Research Report Subliminal Smells Can Guide Social Preferences Wen Li, 1,2 Isabel Moallem, 1 Ken A. Paller, 1,2 and Jay A. Gottfried 1,2,3 1 Cognitive Neurology and Alzheimer’s Disease Center, 2 Department of Psychology, and 3 Department of Neurology, Northwestern University ABSTRACT—It is widely accepted that unconscious pro- cesses can modulate judgments and behavior, but do such influences affect one’s daily interactions with other people? Given that olfactory information has relatively direct ac- cess to cortical and subcortical emotional circuits, we tested whether the affective content of subliminal odors alters social preferences. Participants rated the likeability of neutral faces after smelling pleasant, neutral, or un- pleasant odors delivered below detection thresholds. Odor affect significantly shifted likeability ratings only for those participants lacking conscious awareness of the smells, as verified by chance-level trial-by-trial performance on an odor-detection task. Across participants, the magnitude of this priming effect decreased as sensitivity for odor de- tection increased. In contrast, heart rate responses tracked odor valence independently of odor awareness. These re- sults indicate that social preferences are subject to influ- ences from odors that escape awareness, whereas the avail- ability of conscious odor information may disrupt such effects. The power of smell to influence social preference has been recognized since antiquity: When the women of Lemnos refused to pay tribute to the goddess Aphrodite, they were cursed with such a horrid stench that their husbands took new wives. Widespread use of perfumes centers on the assumption that pleasant fragrances enhance attractiveness. Supraliminal odors (i.e., odors above the threshold of detection) can regulate mood, cognition, and perhaps even mate selection (Herz & Schooler, 2002; Jacob, McClintock, Zelano, & Ober, 2002). Weak ambient scents, which participants do not explicitly notice, can also influence implicit odor memory (Degel & Ko ¨ster, 1999; Ko ¨ster, Degel, & Piper, 2002). However, it is unclear whether odors must be consciously perceptible in order to affect human be- havior, and in particular social preferences. It has long been known that subliminal affective information in visual stimuli can modify social judgments (Fazio, 2001). Studies on affective priming indicate that emotionally charged pictures or words presented subliminally alter subsequent preference judgments in valence-specific directions (e.g., Li, Zinbarg, Boehm, & Paller, in press). Two prominent theories, affective primacy (Zajonc, 1984) and feeling-as-information (Schwarz & Clore, 1996), have been proposed to explain these effects. By either account, an affective stimulus (i.e., prime) evokes an emotional response that is carried over to the pro- cessing of a subsequent stimulus (i.e., target), modifying affec- tive evaluation of the latter stimulus. Given the tendency of smells to induce potent emotional responses (Schiffman, 1974), and given the intimate anatomical connection of the olfactory system with limbic brain regions involved in affective process- ing (Carmichael, Clugnet, & Price, 1994; Gottfried, 2006), it stands to reason that subliminal odors may be particularly ca- pable of influencing social affective evaluations. Nevertheless, recent attempts to demonstrate unconscious effects of odors on social preferences have been unsuccessful. These studies, however, did not test the full range of odor va- lence, and the inclusion of only a pleasant odor (Bensafi et al., 2002) or an affectively ambiguous pheromone-like steroid (Lundstrom & Olsson, 2005) might not allow for clear demon- stration of unconscious affective effects of odors. Additionally, the studies relied on subjective self-reports (Bensafi et al., 2002) and predetermined odor thresholds (Lundstrom & Olsson, 2005) to exclude odors that were consciously perceived, and these methods may not be sufficiently rigorous to curtail sensory awareness (Hannula, Simons, & Cohen, 2005). Conscious odor perception might have occurred on individual trials or in some of the participants, triggering strategic control mechanisms, such as cognitive ‘‘discounting’’ (Kelley, 1973), and thereby dimin- ishing any observable effects. For example, Murphy and Zajonc Address correspondence to Wen Li, Northwestern University Fein- berg School of Medicine, Cognitive Neurology and Alzheimer’s Dis- ease Center, 320 E. Superior St., Searle 11-453, Chicago, IL 60611, e-mail: wenli@northwestern.edu. PSYCHOLOGICAL SCIENCE 1044 Volume 18—Number 12 Copyright r 2007 Association for Psychological Science