IMPORTANCE Spinal cord infarction (SCI) is often disabling, and the diagnosis can be challenging without an inciting event (eg, aortic surgery). Patients with a spontaneous SCI are often misdiagnosed as having transverse myelitis. Diagnostic criteria for SCI are lacking, hindering clinical care and research. OBJECTIVE To describe the characteristics of spontaneous SCI and propose diagnostic criteria. DESIGN, SETTING, AND PARTICIPANTS An institution-based search tool was used to identify patients evaluated at Mayo Clinic, Rochester, Minnesota, from January 1997 to December 2017 with a spontaneous SCI. Patients provided written consent to use their records for research. Participants were 18 years and older with a diagnosis of spontaneous SCI (n = 133), and controls were selected from a database of alternative myelopathy etiologies for validation of the proposed diagnostic criteria (n = 280). MAIN OUTCOMES AND MEASURES A descriptive analysis of SCI was performed and used to propose diagnostic criteria, and the criteria were validated. RESULTS Of 133 included patients with a spontaneous SCI, the median (interquartile range) age at presentation was 60 (52-69) years, and 101 (76%) had vascular risk factors. Rapid onset of severe deficits reaching nadir within 12 hours was typical (102 [77%]); some had a stuttering decline (31 [23%]). Sensory loss occurred in 126 patients (95%), selectively affecting pain/temperature in 49 (39%). Initial magnetic resonance imaging (MRI) spine results were normal in 30 patients (24%). Characteristic MRI T2-hyperintense patterns included owl eyes (82 [65%]) and pencil-like hyperintensity (50 [40%]); gadolinium enhancement (37 of 96 [39%]) was often linear and located in the anterior gray matter. Confirmatory MRI findings included diffusion-weighted imaging/apparent diffusion coefficient restriction (19 of 29 [67%]), adjacent dissection/occlusion (16 of 82 [20%]), and vertebral body infarction (11 [9%]). Cerebrospinal fluid showed mild inflammation in 7 of 89 patients (8%). Diagnostic criteria was proposed for definite, probable, and possible SCI of periprocedural and spontaneous onset. In the validation cohort (n = 280), 9 patients (3%) met criteria for possible SCI, and none met criteria for probable SCI. CONCLUSIONS AND RELEVANCE This large series of spontaneous SCIs provides clinical, laboratory, and MRI clues to SCI diagnosis. The diagnostic criteria proposed here will aid clinicians in making the correct diagnosis and ideally improve future care for patients with SCI. The validation of these criteria supports their utility in the evaluation of acute myelopathy.