1. In order to develop an improved differential sugar absorption test for simultaneously assessing intestinal permeability and lactose intolerance, methods were established for determining raffinose, lactose and L-arabinose in human urine. Using NAD(P)H-coupled enzymatic assays and fluorimetry, each sugar was measurable over a concentration range of approximately 3-300 mumol/l in diluted urine specimens. 2. After an overnight fast, 40 normal volunteers drank an iso-osmotic solution containing raffinose, lactose and L-arabinose. The median 5 h urinary sugar excretion was 0.26% of the ingested raffinose, 0.05% of lactose and 17.5% of L-arabinose. 3. In 143 patients with gastrointestinal disease, excretion of both ingested raffinose and lactose was significantly increased in coeliac disease in relapse or in partial remission and in Crohn's disease, but not in the irritable bowel syndrome, coeliac disease in remission or ulcerative colitis. Excretion of lactose, but not raffinose, was increased in patients with mucosal lactase deficiency, whereas excretion of L-arabinose was reduced in all disease groups except ulcerative colitis. 4. Discrimination between diseases was poor when based on individual sugar recoveries, but improved dramatically when excretion was expressed relative to that of L-arabinose. The raffinose/L-arabinose excretion ratio, an index of intestinal permeability, was greater than 0.08 in 15/15 untreated coeliac patients but less than 0.06 in all normal subjects and in 9/9 lactase-deficient patients, 15/16 recovered coeliac patients, 5/6 patients with ulcerative colitis, 13/16 patients with Crohn's disease and 61/62 patients with irritable bowel syndrome.