OBJECTIVES: Comparative in vitro/in vivo studies show wide differences in functional strain magnitudes and patterns of functional strain distribution in prosthetic devices. The aim of this study was to evaluate the accuracy of three in vitro models for biomechanical studies of prosthetic devices. METHODS: Strain gages were attached to the test RPDs of six subjects. In vitro maxillary models with simple (model A), intermediate (model B) and advanced (model C) features for the anatomical simulation of supporting structures were manufactured for each subject. The dentures were subjected to two separate series of standardized functional and laboratorial loading tests. The principal maximum strain value (epsilon (1)) obtained for each gage and experiment was used to calculate the intra-experiment variance and inter-experiment variance for the in vivo and in vitro experiments. The integrated strain gage difference value between the in vitro and in vivo data (Sigma(diff)) was calculated for every loading test and compared to the in vivo inter-experiment variance with ANOVA and Scheffes test. RESULTS: The bite forces recorded were similar for each subject and loading position and between the two loading sessions. The inter- and intra-experiment variance was found to be higher for in vivo loadings than for in vitro. Results showed that the (Sigma(diff)) values for model B and model C were significantly different from the in vivo strain interexperiment variance (p<0.05). SIGNIFICANCE: This study showed that the anatomical simulation of in vitro models is insufficient to allow for accurate mechanical analyses of maxillary RPDs and that only simple verifications of the strain levels in prosthetic appliances can be attained in vitro.