We applied drugs in aqueous gels to the buccal mucosae of normal volunteers to develop an assay of gastrointestinal irritation potential. We studied effects of pH, concentrations, and formulations. We evaluated irritation by subjects' reports of their sensations and observers' grading of visible reactions. On an irritation log ranking scaled 0 to 5, placebo gels produced virtually no irritation, except those formulated at pH 1, 2, and 14. Albuterol, furosemide, and methyldopa produced essentially no irritation. Minutes of exposure to sodium indomethacin, sodium naproxen, and propranolol (at, respectively, pH 8.2, 9.6, and 4.8 and concentrations of 12%, 40%, and 8%) caused ulceration that took up to weeks to heal. Acid forms of naproxen and indomethacin caused minimal or no irritation. Although irritation models based on one gastrointestinal area have limitations, our results indicate that the minimally invasive buccal assay allows ranking of contact irritation potential of drugs and formulations that could aid in selecting optimum formulations for clinical use.