Statement of problem. Direct bonding of orthodontic brackets to porcelain surfaces has been plagued by failure. Purpose. The purpose of this study was to compare the bond strengths of several different bonding systems when bonding orthodontic brackets to porcelain-fused-to-metal surfaces. Material and methods. Fifty natural glazed feldspathic porcelain-fused-to-noble metal disks 6 mm in diameter and 3 mm in height (1 mm metal and 2 mm porcelain) were fabricated and divided into 5 groups of 10. A different bonding system (GC America Fuji LC, American Ortho Spectrum, 3M Transbond, TP Orthodontics Python, and Kerr Herculite) was assigned to each group, and 50 identical orthodontic brackets were bonded (with the above mentioned systems) to each disk according to each manufacturer's instructions. Each system except TP Orthodontics Python conditioned with phosphoric acid (35% to 37.5%) and all systems were primed with silane before bonding. The specimens were subjected to gradual shear forces up to 123 N in a universal testing machine (Instron Corp, Canton, Mass.) until fracture. The shear bond strength of the bonding systems between the porcelain surface and the bracket was measured in megapascals (MPa). Failures were observed via a Zeiss optical microscope (10×); Tukey's HSD Test and analysis of variance were used to determine significance between the bonding systems at P<.05 level of significance. Results. Failure of all of specimens was adhesive between the porcelain surface and the bonding agents. On the basis of a current literature review, bonding systems were categorized as clinically acceptable if they had a shear bond strength of 6 to 8 MPa. The 3M Transbond Bonding System, American Orthodontics Spectrum Bonding System, and GC America Fuji Ortho LC Bonding System performed within this clinically acceptable range (6 to 8 MPa), whereas Kerr Herculite Bonding System and TP Orthodontics Python Bonding System did not (2 to 4 MPa). The bond strengths of GC America Fuji Ortho LC, 3M Transbond, and American Orthodontics Spectrum were significantly greater (mean = 2.3 times) than TP Orthodontics Python or Kerr Herculite bonding systems. Conclusion. Within the limitations of this study, the results reaffirm the regimen of conditioning with phosphoric acid and priming with silane before bonding orthodontic brackets to feldspathic porcelain fused to noble metal. All products indicated for this purpose may not achieve satisfactory bond strengths; however, because they do not all include the critical steps of conditioning with phosphoric acid and priming with silane. The 3M Transbond Bonding System, American Orthodontics Spectrum Bonding System, and GC America Fuji Ortho LC Bonding System performed within the clinically acceptable range (6 to 8 MPa), whereas Kerr Herculite Bonding System and TP Orthodontics Python Bonding System did not (2 to 4 MPa). (J Prosthet Dent 2003;89:66-69.)