Reading the mind in the nose

Faces tell us a lot about their owners—or at least we like to believe they do (Todorov, 2017). We infer dominance, trustworthiness, attractiveness and emotional states from facial looks. Which parts of the face do we use for such sweeping conclusions? Most research has focused on the eye and mouth regions (e.g., Kim et al., 2022), even though average and expert face observers tend to look somewhere in between (Linka et al., 2022; Peterson & Eckstein, 2012). In fact, we ourselves plead guilty to this oversight (Broda & de Haas, 2022a, 2022b; de Haas et al., 2016, 2021; de Haas & Schwarzkopf, 2018). But there is more to faces. Following intensive examination of ourselves and others, we discovered: Noses. Noses are as individually different as the traits and states of their wearers. They exist in all shapes and sizes, can smell and even run.

Here, we report that participants can rate expressions and traits from isolated nose regions. In an online experiment, 114 participants rated isolated nose, eye and mouth regions of 30 frontal face images that were taken from the fLoc functional localizer package (Stigliani et al., 2015). All participants gave informed consent and were compensated with course credits. The study was approved by the institutional review board and in accordance with the declaration of Helsinki. Ratings of face parts proceeded in random order and along five dimensions. This was followed by ratings of the corresponding full-face images. We correlated the average rating of each face part with that for the corresponding full-face ratings in turn. As can be seen in Figure 1a and b, nasal ratings showed significant consistency with those for full faces (all R²= 20–54%). The extreme examples shown for each dimension may serve to convince the reader of the expressive power of the nose. While we fully concede that some of this effect may be due to parts of the region outside nose proper (such as the cheek), we maintain its expressive force may extend to the nose itself. Consider Figure 1c, which shows the results of a ground truth validation. The senior author employed a method acting approach to enter states of low or high valence by imagining desk rejection or immediate acceptance of this manuscript. This intervention led to remarkable engagement of the central face region, clearly including the nose itself. Observers of the nose were indeed able to track the underlying mental states, as reflected in the stark difference of their valence ratings for the positive vs. negative expression. Readers taking a liking towards null hypothesis significance testing (Cohen, 1994) may be excited to learn that MATLAB 2021a (Natick, MA) was incapable of computing the precise p-value for the corresponding paired t-test (t > 20) because p was too low. The consistency for eyes and mouths tended to be higher than for noses (though this difference failed to reach statistical significance in most cases, probably due to limited power; Table S1). Interestingly, nose consistencies were higher for valence, arousal, and attractiveness than the two personality dimensions. Previous research indicates valence can serve as a proxy for trustworthiness (Oosterhof & Todorov, 2008). Indeed, valence and trustworthiness ratings were correlated for full faces (r = .78, p < .001), mouths (r = .79, p < .001), eyes (r = .67, p < .001), and noses (r = .77, p < .001). Overall, we conclude that nose regions convey a surprising degree of information about their bearers.