Objective: To compare the macro mean shear bond strength (SBS) and failure mode of three cements to two types of metal using mould-enclosed and non-enclosed cement specimens. Methods: Titanium and base metal cobalt-based substrates were finished with 50 mu m aluminium oxide. Two resin-modified glass-ionomers (Riva Luting Plus, Fuji Plus) and one resin cement (RelyX Unicem) were prepared as per manufacturers' instructions. Metal mould-enclosed and non-enclosed cement specimens with a bonding area diameter of 3.5 mm were prepared and stressed to failure using a 2 mm blunt edge shear knife at a cross-head speed of 1 mm/min to determine mean SBS. The shear knife was placed against the surface of the substrate. Failure analysis of the failed interface was performed with a stereo microscope at 40x magnification. Results: Two-way Analysis of Variance demonstrated a significant difference in mean SBS between materials (p = 0.004) and cement (p = 0.001). There was also a significant interaction between method and cement on SBS, F(2, 170) = 7.209, p = 0.003. Post hoc Tukey tests demonstrated no significant difference for either resinmodified glass-ionomer cements (RMGIC) (p = 0.864, p = 0.620) when comparing non-enclosed and mould-enclosed test methods bonded to titanium. There was however a significant difference (p < 0.001) between the mean SBS obtained for the resin cement when comparing the non-enclosed and mould-enclosed test methods. For base metal, the RMGIC's SBS was higher than resin cement but no difference was observed between one of RMGIC's and the resin cement non-enclosed mean SBS. Although not the case for RMGIC's bonded to titanium or one RMGIC bonded to non-precious cobalt-based metal, when comparing the test method on each cement, RMGIC and the resin cement showed significant differences between non- and mould-enclosed specimens mean SBS. In the non-enclosed specimen tests, 71% of all specimens tested exhibited adhesive failure, which was statistically different (p < 0.001) to 91% for the mould-enclosed specimens. Failure mode was not always statistically different within groups, however non-enclosed specimens showed higher frequencies of mixed failures. Conclusion: Within the limits of this study, significant differences were found in comparing the mean SBS between mould-enclosed and non-enclosed specimens. A significant difference was found in failure mode between mould-enclosed and non-mould enclosed specimens. Mould-enclosed specimens bonded to metal exhibited a higher frequency of adhesive failure than non-mould enclosed specimens. Relevance: Mould-enclosed specimens used in place of non-enclosed specimens can be used in SBS testing to give a more valid result when bonding to metal.