The effects of ventilation on air quality, and on the welfare and production performance of dairy ewes were assessed in a 6-wk trial conducted during the winter of 2002. Thirty-six midlactation Comisana ewes were divided into three groups of 12, which were randomly balanced for parity, time of lambing, and number of lambs suckled. Treatments were low (LOV), moderate (MOV), and programmed ventilation regimen (PROV). In LOV and MOV rooms, fans provided 10 ventilation cycles of 40 min each at a fan speed of 1 and 2 m/s, respectively. In the PROV room, the fan was programmed to maintain a 70% relative humidity. Mean ventilation rates were 23, 47, and 73 m3/h per ewe in LOV, MOV, and PROV rooms, respectively. Air concentrations of microorganisms and dust, and of gaseous pollutants were measured twice weekly. Cell-mediated immune response to phytohemagglutinin at d 1, 21 and 42, and humoral response to chicken egg albumin at d 11, 21, 30, and 40 were determined. At d 39, ewes were injected with 2 IU of porcine adreno-corticotropic-hormone/kg body weight(0.75), and subjected to blood sampling for evaluation of cortisol concentrations immediately before and 1, 2, and 4 h after adreno-corticotrophic-hormone injection. Milk yield was recorded daily. Individual milk samples were analyzed weekly for composition, renneting parameters, and somatic cell count. The LOV treatment resulted in higher air concentrations of NH3 and CO2 than the MOV and PROV treatments. Greater amounts of total and respirable dust were found in the PROV room than in the LOV and the MOV rooms. The LOV ewes had lower milk yield than the PROV ewes, lower milk casein content, and higher rate of clot formation than the MOV and PROV ewes. The ventilation regimen did not affect the immune and endocrine responses of the ewes. Results suggest that an intermittent ventilation regimen, providing a mean ventilation rate of 47 m3/h per ewe at a fan speed of 2 m/s, is required to sustain the yield and cheese-making ability of ewe milk during the winter season.