence length given (days)a



Parole length given (days)a



a.  Excludes 11 HRSTU and 2 comparison men who were serving life sentences; difference is statistically significant.

The interview data used in this report came from a series of questions that asked about desistance and change. Men were first asked to report in general how they were feeling about being released. Then they were asked about three key desistance-focused questions: how much they were "planning to desist" (see above; desistance commitment), how much they wanted to desist (desire for desistance), and their current confidence about "going straight" (desistance confidence). Finally, they were shown a pair of 25 cm horizontal lines and asked to indicate on the respective line how much they had changed for the better and how much they had changed for the worse during their current sentence. These results were recorded in centimetres and a net change score was also calculated (based on subtracting the second measurement from the first; range -25.0 to +25.0).

Results and discussion

Turning first to the interview questions, we can see from Box 2 that out of a possible maximum score of six, on average the sample members responded very positively to all of the desistance-related items. Perhaps most notable are the near-maximal ratings to the question "how much are you wanting to go straight?" (desire for desistance). These scores indicate that regardless of whether they undertook intensive psychological treatment (vs. no rehabilitation, or less intensive programmes), and regardless of whether they got out early or at sentence end, most men in the sample had a strong desire to avoid future convictions. Interestingly too, treatment completers and comparisons were equivalent on this rating, though treated men rated their commitment to ("planning to")—and confidence in—desistance more highly, and thought they had changed more for the better and less for the worse than comparison men.

Asterisks indicate means are statistically significantly different: *p<.05; **p<.01. Although secondary desistance involves more than simply avoiding reconviction, we thought it was likely that high-risk prisoners on parole regard being conviction-free at least as a necessary first step. Box 3 shows that reconvictions for any offence (whether including or excluding breaches of parole) were the most strongly and consistently related to both desistance ratings and change ratings, supporting our hypothesis that avoiding any new conviction is the most important outcome for those committed to abstaining from crime (i.e. the most strongly associated with desistance ratings). Interestingly, desistance commitment alone was a significant predictor of all types of reconviction outcomes, though most of the other change and desistance variables predicted at least two outcomes. The exception was the first question: How they were feeling about what life would be like when they got out of prison (rated from "really bad" to "really good"). We put this question into the interview to ascertain that avoiding reconviction was not simply related to any positive rating of the future. We found that it was not: most expected release to be "really good" and this expectation was not related to recidivism.

Box 2
Interview data

HRSTU completers

Comparison sample

Combined samples

General feeling about releasea




Going "straight"a

Commitment to?**




Wanting to?








Change for betterb**




Change for worseb**




Net changec**




a. Range 1-6; b. Range 0-25; c. Range -25-+25.

Box 3.
Correlations between interview ratings and type of reconviction


Reconviction incl. breach

Reconviction excl. breach



General feeling about releasea






Going "straight"a

Commitment to?






Wanting to?












Changed for betterb






Changed for worseb






Net changec






a. Range 1-6; b. Range 0-25; c. Range -25-+25.
Asterisks indicate means are statistically significantly different: *p<.05; **p<.01.

Lastly, because correlational analyses not presented here showed that, as expected, desistance ratings and change ratings were all somewhat related to each other too, we looked more closely at which variables in combination could best predict reconviction for any offence (breach or new offending combined) in the 12 months after release (see Box 4 for the statistical details). In essence, the combination of how strongly they were committed to desistance (the desistance commitment question ratings only) and the net amount of change they thought they had made in prison predicted recidivism the best. However, when we ran additional analyses we found that the contribution of net change (i.e., positive change – negative change) was actually almost entirely driven by how much they thought they had changed negatively during the current prison sentence. And of course comparison men gave significantly higher ratings to this variable, which in part accounts for their higher rates of reconviction after release3.

What should we make of these findings? First, as we expected, they show that offenders believe that they have changed during their prison sentence in ways that are meaningful for desistance, and that many are strongly committed to desistance. Desistance commitment was stronger in the HRSTU sample but is not exclusively a product of intensive psychological treatment. However, treatment may lead a commitment to desistance to be more successful: comparison men reported that they wanted to "go straight" as much as HRSTU completers, but were not as confident about achieving desistance. Comparison men were also less likely to desist after release; as we note here (see Footnote 2) and elsewhere, in the Parole Project research they were more likely to be reconvicted for all types of outcomes than the HRSTU graduates (see Polaschek et al., 2014, 2015). Similarly, comparison men thought they had made smaller changes for the better and were more likely to rate themselves as having deteriorated in prison than treated men.

Although we might expect desistance ratings and change in prison to be related (i.e., those who think they have changed a lot for the better may expect more strongly to desist on release), and they were somewhat, still these variables were independent enough of each other to offer unique contributions to the prediction of reconviction. Intriguingly for self-report sceptics, these high-risk men’s ratings of their progress in prison and of their desistance commitment in particular, represented valid assessments in that they predict future recidivism. Scepticism about offender self-report persists amongst correctional staff despite significant previous research showing that when offenders are asked to make ratings on the "right things" (i.e., on topics that are themselves predictive of reconviction), overall their ratings are as valid in predicting recidivism as other more objective methods (see for example, Walters, 2006). However, although self-reports themselves are predictive, measures of change based on self-reported data are often not predictively valid (Serin, Lloyd, Helmus, Derkzen, & Luong, 2013), so finding here that both positive and negative change appraisals using a simple visual analogue scale were predictive of reconviction is surprising and encouraging. We think the finding about the importance of predictions of negative change is particularly striking, given that overall, 43 percent of the sample gave a rating of "0" (i.e., no "change for the worse"). This result would be worthy of further investigation, since in this study we did not collect much detail on the nature of these negative changes. However, if it holds up, it suggests that assessors may want to ask about perceptions of getting worse in their pre-release assessments, if they do not already do so.

