Resolution of the meaning of a semantically ambiguous word requires knowledge about the space of possible meanings of that word, and the selection of a meaning in the light of available evidence and given situational constraints. As such, ambiguity resolution bears many similarities to decision making scenarios more generally. We report on an experiment exploring this analogy by applying some standard manipulations from the decision making literature to a semantic disambiguation task. We explore two particular proposals: (1) that depth of semantic processing can be cast as strategy selection reflecting a risk-sensitive effort-accuracy tradeoff, and (2) that thresholds for inference about meaning in context are situationally flexible and learnable via feedback. One robust property of decision making is people’s ability to use feedback in order to adjust responses to maximize payoffs. Participants completed a semantic entailment judgment task in which they received trial-by-trial feedback, and payoff matrices and decision thresholds were manipulated across conditions. We find an effect of risk, with participants employing different comprehension strategies depending on relative gains and losses. We also find that participants were in fact sensitive to varying decision thresholds and accurately adjusted their behavior to match the constraints on what qualified as a true conclusion in different conditions. We take these findings as preliminary evidence that ambiguity resolution in language can be modeled, at least in part, as involving more general decision processes.