Strength and tempo of directional selection in the wild

Abstract

Directional selection is a major force driving adaptation and evolutionary change. However, the distribution, strength, and tempo of phenotypic selection acting on quantitative traits in natural populations remain unclear across different study systems. We reviewed the literature (1984-1997) that reported the strength of directional selection as indexed by standardized linear selection gradients (beta). We asked how strong are viability and sexual selection, and whether strength of selection is correlated with the time scale over which it was measured. Estimates of the magnitude of directional selection (absolute value of beta) were exponentially distributed, with few estimates greater than 0.50 and most estimates less than 0.15. Sexual selection (measured by mating success) appeared stronger than viability selection (measured by survival). Viability selection that was measured over short periods (days) was typically stronger than selection measured over longer periods (months and years), but the strength of sexual selection did not vary with duration of selection episodes; as a result, sexual selection was stronger than viability selection over longer time scales (months and years), but not over short time scales (days).

Figure 1

Frequency distribution of estimated linear selection gradients (|β|) grouped by 0.05 β units. Selection gradients (n = 993) were compiled from 63 studies (median |β| = 0.150; mean |β| = 0.220). Two outlying estimates of |β| > 5 were excluded from analyses. This distribution was similar to an exponential distribution (P = 0.091).

Figure 2

Frequency distributions showing relative strength of sexual and natural selection. Magnitudes of linear selection gradients were grouped by 0.05 |β| units. Frequency distributions (standardized to account for sample size differences between groups) were significantly different (P < 0.001). More than 33% of studies that used survival as a fitness component reported |β| ≤ 0.05. Median and mean |β| based on measures of survival were 0.153 and 0.088, respectively; median and mean |β| based on mating success were 0.250 and 0.180, respectively.

Figure 3

(a) Frequency distributions of strengths of selection based on survival measured over episode lengths of days (<30 days), months (31–364 days), and years (>365 days). Frequency distributions (standardized to account for sample size differences among groups) of linear selection gradients (|β|) differed significantly (P < 0.002). More than half (54%) of βs measured over a year reported |β| ≤ 0.1, whereas only 36% of βs measured over days reported |β| ≤ 0.1. (b) Frequency distributions of strengths of selection based on mating success measured over the same episode length categories. Frequency distributions (standardized to account for sample size differences among groups) of linear selection gradients (|β|) were not significantly different (P > 0.75). In both figures, selection gradients were grouped by 0.05 |β| units.