The primary aim of the current study was to determine the prevalence and usage of the diagnosis of personality disorder not otherwise specified (PDNOS). As a secondary objective, this study explores the impact of assessment method and definition of PDNOS on observed prevalence. A meta-analysis is reported of 51 studies reporting separate rates for the overall prevalence of specific personality disorders and PDNOS. The best estimate of the absolute prevalence of PDNOS in patient samples is in the range of 8 to 13% and the best estimate of the relative prevalence of PDNOS (i.e., prevalence of PDNOS divided over the overall Axis II prevalence excluding PDNOS) is in the range of 21-49%. In structured interview studies PDNOS is the third most frequently used personality disorder diagnosis, whereas in nonstructured interview studies, PDNOS is often the single most frequently used diagnosis. PDE/IPDE yielded higher PDNOS prevalences than either the SCID or SIDP. Only slightly more than half of the studies provided an operational definition of PDNOS, and the most frequently occurring definition is "mixed" personality disorder. PDNOS would perhaps be the most frequently used diagnosis if it was used in a manner consistent with its definition in DSM-IV. The clinical relevance of future classification, assessment, and research could substantially benefit from validated, explicit, and uniform guidelines for the assessment of PDNOS.