In many natural domains, risks and rewards are inversely related (Pleskac & Hertwig, 2014). We sought to understand how people might use this relationship in choosing among risky gambles. To do so we, manipulated risk-reward structures of monetary gambles to be either negatively or positively correlated, or uncorrelated. After substantial exposure to these environments, participants completed a speeded choice task among non-dominated gambles. Eye-tracking data from this task suggests that participants often shifted their attention to mainly one attribute in the correlated conditions, in which the risk-reward relationship was present. This was an adaptive strategy that resulted in a similar proportion of expected-value maximizing choices, compared to a more compensatory processing strategy.