Emotions can be assessed by means of different diagnostic methods, for example, by self-report instruments, ratings of facial expressions, or by projective techniques. This study presents an alternative approach: a computerized investigation of verbally expressed emotions by means of the Affective Dictionary Ulm (ADU; Dahl, Hölzer, & Berry, 1992), which was applied to responses in the Holtzman Inkblot Technique (HIT; Holtzman, 1988; Holtzman, Thorpe, Swartz, & Herron, 1961). A normal group (n = 30), patients with neurotic disorders (n = 30), borderline patients (n = 30), acute schizophrenics (n = 25), and chronic schizophrenics (n = 25) were compared in regard to verbally expressed emotions. According to the results, patients with neurotic disorders did not differ from the normal group in regard to verbally expressed emotions. Borderline patients expressed fear and emotions in general significantly more frequently than all other diagnostic groups. Furthermore, borderline patients differed in regard to specific emotions from patients with neurotic disorders, acute schizophrenics, and chronic schizophrenics. Acute schizophrenics did not differ from the normal group in regard to the expression of emotions, whereas chronic schizophrenics expressed anger, fear, anxiety, and emotions in general significantly less frequently than normals. By a discriminant analysis using verbally expressed emotions as predictors of the diagnosis, hit rates between 87% and 100% could be achieved. Furthermore, hypotheses about correlations between emotions on the one hand and internalized primitive object relations and bizarre-idiosyncratic thinking were tested empirically. Significant correlations could be demonstrated. These results support the validity of assessing emotions through a lexical content analysis of the HIT by use of the ADU.