In spring 1996, an outbreak of Q fever occurred among residents of a rural town (population: 300) in Germany. A retrospective cohort study was conducted to ascertain the extent of the outbreak and to assess potential risk factors for illness. In July 1996, all residents > or =15 years received a self-administered questionnaire and were offered Coxiella burnetii antibody testing. Residents were considered to have probable Q fever if they had a positive result for IgM C. burnetii antibodies by ELISA or possible Q fever if they had fever > or =39 degrees C lasting >2 days and > or =3 symptoms (chills, sweats, severe headache, cough, aching muscles/joints, back pain, fatigue, or feeling ill) after 1 January 1996. Two hundred (84%) of the 239 residents aged > or =15 years either completed the questionnaire or submitted blood for antibody testing. Forty-five (23%) of these 200 met the probable or possible case definitions. Onsets of illness occurred in January-June 1996. Cases were geographically distributed throughout the town. Persons reporting contact with sheep (32% vs 18%, RR: 1.8, 95% CI: 1.1-2.9) and walking near a large sheep farm located next the town (34% vs 8%, RR: 4.5, 95% CI: 1.7-12.2) were more likely to have met the case definition than those without these exposures. Fifteen of 20 samples from the large sheep flock were positive for C. burnetii antibodies. The sheep had lambed outdoors in December 1995-January 1996 while the weather was extremely dry. The timing of the outbreak after lambing, the uniform distribution of cases throughout the town and the absence of risk factors among most case-persons suggest airborne transmission of C. burnetii from the large sheep farm.