OBJECTIVE To measure the 'repair' strength of various combinations of composite using four manufacturers' adhesive systems, to compare the bond strengths with the cohesive strength of the original, unrepaired products and to assess whether the chemical nature of the resin matrix influenced the repair strength. METHODS Specimens were prepared of three composite materials Durafill, Heraeus Kulzer; P90 (Silorane) 3M ESPE; Z250 (3M ESPE) and aged in water at 60°C for 1 month. One surface of each specimen was faced with 80-grit silicone carbide paper, one of four adhesives placed (Ecusit, DMG; Clearfil Repair, Kuraray; P90 System Adhesive; Single Bond 2, 3M ESPE) and 'repair' composite added of the same type as above, such that all combinations of original and repair composite and adhesive were used. 'Stick' samples, approximately 6 mm × 0.8 mm × 4 mm were prepared from each repair specimen, a neck created at the junction of original and repair composites and the hour-glass sample tested in tension at 1 mm/min. The microtensile bond strength of the repair was calculated and the mode of failure (adhesive; cohesive in the original composite; cohesive in the repair composite) recorded. RESULTS There was no significant difference between the cohesive strengths of Filtek P90 and Filtek Z250 (both ≈106 MPa); both were significantly stronger than Durafill (67.0 MPa). For bonding to Durafill the bond strengths ranged from 17.6 MPa to 50.9 MPa; for bonding to P90, the bond strengths ranged from 5.0 MPa to 54.2 MPa; for bonding to Z250, the bond strengths ranged from 17.2 MPa to 75.4 MPa. Clearfil Repair appeared to provide the most consistently high bond strengths, followed by the P90 System Adhesive, Single Bond 2 and Ecusit. Overall, the majority of failures (74%) was adhesive. CONCLUSIONS It appears that bonding of new dimethacrylate-based composite to old dimethacrylate-based composite can be a viable clinical procedure. However, if the original composite is silorane-based (e.g., P90), then using the silane-based adhesive may be the best repair option, and similarly if it is planned to effect a repair with a silorane-based composite, using a silane-based adhesive may give the best outcome. The null hypotheses are thus rejected.