Do relationships matter? Examining the quality of probation officers’ interactions with parolees in preventing recidivism

Devon L. L. Polaschek
Victoria University of Wellington

Author biography

Devon Polaschek, PhD DipClinPsych 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.

Criminal justice supervision of offenders in the community – whether post-custody or as a sentence in its own right – has changed in nature over generations and jursidictions according to the political climate of the day. It can be anything from a primarily punishment-based experience, to intensive surveillance designed to detect any act of criminality or non-compliance, to social work-based case management. Quite recently, the role has been expanded into core correctional practices based on psychological research on how to change human behaviour. Depending on the main goals of supervision, the importance of the relationship between the staff member and the offender has also varied. For example if the main function of a probation officer is to detect non-compliant behaviour, probably neither party has an expectation of high relationship quality. But what if the goal is to shape the offender’s behaviour through established principles of social influence?

For psychologists, especially those trained to practice with people with psychological difficulties or distress, the relationship between the professional and the person engaged with the service has long been understood to be an important, even essential ingredient for positive outcomes. In fact, research on psychotherapy shows that what are called “non-specific factors” – elements shared across all types of therapy – make the biggest contribution to successful therapy outcomes: bigger than the therapy “brand”* . Relationship quality is the most easily addressed component of these common factors.

But this relationship is not just like any relationship. The most popular theoretical model calls it the working alliance and outlines three elements: an agreement on the goals that the therapist and client will work on, collaboration on the tasks that will be used to achieve the goals and an overall bond – positive feelings about, or attachment to, each other – that facilitates the collaboration between therapist and client (Bordin, 1979).

Little research has been done on the working alliance in programmes with offenders. We have proposed an adaptation of the model for offender rehabilitation (Ross, Polaschek, & Ward, 2008), and demonstrated in Te Whare Manaakitanga Special Treatment Unit (Rimutaka Prison) that it was not so much how good the relationship was early in treatment, but how much it strengthened over the course of the programme that predicted how much offenders changed on their dynamic risk factors (Polaschek & Ross, 2010).

Since the landmark Manitoba “black box” study (Bonta, Rugge, Scott, Bourgon, & Yessine, 2008), there has been a surge of interest in training probation officers to use their time with offenders to focus on criminogenic needs, and even to conduct micro-interventions. Improving this aspect of practice leads to reductions in recidivism (Davies, 2016). However, the working alliance concept is not exclusively relevant to the treatment or programme context. Many roles in Corrections require staff and offenders to work together to achieve the best outcome for the offender (e.g., employment, education, pre-sentence assessments, release planning), making the concept of the working alliance broadly relevant. It stands to reason that while we might hope that offenders would easily form a good alliance with us, many won’t naturally do so. So it is up to staff to take the initiative in developing a good working relationship. Until recently, research interest has not been directed toward the quality of interactions and bonds between correctional staff and offenders, and whether they do matter for recidivism. This lack of interest is a bit surprising really, when we consider the amount of time probation officers and prison officers, for example, spend in contact with offenders. Perhaps there has been some complacency that good relationship building is second nature for staff who often are recruited from social work and other human services backgrounds. And that may be true, but at least three other possibilities exist.

The first is that not all staff may actually have high levels of skill in forming a constructive working relationship with offenders, and others may have views about how to relate effectively that are misguided. For example, some staff may believe a tough, authoritarian approach is best. Research with probation officers (POs) has shown that the “toughness” component of relationship quality predicted probation violations, sentence revocation and rearrest (Kennealy, Skeem, Manchak, & Louden, 2012; Skeem, Eno Louden, Polaschek, & Camp, 2007).

Second, probation officers have complex roles with the offenders they oversee. In New Zealand, at a minimum, they have a monitoring and enforcement role alongside a supportive, helping and brokerage role. These complexities are a challenge to developing a good working relationship as both a “counsellor” and “cop” (Kennealy et al., 2012). Third, even if relationship building skills are high; as Bonta and colleagues speculate, the relationship itself may be a necessary but not sufficient condition to reduce recidivism. In other words, high relationship quality may only be important through its ability to influence the offender with regard to changes in criminogenic needs (Andrews & Bonta, 2010).

