Background: Narcissism is seen as a normal but heterogeneously formed personality variable, ranging from ‘grandiosity-exhibitionism’ to ‘vulnerability-sensitivity’. This article reports the development and factorial validation of a short version of a narcissism inventory. Sampling and Methods: The sample includes data of 4,509 consecutive psychosomatic inpatients. The overall sample was divided in 2 equally sized randomized subsamples. One sample (n = 2,262) was used for exploratory factor analysis (principal component analysis). The other sample (n = 2,265) was used for confirmatory tests of the model fit of the newly built NI-20 version, and to analyze the model fit separately for men and women using structural equation modeling with AMOS software. Results: The short version (NI-20) consists of 20 items, with items representing almost all of the original 18 subscales and 4 second-order dimensions. The NI-20 possesses properties similar to the NI-90, with a considerable gain in test economy. The 4-factor structure of the NI-20 was confirmed, and reaches good fit indices. Conclusions: The NI-20 is an economical instrument with acceptable psychometric characteristics that reflects the heterogeneous aspects of narcissism. A methodological limitation is that the interactions between sociodemographic variables were not included as potential predictors.