Immunohistochemical staining for mismatch repair proteins has during recent years been established as a routine analysis in many pathology laboratories with the aim to identify tumors linked to the hereditary nonpolyposis colorectal cancer syndrome. Despite widespread application, data on reliability are lacking. We therefore evaluated interobserver variability among 6 pathologists, 3 experienced gastrointestinal pathologists and 3 residents. In total, 225 immunohistochemically stained colorectal cancers were evaluated as having normal, weak, loss of, or nonevaluable mismatch repair protein staining. Full consensus was achieved in 51% of the stainings for MLH1, 61% for PMS2, 83% for MSH2, and 45% for MSH6. Weak stainings were the main cause of reduced consensus, whereas contradictory evaluations with normal as well as loss of staining were reported in 2% to 6% of the tumors. Interobserver variability was considerable, though experienced pathologists and residents reached the same level of consensus. Because results from immunohistochemical mismatch repair protein stainings are used for decisions on mutation analysis and as an aid in the interpretation of gene variants of unknown significance in hereditary nonpolyposis colorectal cancer, the interobserver variability identified highlights the need for quality assessment programs, including guidelines for classification of different expression patterns.