After decades of decline in prevalence of complete tooth loss (edentulism), the trend continues to be misinterpreted, producing flawed projections and misdirected health goals. We investigated population trends in edentulism among U.S. adults aged ≥15 yr by creating time-series data from 5 national cross-sectional health surveys: 1957-1958 (n ≈ 100,000 adults), 1971-1975 (n = 14,655 adults), 1988-1998 (n = 18,011 adults), 1999-2002 (n = 12,336 adults), and 2009-2012 (n = 10,522 adults). Birth cohort analysis was used to isolate age and cohort effects. Geographic and sociodemographic variation in prevalence was investigated with a sixth U.S. survey of 432,519 adults conducted in 2010. Prevalence through 2050 was projected with age-cohort regression models using Monte-Carlo simulation of prediction intervals. Across the 5-decade observation period, edentulism prevalence declined from 18.9% in 1957-1958 (95% confidence limits: 18.4%, 19.4%) to 4.9% in 2009-2012 (95% confidence limits: 4.0%, 5.8%). The most influential determinant of the decline was the passing of generations born before the 1940s, whose rate of edentulism incidence (5%-6% per decade of age) far exceeded later cohorts (1%-3% per decade of age). High-income households experienced a greater relative decline, although a smaller absolute decline, than low-income households. By 2010, edentulism was a rare condition in high-income households, and it had contracted geographically to states with disproportionately high poverty. With the passing of generations born in the mid-20th century, the rate of decline in edentulism is projected to slow, reaching 2.6% (95% prediction limits: 2.1%, 3.1%) by 2050. The continuing decline will be offset only partially by population growth and population aging such that the predicted number of edentulous people in 2050 (8.6 million; 95% prediction limits: 6.8 million, 10.3 million) will be 30% lower than the 12.2 million edentulous people in 2010.