To select promising health messages, formative research has often relied on perceived message effectiveness (PME) scales assessing either of two related constructs, message perceptions (persuasive potential) and effects perceptions (potential for behavioral impact). We sought to examine their incremental criterion validity within a comparative framework. Participants were 703 U.S. adult smokers (ages \(\ge\) 21) who received anti-smoking or comparable control (littering) messages on their cigarette packs for 3 weeks. Structural equation models examined both PME constructs as simultaneous correlates of outcomes from the UNC Tobacco Warnings Model. Message perceptions demonstrated incremental criterion validity with attention, an early behavioral antecedent (\(\beta\) = 0.82, p < .001). Effects perceptions demonstrated incremental criterion validity with later behavioral antecedents (range \(\beta\) = 0.74–0.87, all p < .01) and quitting behaviors (\(\beta\) = 0.36–0.66, all p < .001). Formative research on anti-smoking messages may benefit from focusing on effects perceptions to characterize potential for behavior change.