Confocal laser endomicroscopy (CLE) is a technique that permits real-time detection and quantification of changes in intestinal tissues and cells, including increases in intraepithelial lymphocytes and fluid extravasation through epithelial leaks. Using CLE analysis of patients with irritable bowel syndrome (IBS), we found that more than half have responses to specific food components. Exclusion of the defined food led to long-term symptom relief. We used the results of CLE to detect reactions to food in a larger patient population and analyzed duodenal biopsy samples and fluid from patients to investigate mechanisms of these reactions. In a prospective study, 155 patients with IBS received 4 challenges with each of 4 common food components via the endoscope, followed by CLE, at a tertiary medical center. Classical food allergies were excluded by negative results from immunoglobulin E serology analysis and skin tests for common food antigens. Duodenal biopsy samples and fluid were collected 2 weeks before and immediately after CLE and were analyzed by histology, immunohistochemistry, reverse transcription polymerase chain reaction, and immunoblots. Results from patients who had a response to food during CLE (CLE+) were compared with results from patients who did not have a reaction during CLE (CLE-) or healthy individuals (controls). Of the 108 patients who completed the study, 76 were CLE+ (70%), and 46 of these (61%) reacted to wheat. CLE+ patients had a 4-fold increase in prevalence of atopic disorders compared with controls (P= .001). Numbers of intraepithelial lymphocytes were significantly higher in duodenal biopsy samples from CLE+ vs CLE- patients or controls (P= .001). Expression of claudin-2 increased from crypt to villus tip (P < .001) and was up-regulated in CLE+ patients compared with CLE- patients or controls (P= .023). Levels of occludin were lower in duodenal biopsy samples from CLE+ patients vs controls (P= .022) and were lowest in villus tips (P< .001). Levels of messenger RNAs encoding inflammatory cytokines were unchanged in duodenal tissues after CLE challenge, but eosinophil degranulation increased, and levels of eosinophilic cationic protein were higher in duodenal fluid from CLE+ patients than controls (P= .03). In a CLE analysis of patients with IBS, we found that more than 50% of patients could have nonclassical food allergy, with immediate disruption of the intestinal barrier upon exposure to food antigens. Duodenal tissues from patients with responses to food components during CLE had immediate increases inexpression of claudin-2 and decreases in occludin. CLE+patients also had increased eosinophil degranulation, indicating an atypical food allergy characterized by eosinophil activation.