Magnetic resonance (MR) imaging characteristics are commonly used to help predict intracranial disease categories in dogs, however, few large studies have objectively evaluated these characteristics. The purpose of this retrospective study was to evaluate MR characteristics that have been used to differentiate neoplastic, inflammatory, and vascular intracranial diseases in a large, multi-institutional population of dogs. Medical records from three veterinary teaching hospitals were searched over a 6-year period for dogs that had diagnostic quality brain MR scans and histologically confirmed intracranial disease. Three examiners who were unaware of histologic diagnosis independently evaluated 19 MR lesion characteristics totaling 57 possible responses. A total of 75 dogs with histologically confirmed intracranial disease were included in analyses: 51 with neoplasia, 18 with inflammatory disease, and six with cerebrovascular disease. Only strong contrast enhancement was more common in neoplasia than other disease categories. A multivariable statistical model suggested that extra-axial origin, T2-FLAIR mixed intensity, and defined lesion margins were also predictive of neoplasia. Meningeal enhancement, irregular lesion shape, and multifocal location distinguished inflammatory diseases from the other disease categories. No MR characteristics distinguished vascular lesions and these appeared most similar to neoplasia. These results differed from a previous report describing seven MR characteristics that were predictive of neoplasia in dogs and cats. Findings from the current study indicated that the high performance of MR for diagnosing canine intracranial diseases might be due to evaluator recognition of combinations of MR characteristics vs. relying on any one MR characteristic alone.