I explore different aspects of face perception to contribute to the growing evidence on how faces are processed, encoded, stored and retrieved from memory.
I defended my doctoral thesis at the (International Max Planck Research School) in February 2011 and then continued working as a post-doctoral researcher in the group of the department for another year.
Until September 2013, I was then working at the , Seoul, Korea, in the Lab, studying the perception of own- and other race faces and the role of cultural differences in face perception.
From October 2013 to April 2016, I worked with , at and then at, as part of an , on the cognitive processes involved in the perception and recognition of familiar and unfamiliar faces (and why these relate to different processes).
Currently I am back at the MPI for a project on face memory, working with Isabelle Bülthoff and Mintao Zhao in the .
Eye Movements in Face Comparison: The Role of Sex, Task, and Symmetry
Armann, R., & Bülthoff, I. (2009). Gaze behavior in face comparison: The roles of sex, task, and symmetry. Attention, Perception and Psychophysics 71(5) 1107-1126.
Knowing where people look on a face provides an objective insight into the information entering the visual system and allows us to infer the cognitive processes involved in face perception. We recorded eye movements of human participants to identify the facial features that they considered informative when performing different face comparison tasks.
Fixations to the different features in a face are dependent on the task an observer is performing . Contrary to earlier studies, we only used unfamiliar face identities and while comparing two faces, observers could find all the information necessary to solve the task on the screen. Thereby, we wanted to investigate if the facial features that are considered diagnostic in single-face studies continue to be so, even when an internal memory representation of the faces under investigation is not required. Second, we were interested in how participants would compare two faces presented next to each other, e.g., using a feature-by-feature, versus a more holistic, strategy. Third, since there is evidence for sex-related behavioral differences in face perception , we compared male and female participants eye movements.
Observers viewing behavior and performance were examined in two tasks and using two types of face stimuli (sex morphs and identity morphs). Within each task, we parametrically varied the difficulty by varying the similarity of two faces in a pair. Participants either decided if two faces of a pair were same or different, or which face of a pair was more feminine than the other. Frequency, duration, and temporal sequence of fixations on previously defined areas of interest in the faces were analyzed (see Figure 1). Behavioral data and eye movements were then compared regarding tasks (same-different, which one is more feminine), facial dimensions under investigation (sex, identity), and observers sex.
Results and Conclusion
Our results reveal that the same parts of a face are considered informative in single-face studies and in face comparison in all tasks (i.e., diagnostic features differed between tasks, as has been shown before).
Eye movements and performance varied in both sex-related tasks that primarily differed in terms of instructions (same-different, or which one is more feminine). We propose that observers used different scanning strategies in both tasks a more feature-based strategy in the former one, while the faces were compared in a more holistic way in the latter one. Furthermore, surprising large asymmetries were found in viewing behavior as observers constantly compared the inner halves of both face stimuli, suggesting that they regarded the faces as symmetric or that at least they found this approach a useful heuristic. These results are in contrast to the general left hemi-field bias reported in numerous face recognition tasks. As to the sex differences, our data reveal that male observers looked more often at the cheeks and female observers scanned the eyes more, but only when the sex of the faces was a relevant feature in the task they were performing (see Figure 2). Such sex-differences in eye movements have not been reported before.
1. Pearson, A.M., Henderson, J.M., Schyns, P.G., & Gosselin, F. Task-dependent Eye Movements During Face Perception. Abstracts of the Psychonomic Society, 8, 84 (2003).
2. Rehnmann, J., & Herlitz, A. Women remember more faces than men do. Acta Psychologica, 124(3), 244-255 (2007).
Categorical Perception of Male and Female Faces and the Single-Route Hypothesis
Armann, R., & Bülthoff, I. (2012). Male and female faces are only perceived categorically when linked to familiar identities - and when in doubt, he is a male. Vision Research 63, 69-80.
Categorical perception (CP) is a fundamental cognitive process that enables us to sort similar objects in the world into meaningful categories with clear boundaries between them. CP has been found for high-level stimuli like human faces, more precisely, for the perception of face identity, expression and ethnicity [1,2,3]. For sex however, which represents another important and biologically relevant dimension of human faces, results have been equivocal so far [4,5].
Here, we reinvestigate CP for sex using newly created face stimuli to control two factors that in our opinion might have influenced the results in earlier studies. Our new stimuli are (a) derived from single face identities, so that changes of sex are not confounded with changes of identity information, and (b) normalized in their degree of perceived maleness and femaleness. The latter aspect is important because natural variations of perceived masculinity and femininity of faces might hide evidence of categorical perception, as they result in differing locations of the categorical boundary for each identity (see Figure 1).
Prior to the experiments, we collected extensive ratings of numerous original faces and sex morphs to create face stimuli with similarly perceived degrees of maleness and femaleness.
CP of sex was then tested with classical discrimination and classification tasks using the controlled male-female continua derived from those faces (Figure 1). To further investigate the role of face identity information in sex perception, participants were either naïve or went through a familiarization phase before testing. Familiarization was done with (a) average faces of both sexes (familiarization with sex information only), (b) with male and female faces not used in the CP tasks (sex and identity learned, but non-relevant identity information), or (c) with the face stimuli used in the subsequent CP tasks (relevant sex and identity information).
