Due to the mediating role of body size in determining fitness, the ‘bigger is better’ hypothesis still pervades evolutionary ecology despite evidence that natural selection on phenotypic traits varies in time and space. For Pacific salmon (genus Oncorhynchus), most individual studies quantify selection across a narrow range of sizes and ages; therefore, uncertainties remain concerning how selection on size may differ among diverse life-histories. Here, we quantify the direction and magnitude of natural selection on body size among age-classes of multiple marine cohorts of O. nerka (sockeye salmon). Across four cohorts of seaward migrants, we calculated standardized selection differentials by comparing observed size distributions of out-migrating juvenile salmon to back-calculated smolt length from the scales of surviving, returning adults. Results reveal the magnitude of selection on size was very strong (> 90^th percentile compared to a database of 3,759 linear selection differentials) and consistent among years. However, the direction of selection on size consistently varied among age-classes. Selection was positive for fish migrating to sea after two years in freshwater (age 2) and in their first year of life (age 0), but negative for fish migrating after 1 year in freshwater (age 1). The absolute magnitude of selection was negatively correlated to mean ocean-entry timing, which may underpin negative selection favoring small age-1 fish, given associations between size and timing of seaward migration. Collectively, these results indicate that ‘bigger is not always better’ in terms of survival and emphasize trade-offs that may exist between fitness components for organisms with similarly diverse migratory life-histories.

O. nerka smolts were sampled at ocean entry by the Alaska Department of Fish & Game (ADF&G) from May 20 – July 30, 1990; May 11 – August 4, 1991; May 5 – July 31, 1992; and May 10 – August 6, 1993 using a Canadian fan trap 1.6 km downstream of Lower Olga Lake. A smear of scales was pulled from individuals and mounted on glass slides; fork length was measured and smolt scales were later used for age-determination.

Adult O. nerka have been enumerated and sampled for age and length (mid-eye to tail fork) at the Upper Station weir since 1928. Based on stock-specific run timing, fish returning through July 15 are considered the Early Run, where fish returning after July 15 are assumed to be the Late Run. Run reconstructions based on escapement and scale pattern analysis are available beginning in 1969 for the Early Run and 1970 for the Late Run. As such, the collection of archived adult scales begins in 1969-1970 and continues annually by ADF&G.

Archived scales sampled by ADF&G from returning adult fish during 1969 – 2016 were preserved as impressions in acetate cards and include corresponding information on length, age, and sex (based on physical examination of external characteristics). Scales and impressions were included in the study based on the following criteria: 1) we agreed with ADF&G age determination, 2) annuli are clearly defined and not affected by regeneration or reabsorption of the scale, and 3) the shape of the scale indicates it was taken from the preferred area, which is immediately above the lateral line and slightly forward of the adipose fin. Smolt scales were randomly sampled (n = 1300) for each age-class (0 – 3) from when smolts were sampled (1990 – 1993) with stratified random sampling for each age (0 – 2) and year (1990 – 1993). A random sample of age-3 smolts were sampled across all years due to a low number of available scales (n = 100). A random sample of 50 adult scales was selected from each returning age-class that entered the ocean in 1990 – 1993; in addition, age-classes were included in the analysis if 25 readable scales of each sex could be obtained, with a total of 50 scales per age-class in a given year. Acetate scale impressions were digitized using a Z-Scan 46-II microfiche reader attached to a 19.3 mm zoom lens and images were exported at high resolution (3352 x 4425 pixels) to Image-Pro software® for accurate measurements of scale patterns. We measured the distance (mm) on the longest axis from the focus of the scale to the last circulus at the end of freshwater growth. In addition, we measured the total distance from the focus of the scale to the edge of the scale for both adult and smolt scales. Random checks between two trained readers were done to assure consistency in measurements for adult scales (~5% of samples).