Conjoint analysis, which aims to uncover the optimal combination of attributes influencing customer choice, is widely used by marketers to predict the success of new product and service introductions. In recent years, researchers have incorporated considerable mathematical sophistication into conjoint models and extended its domain to diverse areas such as pricing, market share, profitability, product positioning, distribution channels, and advertising. Despite these advances, the predictive power of conjoint applications is often compromised by response biases and measurement errors. The purpose of this research is to isolate and investigate the impact of one such bias that arises from the manner in which stimuli are presented to respondents. Based upon an appraisal of over four decades of conjoint studies in the major marketing journals, the authors make a case for the possible existence of two types of biases, i.e.: (1) stimulus joint presentation bias, when concept cards are shown simultaneously (side by side) to respondents, and (2) stimulus separate presentation bias, where cards are presented separately (one at a time). Two conjoint experiments were designed to investigate the effects of these biases on respondent choices. Results indicate that bias manifests itself in conjoint designs when there is a mismatch between presentation mode and respondents’ cognitive (evaluable) burden. Left unaddressed, stimulus presentation mode bias may: (1) have a deleterious effect on respondents’ choice behavior; and (2) compromize the predictive accuracy of conjoint models. The authors discuss several approaches that can account for and mitigate the negative impact of presentation mode biases on conjoint outcomes.