This study was conducted to evaluate the surface effects of the air-powder polisher on direct restorative materials in vivo. Hybrid composite, microfilled composite, amalgam, and glass ionomer restorations were polished for 10 seconds. Replicas of the pre-treatment and post-treatment surfaces were examined in the SEM. A visual and written rating scale was developed to evaluate surface roughness, i.e., 1 = smoothest and 7 = roughest. Micrographs (x200) of the pre- and post-treatment surfaces were evaluated as (1) randomly ordered samples and (2) paired samples. The pre-treatment ratings from the randomly ordered samples ranged from 1-7 for the microfilled and hybrid composite, and 3-7 for the glass ionomer restorations. The range of post-treatment ratings was slightly higher for the hybrid and amalgam (2-7) and the glass ionomer (4-7). Results were analyzed by the Wilcoxon ranks test, which showed that the differences between the ratings of pre and post-treatment surfaces were not significant (P < or = 0.05). A comparison between the two evaluation methods using the Spearman rank correlation (Rho) test showed a good correlation (P < or = 0.05) between the random versus paired evaluation methods and suggested that the random rating method was reliable. Thus, a 10-second exposure with the air-powder polisher does not significantly roughen the surface of the materials tested. Visual observations showed that surface changes were dependent on initial conditions, with smooth surfaces becoming rougher, and surfaces that were extremely rough generally becoming smoother.(ABSTRACT TRUNCATED AT 250 WORDS)