Limits...
How category structure influences the perception of object similarity: the atypicality bias.

Tanaka JW, Kantner J, Bartlett M - Front Psychol (2012)

Bottom Line: If similarity judgments are guided purely by their physical properties, the morph should be perceived to be equally similar to its typical parent as its atypical parent.In this framework, atypical stimuli are located in a sparser region of the space where there is less competition for recognition and therefore, these representations capture a broader range of inputs.These results suggest that the perceived likeness of an object is influenced by the organization of surrounding exemplars in the category space.

View Article: PubMed Central - PubMed

Affiliation: Department of Psychology, University of Victoria Victoria, BC, Canada.

ABSTRACT
Why do some faces appear more similar than others? Beyond structural factors, we speculate that similarity is governed by the organization of faces located in a multi-dimensional face space. To test this hypothesis, we morphed a typical face with an atypical face. If similarity judgments are guided purely by their physical properties, the morph should be perceived to be equally similar to its typical parent as its atypical parent. However, contrary to the structural prediction, our results showed that the morph face was perceived to be more similar to the atypical face than the typical face. Our empirical studies show that the atypicality bias is not limited to faces, but extends to other object categories (birds) whose members share common shape properties. We also demonstrate atypicality bias is malleable and can change subject to category learning and experience. Collectively, the empirical evidence indicates that perceptions of face and object similarity are affected by the distribution of stimuli in a face or object space. In this framework, atypical stimuli are located in a sparser region of the space where there is less competition for recognition and therefore, these representations capture a broader range of inputs. In contrast, typical stimuli are located in a denser region of category space where there is increased competition for recognition and hence, these representation draw a more restricted range of face inputs. These results suggest that the perceived likeness of an object is influenced by the organization of surrounding exemplars in the category space.

No MeSH data available.


Related in: MedlinePlus

Information maximization account of atypicality bias. (A) Probability density function for typical and atypical faces. (B) Perceptual response for typical, atypical and the typical-atypical morph faces as predicted by adaptive transfer function.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC3368386&req=5

Figure 7: Information maximization account of atypicality bias. (A) Probability density function for typical and atypical faces. (B) Perceptual response for typical, atypical and the typical-atypical morph faces as predicted by adaptive transfer function.

Mentions: Here we show how the information maximization principle of optimal perceptual coding would account for face typicality effects, and in particular the atypicality bias. We begin with the assumption that typical faces are from high density regions of the probability distribution for a set of physical characteristics, and that atypical faces have characteristics with lower probability on at least some dimensions, as illustrated in Figure 7A. For example, a particular physical characteristic might be distance between the eyes. Figure 7B illustrates the shape of the perceptual response function under the information maximization principle. In our example, this response function could be JND for facial identity as a function of interocular distance. The typical face is mapped to a region of the response function that is steep, and the atypical face is mapped to a region of the response function that is shallow. Due to the shape of the response function, the physical 50% morph between the typical and atypical face is projected closer to the atypical parent in the perceptual response, as shown by the dashed line onto Y-axis.


How category structure influences the perception of object similarity: the atypicality bias.

Tanaka JW, Kantner J, Bartlett M - Front Psychol (2012)

Information maximization account of atypicality bias. (A) Probability density function for typical and atypical faces. (B) Perceptual response for typical, atypical and the typical-atypical morph faces as predicted by adaptive transfer function.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC3368386&req=5

Figure 7: Information maximization account of atypicality bias. (A) Probability density function for typical and atypical faces. (B) Perceptual response for typical, atypical and the typical-atypical morph faces as predicted by adaptive transfer function.
Mentions: Here we show how the information maximization principle of optimal perceptual coding would account for face typicality effects, and in particular the atypicality bias. We begin with the assumption that typical faces are from high density regions of the probability distribution for a set of physical characteristics, and that atypical faces have characteristics with lower probability on at least some dimensions, as illustrated in Figure 7A. For example, a particular physical characteristic might be distance between the eyes. Figure 7B illustrates the shape of the perceptual response function under the information maximization principle. In our example, this response function could be JND for facial identity as a function of interocular distance. The typical face is mapped to a region of the response function that is steep, and the atypical face is mapped to a region of the response function that is shallow. Due to the shape of the response function, the physical 50% morph between the typical and atypical face is projected closer to the atypical parent in the perceptual response, as shown by the dashed line onto Y-axis.

Bottom Line: If similarity judgments are guided purely by their physical properties, the morph should be perceived to be equally similar to its typical parent as its atypical parent.In this framework, atypical stimuli are located in a sparser region of the space where there is less competition for recognition and therefore, these representations capture a broader range of inputs.These results suggest that the perceived likeness of an object is influenced by the organization of surrounding exemplars in the category space.

View Article: PubMed Central - PubMed

Affiliation: Department of Psychology, University of Victoria Victoria, BC, Canada.

ABSTRACT
Why do some faces appear more similar than others? Beyond structural factors, we speculate that similarity is governed by the organization of faces located in a multi-dimensional face space. To test this hypothesis, we morphed a typical face with an atypical face. If similarity judgments are guided purely by their physical properties, the morph should be perceived to be equally similar to its typical parent as its atypical parent. However, contrary to the structural prediction, our results showed that the morph face was perceived to be more similar to the atypical face than the typical face. Our empirical studies show that the atypicality bias is not limited to faces, but extends to other object categories (birds) whose members share common shape properties. We also demonstrate atypicality bias is malleable and can change subject to category learning and experience. Collectively, the empirical evidence indicates that perceptions of face and object similarity are affected by the distribution of stimuli in a face or object space. In this framework, atypical stimuli are located in a sparser region of the space where there is less competition for recognition and therefore, these representations capture a broader range of inputs. In contrast, typical stimuli are located in a denser region of category space where there is increased competition for recognition and hence, these representation draw a more restricted range of face inputs. These results suggest that the perceived likeness of an object is influenced by the organization of surrounding exemplars in the category space.

No MeSH data available.


Related in: MedlinePlus