Classic decision theories assume that gamble payoff and probability are independent in subjective evaluation. This study argues that when a simple gamble is paired with a sure outcome, the independence assumption is violated. Important aspects of this study, unlike most others in the literature, were (1) no parametric assumptions were made in evaluating independence, and (2) participants were paid according their performance in an attempt to motivate to seek maximum gain and minimum loss. Experiment 1 used a new procedure, relays, to test independence. Then three dependent models, which used functional expressions, were proposed and evaluated in Experiments 1 and 2 to pinpoint a two-way interaction pattern in which the contrast along one attribute of a gamble, either payoff or probability, influences the evaluation of the other. The results showed that the contrast between a gamble payoff and a sure outcome is more likely to affect the evaluation of probability than the other way around. This phenomenon and the accompanying uncommon value and weighting functions may be explained by the task by which participants were encouraged to gain much and lose less by contingent final payment.