The objective of this study was to compare 3 orthodontic adhesives in the areas of shear-peel bond strength, location of adhesive failure, and extent of enamel cracking before bonding and after debonding of orthodontic brackets. The adhesives included a composite resin control (Transbond XT; 3M/Unitek, St Paul, Minn), a resin-modified glass ionomer cement (Fuji Ortho LC; GC America Corp, Alsip, Ill), and a polyacid-modified composite resin under dry and saliva-contaminated conditions (Assure; Reliance Orthodontic Products Inc, Itasca, Ill). Metal brackets were bonded to the buccal surfaces of 160 (4 groups of 40) human premolars. The bonded teeth were stored in deionized water at 37 degrees C for 30 days and thermocycled for 24 hours before debonding with a Universal Instron (Instron Corp, Canton, Mass) testing machine. The extent of cracking in the buccal surfaces was evaluated under 16x magnification before bonding and after debonding. Although the bond strength of the composite resin control (20.19 MPa) was significantly greater (P <.05) than that of the adhesives in the other groups, clinically acceptable shear-peel bond strengths were found for all adhesives (Fuji Ortho LC = 13.57 MPa, Assure-dry = 10.74 MPa, Assure-wet = 10.99 MPa). The bond strength for the Assure adhesive was not significantly affected by saliva contamination. The sample of extracted premolars used in this study displayed a greater frequency of buccal surface enamel cracking (46.7%) than that reported in the literature for in vivo premolars (7.8%-10.2%), which was possibly due to the extraction process. The frequency of enamel cracking in a subset of this sample (n = 34) increased from 46.4% at prebonding to 62.4% at postdebonding as a result of the forces of debonding.