A seminatural, factorial‐design experiment was used to quantify dynamics of the pathogen Mycoplasma agassizii and upper respiratory tract disease in the Mojave desert tortoise (Gopherus agassizii) over 2 years. Groups of initially healthy animals were separated into serologically positive (seropositive), seronegative, and artificially infected groups and paired into 23 pens. We found no evidence of long‐term immune protection to M. agassizii or of immunological memory. Initially seronegative, healthy tortoises experienced an equal amount of disease when paired with other seronegative groups as when paired with seropositive and artificially infected groups—suggesting that recrudescence is as significant as transmission in introducing disease in individuals in this host–pathogen system. Artificially infected groups of tortoises showed reduced levels of morbidity when paired with initially seronegative animals—suggesting either a dilution effect or a strong effect of pathogen load in this system. Physiological dynamics within the host appear to be instrumental in producing morbidity, recrudescence, and infectiousness, and thus of population‐level dynamics. We suggest new avenues for studying diseases in long‐lived ectothermic vertebrates and a shift in modeling such diseases.