Four arch length prediction equations (Nance, Johnston-Tanaka, Moyers, and Hixon-Oldfather) were compared by examining pretreatment casts, pretreatment intraoral radiographs, and posttreatment casts of forty-one patients of mixed-dentition age. 2. A comparison of correlation coefficients and slopes of the predicted arch length versus the actual arch lengths revealed that the Hixon-Oldfather method conformed closest to the ideal. 3. No combination of the four methods produced a more accurate equation than the single most accurate method. 4. Neither the sex of the patient nor the type of occlusion affected the prediction accuracy of any of the four equations. 5. All methods tend to overpredict the arch length size by 1 to 3 mm., with the exception of the Hixon-Oldfather equation, which underpredicted by approximately 0.5 mm. 6. An analysis of the intrainvestigator error showed a very low standard error of estimate for individual tooth measurements and for the prediction values. 7. A variance analysis showed that most of the variation was due to arch length (85%), a slight amount was due to the prediction method (8%), and 6% of the variation was due to the rater. 8. A low correlation was found between space available versus actual discrepancy and space available versus actual arch length. 9. High correlation coefficients were found for the predicted arch lengths when compared with the actual arch lengths. As expected, the correlation coefficients for the predicted widths of only the canines and premolars compared with the actual widths were not quite as high.
