Surprising spatiotemporal stability and frequency-independence across multiple fitness peaks driving adaptive radiation in the wild

The effect of the environment on fitness in natural populations is a fundamental question in evolutionary biology. However, experimental manipulations of environment and phenotype are rare. Thus, the relative importance of the competitive environment versus intrinsic organismal performance in shaping the location, height, and fluidity of fitness peaks and valleys remains largely unknown. We experimentally tested the effect of competitive environment on the fitness landscape driving the evolution of novelty in a sympatric adaptive radiation of a generalist and two trophic specialist pupfishes, a scale-eater and molluscivore, endemic to San Salvador Island, Bahamas. We manipulated phenotypes, by generating 2,611 F4/F5 lab-reared hybrids, and competitive environment, by altering frequencies of rare phenotypes between high- and low-frequency field enclosures, then tracked hybrid survival in two natural lake populations on San Salvador. We found no evidence of frequency-dependent effects on survival fitness landscapes, indicating robustness to the competitive environment. Although survival surfaces favored alternate phenotypes between lakes, joint fitness estimation across lake environments supported multiple fitness peaks for generalist and molluscivore phenotypes and a large fitness valley isolating the most divergent scale-eater phenotype, strikingly similar to a previous independent field experiment. The consistency of this complex fitness landscape across competitive environments, multivariate trait axes, and spatiotemporal heterogeneity provides surprising evidence of stasis in major features of fitness landscapes despite substantial environmental variance, possibly due to absolute biomechanical constraints on diverse prey capture strategies within this radiation. These results challenge competitive speciation theory and highlight the interplay between organism and environment underlying static and dynamic features of the adaptive landscape.