The determination of HER2 amplification is critical to selecting appropriate patients for HER2 targeted therapy in breast cancer. Dual in situ hybridization (DISH), an alternative to fluorescence in situ hybridization (FISH) and immunohistochemistry, is now available. To compare the FISH and DISH methods, we tested 251 samples enriched for common or difficult-to-assess HER2 anomalies. Seven samples failed DISH testing. There was a 64% (156/244) concordance between FISH and DISH by anomaly (κ = 0.58, 95% confidence interval, 0.51-0.65; P < .0001) and an 83% (203/244) concordance by amplification status (κ = 0.58; 95% confidence interval, 0.47-0.69; P < .0001). DISH resulted in lower estimates of HER2/ centromere 17 ratios than FISH, and many cases that were equivocal with FISH were normal with DISH. DISH did not detect any case with coamplification of HER2 and centromere 17. Using a cohort of difficult specimens, we observed less than 95% concordance between FISH and DISH. DISH may underestimate the HER2/chromosome 17 ratio, or FISH may overestimate this ratio.