BACKGROUND It is unclear if brief episodes of paroxysmal atrial fibrillation (PAF) detected by prolonged cardiac monitoring are an occult of cause of cryptogenic strokes (CS). We compared the incidence of PAF in patients with CS and patients with stroke of known cause (SKC) using prolonged ambulatory cardiac monitoring. METHODS We prospectively enrolled patients within 3 months of ischemic stroke to undergo noninvasive cardiac monitoring for 3 weeks. Primary end point was PAF detection independently confirmed by 2 blinded cardiologists. RESULTS The study consisted of 132 patients, 66 had CS and 66 had SKC. Episodes of PAF were detected in 16 of 64 (25%) patients with CS and 9 of 64 (14%) patients with SKC (P=.12). Duration and number of PAF episodes, PAF burden, and time of first PAF detection did not differ significantly between the 2 groups (P>.05 for all). In patients younger than 65 years, PAF was more common in the CS group (22% versus 3%; P=.07), whereas in patients 65 years or older, the rates of detection were similar (27% in CS versus 25% in SKC; P=.9). Among patients younger than 65 years with embolic imaging pattern, PAF was only observed in the CS group (21% versus 0%; P=.03). CONCLUSIONS Very short episodes of PAF are common in patients with CS and with SKC, but their pathogenic significance is unclear. Predominance of PAF in younger patients with CS and embolic infarct pattern suggests a causative role in these cases. More research is needed before prolonged cardiac rhythm monitoring can be recommended to guide anticoagulation in CS patients.