The association between Mycobacterium avium ssp. paratuberculosis (MAP) and milk production was estimated on 2 California dairies using longitudinal data from 5,926 cows. Both study herds had moderate MAP seroprevalence, housed cows in freestalls, and had Johne's disease control programs. Cow MAP status was determined using both serum ELISA and fecal culture results from cows tested at dry-off and from whole-herd tests. Potential confounders were evaluated based on a causal diagram. Mixed models with 2 functions (splines) for days in milk (DIM) representing milk production pre- and postpeak used in similar studies were further modified to use each cow's observed DIM at peak and lactation length. Cows that were seropositive produced 2.5kg less 4% fat-corrected milk (FCM) per day than their seronegative herdmates. In addition, cows that were fecal-culture positive by liquid culture and confirmed by PCR produced 2.2kg less 4% FCM per day than their fecal-culture negative herdmates. The decrease in milk production in MAP test-positive compared with test-negative cows started in the second lactation. A switch in MAP status in either ELISA or fecal culture results from positive to negative had no significant association with milk production. Modified DIM functions that used the observed DIM at peak had better model fit than another function that assumed a fixed peak at 60 DIM. Cows that tested positive for MAP on serum ELISA or fecal culture produced less milk than cows that tested negative, and the association between MAP and milk production was not confounded by mastitis, elevated somatic cell counts, or uterine or metabolic cow conditions.