This literature on positive working alliances assumes that, rather than the staff member’s relationship approach to working with the offender being reactive to the offender, the staff member works hard to achieve a positive approach to all offenders, no matter how challenging. For example, rather than showing liking only toward likeable clients, we cultivate and show positive regard – perhaps by searching for something to like – even for the most difficult clients; and that we go out of our way to be fair with offenders who have not necessarily been respectful to us as staff. In practice though, perhaps this expectation is aspirational rather than completely realistic. Again, research suggests that PO relationship behaviour is poorer with offenders who have more challenging pre-existing characteristics e.g., stronger, more unstable negative emotions, higher criminal risk (Kennealy et al., 2012).

This article explores the quality of relationships between POs and high-risk male parolees, using data from the Parole Project** . Our main research aims were to examine (a) levels of PO relationship quality as judged by both the PO and the parolee; (b) whether PO relationship behaviour predicted recidivism; and (c) the extent to which relationship quality was simply a reaction to the offender, or somewhat independent of pre-existing offender characteristics that may make it harder both to form an alliance, and more likely the offender will re-offend. Evidence of independence would suggest that POs were being proactive in relationship building, rather than just working well with those who worked well with them.



The Parole Project sample consists of about 300 men with a RoC*RoI of at least 0.65 who were recruited between 2010 and 2014, just prior to their release from prison sentences of at least two years. Half of these men had completed one of the High Risk Special Treatment Unit Rehabilitation Programmes (STURP) (the Treatment sample). Of the others (the Comparison sample) about 70% had completed some form of treatment or programme (e.g., Medium Intensity Rehabilitation Programme (MIRP), Drug Treatment Unit (DTU), individual psychological treatment).


Following release from prison, all offenders commenced a period of parole***. About two months after release we contacted and interviewed sample members and their probation officers. Part of that interview included a rating of relationship quality. The rating was a shortened version of the Dual Role Relationship Inventory (DRI-R; Skeem et al., 2007), developed for investigating relationship quality in staff supervising mentally disordered US probationers (see Box 1).


Men in the parole project were recruited and interviewed in prison just prior to release; those data are not presented here. They were then interviewed again about two months after release, as were their POs. Relationship quality was rated in these phone interviews.

We measured a number of other variables just prior to release that were useful for examining, or ruling out, the influence of pre-existing offender characteristics (e.g., being higher risk, and less engaged in change or less committed to desistance prior to release) that could both make relating more difficult, leading to lower ratings both by parolees and probation officers, and could also predict recidivism

Results and discussion

Overall, relationship quality ratings were moderately high and similar, regardless of perspectivea. On average, PO ratings of relationship quality two to three months after the man’s release were the equivalent of 4.6 (out of 7; SD=.73) and 4.9 (out of 7; SD=1.4) for parolees’ ratings of their PO. They were significantly higher for the treatment sample than comparison men, and for men who were paroled before sentence end (cf. those who ‘maxed out’), but were unrelated to RoC*RoI. Quality of release planning and readiness for release, dynamic risk for violence and current engagement in change measured just prior to releaseb were all related to PO ratings but not to parolee ratings. This pattern suggests that parolees at higher dynamic risk and less prepared for release are not, for example, more likely to view their PO’s relationship behaviour in a more negative light than others do.

Dynamic Risk Assessment for Offender Re-entry (DRAOR) ratings (stable, protective, internal and external scales) averaged over the first 100 days after release were related both to PO and parolee ratings of relationship quality, raising the possibility that men whose PO rated them higher risk were treated a little less well by their staff member. With regard to offending, those who had a breach conviction, a conviction for a new offence, or were reimprisoned within two months of release had poorer ratings from both perspectives.

To examine the relationship between relationship quality and recidivism, I used survival analyses for the full time the offenders had been in the community since released. Ratings from both rating perspectives were significant predictors of reconviction and reimprisonment.

However, as I just noted, the quality of a POs’ relationship behaviour – especially when rated by POs themselves – is poorer for men with higher dynamic risk, who are less engaged in change, are less ready for release when paroled, have not been treated in a STURP, and have higher concurrent ratings of dynamic risk and protective factors (DRAOR). All of these factors also predict recidivism. Therefore the final analyses were to determine whether the way POs treated parolees in the first couple of months was simply a consequence of these differences; or more positively, related to outcomes even when these other factors are taken into account.

