OBJECTIVE To evaluate the clinical features and pathological joint changes in dogs with erosive immune-mediated polyarthritis (IMPA). DESIGN Retrospective case series. ANIMALS 13 dogs with erosive IMPA and 66 dogs with nonerosive IMPA. PROCEDURES The medical record database of a veterinary teaching hospital was reviewed to identify dogs with IMPA that were examined between October 2004 and December 2012. For each IMPA-affected dog, information extracted from the medical record included signalment, diagnostic test results, radiographic findings, and treatments administered. Dogs were classified as having erosive IMPA if review of radiographs revealed the presence of bone lysis in multiple joints, and descriptive data were generated for those dogs. All available direct smears of synovial fluid samples underwent cytologic evaluation. The synovial fluid total nucleated cell count and WBC differential were estimated and compared between dogs with erosive IMPA and dogs with nonerosive IMPA. RESULTS 13 of 79 (16%) dogs had erosive IMPA. Dogs with erosive IMPA had a mean ± SD age of 7.1 ± 2.4 years and bodyweight of 8.3 ± 3.4 kg (18.3 ± 7.5 lb). All 13 dogs had erosive lesions in their carpal joints. The estimated median synovial fluid lymphocyte count for dogs with erosive IMPA was significantly greater than that for dogs with nonerosive IMPA. All dogs received immunosuppressive therapy with leflunomide (n = 9), prednisone (3), or prednisone-azathioprine (1). CONCLUSIONS AND CLINICAL RELEVANCE Results indicated erosive IMPA most commonly affected the carpal joints of middle-aged small-breed dogs. Further genetic analyses and analysis of lymphocyte-subsets is warranted for dogs with erosive IMPA. Word count = 256