OBJECTIVES To compare, in vitro, the fluoride release from a conventional glass ionomer cement (Ketac-Cem), a resin-modified glass ionomer cement (3M-Multicure) and a polyacid modified composite (Ultra Band-Lok) using a banded tooth model and a disc model with the same mean cement weight. METHODS Forty pairs of caries-free third molars were collected and divided into two groups, each of 20 teeth. One tooth from each pair was banded with Ketac-Cem and the other with Ultra Band-Lok or 3M-Multicure; the average band size for each cement group was the same. Two coats of nail varnish were painted on each tooth to within 1mm of the band margin. Five discs (4.5mm diameter and 2mm depth) were prepared for each cement, these dimensions having been calculated so that the mean cement weight of the banded tooth model matched that of the disc model for each cement. The fluoride released into 2ml of deionised water, from each banded tooth or disc, was measured at regular intervals over 30 days using an Orion ion-selective electrode connected to an ion analyser. RESULTS At 30 days, for both banded tooth and disc models, the mean cumulative fluoride release was greatest from 3M-Multicure followed by Ketac-Cem, which in turn released more fluoride than Ultra Band-Lok. These differences were all significant (p<0.05). Despite having the same mean cement weight, the banded tooth model for Ketac-Cem and 3M-Multicure released approximately 3-4 times more cumulative fluoride than the disc model after 30 days (p<001). For Ultra Band-Lok, both models released comparable levels of fluoride (p>0.05). CONCLUSIONS Cement type, specimen geometry and surface area appear to influence significantly fluoride release characteristics.