The factor analytic studies that have been done with the Millon Clinical Multiaxial Inventory (MCMI), a 175-item paper-and-pencil test designed to evaluate symptomology in psychiatric patients, have been problematic in that the test's high inter-scale correlation could lead to an artificially robust factor structure. The present investigation avoided the issue of scale overlap by working with the patients' answers to the individual items of the inventory rather than the overlapping scales. The subjects were 2,129 male psychiatric inpatients who completed valid protocols. Principal component analysis led to 17 interpretable factors that accounted for 41% of the variance. Thirteen of these factors contained items resembling scales from the MCMI. These factors represented basic personality scales (e.g., Avoidant), as well as clinical symptom scales (e.g., Dysthymic). Other factors represented more circumscribed characteristics such as suicidality, weepiness, and lability. The MCMI scales most poorly represented in the factor structure were those assessing psychoticism and severely maladaptive personality styles, a surprising finding in a sample of psychiatric inpatients. Also, in contrast to scale-based factor analyses, no general maladjustment factor emerged. (Author)