To conduct a prospective case-control study about causative factors of severe bullous erythema multiforme, Stevens-Johnson syndrome, and toxic epidermal necrolysis, we needed to define criteria for classifying the cases and standardize the collection of data so that cases could be reliably diagnosed according to this classification. Based on review of case histories and photographs of patients, a group of experts proposed a classification based on the pattern of erythema multiforme-like lesions (categorized as typical targets, raised or flat atypical targets, and purpuric macules) and on the extent of epidermal detachment. An atlas illustrating this classification that included photographs and schematic drawings was developed. We compared the evaluations of 28 cases by four nonphysicians relying on the atlas with the evaluations of the same cases by five experts not using the atlas to determine the usefulness of this atlas for classifying cases according to our nosologic schema. The following consensus classification in five categories was proposed: bullous erythema multiforme, detachment below 10% of the body surface area plus localized "typical targets" or "raised atypical targets"; Stevens-Johnson syndrome, detachment below 10% of the body surface area plus widespread erythematous or purpuric macules or flat atypical targets; overlap Stevens-Johnson syndrome-toxic epidermal necrolysis, detachment between 10% and 30% of the body surface area plus widespread purpuric macules or flat atypical targets; toxic epidermal necrolysis with spots, detachment above 30% of the body surface area plus widespread purpuric macules or flat atypical targets; and toxic epidermal necrolysis without spots, detachment above 10% of the body surface area with large epidermal sheets and without any purpuric macule or target. Using the atlas, the nonexperts showed excellent agreement with the experts. This study suggests that an illustrated atlas is a useful tool for standardizing the diagnosis of acute severe bullous disorders that are attributed to drugs or infectious agents. Whether the five categories proposed represent distinct etiopathologic entities will require further epidemiologic and laboratory investigations.