Results and Conclusion
We found no significant CP for the sex of faces in naïve participants. When faces are only manipulated in their sex  but not across identities , CP for sex was not found, even when the perceived degree of maleness and femaleness of the faces is controlled. Familiarization with sex information provided by faces other than those used in the testing did not result in significant CP, while familiarization with the faces used in the test did . Thus CP for sex appeared only when participants got familiarized with the identities but not with the perceptual sex range used in the subsequent tests. These results support the single-route hypothesis [7,8] stating that sex and identity information in faces is not processed in parallel (as was suggested in the classical Bruce & Young model of face perception ), but rather linked in a single route. Sex information might therefore not be processed independently from identity information. Interestingly, although the endpoint faces used to create the face stimuli had been normalized prior to the experiment, we still observed a consistent male bias in all categorization tasks. Thus, there might be other (more cognitive) causes for this bias, in addition to the stimulus-driven bias that is usually found in face stimuli deprived of non-face information such as hair.
1. Beale, J.M., & Keil, F.C. (1995). Categorical effects in the perception of faces. Cognition, 57(3), 217-239.
2. Calder, A.J., Young, A.W., Perrett, D.I., Etcoff, N.L., Rowland, D. (1996). Categorical perception of morphed facial expression. Visual Cognition. 3, 81-117.
3. Levin, D.T., Beale, J.M. (2000). Categorical perception occurs in newly learned faces, other-race faces, and inverted faces. Perception & Psychophysics 62(2), 386-401.
4. Campanella, S., Chrysochoos, A., & Bruyer, R. (2001). Categorical perception of facial gender information: Behavioural evidence and the face-space metaphor. Visual Cognition 8(2), 237-262.
5. Bülthoff, I., & Newell, F. N. (2004). Categorical perception of sex occurs in familiar but not unfamiliar faces. Visual Cognition, 11(7), 823-855.
6. Bruce, V., & Young, A. (1986). Understanding face recognition. The British Journal of Psychology, 77(3), 305-327.
7. Rossion, B. (2002). Is sex categorization from faces really parallel to face recognition? Visual Cognition, 9(8), 1003-1020.
8. Ganel, T., & Goshen-Gottstein, Y. (2002). Perceptual integrality of sex and identity of faces: further evidence for the single-route hypothesis. Journal of Experimental Psychology: Human Perception & Performance, 28, 85486.
9. Armann, R., & Bülthoff, I., under revision in Vision Research.
Sex Morph Continuum. Controlled morph continuum created by applying the sex vector on one single (here: female) face identity. Numbers indicate the percentage of contribution of the female face to create the morph, i.e., 100% is an original female face, 0% is a male face. Note that the female faces have been feminized prior to the experiment, based on rating experiments, hence the 0 - 140 (rather than 0 - 100) range.
Are faces coded using race-specific norms?
Armann, R., Jeffery, L., Calder, A.J., & Rhodes, G. (2011). Race-specific norms for coding face identity and a functional role for norms. Journal of Vision, 11(13:9) 1-14.
Recent models of face perception often adopt a framework in which faces are represented as points or vectors in a multidimensional space, relative to the average face that serves as a norm. This has been shown using electrophysiology and brain imaging, but also high-level face adaptation, a technique adopted from low-level psychophysics: Adaptation to complex stimuli like faces sharing a particular physical characteristic can induce a perceptual aftereffect in a similar manner to other more low-level aftereffects found for example for color or motion perception. Visual aftereffects have been observed following exposure to faces varying in identity , sex , ethnicity or race , expression , and to distorted faces .
The face space framework with the assumption of a norm-based representation can explain a number of empirical phenomena in face perception one of them is the other race effect. This effect is based on the finding that faces are easily and quickly classified by ethnicity , but that we are inefficient at discriminating between people whose faces do not share the morphological features we are used to see in everyday live [e.g., 6]. It suggests that our face processing system has been tuned very finely through visual experience: The mechanisms developed to recognize individual faces of our own race cannot be generalized efficiently to faces of another race. This is not surprising if one considers that face expertise needs long training and that adult face processing abilities are not fully developed until adolescence, as shown by a number of studies of face processing development [e.g., 7]. Valentines framework explains the other-race phenomenon in terms of face space dimensions that are chosen as most appropriate to discriminate efficiently between faces someone encounters during development. If one sees mainly faces of one race, one ends up with a face space defined by dimensions inappropriate for discriminating between faces from another race because those other-race faces might be best characterized by other typical dimensions. Consequently, in the absence of appropriate dimensions, other-race faces build a dense cluster of faces distant from the center, and are thus perceptually very similar to each other. The question now arises whether faces are in fact encoded relative to the average face of the whole population, or whether other-race faces might instead be grouped around an average face of their own.
The aim of the present experiment is to test whether race-selective norms (rather than a common general norm) are used to code the identity of a face. Jaquet and colleagues [8,9] could induce opposite aftereffects (by exposure to either compressed or expanded facial features) simultaneously in Asian and Caucasian faces, and reasoned that Asian and Caucasian faces belong to perceptually distinct subgroups and thus might be encoded relative to distinct norms.
