Abstract.-Because many stocks of Pacific salmon Oncorhynchus spp. are listed under the U.S. Endangered Species Act (ESA), research has focused on predicting the future population dynamics for these low-abundance stocks. One method used to make predictions is known as population viability analysis. Pacific salmon populations exhibit much higher apparent variability than other ESA-listed vertebrates, and high variability increases the probability of extinction. If the high variability is primarily due to counting methods, it could be reduced in model predictions by using methods that correct for measurement error, sampling error, or both. Using data from British Columbia pink salmon O. gorbuscha and Snake River springor summer-run Chinook salmon O. tshawytscha and several modeling approaches (Ricker, Dennis, and statespace models), we compared repeated counts of the same population (e.g., spawner and fry, dam and redd counts). We applied the methods to the first half of the time series and compared the predictions with the last half of the time series. The results demonstrated that having counts of all life stages of a Pacific salmon population is no guarantee that variability will be markedly reduced. Measurement error is not the primary cause of high variability in empirical estimates of abundance or in predicted future abundance for the stocks analyzed. The very wide bounds on predicted abundance limit the utility of the model predictions for making management decisions. Furthermore, obtaining more accurate or complete measurements of population abundance is unlikely to reduce the wide error bounds in predictions of future abundances.Over the past 15 years many stocks of Pacific salmon Oncorhynchus spp. have been listed as threatened or endangered under the U.S. Endangered Species Act (ESA) (NOAA Fisheries 2005). This has focused researchers' attention on a difficult problem: how best to predict the future course of population dynamics for listed stocks whose abundances are often greatly reduced from historic levels. One method used to make such predictions is known as population viability analysis (PVA). Several PVA methods have been used for modeling listed salmon populations, beginning with spawner-recruit (SR) models familiar to many fisheries biologists (e.g., Marmorek et al. 1998;Paulsen and Hinrichsen 2002). More recently, researchers, drawing on methods used for other ESAlisted vertebrate species (e.g., northern spotted owl Strix occidentalis caurina, loggerhead sea turtle Caretta caretta, snail kite Rostrhamus sociabilis, and Bachman's sparrow Aimophifa aestivulis; Morris et al. 2002), have employed diffusion approximation methods (e.g., Holmes 2004;Holmes and Semmens 2004). These methods have also been applied for long-term projections by management agencies at the evolutionarily significant unit (ESU) and population levels (NOAA 2000). Unlike spawner-recruit models diffusion approximation methods do not require information on the age of returning progeny (recruits) and, therefore, can be employed for...