A series of Cox regressions showed that POs’ own ratings of their relationship behaviour predicted reconviction excluding breaches of parole, when risk characteristics were first taken into account. Parolee ratings were close to significance. But for reimprisonment, it was the parolee’s perspective that was predictive after controlling for the pre-release and concurrent risk-related differences; the PO rating was non-significant.

Conclusions, limitations and implications for future research and practice

There are several interesting conclusions and practice implications that can be drawn from this research. First, POs’ relationship quality-related behaviour in the first two months of parole was moderately good overall, according to both the staff members themselves and their parolees. This is a commendable result: we interviewed all of the parolees and some were certainly challenging to relate to.

The correlation between the pairs of ratings of the same PO suggests that parolees provide an important external source of information about PO behaviour, and the predictive validity of the parolee ratings  – they predicted some outcomes that PO ratings did not – further supports the value of offenders’ perspectives.

Creating the capacity to influence a parolee through positive relationship behaviour requires that POs rise above the simple reactions we all have to working with people who can be hostile, unreliable, deceptive, disinterested disrespectful and committedly antisocial. There was some evidence here that POs are not entirely successful in doing so; relationship quality ratings were poorer for higher risk, less motivated men with fewer protective factors and poorer release plans, and for men who re-offended before the rating was made. We in New Zealand are not alone in showing this reactive pattern: similar evidence was found in a previous US sample (Kennealy et al., 2012). These data collectively suggest some reactivity in how POs work with offenders, and perhaps to expect otherwise is not realistic; first and foremost we are ordinary people, after all.

But there is also evidence that staff are having some success with the professional skills of maintaining a positive, constructive approach, despite offender characteristics. Analyses that controlled for these characteristics found evidence that relationship quality was independently related to recidivism. These are encouraging findings.

There are several limitations to this study, most of which are gateways to valuable future research. The most notable is that we didn’t directly measure how POs behaved towards their parolees; instead we measured each party’s perceptions of that behaviour. Measuring perceptions is valid if relationship quality is important because of its effects on perceptions, which is probably the case. But an independent assessment of behaviour would be useful in assessing the actual skills of POs in this area, which would be important for assessing training needs and current competencies.

Similarly we used proxy variables such as dynamic risk levels and current engagement in change to stand in for the actual way that offenders behave toward POs. Direct measurement of offenders’ behaviour with POs would also help identify how best to improve relationship skills.

Second, we didn’t examine whether POs also used their time with parolees in an effective manner (e.g., to address criminogenic needs; Davies, 2016). This is an important area in its own right, but our omission means that a competing explanation is that those who cultivate good relationships also are better at spending their time with parolees in a more change-inducing manner.

Third, we didn’t examine PO characteristics. There is plenty of research on the influence of not just the characteristics of the client, but also of the staff member, on relationship quality. And the extent to which good relationship behaviour can be trained is an unresolved issue: examining PO characteristics and attitudes in relation to relationship behaviour could shed more light on this important question, and provide a way of monitoring training gains.

One other issue that arose in the study may have implications for practice support. Ratings of relationship quality were obtained during interviews with a member of our research team. We noticed that many probation officers were uncomfortable about answering the question “I care about ___ as a person” (see Box 1), often commenting that to answer this question positively would seem to imply some impropriety in the nature of their relationships with parolees. But the research on core correctional practices and positive working alliances in general suggests that showing liking and respect for offenders is desirable. Perhaps this is an area where more professional development is indicated.

But overall these results are encouraging, and suggest the value of continuing to investigate staff core correctional practices and their interactions with the behaviour of high-risk offenders serving sentences with a supervision component in the community. They also suggest the value of extending this research approach to the interactions of other correctional staff in direct contact with offenders.

ReferencesAndrews, D. A., & Bonta, J. (2010). The psychology of criminal conduct (5th ed.). Newark, N.J.: Matthew Bender.

Bonta, J., Rugge, T., Scott, T., Bourgon, G., & Yessine, A. (2008). Exploring the black box of community supervision. Journal of Offender Rehabilitation, 47, 248-270.

Bordin, E. (1979). The generalizability of the psychoanalytic concept of the working alliance. Psychotherapy, 16, 252-260.

Davies, S. (2016). Improving the effectiveness of probation officer supervision through structured training programs. In D. L. L. Polaschek, A. Day, & C. R. Hollin (Eds.), The Handbook of Psychology in Corrections (Vol. 2. Chapter under Review). Chichester, UK: Wiley.