If a face is described in terms of its deviations from an average face, for example as having a longer nose and bigger eyes than average, then its opposite face (in terms of the face space framework), is a face with an inverse description for all its features, e.g., smaller nose and smaller eyes than average. One intriguing finding from face adaptation studies is that exposure to such an antiface shifts what observers perceive as average (i.e., midway between original face and antiface) towards that antiface with the effect that the actual average face is perceived as closer to the original face after adaptation [e.g.,10]. Testing the increase in perceived similarity between original face and average before and after adaptation to an antiface provides a measure of the adaptation aftereffect.
If faces are coded relative to race-specific norms, then the original face and the antiface derived by morphing through a race-specific average face should truly lie opposite in face space. Consequently, we would expect observers to show more pronounced aftereffects following adaptation to an antiface derived from a race-specific average face than to the one derived from a mixed average or an average of the other race . Alternatively, if faces are coded relative to a generic norm, then antifaces generated from the mixed average face will be opposite in face-space and their aftereffects will be largest.
We measured identity aftereffects for target faces and corresponding anti-faces (i.e., the faces that participants are being adapted to) that were opposite race-specific (Asian and Caucasian) averages and pairs that were opposite a generic average (both races morphed together).
Results and Conclusion
Aftereffects were larger for race-specific than for generic anti-faces. Since adapt-test pairs that lie opposite each other in face space generate larger aftereffects than non-opposite test pairs, these results suggest that Asian and Caucasian faces are coded using race-specific norms. Moreover, identification (at low identity strength) of the target faces was easier around the race-specific norms than around the generic norm, indicating that norms also have a functional role in face processing.
1. Leopold, D.A., OToole, A.J., Vetter, T., & Blanz, V. (2001). Prototype-referenced shape encoding revealed by high-level aftereffects, Nature Neuroscience 4, 8994.
2. Webster, M.A., Kaping, D., Mizokami, Y. Duhamel, P. (2004). Adaptation to natural face categories, Nature 428, 557560.
3. Rhodes, G., Jeffery, L., Watson, T., Jaquet, E., Winkler, C., & Clifford, C.W.G. (2004). Orientation-contingent face aftereffects and implications for face coding mechanisms. Current Biology 14, 2119-2123.
4. Webster, M.A., & MacLin, O.H. (1999). Figural aftereffects in the perception of faces. Psychonomic Bulletin & Review, 6, 647-653.
5. Levin, D.T. (1996). Classifying faces by race: the structure of face categories. Journal of Experimental Psychology 22(6), 1364-1382.
6. Bothwell, R.K., Brigham, J.C., & Malpass, R.S. (1989). Cross-racial identification. Personality and Social Psychology Bulletin 15(1), 19-25.
7. Schwarzer, G. (1997). Development of face categorization: The role of conceptual knowledge. Sprache & Kognition 16(1), 14-30.
8. Jaquet, E., Rhodes, G., & Hayward, W.G. (2007). Opposite aftereffects for Chinese and Caucasian faces are selective for social category information and not just physical face differences. Quarterly Journal of Experimental Psychology 60, 1457-1467.
9. Jaquet, E., Rhodes, G., & Hayward, W.G. (2008). Race-contingent aftereffects suggest distinct perceptual norms for different race faces. Visual Cognition 16, 734-753.
10. Rhodes G., & Jeffery, L. (2006). Adaptive norm-based coding of facial identity, Vision Research 46, 29772987.
* 12.09.1980 in Wuerzburg
- 07/2000 Abitur Uhlandgymnasium Tuebingen, Germany
- 04/2001-09/2005 Studies of biology, Johannes-Gutenberg University, Mainz, Germany Major: neurobiology, minors: developmental genetics, philosophy (ethics)
- 10/2003-08/2004 Exchange student at the Paul-Sabatier University, Toulouse, France (Erasmus-socrates-stipendiary) Program: neurosciences; internship in the Centre de Recherches sur la Cognition Animale, CRCA/CNRS
- 10/2005-7/2006 Diploma thesis at the Max-Planck-Institute for Biological Cybernetics, Tübingen, Germany (Department: Cognitive and Computational Psychophysics): "The Role of Eye Movement in Face Recognition"
- 2007-2010 Phd student at the Max-Planck-Institute for Biological Cybernetics, Tübingen, Germany (Department: Cognitive and Computational Psychophysics)
- 2011 PostDoc at the Max-Planck-Institute for Biological Cybernetics, Tübingen, Germany (Department: Cognitive and Computational Psychophysics)
- 2012-2013 PostDoc at the Department of Brain and Cognitive Engineering, Biological Cybernetics Lab, Korea University, Seoul
- 2013-2016 Research Fellow at Aberdeen University / York University, working with Mike Burton as part of the FACEVAR ERC grant
Organizational Unit (Department, Group, Facility):
- Alumni of the Department Human Perception, Cognition & Action
- Alumni of the Group Recognition & Categorization