Radial bias in face identification

Abstract

Peripheral vision in humans is characterized by a preference for radially oriented stimuli (pointing towards the fovea), which tend to be better perceived than tangentially oriented stimuli in a wide variety of tasks (e.g., experiments measuring contrast, spatial frequency, phase discrimination or orientation discrimination thresholds). This is consistent with the neurophysiological findings that the receptive fields of visual cells at low-level of processing are radially elongated, starting with the retinal ganglion cells. Despite evidence that the human visual system is radially biased, it is not known whether this bias also influences the visual recognition of meaningful objects. To address this question, we used the case of face identification, which have been shown to rely primarily on horizontally oriented content (e.g., edges of eyes and eyebrows, base of nose, mouth). We reasoned that, as a result of the radial bias, humans should be better at recognizing their conspecifics' faces when perceived along the horizontal meridian, where the encoding of horizontal orientation is facilitated. In a psychophysical experiment, we trained two samples of participants to recognize upright and inverted faces appearing briefly on either the horizontal or vertical meridian. We measured the effect of face inversion, taken here as a marker of the high-level ‘specificity’ of face identity processing, across a range of visibility levels. We found that the face inversion effect was larger when faces were perceived on the horizontal compared to the vertical meridian. Our results support the hypothesis that the high-level mechanisms in the face processing system are more easily engaged when faces are perceived on the horizontal meridian, where horizontal cues in the face align with the radial bias. This result suggests that low level radial biases in peripheral vision can influence the recognition of complex objects.

The resolution of face perception varies systematically across the visual field

“…Note first that the presence of a consistent inversion effect indicates that performance on our gender acuity task derives from processing within face-selective systems (at least in part), rather than stemming purely from lower-level limitations on vision. However, the size of this inversion effect did not vary significantly around the visual field, as one might expect if face-selective processes were themselves to vary, and contrary to the recent finding of larger inversion effects in an identity-recognition task on the horizontal meridian than the vertical [58]. It could be that while the strength of face-selective processing varies around the visual field (leading to variable inversion effects with fixed-size stimuli), the resolution of these face-selective processes does not.…”

“…Nevertheless, the anisotropies that we observe for face perception could be partly driven by variations in featural selectivity, including properties like the radial bias, whereby peripheral contrast sensitivity and orientation discrimination are better for stimuli oriented towards the fovea [53,54]. Because horizontal information is particularly informative for face recognition [55][56][57], this could improve performance on the horizontal vs. vertical meridian-either passively through the pooling of variations in low-level information, or more actively by boosting the response to the optimal orientations for high-level processing, as argued recently [58]. However, though these factors could contribute to the horizontal-vertical anisotropy, these radial variations are matched across the upper and lower fields, making them unable to explain the upper-lower anisotropy.…”

Radial bias in face identification

“…Raw data and analysis code are available at https://doi.org/10. 17605/OSF.IO/9VUT4 [52]. The same approach was used to analyse data from both groups of participants (left-up group and right-low group).…”

“…Data accessibility. The datasets and analysis scripts supporting this article can be downloaded from a public Open Science Framework repository: https://doi.org/10.17605/OSF.IO/9VUT4 [52].…”

The resolution of face perception varies systematically across the visual field

“…Nevertheless, the anisotropies that we observe for face perception could be partly driven by variations in featural selectivity, including properties like the radial bias, whereby peripheral contrast sensitivity and orientation discrimination are better for stimuli oriented towards the fovea [53,54]. Because horizontal information is particularly informative for face recognition [55][56][57], this could improve performance on the horizontal vs. vertical meridian -either passively through the pooling of variations in low-level information, or more actively by boosting the response to the optimal orientations for high-level processing, as argued recently [58]. However, though these factors could contribute to the horizontal-vertical anisotropy, these radial variations are matched across the upper and lower fields, making them unable to explain the upperlower anisotropy.…”