This study aimed to evaluate the reproducibility of a new grading system for lumbar foraminal stenosis. Four grades were developed for lumbar foraminal stenosis on the basis of sagittal MRI. Grade 0 refers to the absence of foraminal stenosis; grade 1 refers to mild foraminal stenosis showing perineural fat obliteration in the two opposing directions, vertical or transverse; grade 2 refers to moderate foraminal stenosis showing perineural fat obliteration in the four directions without morphologic change, both vertical and transverse directions; and grade 3 refers to severe foraminal stenosis showing nerve root collapse or morphologic change. A total of 576 foramina in 96 patients were analyzed (from L3-L4 to L5-S1). Two experienced radiologists independently assessed the sagittal MR images. Interobserver agreement between the two radiologists and intraobserver agreement by one reader were analyzed using kappa statistics. According to reader 1, grade 1 foraminal stenosis was found in 33 foramina, grade 2 in six, and grade 3 in seven. According to reader 2, grade 1 foraminal stenosis was found in 32 foramina, grade 2 in six, and grade 3 in eight. Interobserver agreement in the grading of foraminal stenosis between the two readers was found to be nearly perfect (kappa value: right L3-L4, 1.0; left L3-L4, 0.905; right L4-L5, 0.929; left L4-L5, 0.942; right L5-S1, 0.919; and left L5-S1, 0.909). In intraobserver agreement by reader 1, grade 1 foraminal stenosis was found in 34 foramina, grade 2 in eight, and grade 3 in seven. Intraobserver agreement in the grading of foraminal stenosis was also found to be nearly perfect (kappa value: right L3-L4, 0.883; left L3-L4, 1.00; right L4-L5, 0.957; left L4-L5, 0.885; right L5-S1, 0.800; and left L5-S1, 0.905). The new grading system for foraminal stenosis of the lumbar spine showed nearly perfect interobserver and intraobserver agreement and would be helpful for clinical study and routine practice.