Background. Outbreaks of Yersinia pseudotuberculosis infection have been epidemiologically linked to fresh produce, but the bacterium has not been recovered from the food items implicated. In May 2003, a cluster of gastrointestinal illness and erythema nodosum was detected among schoolchildren who had eaten lunches prepared by the same institutional kitchen. Methods. We conducted a case-control study and trace-back, environmental, and laboratory investigations. Case patients had culture-confirmed Y. pseudotuberculosis O:1 infection, erythema nodosum, or reactive arthritis. Bacterial isolates from clinical and environmental samples were compared using pulsed-field gel electrophoresis (PFGE). Results. Of 7392 persons at risk, 111 (1.5%) met the case definition; 76 case patients and 172 healthy control subjects were enrolled in the case-control study. Only raw grated carrots were significantly associated with illness in a logistic-regression model (multivariable odds ratio, 5.7 [95% confidence interval, 1.7-19.5]); a dose response was found for increasing amount of consumption. Y. pseudotuberculosis O:1 isolates from 39 stool specimens and from 5 (42%) of 12 soil samples that contained carrot residue and were obtained from peeling and washing equipment at the production farm were indistinguishable by PFGE. Conclusions. Carrots contaminated early in the production process caused a large point-source outbreak. Our findings enable the development of evidence-based strategies to prevent outbreaks of this emerging foodborne pathogen.