L’évaluation de la psychopathie en milieu carcéral est importante et le recours à plusieurs outils diagnostiques est conseillé. Notre étude évalue la validité prédictive de huit variables du test de Szondi pour les trois scores de la Psychopathy CheckList-Revised sur un échantillon de 30 détenus belges. Les résultats indiquent qu’une seule variable est significativement corrélée (r=0,44) avec les scores de la PCL-R mais pas les sept autres. Ce résultat intéressant invite à la réplication. En attendant, nous proposons d’être très prudent dans l’utilisation du test de Szondi pour évaluer les caractéristiques de la psychopathie.

The assessment of psychopathy in prison is important and the use of several diagnostic tools is recommended. Besides the use of the Psychopathy CheckList-Revised (PCL-R), projective tests may be part of the personality assessment. Among these projective tests, the Szondi test lacks psychometric data concerning its predictive validity. We have found no quantitative study that gives this kind of information.

We chose eight Szondi test variables supposed to assess psychopathic features according to the Szondi theory: (a) social index, (b) number of accentuations, (c) acting index, (d) number of [e –] reactions, (e) pleasure principle, (f) number of empty reactions in the central vectors, (g) number of negative reactions in the central vectors and (h) number of empty reactions in the test foreground. Our study evaluates the predictive validity of these eight variables of the Szondi test for the three scores of the PCL-R (total score of the PCL-R, score of the emotional factor 1 and score of the antisocial factor 2). Our sample was made of 30 Belgian prisoners. We received a psychosocial report containing a thorough clinical assessment, which was used to score the PCL-R, and the results of the Szondi test for each prisoner. We scored the PCL-R prior to receiving the Szondi test protocol. We compared the two data sets with each other using a non-parametric correlation index (Spearman's r).

The results showed that only one variable of the eight was significantly correlated (r=0.44) with PCL-R scores. This significant variable was the pleasure principle and refers to simultaneous positive reactions in three vectors: [p+], [d+] and [m+].

Our result casts doubt on the predictive utility of the Szondi test in predicting psychopathy on its behavioral side as well as on its emotional side. With regard to this interesting result, we clearly need to repeat this analysis. In the meanwhile, we propose to be very careful in using the Szondi test to evaluate psychopathic patterns.