In dendritic river systems, gene flow is expected to occur primarily within watersheds. Yet, rare cross‐watershed transfers can also occur, whether mediated by (often historical) geological events or (often contemporary) human activities. We explored these events and their potential evolutionary consequences by analyzing patterns of neutral genetic variation (microsatellites) and adaptive phenotypic variation (male color) in wild guppies (Poecilia reticulata) distributed across two watersheds in northern Trinidad. We found the expected signatures of within‐watershed gene flow; yet we also inferred at least two instances of cross‐watershed gene flow—one in the upstream reaches and one further downstream. The upstream cross‐watershed event appears to be very recent (41 ± 13 years), suggesting dispersal via recent flooding or undocumented human‐mediated transport. The downstream cross‐watershed event appears to be considerably older (577 ± 265 years), suggesting a role for rare geological or climatological events. Alongside these strong signatures of both contemporary and historical gene flow, we found little evidence of impacts on presumably adaptive phenotypic differentiation, except perhaps in the one instance of very recent cross‐watershed gene flow. Selection in this system seems to overpower gene flow—at least on the spatiotemporal scales investigated here.
Populations that are asymmetrically isolated, such as above waterfalls, can sometimes export emigrants in a direction from which they do not receive immigrants, and thus provide an excellent opportunity to study the evolution of dispersal traits. We investigated the rheotaxis of guppies above barrier waterfalls in the Aripo and Turure rivers in Trinidad—the later having been introduced in 1957 from a below-waterfall population in another drainage. We predicted that, as a result of strong selection against downstream emigration, both of these above-waterfall populations should show strong positive rheotaxis. Matching these expectations, both populations expressed high levels of positive rheotaxis, possibly reflecting contemporary (rapid) evolution in the introduced Turure population. However, the two populations used different behaviors to achieve the same performance of strong positive rheotaxis, as has been predicted in the case of multiple potential evolutionary solutions to the same functional challenge (i.e., “many-to-one mapping”). By contrast, we did not find any difference in rheotactic behavior above versus below waterfalls on a small scale within either river, suggesting constraints on adaptive divergence on such scales.