Ninety-nine extracted human molar teeth were used in this study comparing the shear bond strengths on dentine of one light-cured and two auto-cured polyalkenoate (glass ionomer) cements. Bond strength can be influenced by differences in tooth structure. A balanced-incomplete block design (Hull and Nie, 1981) was used to reduce variation attributable to such differences. Cements were applied to paired dentine surfaces in combinations such that 66 tooth sides were treated with each material. A light-cured dentinal adhesive and composite resin restorative material were then placed and shear bond strength testing was conducted exactly 24 h after the completion of each specimen. Mean forces (MPa) for the three materials were compared using an appropriate analysis of variance model (balanced-incomplete-blocks) The shear bond strengths (MPa) of the light-cured liner (Espe, Seefeld/Oberbay, FRG) was 4.71 +/- 1.16. Vitrabond showed the greatest variance of all three materials tested, however this material's average bond strength was greater than the maximum achieved for the other materials. Student-Newman-Keuls comparison of means showed that all cements differed significantly from each other (alpha = 0.05). It is concluded that the light-cured glass ionomer liner exhibited significantly better shear bond strength performance than the two auto-cured glass ionomers tested.