Search
Presentation Details
| Poster #Cognition Reevaluating Odor Mixtures: Evidence for Predominant Linearity |
| Robert Pellegrino1, Jennifer Margolis1, Carissa Evans1, Mathew Andres1, Emily Mayhew2, Alex Wiltschko3, Rick Gerkin3, Joel Mainland1,4 1Monell Chemical Senses Center, Philadelphia, PA, United States 2Michigan State University, East Lansing, MI, United States 3Osmo, New York, NY, United States 4Department of Neuroscience, University of Pennsylvania, Philadelphia, PA, United States |
Recent models have shown that physicochemical properties of individual molecules can be used to accurately predict perception. But predicting the perception of natural odors, which often consist of complex mixtures of chemicals, remains elusive. It has been reported that odor quality can shift dramatically depending on the specific combination of odorants, even when individual components remain constant, with one example being configural perception, where the combination of odorants gives rise to a new, emergent perceptual quality distinct from the qualities of the individual components. However, linear models, which assume no emergent effects, have performed well at predicting odor mixture similarity in human behavioral studies, raising the question of how often odor mixtures exhibit linear behavior. To test this, we collected descriptive ratings for a large, diverse set of mixtures (N = 706) and their components (N = 524). Only one out of 706 mixtures (<0.14%) had a quality that could not be explained by a linear combination of the qualities of the components. To further challenge the linear model, we collected descriptive ratings for eight previously reported mixtures showing configural processing and their component odorants (e.g., caramel + strawberry = pineapple). These mixtures showed linear behavior and a linear model predicted odor perception with high accuracy (r = 0.82). We conclude that most odor mixtures fall within a region defined by linear mixing of their components, challenging prior assumptions in the field. This has broad implications for industries such as flavor and fragrance, as well as environmental monitoring and health, and suggests that good mixture perception models are possible without building in configural assumptions. |