The authors have declared that no competing interests exist.
Conceived and designed the experiments: SHMvG. Performed the experiments: KH KLB. Analyzed the data: KH SCM SHMvG. Wrote the paper: KH SHMvG.
Children with antisocial behaviour show deficits in the perception of emotional expressions in others that may contribute to the development and persistence of antisocial and aggressive behaviour. Current treatments for antisocial youngsters are limited in effectiveness. It has been argued that more attention should be devoted to interventions that target neuropsychological correlates of antisocial behaviour. This study examined the effect of emotion recognition training on criminal behaviour.
Emotion recognition and crime levels were studied in 50 juvenile offenders. Whilst all young offenders received their statutory interventions as the study was conducted, a subgroup of twenty-four offenders also took part in a facial affect training aimed at improving emotion recognition. Offenders in the training and control groups were matched for age, SES, IQ and lifetime crime level. All offenders were tested twice for emotion recognition performance, and recent crime data were collected after the testing had been completed.
Before the training there were no differences between the groups in emotion recognition, with both groups displaying poor fear, sadness and anger recognition. After the training fear, sadness and anger recognition improved significantly in juvenile offenders in the training group. Although crime rates dropped in all offenders in the 6 months following emotion testing, only the group of offenders who had received the emotion training showed a significant reduction in the severity of the crimes they committed.
The study indicates that emotion recognition can be relatively easily improved in youths who engage in serious antisocial and criminal behavior. The results suggest that improved emotion recognition has the potential to reduce the severity of reoffending.
Antisocial behaviour in childhood and adolescence is associated with a range of negative outcomes in adulthood. It predicts future antisocial behaviour (arrests, severity of crimes, conviction rates and length of prison sentences [
The short- and long-term effectiveness of individual-focused preventions (e.g., child skills training) and family-focused preventions (e.g., Multi-Systemic Therapy) in reducing antisocial behaviours has been demonstrated [
One of the best-replicated findings is that individuals who exhibit inappropriate interpersonal and antisocial behaviour have problems in facial emotion recognition, particularly fear and sadness [
Finally, it has been shown that delinquents are more likely than non-delinquents to misinterpret expression of disgust as anger [
Training programs that teach participants to pay more attention to important facial features using explicit methods [
If young offenders’ facial emotion recognition of key distress emotions like fear and sadness can be improved through training, then according to the IES model such interventions could alter how young offenders respond and interact with potential victims. Similarly, by improving the detection of anger, young offenders may pick up early warning signals of social punishment more effectively and refrain from continuing their negative actions. On the other hand, an improved ability to detect anger may reduce misinterpretation of other emotions, such as disgust, and lead to less threatening responses in otherwise non-threatening situations. Consequently, the training of these emotions could have positive effects on future antisocial behaviour, ultimately leading to a reduction in crime.
The current study assessed facial affect recognition and objectively recorded crime levels in a group of juvenile male offenders prior to and after completion of an emotion recognition training intervention and compared their data to those of juvenile male offenders who did not complete the training intervention. We expected (a) that the training would result in a significant improvement in the recognition of those negative emotions that were trained (fear, sadness, anger), and (b) that offenders in the training condition would show a greater reduction in reoffending levels compared to those receiving the usual treatment up to six months post training.
Male young offenders (YOs) aged 12–18 years (mean = 16.21) took part and were recruited from the Cardiff and Vale of Glamorgan Youth Offending Services (YOS). Young offenders were recruited with the help of YOS caseworkers who recommended suitable participants. All participants completed the emotion recognition test twice (average time between tests = 23 days). YOs were eligible to participate if they had been convicted of an offence and received a court order to attend the YOS. To be included in the analyses participants were required to have completed the pre- and post- facial recognition measures in the required time frame without incarceration; however, later incarceration was not a reason for exclusion. All participants met these eligibility requirements and no participants dropped out from the study. A group of 24 offenders (Training group) completed the emotion training in the interval between the first and second emotion recognition test; another group of 26 offenders (Control group) was tested twice during the same time period, but did not receive the emotion training. Group allocation was strongly influenced by the opportunity and availability of the offenders to attend the YOS offices where testing took place for the required number of sessions. This was discussed with caseworkers before participants were officially approached about the study. Participants were specifically asked to take part in the Training group or the Control group, and were not given the option to choose between conditions. All parts of the study were completed at YOS offices and conducted by trained researchers (KH, KLB). Sample size (n = 50) was based on a previous study comparing YOs and non-offending matched controls [
The study was approved by the School of Psychology Research Ethics Committee at Cardiff University. All participants (n = 50) and their parents/guardians provided written informed consent.
