Examines the reliability (internal consistency) and the factorial validity, as well as the construct validity and the criterion-oriented validity (criterion: Delinquent Behavior) of the Arnett Inventory of Sensation Seeking (AISS; Arnett, 1994). The study analyzed a sample of German adolescents aged 14 to 16 years (N = 1236) who, in addition to the AISS, also completed the NEO-Five Factor Inventory (NEO-FFI) and the Delinquent Behavior subscale of the Youth Self-Report (YSR). The results confirm the hypothesized two-factor structure of the AISS (novelty vs. intensity of stimulation), however, as indicated by modification indices and standardized expected parameter changes, some items had to be excluded from the model to provide an adequate fit for the data. Nevertheless, the internal consistencies of the shortened two-factor scale are considerably low, ranging from .46 to .58. As an index of construct validity, males scored higher than females on both the Intensity subscale and the Total scale. Correlation with the scales of the NEO-FFI revealed that the convergent and discriminant validities of the AISS scales were quite satisfactory. Indications of criterion-oriented validity were obtained by prediction of delinquent behavior (setting fire, stealing, and playing truant) using the Intensity subscale. On the whole, the results yielded sufficient support for the validity of the AISS, while pointing to certain desirable modifications.