A seasonal peak in hematologic malignancies may support hypotheses of infection-related precipitating events. Moderately increased incidence rates have been observed during the spring for leukemias and Hodgkin's disease but few studies have been conducted of non-Hodgkin's lymphoma (NHL). Our study consisted of 77,173 NHL patients in the Surveillance, Epidemiology, and End Results database diagnosed during 1973 - 99. Chi-square analyses showed excess observed-vs.-expected diagnoses during March, April, and June (P < 0.0001). B-cell origin subtype, but not T-cell/NK, was diagnosed more frequently in March. Controlling for age, sex, geographical location, and diagnosis year, multivariate Poisson regression revealed peaks in both March and April (P < 0.0001). Excluding cases in December, due to consistent troughs, regression uncovered greater-than-expected incidence during spring months for patients aged 20 - 39 years and 40 - 64 years (P = 0.043, P = 0.0001) but not among patients >or= 65 years. Future studies are needed to discern if a spring peak is due to diagnostic bias or other uncontrolled factors.