Box 4.

Best combination of predictors of reconviction including breaches, in the first 12 months following release: Logistic regression results.

Logistic regression was chosen because it is the type of regression best suited to dichotomous outcome variables (in this case "reconvicted" or "not"). The dependent variable was whether or not the parolee had any reconviction, including parole breaches, for behaviour that occurred in the first 12 months after release on parole.

The first model—based on two variables: (a) strength of desistance commitment —the rating made on the "planning to go straight" question, and (b) net change in prison—significantly predicted reconviction: X2(2, N=287) = 18.4, p<.01; Nagelkerke pseudo-R2=8.7%. Both variables contributed significantly to the resulting model: For desistance commitment, Wald (1)=4.04, p=.045, Odds Ratio=.76 (95% CI = .58, .99); meaning that for every 1-point increase in this rating, the likelihood of reconviction decreases by 24 percent. For net change, Wald (1)=6.68, p=.01, OR=.95, (95 percent CI = .92, .99); for every 1 point increase in net change, the likelihood of recidivism decreases by 5 percent.

The second model was exactly the same except that instead of net change, we inserted the rating of how much the prisoner thought he had "changed for the worse" (negative change): the negative half of the net change rating. The results were similar: the overall model significantly predicted any reconviction X2(2, N=287) = 16.6, p<.01; Nagelkerke pseudo-R2=7.9%. Again, both variables contributed significantly to the resulting model: For desistance commitment, Wald (1)=5.6, p=.018, Odds Ratio=.73 (95 percent CI = .56, .95); meaning that for every 1-point increase in this rating, the likelihood of reconviction decreases by 27 percent. For negative change, Wald (1)=5.03, p=.025, OR=.1.1 (95% CI = .1.00, .1.13); for every 1-point increase in negative change rated, the likelihood of recidivism increased by 6 percent.

This is a relatively old, high-risk parolee sample in which we think that most of the members are, at the point of release, engaged in what Maruna and Farrall (2004) referred to as proto-secondary desistance: that is, they are trying to become "desisters" but some or perhaps many may still return to offending over the longer term. The results have implications for desistance theory. They suggest that rather than requiring of offenders a period of crime-free activity in the community before we define them as being "in desistance", some already meet the psychological conditions in Maruna and Farrall’s definition of secondary desistance (self-conscious awareness of desisting and a sense of being a "changed person") while still in prison. And in turn, these psychological conditions predict reconviction. Anecdotally, for many of those who were interviewed the turning point was being imprisoned "yet again", a finding that is relevant to the phase of decisive momentum in the Gobbels et al. (2012) theory of desistance in sexual offenders. The results also tentatively supported their theory in that those with the most rehabilitation (the HRSTU completer sample) were more committed to, and more confident about desistance.

The findings reported here may be helpful for those involved in assessing and making recommendations, plans and decisions about prisoner release, but this report is our first examination of these ratings. In future analyses of this data set we will be examining links between these pre-release interview ratings and other post-release variables such as DRAOR scores on parole, and information the parolees and their probation officers gave us at follow-up interviews.


  • Burnett, R. (2004). To reoffend or not to reoffend? The ambivalence of convicted property offenders. In S. Maruna & R. Immarigeon (Eds.), After crime and punishment: Pathways to offender reintegration (pp. 152-180). Cullumpton, Devon: Willan.
  • Gobbels, S., Ward, T., & Willis, G. M. (2012). An integrative theory of desistance from sex offending. Aggression and Violent Behavior, 17, 453-462.
  • Kazemian, L. (2007). Desistance from crime: Theoretical, empirical, methodological, and policy considerations. Journal of Contemporary Criminal Justice, 23, 5-27.
  • Maruna, S., & Farrall, S. (2004). Desistance from crime: A theoretical reformulation. Kolner Zeitschrift fur Soziologie und Sozialpsychologie, 42, 1-24.
  • Polaschek, D. L. L., Yesberg, J. A., Dickson, S. R., Casey, A. and Bell, R. K. (2014). Does intensive psychological treatment lead to reductions in recidivism in high-risk prisoners? A quasi-experimental investigation. Manuscript in preparation.
  • Polaschek, D. L. L., Yesberg, J. A., & Chauhan, P. (2015). How does psychological treatment work? An examination of factors that predict short-term reconvictions in high-risk samples of treated and untreated prisoners. Manuscript under review.
  • Serin, R. C., Lloyd, C. D., Helmus, L. , Derkzen, D. M., & Luong, D. (2013). Does intra-individual change predict offender recidivism? Searching for the Holy Grail in assessing offender change. Aggression & Violent Behavior, 18, 32-53.
  • Walters, G. D. (2006). Risk-appraisal versus self-report in the prediction of criminal justice outcomes Criminal Justice and Behavior, 33, 279-304.

1.  Also known as the ‘STURP’ (Special Treatment Unit Rehabilitation Programmes): Puna Tatari (Spring Hill Corrections Facility), Karaka (Waikeria Prison), Te Whare Manaakitanga (Rimutaka Prison), and Matapuna (Christchurch Men’s Prison).

2.  These men are called ‘comparison’ rather then ‘untreated’ because about 70 percent had attended some form of programme: for example, a Dependency Treatment Unit or the Medium Intensity Programme.

3.  The reconviction rate for new offences for treated men was 51 percent and 65 percent for comparison men. When breaches of parole were included, the rates were 62 percent and 74 percent respectively; rates of breaches were 34 percent and 50 percent.