Data on the number and severity of offences that YOs had committed and which had led to criminal prosecutions by a court were compiled from YOS databases after the completion of the emotion recognition/training study. For each offender we collected crime data covering three time periods: (1) lifetime crime data; these covered all crimes ever committed up to 12 months ago; (2) pretest crime data; all crimes committed in the 6 months leading up to the first emotion recognition test, and (3) posttest crime data; all crimes committed in the 6 months following the second emotion recognition test. Juvenile offenders had been involved in the following types of offence: aggravated burglary, aggravated taking, arson, assault, attempted robbery, attempted theft, breach of order, burglary, public order (harassment, affray), public order (common assault), criminal damage, motor/traffic offences, obstructing police, drug offences, robbery, shoplifting, theft, public order (threatening, abusive), trespassing, TWOC (taking [vehicle] without owner’s consent), wounding with intent to do GBH (grievous bodily harm; i.e., assault). The YOS also assigns each offence a severity score ranging from 1 (e.g., minor public order offences) to 8 (e.g., murder). In case of multiple offences, the highest severity score (the most serious crime committed) was recorded. The mean number of offences in our sample before the start of our study (lifetime offence rate) was 6.68 (SD = 8.75; range 1–49), the mean severity score was 3.20 (SD = 1.09; range 1–6), and the mean score of the most severe offence committed was 4.82 (SD = 1.66; range 1–7).
Many of the offenders were not in statutory full time education, but received some form of education provided by the YOS. Because of this it was considered to be more appropriate to use measures of IQ to assess educational ability. The two-subset form of the Wechsler Abbreviated Scale of Intelligence (WASI; [
The Youth Psychopathy Inventory (YPI [
The Facial Emotion Recognition (FER; see [
Participants were known to the YOS, had their own caseworker and continued to receive their statutory interventions and allocated contact time with the YOS for the duration of the study (including the 6 months follow-up). The type of intervention and amount of contact with the caseworker varied between offenders depending on the court order received and their risk of reoffending.
Ten YOs (training = 4; control = 6) belonged to preventative or early intervention programmes. Offenders on these orders are considered at low level of risk of reoffending and are seen less regularly. Thirty-four participants had a Youth Rehabilitation Order (training = 16; control = 18), which is the standard community sentence for the majority of offenders. The specific requirements of the order and the amount of contact with the YOS follow an individual risk- and needs-based approach. Six participants were on probation after a custodial sentence or on bail (training = 4, control = 2). Offenders on these orders are considered at high risk and are seen 3–5 times a week. Although all interventions are tailored to each individual, most orders encourage engagement with statutory education and will work with the young person to put this in place if it is not currently happening (for example, through training, work placements, apprenticeships and college), as well as with health, substance misuse and family support services, and focus on restorative justice and encourage victim work.
