We used 13 years (1995-2007) of capture-mark-recapture data to assess population dynamics of endangered Lost River suckers Deltistes luxatus and shortnose suckers Chasmistes brevirostris in Upper Klamath Lake, Oregon. The Cormack-Jolly-Seber method was used to estimate survival, and information theoretic modeling was used to assess variation due to time, gender, species, and spawning subpopulations. Length data were used to detect multiple year-class failures and events of high recruitment into adult spawning populations. Average annual survival probability was 0.88 for Lost River suckers and 0.76 for shortnose suckers. Mean life span estimates based on these survival rates indicated that Lost River suckers survived long enough on average to attempt reproduction eight times, whereas shortnose suckers only survived to spawn three to four times. Shortnose sucker survival was not only poor in years of fish kills (1995-1997) but also was low in years without fish kills (i.e., 2002 and 2004). This suggests that high mortality occurs in some years but is not necessarily associated with fish kills. Annual survival probabilities were not only different between the two species but also differed between two spawning subpopulations of Lost River suckers. Length composition data indicated that recruitment into spawning populations only occurred intermittently. Populations of both species transitioned from primarily old individuals with little size diversity and consistently poor recruitment in the late 1980s and early 1990s to mostly small, recruit-sized fish by the late 1990s. A better understanding of the factors influencing adult survival and recruitment into spawning populations is needed. Monitoring these vital parameters will provide a quantitative means to evaluate population status and assess the effectiveness of conservation and recovery efforts.