English  Sprachen Icon  |  Gebärdensprache  |  Leichte Sprache  |  Kontakt


Many faces of face perception

Von Dr. Dragan Rangelov (14.10.2014)

Analytical processing is one of the most pertinent properties of cognition. Already at the very early processing stages, sensory input is processed in terms of presence/absence of a limited set of features, e.g., something red, blue, bright, moving or stationary. Only at the later, integrative stages do individual features get combined into complex, multi-feature percepts of, e.g., a vehicle. Analytical sensory processing is crucial for permitting the cognitive system to accurately process myriads of real-life objects by using a limited set of basic stimulus features and combining them flexibly. With the apparent benefits of processing sensory input analytically, it is rather surprising that some complex stimuli appear not to be processed in this way. One such stimulus class, at least as far humans are concerned, are human faces. In my work, I investigate which cognitive processes are involved in processing faces as compared to comparably complex stimuli.

Even a simple introspection demonstrates that we can discriminate two faces faster and more accurately, as well as remember faces more easily, relative to non-faces.  Over the past decades, numerous studies have confirmed this intuitive notion that human faces are truly a special stimulus class, superiorly processed relative to other complex stimuli. Recent studies have shown that faces are also able to capture attention automatically, suggesting that early sensory processes preceding focal attention are already modulated by having to process faces. In the visual attention literature the ability of a stimulus property to capture attention automatically implies existence of neural structures specifically tuned to this particular property; showing that a single red bar amongst blue bars captures attention implies existence of neurons specifically tuned to red. Consequently, showing that a single face amongst non-faces captures attention suggested that faces are processed similar to basic stimulus features, e.g., red. However, unlike basic features, faces are complex stimuli.

I investigated whether the neural correlates that support attentional capture of faces (i.e., face-detector neurons) encode only the global or both global and local face properties. If the face-detectors encoded both the global and local attributes, than a face-target differing only in its local features, e.g., a different mouth should also capture attention when presented amongst other faces. By contrast, for face-detectors encoding only the global properties, presenting many faces would activate many detectors, increasing overall activity levels. Importantly, this activity would not help solve the task and, consequently, it should be more difficult to find the face-target compared to non-faces for which no detectors exist. Figure 1 shows the paradigm in which a stimulus array is briefly presented comprising either faces or non-face stimuli. The stimuli can be presented in their normal configuration or scrambled. On some trials, one stimulus (target) differs from the rest in its local property, e.g., mouth/door color, the task being to detect target’s presence/absence.  In a series of behavioral studies I showed that detecting target is less accurate (as well as slower) for normal relative to scrambled face stimuli. By contrast, this local feature suppression effect is absent for non-face stimuli. Furthermore, electrophysiological investigation shows that early processes of target selection (PCN component) are slower for normal than scrambled faces. Also, post-selective distractor inhibition (Ptc) is weaker and target processing (SPCN) is more effortful for normal than scrambled faces. Again, no such effects are observed for non-face stimuli.

Rangelov: Fig. 1[Bildunterschrift / Subline]: Figure 1. Summary of the main research findings of several behavioral and electrophysiological studies.

Taken together, my findings show that only global, rather than both to global and local face properties attract attention automatically. This finding is important for two reasons. Firstly, it supports the notion that there are specialized neural structures specifically sensitive to combinations of features, similar to cells tuned to a single visual feature. Secondly, since globally-tuned face detectors cannot support superior face discrimination and memory for individual faces, my findings also imply that face processing is different from processing of non-faces in at least two ways. Face-detectors permit quick orientating of visual attention to any face in the visual field relative to non-faces. Furthermore, different neural structures would permit superior memory for individual faces relative to non-faces.

Finally, the local-feature suppression effect for faces I demonstrated might also have clinical applications. Prosopagnosia is a clinical condition where people have difficulties in identifying faces which usually adversely affects social lives of people with this condition. Importantly, the inability to process faces would predict weaker local-feature suppression for faces in the paradigm I developed. Put differently, clinical population should perform better in my task relative to control group. We are presently pursuing this line of research which, should it prove successful, might result in a diagnostic technique where clinical population achieves better scores relative to control group. Additionally, devising a behavioral measure sensitive to prosopagnosia would also permit finding neural correlates of this measure, and, by extension, help describe neural underpinnings of inability to process faces.

Scientific career
  • 2000-2005
  • Diploma in Psychology,University of Belgrade, Serbia
  • 2005-2007
  • M.Sc. in Neuro-Cognitive Psychology, Ludwig-Maximilians Universität, Munich
  • 2007-2010
  • PhD at Graduate School of Systemic Neuroscience, Ludwig-Maximilians-Universität, München
  • 2010-2012
  • Lecturer, Department of Psychology, Ludwig-Maximilians-Universität, Munich
  • since 2012
  • Temporary Principal Investigator: DFG project "Einfluss aufgabenirrlevanter Merkmale auf kognitive Kontrollprozesse"

Publikationen (Auszug)
  • * Rangelov. D., & Zeki, S. (2014). Non-binding relationship between visual features. Frontiers in. Human Neuroscience 8:749. doi: 10.3389/fnhum.2014.00749
  • * Zinchenko, A., Kim, H., Danek, A., Müller, H.J., & Rangelov, D. (2014). Local feature suppression effect in face and non-face stimuli, Psychological Research, 1 – 12, doi: 10.1007/s00426-014-0548-6
  • * Rangelov, D., Töllner, T., Müller, H.J., & Zehetleitner, M. (2013). What are task-sets: A single, integrated representation or a collection of multiple control representations?, Frontiers in Human Neuroscience, 7, doi: 10.3389/fnhum.2013.00524
  • * Rangelov, D., Müller, H. J., & Zehetleitner, M. (2013). Visual search for feature singletons: Multiple mechanisms produce sequence effects in visual search. Journal of Vision, 13(3):22, 1–16, doi:10.1167/13.3.22