Agassiz’s desert tortoise (Gopherus agassizii), a threatened species of the southwestern United States, has severely declined to the point where 76% of populations in critical habitat (Tortoise Conservation Areas) are below viability. The potential for rapid recovery of wild populations is low because females require 12–20 years to reach reproductive maturity and produce few eggs annually. We report on a 34‐year mark‐recapture study of tortoises initiated in 1979 at the Desert Tortoise Research Natural Area in the western Mojave Desert, California, USA, and provide substantive data on challenges faced by the species. In 1980, the United States Congress designated the Research Natural Area and protected the land from recreational vehicles, livestock grazing, and mining with a wildlife‐permeable fence. The 7.77‐km² study area, centered on interpretive facilities, included land both within the Natural Area and outside the fence. We expected greater benefits to accrue to the tortoises and habitat inside compared to outside. Our objectives were to conduct a demographic study, analyze and model changes in the tortoise population and habitat, and compare the effectiveness of fencing to protect populations and habitat inside the fence versus outside, where populations and habitat were unprotected. We conducted surveys in spring in each of 7 survey years from 1979, when the fence was under construction, through 2012. We compared populations inside to those outside the fence by survey year for changes in distribution, structure by size and relative age, sex ratios, death rates of adults, and causes of death for all sizes of tortoises. We used a Bayesian implementation of a Jolly Seber model for mark‐recapture data. We modeled detection, density, growth and transition of tortoises to larger size‐age classes, movements from inside the protective fence to outside and vice versa, and survival. After the second and subsequent survey years, we added surveys to monitor vegetation and habitat changes, conduct health assessments, and collect data on counts of predators and predator sign. At the beginning of the study, counts and densities for all sizes of tortoises were high, but densities were approximately 24% higher inside the fence than outside. By 2002, the low point in densities, densities had declined 90% inside the fence and 95% outside. Between 2002 and 2012, the population inside the fence showed signs of improving with a 54% increase in density. Outside the fence, densities remained low. At the end of the study, when we considered the initial differences in location, densities inside the fence were roughly 2.5 times higher than outside. The pattern of densities was similar for male and female adults. When evaluating survival by blocks of years, survivorship was higher in 1979–1989 than in 1989–2002 (the low point) and highest from 2002 to 2012. Recruitment and survival of adult females into the population was important for growing the population, but survival of all sizes, including juveniles, was also critical.