Limits...
Stimulus requirements for face perception: an analysis based on "totem poles".

Paras CL, Webster MA - Front Psychol (2013)

Bottom Line: This allowed us to examine the prominence and properties of different features and their necessary configurations.Moreover, the prominence of eyes depended primarily on their luminance contrast and showed little influence of chromatic contrast.This suggests that the requisite trigger features are sufficient to holistically "capture" the surrounding noise structure to form the facial representation.

View Article: PubMed Central - PubMed

Affiliation: Department of Psychology, University of Nevada Reno, NV, USA.

ABSTRACT
The stimulus requirements for perceiving a face are not well defined but are presumably simple, for vivid faces can often by seen in random or natural images such as cloud or rock formations. To characterize these requirements, we measured where observers reported the impression of faces in images defined by symmetric 1/f noise. This allowed us to examine the prominence and properties of different features and their necessary configurations. In these stimuli many faces can be perceived along the vertical midline, and appear stacked at multiple scales, reminiscent of "totem poles." In addition to symmetry, the faces in noise are invariably upright and thus reveal the inversion effects that are thought to be a defining property of configural face processing. To a large extent, seeing a face required seeing eyes, and these were largely restricted to dark regions in the images. Other features were more subordinate and showed relatively little bias in polarity. Moreover, the prominence of eyes depended primarily on their luminance contrast and showed little influence of chromatic contrast. Notably, most faces were rated as clearly defined with highly distinctive attributes, suggesting that once an image area is coded as a face it is perceptually completed consistent with this interpretation. This suggests that the requisite trigger features are sufficient to holistically "capture" the surrounding noise structure to form the facial representation. Yet despite these well articulated percepts, we show in further experiments that while a pair of dark spots added to noise images appears face-like, these impressions fail to elicit other signatures of face processing, and in particular, fail to elicit an N170 or fixation patterns typical for images of actual faces. These results suggest that very simple stimulus configurations are sufficient to invoke many aspects of holistic and configural face perception while nevertheless failing to fully engage the neural machinery of face coding, implying that that different signatures of face processing may have different stimulus requirements.

No MeSH data available.


Related in: MedlinePlus

An example of images with symmetric luminance contrast and random chromatic contrast (left), or symmetric color and random luminance (right).
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC3569666&req=5

Figure 8: An example of images with symmetric luminance contrast and random chromatic contrast (left), or symmetric color and random luminance (right).

Mentions: We also generated a further set of images to examine whether the chromatic contrast of the features influenced the face percepts. In this case the noise was created by randomly sampling the independent RGB values of each pixel, and then filtering the amplitude spectra to 1/f as before. The images were constructed to be symmetric in luminance while random in color or vice versa (Figure 8). A third set was constrained to be photometrically equiluminant. Nine new undergraduate observers were tested with this stimulus set. For these images the face percepts were very strongly dominated by the luminance contrast, such that observers again chose eyes based on regions of dark luminance levels (Figure 9). As a result, the feature locations were strongly symmetric when the luminance was symmetric, and asymmetric when the symmetry was instead confined to chromatic variations. For example, in Figure 8, a strong impression of symmetric faces is evident in the luminance symmetric noise, and is seemingly unaffected by the salient yet random variations in chromatic contrast. Conversely, this percept again breaks down when the symmetry is instead carried by the chromatic contrast. Nevertheless, when luminance was nominally removed (in the equiluminant images), the faces seen were again aligned to the axis of symmetry (Figure 9 right panel). This suggests that color itself could support some weak perception of faces. However, another possibility is that the choices were, instead, based on residual luminance variations in the images. Either way, the poor cue provided by color is consistent with deficits in face recognition as well as many other higher-order spatial judgments including shading and symmetry detection, when the information is carried by color rather than luminance (Gregory, 1977; Cavanagh, 1987; Morales and Pashler, 1999).


Stimulus requirements for face perception: an analysis based on "totem poles".

Paras CL, Webster MA - Front Psychol (2013)

An example of images with symmetric luminance contrast and random chromatic contrast (left), or symmetric color and random luminance (right).
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC3569666&req=5

Figure 8: An example of images with symmetric luminance contrast and random chromatic contrast (left), or symmetric color and random luminance (right).
Mentions: We also generated a further set of images to examine whether the chromatic contrast of the features influenced the face percepts. In this case the noise was created by randomly sampling the independent RGB values of each pixel, and then filtering the amplitude spectra to 1/f as before. The images were constructed to be symmetric in luminance while random in color or vice versa (Figure 8). A third set was constrained to be photometrically equiluminant. Nine new undergraduate observers were tested with this stimulus set. For these images the face percepts were very strongly dominated by the luminance contrast, such that observers again chose eyes based on regions of dark luminance levels (Figure 9). As a result, the feature locations were strongly symmetric when the luminance was symmetric, and asymmetric when the symmetry was instead confined to chromatic variations. For example, in Figure 8, a strong impression of symmetric faces is evident in the luminance symmetric noise, and is seemingly unaffected by the salient yet random variations in chromatic contrast. Conversely, this percept again breaks down when the symmetry is instead carried by the chromatic contrast. Nevertheless, when luminance was nominally removed (in the equiluminant images), the faces seen were again aligned to the axis of symmetry (Figure 9 right panel). This suggests that color itself could support some weak perception of faces. However, another possibility is that the choices were, instead, based on residual luminance variations in the images. Either way, the poor cue provided by color is consistent with deficits in face recognition as well as many other higher-order spatial judgments including shading and symmetry detection, when the information is carried by color rather than luminance (Gregory, 1977; Cavanagh, 1987; Morales and Pashler, 1999).

Bottom Line: This allowed us to examine the prominence and properties of different features and their necessary configurations.Moreover, the prominence of eyes depended primarily on their luminance contrast and showed little influence of chromatic contrast.This suggests that the requisite trigger features are sufficient to holistically "capture" the surrounding noise structure to form the facial representation.

View Article: PubMed Central - PubMed

Affiliation: Department of Psychology, University of Nevada Reno, NV, USA.

ABSTRACT
The stimulus requirements for perceiving a face are not well defined but are presumably simple, for vivid faces can often by seen in random or natural images such as cloud or rock formations. To characterize these requirements, we measured where observers reported the impression of faces in images defined by symmetric 1/f noise. This allowed us to examine the prominence and properties of different features and their necessary configurations. In these stimuli many faces can be perceived along the vertical midline, and appear stacked at multiple scales, reminiscent of "totem poles." In addition to symmetry, the faces in noise are invariably upright and thus reveal the inversion effects that are thought to be a defining property of configural face processing. To a large extent, seeing a face required seeing eyes, and these were largely restricted to dark regions in the images. Other features were more subordinate and showed relatively little bias in polarity. Moreover, the prominence of eyes depended primarily on their luminance contrast and showed little influence of chromatic contrast. Notably, most faces were rated as clearly defined with highly distinctive attributes, suggesting that once an image area is coded as a face it is perceptually completed consistent with this interpretation. This suggests that the requisite trigger features are sufficient to holistically "capture" the surrounding noise structure to form the facial representation. Yet despite these well articulated percepts, we show in further experiments that while a pair of dark spots added to noise images appears face-like, these impressions fail to elicit other signatures of face processing, and in particular, fail to elicit an N170 or fixation patterns typical for images of actual faces. These results suggest that very simple stimulus configurations are sufficient to invoke many aspects of holistic and configural face perception while nevertheless failing to fully engage the neural machinery of face coding, implying that that different signatures of face processing may have different stimulus requirements.

No MeSH data available.


Related in: MedlinePlus