The Frykman, Melone, Mayo, and AO classification systems for distal radius fractures were evaluated for interobserver reliability and intraobserver reproducibility in a clinical setting using initial plain radiographs. Two attending orthopedic hand surgeons and two attending radiologists classified 55 sets of distal radius fractures. kappa-statistics were used to establish a relative level of agreement between observers for the two readings and between separate readings by the same observer. Interobserver agreement was rated as moderate for the Mayo classification and fair for the Frykman, Melone, and AO classifications. Intraobserver agreement was substantial for only one of four observers for each of the Frykman, Melone, and Mayo, while the remaining three observers achieved only fair to moderate reproducibility. Intraobserver agreement for the AO classification was fair for all four of the observers. Neither interobserver or intraobserver agreement was affected by combining similar subclasses in the Melone classification or by reducing the number of categories in the AO system from 27 to 9. However, further reducing the AO system to its three main types brought agreement to the "substantial" level. No difference was found in interobserver agreement between the first and second readings or in interobserver or intraobserver agreement between orthopedic hand surgeons and radiologists. Understanding the limitations of fracture classifications based solely on plain radiographs can help avoid undue reliance on them. Given the low degree of interobserver and intraobserver agreement for each of the distal radius fracture classifications in this study, their use as the sole means for determining the direction of treatment or for the direct comparison of results among different studies is not warranted.