It was suggested that the dual process models of attitude change should be extended to include the biased processing of strong attitudes. The main hypothesis of the extended model is that too much involvement intrinsic in strong attitudes may hinder objective processing, resulting in resistance to change even under strong message. Both attitude change and cognitive response measures in a 3 (attitude strength) x 2 (message quality) factorial design experiment supported the extended model. Only the holders of moderate attitudes showed greater attitude change when given a strong, rather than a weak, message. When given a strong message, holders of strong attitudes showed a boomerang effect by generating relatively greater counter-arguments (implying a central but biased processing with high motivation) in contrast with holders of weak attitudes who generated indifferent appeals and greater change in attitude regardless of the quality of the argument (implying a peripheral processing with low motivation).