Loss aversion was long regarded as one of the most robust phenomena in the social sciences (Kahneman, 2011). Recent research questions the generalizability of loss aversion (e.g., Gal & Rucker, 2018). While loss aversion has mostly been tested in binary choice or accept/reject tasks, it appears less pronounced in decisions from experience (Ert & Erev, 2008; Yechiam & Hochman, 2013). In two experiments, we tested whether loss aversion generalizes to decision-from-experience paradigms in which participants sampled discretized draws from underlying continuous distributions. In the first experiment, we used subjective valuation tasks in which participants evaluated individual mixed lotteries containing gains and losses. In the second experiment, we used a binary choice paradigm in which participants sampled draws from two lotteries before choosing one of them to play. Lotteries were paired to have matched expected values but different outcome ranges, producing positive–positive, mixed–positive, and mixed–mixed combinations. We did not observe higher weights for losses than for gains in the valuation-from-experience task. In contrast, in choices from experience, participants preferentially selected positive-only lotteries and avoided those containing losses. Our findings suggest that loss aversion depends on the elicitation method (experienced-based choice versus valuation) adding to the literature documenting differences in decision behavior depending on the elicitation method (Bouchouicha et al., 2024; Lichtenstein & Slovic, 1971).