How radial biases influence fast saccades towards faces in the periphery

Saccadic choice studies showed that humans initiate the fastest and most accurate saccades towards faces compared to other visual categories such as animals, objects and scenes. These studies typically present images along the horizontal meridian. Interestingly, the variations in contrast that specifically characterize a human face are oriented horizontally. In peripheral vision, the face-specialized visual mechanisms engage most strongly along the horizontal meridian due to radial bias facilitating access to the characteristic horizontal structure of the face. We therefore hypothesized that the horizontal set-up in past saccadic choice studies may have privileged the access to the horizontal structure characteristic of the human face, and therefore (partly) account for the saccadic advantage for faces reported so far. In a saccadic choice task, 14 human participants had to saccade towards a target (a face or a vehicle) presented at 15° of eccentricity along the horizontal and the vertical meridians. We expected to corroborate the saccadic advantage for faces along the horizontal meridian where the radial bias facilitates access to horizontal content, and to see a reduction of this advantage along the vertical meridian. Our preliminary analyses of the proportion and latency of correct saccades (saccadic reaction time, SRT; in milliseconds from the onset of stimuli) show that, across meridians, saccades towards faces were more accurate than saccades towards vehicles. While SRT were on average faster for faces than vehicles along both meridians, this mean SRT advantage for faces was largest along the horizontal meridian. Along the horizontal meridian, participants performed more correct than incorrect saccades towards faces from 140 ms poststimulus onset; they needed a longer time to correctly saccade towards vehicles (minimum SRT of 160ms). Saccadic accuracy dropped overall along the vertical meridian making minimum SRT results unreliable in this condition, especially with the small size of our sample. These preliminary findings suggest that the reduced access to horizontal information in the vertical meridian could decrease the saccadic advantage for faces. The horizontal set-up used in most past saccadic choice studies may have accentuated the saccadic advantage for faces due to the radial bias influencing the access to the oriented structure of the human faces. However, further data is needed to challenge the widely shared assumption of a general (non radially-biased) saccadic advantage for the processing of faces.