This study investigated modulation of risky-choice framing (RCF) and attribute framing (AF) effects by numeral modi-fication. In Experiment 1, in which the numerals were mod-ified with (the German equivalent of) exactly to enforce a precise reading, there were significant RCF and AF effects. Experiment 2 and 3 addressed the effects of (the German equivalents of) at most and up to. Both modifiers set an upper bound. Yet, they exhibit a sharp contrast in evalua-tive contexts. In Experiment 2, there was a significant inter-action of modifier and frame for RCF, with a reversed framing effect for at most and a standard framing effect for up to. The modifier-by-frame interaction effect was repli-cated in Experiment 3 for AF. To explain framing effects with bare and modified numerals, we propose a semantic-pragmatic account in terms of salience and valence.