Natural selection has almost certainly shaped many evolutionary trajectories documented in fossil lineages, but it has proven difficult to demonstrate this claim by analyzing sequences of evolutionary changes. In a recently published and particularly promising test case, an evolutionary time series of populations displaying armor reduction in a fossil stickleback lineage could not be consistently distinguished from a null model of neutral drift, despite excellent temporal resolution and an abundance of indirect evidence implicating natural selection. Here, we revisit this case study, applying analyses that differ from standard approaches in that: (1) we do not treat genetic drift as a null model, and instead assess neutral and adaptive explanations on equal footing using the Akaike Information Criterion; and (2) rather than constant directional selection, the adaptive scenario we consider is that of a population ascending a peak on the adaptive landscape, modeled as an Orstein-Uhlenbeck process. For all three skeletal features measured in the stickleback lineage, the adaptive model decisively outperforms neutral evolution, supporting a role for natural selection in the evolution of these traits. These results demonstrate that, at least under favorable circumstances, it is possible to infer in fossil lineages the relationship between evolutionary change and features of the adaptive landscape.

Inferring the causes for change in the fossil record has been a persistent problem in evolutionary biology. Three independent lines of evidence indicate that a lineage of the fossil stickleback fish Gasterosteus doryssus experienced directional natural selection for reduction of armor. Nonetheless, application to this lineage of three methods to infer natural selection in the fossil record could not exclude random process as the cause for armor change. Excluding stabilizing selection and genetic drift as the mechanisms for biostratigraphic patterns in the fossil record when directional natural selection was the actual cause is very difficult. Biostratigraphic sequences with extremely fine temporal resolution among samples and other favorable properties must be used to infer directional selection in the fossil record.

No abstract

M odels, experiments, and field studies provide evidence of the ecological controls on evolution, but extrapolating results over longer time scales is a perennial problem in evolutionary biology. Trophic ecology and competition for food, for example, are thought to drive speciation through niche differentiation, character displacement, and phenotypic divergence (1). Yet direct evidence that feeding controls evolution over extended time scales, available only from the fossil record, is difficult to obtain because it is rarely possible to directly analyze dietary change in long-dead animals. Functional changes must be inferred from changes in morphology, and attempts to determine whether morphological changes were caused by shifts in feeding can become circular.Here, we report an investigation of trophic resource use in a fossil sequence preserving an evolving lineage of threespine stickleback (Gasterosteus). We focus on stickleback for two reasons. First, perhaps the best-known work on speciation in fishes concerns stickleback in postglacial coastal lakes in Canada, where planktivores and benthic feeders coexist as two reproductively isolated and phenotypically distinct trophic forms. The differences between these forms result from competition for food (1, 2). Second, fossil stickleback from the Miocene Truckee Formation (Nevada) provide a detailed, high-resolution record of evolutionary change within a lineage spanning tens of thousands of years (3).We investigated the relationship between trophic resource use and evolutionary change through quantitative analysis of dental microwear (4). Laboratory feeding experiments and analyses of wild stickleback populations show that microwear exhibits a progressive shift from planktivores to benthic feeders (Fig. 1, A and B) (5). Discriminant analysis using feature length and density indicates that scores for the first discriminant function (DF) are a good predictor of trophic ecology. For wild fish populations (n = 4), mean scores were significantly correlated with diet (r = 0.95, P = 0.05) and gill raker number (r = -0.996, P = 0.004).Analysis of fossil stickleback teeth revealed an overall range and pattern of feature densities and lengths similar to that of extant fish (Fig. 1C), suggesting that the fossil microwear records a similar benthic-planktonic feeding spectrum. This was supported by application of the DF derived from wild fish to the fossils: DF scores vary significantly between samples (F = 10.8, df of 7 and 87, P = 0.0001), and a Tukey-Kramer procedure revealed significant pairwise differences. This procedure also grouped some fossil samples with benthic-feeding wild populations (samples 19.6, 19.7, 19.6, and 6.9), others with planktivore populations (21.5), with some placed between (2.9 and 3.8).These changes in inferred trophic ecology are significantly correlated with evolutionary changes in armor phenotype through time (3) (Fig. 1E). DF scores are correlated with dorsal [nonparametric Spearman rank correlation (r s ) = 0.23, P = 0.03, n = 89 fish] and pe...