This dissertation aims to differentiate two types of boomerang effects on belief and attitude change: a boomerang effect under a proattitudinal message and a boomerang effect under a counterattitudinal message. By employing a 2 (Message valence: anti-policy vs. pro-policy) × 2 (Issues: legal age for drinking vs. legal age of marriage) × 2 (Threat to freedom: low threat vs. high threat) × 2 (Argument quality: low quality vs. high quality) plus 2 (Control groups: no-message control for the two issues) cross-sectional factorial design (N = 458), antecedents and mediators that bring about the two types of boomerang effect were examined. Under a counterattitudinal message, both argument quality and prior belief strength predicted a boomerang effect: Those receiving a low-quality argument or those with a strong prior belief, as compared with the control group, exhibited a boomerang on belief and attitude. The dominant mechanism that explained the relationship between argument quality and belief position boomerang was counterarguing (vs. anger). Under a proattitudinal message, there was an indirect effect of trait reactance on belief boomerang through anger (vs. negative cognitions). But the perceived threat to attitudinal freedom did not predict a boomerang effect. These results contribute to attitude change research by empirically separating cognitive and affective mechanisms for boomerang effects. Furthermore, this study refines the construct of negative cognitions and integrates reactance theory and the cognitive response perspective on boomerang effects. Both structural equation models and confirmatory factor analysis suggested that counterarguments and nonrefutational thoughts were two distinct types of negative cognitions. The two constructs were caused by different sets of antecedents and had different outcomes: Poor argument quality caused counterarguments, whereas perceived threat and trait reactance caused nonrefutational thoughts. Only counterarguments mediated the effects of argument quality on the boomerang effects for belief (e.g., the extent to which the legal drinking age should be decreased on a magnitude scale) and belief position (e.g., the legal age for drinking), which subsequently predicted the boomerang effect on attitude (e.g., the extent to which the legal drinking age is liked). This dissertation expands the theoretical scope of belief and attitude change research. Future research should explore the persuasive appeals for mitigating the cognitive or affective process resulting in a boomerang effect. Among those who are more prone to boomerang on certain issues, a boomerang appeal can be employed to persuade.