Background: The Narcissism Inventory (NI) is a frequently used German inventory for measuring narcissism in clinical settings; an additional short version (NI-90) also exists. Psychometric properties of the NI-90 scales were examined in clinical and non-clinical adolescent samples. Methods: Two adolescent samples were assessed with the NI-90: a non-clinical sample (n = 439, mean age ± SD = 15.05 ± 1.77 years) and a clinical sample (n = 235, 18.26 ± 0.77 years). Confirmatory factor analysis and principle component analysis were used to scrutinize the structure of the scales. Multiple regression analysis was used to predict the scores on two scales (helpless self; negative body self). Results: This study revealed heterogeneity in the NI-90 scales, which in turn explains the wide range seen in Cronbach's α (from 0.53 to 0.93). The postulated 4-factor structure could not be replicated in both samples. Multiple regression analysis revealed that personality disorder did not significantly predict negative body self or helpless self scores, whereas eating, mood, as well as somatoform and conversion disorders did. One NI-90 scale (greedy for praise and reassurance) showed sufficient psychometric quality for the measurement of narcissism in both samples. Conclusion: Based on the results, the authors recommend revising the NI-90. Items that may be useful for measuring aspects related to affective and body image complaints are presented. The greedy for praise and reassurance scale may be valuable for measuring features of 'overt' narcissism.