We used an adapted version of the Facial Affect Recognition (FAR) intervention [
The FER variables are mean correct recognition scores for happy, sad, fearful, disgust and angry expressions, at both pre- and post-test. A mixed-design ANOVA was used with emotion (5 levels), and time (2 levels) as within-subject factors, and group (Training vs. Control) as between-subject factors. Where the assumption of sphericity was violated the Huynh-Feldt correction was used. Planned follow-up tests explored the effect of training on each emotion separately and used the Bonferonni correction. One-way ANOVAs examined differences in age, IQ, SES and personality measures between the experimental groups. Time to reoffend was analysed using the Cox Proportional Hazards Model [
Chi-square analysis revealed no differences between the Control and Training groups with respect to type of orders and interventions received, χ2(2) = 1.1,
Variable | Training | Control |
---|---|---|
Age (years) | 16.08 (1.2) | 16.35 (1.2) |
IQ | 80.89 (9.64) | 83.77 (7.64) |
SES (mean) | 1.3 (0.62) | 1.6 (0.86) |
Low (= 1) | 79% (19) | 65% (17) |
Middle (= 2) | 13% (3) | 12% (3) |
High (= 3) | 8% (2) | 23% (6) |
YPI | 115.2 (16.5) | 125.8 (22.8) |
CU traits | 34.2 (6.6) | 35.5 (8.4) |
Age at first offence | 14.17 (1.7) | 13.24 (1.7) |
Lifetime offence rate | 6.42 (6.24) | 6.92 (10.68) |
Pretest offence rate (6m pretest) | 3.63 (4.95) | 1.92 (1.67) |
Pretest offence most severe (6m pretest) | 3.75 (2.23) | 2.62 (1.84) |
Pretest offence mean severity (6m pretest) | 2.92 (1.82) | 2.24 (1.53) |
Re-offence rate (6m posttest) | 1.67 (2.58) | 0.92 (1.44) |
Re-offence most severe (6m posttest) | 2.08 (2.28) | 1.85 (2.53) |
Re-offence mean severity (6m posttest) | 1.76 (1.89) | 1.57 (2.11) |
Table entries show mean values (standard deviations in parentheses), or % of group (with numbers in parentheses); IQ = intelligence quotient; SES = socio-economic status; YPI = Youth Psychopathy Inventory; CU = Callous and Unemotional.
Error bars show +1 standard deviation.
There was a significant time by group interaction,
There was a significant time by group interaction,
There was a significant time by group interaction,
There was a significant main effect of group,
There was no main effect of time,
Posttest reoffending data are shown in
The present study sought to establish whether emotion recognition abilities could be improved in young offenders, a group that has been found to have particular problems in recognising negative emotions [
The findings show that juvenile offenders in the Training and Control groups displayed statistically equivalent recognition abilities at pretest and confirm previous findings of poor recognition of negative emotions in offenders. Bowen et al. [
Previous studies have found disgust recognition impairments in antisocial children [
Turning to the crime data, we found that offenders in the Training and Control groups were similar with respect to lifetime crime data (as well as age, IQ, SES and CU traits) before the study, but differed in offence severity in the 6 months before the training, with those who went on to take part in the emotion training having committed more severe offences (
It should be noted that the observed effects, in terms of offences severity, were relatively small and of moderate effect size. This is understandable in a sample as small and complex as the current one. Most of the young offenders were persistent offenders (see
The intervention was not randomised. Because of logistical issues within the YOS it was not possible to randomly allocate offenders to group; consequently, a quasi-experimental design was used based on the availability of young offenders to attend the YOS offices. With the help of caseworkers it was decided in advance whether offenders would be able to attend for the number of sessions required to complete training. Those who were unlikely to be able to attend all sessions formed the control group. It could be argued that this could have biased the sample; however, we have shown that the two groups did not differ on a number of key variables suggesting the groups were equal. It should be noted that there are other possible confounding variables, which we did not control for and that could have impacted upon our data, such as substance use, self-reported aggression, opportunity, and maltreatment. Our findings should therefore be interpreted with this caveat in mind. Despite the drawback of non-randomised designs, evidence suggests that there are no differences in magnitude of effect sizes between randomised control trials and quasi-experimental designs [
We have shown that emotion recognition can be improved in youths who come into contact with the police for a wide range of different types of antisocial behaviour problems by administering a relatively brief, targeted intervention. Importantly, the training subsequently had a positive effect on criminal behaviour by reducing reoffence severity. This suggests that interventions that target neurobiological impairments are not only relatively easily achievable, but also have a beneficial impact on the lives of young people and their communities.
Within the autism literature compensatory changes in neural activity, measured by fMRI, have been observed alongside improved recognition in those with autism trained to attend and interpret emotional faces [
(XLS)
We thank staff and boys at the Cardiff and Vale of Glamorgan Youth Offending Services, and Joanne Morgan, Eva Syngelaki and Scott Rogers for help with collecting the data. The research was supported by two ESRC UK studentships (ES/J500197/1 & ES/GO17840/1).