Kennealy, P. J., Skeem, J. L., Manchak, S. M., & Louden, J. E. (2012). Firm, fair, caring officer-offender relationships protect against supervision failure. Law and Human Behavior, 36, 496-505.

Polaschek, D. L. L., & Ross, E. C. (2010). Do early therapeutic alliance, motivation, and change readiness predict therapy outcomes for high risk violent prisoners? Criminal Behaviour and Mental Health, 20, 100-111.

Polaschek, D. L. L., & Yesberg, J. A. (2015). Desistance in high-risk prisoners: Pre-release self-reported desistance commitment and perceptions of change predict 12-month survival. Practice: The New Zealand Corrections Journal, 3(1), 24-29.

Ross, E. C., Polaschek, D. L. L., & Ward, T. (2008). The therapeutic alliance: A theoretical revision for offender rehabilitation. Aggression and Violent Behavior, 13, 462-480.

Skeem, J. L., Eno Louden, J., Polaschek, D. L. L., & Camp, J. (2007). Assessing relationship quality in mandated community treatment: Blending care with control. Psychological Assessment, 19, 397-410.

Wilson, N. J. (2011). Release proposal feasibiity assessment – revised (RPFA-R) manual and rating booklet Version 4. Department of Corrections. Wellington, New Zealand.

Wong, S. C. P., & Gordon, A. (1998-2003). Violence Risk Scale. Unpublished manuscript, Saskatoon, Saskatchewan, Canada.

This research was made possible by the considerable efforts of the other members of the research team: especially Drs Rebecca Bell, Julia Yesberg, Sophie Dixon, and Allanah Casey, the many Departmental staff and the offender participants who helped with the data collection, and funding support from the Department and Victoria University of Wellington.

*There are a few exceptions, and one is therapy work with offenders, where approaches other than those underpinned by cognitive or behavioural psychological science have not been demonstrated to reduce re-offending.

**For more details see Polaschek and Yesberg, 2015.

***We use the term “parole” throughout to refer both to the six-month period required for those released at their statutory release date on conditions, and the longer periods that apply to prisoners released at the discretion of the Parole Board, before their statutory release date.

Box 1.
Relationship quality scale: Probation officer version

Response options

Never RarelyOccasionally Sometimes Often Very often Always 
  1. I treat _____ fairly.
  2. I care about ____ as a person.
  3. I take the time required to really understand ___.
  4. I take all of _____’s needs into account
  5. ___ seems to feel safe enough to be open and honest with me.
  6. ___ seems to feel I am someone he can trust.
  7. ___ seems worried that I am looking to punish him.
  8. I expect ___ to do things independently, and don’t help him out.

Parolees completed a parallel version, rating the probation officer on the same 7-point rating scale. For example Item 1 was “How often do you think that [your probation officer] treats you fairly?”

Items were taken from the DRI-R (Skeem et al., 2007) with permission.

Box 2.
Technical notes on data analysis

a. Relationship quality data were available for 254 probation officers and 205 parolees, with ratings from both in 176 cases. Ratings were made in the two month follow-up telephone interviews. To maximise statistical power, n=176 when the two ratings are compared with each other. But most analyses use one or other rating. In these cases, n=254 for analyses with PO ratings, and 205 for offender-based ratings. Internal reliability analyses showed that (a) for POs, only the first 7 items in Box 1 were internally consistent (=.84) and (b) for parolees, only the first 6 items were internally consistent (=.94). To compare PO and parolee ratings, a mean item rating was calculated, but it was based in each case only on the internally consistent items.

b. These measures were the dynamic and stage of change items from the Violence Risk Scale (Wong & Gordon, 1998-2003), the Release Proposal Feasibility Assessment-Revised (Wilson, 2011), and the Release Plan Quality scale (developed for the Parole Project).

c. These men were removed from the sample for the 12 month recidivism analyses because they were reconvicted for new offences committed before the relationship rating was made.

d. For survival analyses the mean follow-up time was 808 days (SD=318).

e. To maximise sample size when PO and parolee ratings were analysed separately, all available ratings from that perspective was used (see Technical Note a above).

f. Interaction terms were also entered into each of the Cox regressions to determine whether dynamic risk for violence was moderating ratings of relationship quality. No interactions were found.