Empathy and social desirability: a comparison of delinquent and non-delinquent participants using direct and indirect measures

Abstract

Empathy has repeatedly been shown to be relevant for pro-social behaviour, while a meta-analysis reported lower levels of empathy among delinquent compared to control subjects. However, most findings are based on self-reports of the ability or the willingness of persons to empathize with others. The present study measures empathy and attitudes towards empathy in a sample of incarcerated delinquent subjects and a non-delinquent control group using direct measures and a Single-Target Implicit Association Test (ST-IAT).

Higher social desirability was associated with higher levels of self-reported empathy in the delinquent sample whereas the indirect measure revealed a less strong positive connotation of empathy for the delinquent group compared to the control group. In a discriminant function analysis, delinquency was predicted by lower ST-IAT scores, higher cognitive concern, and higher social desirability. Different correlational patterns of direct and indirect measures are observed for delinquent and non-delinquent subjects.

Whereas differences in the indirect measure correspond to our expectations, results for the self-report measures contradict meta-analytical findings of lower levels of empathy amongst delinquent compared to control subjects. Findings are discussed with regard to effects of incarceration and conceptualization of delinquency.

Keywords:

• empathy
• delinquency
• indirect measures
• Single-Target IAT
• social desirability

Notes

1. We checked for differences between the non-student subjects and the student subjects within the control group regarding several variables (including the criteria) but did not find any significant discrepancy.

2. We name the condition ‘congruent’, in which empathy is associated with positive valence, because it well reflects the general desirability of empathy (McGrath, Cann, & Konopasky, 1998).

3. Two items (numbers 12 and 16) were omitted from the new item-factor assignment due to low item–scale correlations (Leibetseder et al., 2007).

4. Between-subject variation of block order in the ST-IAT had no significant influence on the outcome of the ST-IAT. There was neither an effect of sequence, F(1,110) = 0.66, p = 0.42, nor a significant interaction (F(1,110) = 0.12, p = 0.73).

5. Dropping the remaining 18 control subjects that reported having experienced an encounter with the police resulted in highly comparable results. However, the former significant difference in social desirability slightly decreased (Hedge's g=0.37) and failed to reach significance due to reduced sample size.

6. We additionally conducted a discriminant function analysis without the suspect 18 control subjects mentioned above. Results were highly comparable to findings for the total control sample: the discriminant function was again significant [Wilks’ λ of 0.74 with χ ² (7, n=92) = 25.59, p=0.001] and now revealed a slightly reduced canonical correlation with delinquency of 0.51. The mean percentage of correctly classified cases was highly comparable to the analysis including the full sample (77%). However, now there was a considerable misbalance in correctness of classification between both groups, with a higher rate of control subjects falsely classified as belonging to the delinquent group (40.6%) compared to only 59.4% correctly classified control cases and a slightly reduced rate of falsely classified delinquent cases (13.3%) compared to 86.7% correctly classified delinquent cases. The standardized discriminant coefficients were as follows: Cognitive Sensitivity (0.18), Emotional Sensitivity (−0.56), Cognitive Concern (1.03), Emotional Concern (−0.75), SES-10 (0.16), ST-IAT (−0.49), and Age (0.50). Again, Cognitive Concern (0.43), ST-IAT (−0.42), Age (0.36), and SES-10 (0.31) contributed most to discriminating between groups. Group centroids were −0.80 for the control group and 0.42 for the delinquent group.