Human Face Recognition

Holistic or Feature Based?

After decades of research, it is still not entirely clear whether faces are recognized holistically or based on specific features (Brewer, 2010). There is evidence that it may in fact be both.
It has been shown that faces are a lot harder to recognize when the are presented upside down (Farah et al., 1998).


image
Do you recognize this person when her face is presented upside down? Click on the image to rotate it!

If face recognition were purely feature-based, that shouldn’t be the case since features remained intact and it’s just their presentation that changed. Additionally, in a study by Young et al. participants performed poorly at trying to recognize a celebrity based on just the top half of the face when it was combined with the bottom half of a different face (1987). Moreover, participants performed better at recognizing the top half when the two halves were misaligned, suggesting that we can't just tune out some of the features of a face and focus on others.

It's harder for people to recognize the top half of the face if the two halves are aligned.

However, it is also known that some facial features are more important depending on if the face is familiar or unfamiliar:
in the former case, internal features, such as shape of nose or eye color, have been shown to be more important, while external features such as face outline or hairstyle have been shown to be more important for unfamiliar faces (Ellis et al., 1979).

Stages of Face Recognition

Tsao & Livingstone (2008) suggest that human face recognition is comprised of the following stages:

  1. Face detection: Recognizing that the object in front of you is a face and not something else. Requires holistic processing accomplished by extracting features that a face has in common with other faces.
  2. Measurement and Categorization: Features of a face are measured and represented in a way that will allow for easy identification. May be compared to templates stored in memory
  3. Identification: once the features have been measured, the face’s identity is established.

Individual differences

image

Not all people are equally good at recognizing faces - with people who don’t have the ability to identify people by looking at their face at all (this condition is called prosopagnosia, and famous people such as Oliver Sacks and Jane Goodall had it) on one end and so-called super-recognizers (Remember supertasters? Same deal but with face recognition!), who are able to recognize faces after only having briefly seen them once. Duchaine et al.’s study provides evidence that face recognition ability is genetic: face recognition ability in identical twins was strongly correlated (0.70) whereas the correlation in fraternal twins was significantly lower (0.30).


Just like some people are better at recognizing faces than others, some faces lend themselves better to being recognize than others by being more distinctive. Valentine (1991) introduced the idea of ‘face-space’: faces are arranged in this space by the ease with which they are recognized. In the very center of the space is a person whose features are average on all parameters and the more distinctive someone is based on their features, the further away they are from the center of the space.

Mechanisms of Face Recognition

Much about the neural processes involved in face recognition can be learned by conducting studies with subjects whose visual processing is impaired in some way because of brain damage. It was determined this way that face recognition partially depends on the same mechanisms as other types of visual recognition such as object and scene recognition.