Caitlin Long¹, Lei Yuan¹, Claudia Wu¹, Ipek Oruc¹; ¹The University of British Columbia

Face recognition accuracy under blur improves when faces are presented at smaller sizes. We term this unexpected advantage the blur paradox, which has been replicated in studies where digitally scaled and blurred face images produce identity adaptation aftereffects at small, but not large, sizes. Furthermore, the blur paradox has been demonstrated in real-world settings, where intact images are viewed through physical blurring filters and viewing distance is varied to adjust for size. Here, we examine whether the blur paradox arises in a linear or categorical manner by testing over a broader range of viewing distances. Forty-four participants (34 females, mean age = 32.7±14.3, range: 19-74 years) completed a face recognition task in which celebrity faces were presented at four viewing distances – 160cm, 295cm, 543cm, and 1000cm – corresponding to face widths of 10.4°, 5.6°, 3.1°, and 1.7°, respectively. Trials were blocked by distance with 25 of 100 celebrities viewed at each distance and the order counterbalanced across participants. Faces were viewed through a Luminit holographic 3.5-degree Light Shaping Diffuser® positioned at a fixed distance in front of the screen, ensuring a consistent blur level. Participants attempted identifying the celebrity and if unsuccessful, were shown an intact image. Trials where the celebrity was unfamiliar to the participant were excluded from analysis. We found a significant main effect of viewing distance (F(3,129)=16.06,p<<.001,η_p^2=0.27,f=0.61), with highest accuracy at the farthest (M=70.80%) and lowest accuracy at the closest distance (M=54.42%,p<<.001,d=0.91), with a gradual, rather than categorical transition in between. Exploration of participant demographics suggested a stronger effect among older participants (>50 years) and a weaker effect among left-handed participants. No significant sex differences were observed. These findings confirm the small-size advantage for recognition under blur and its persistence in ecologically valid conditions, with accuracy improving gradually over a wide range of distances.

Acknowledgements: This work was supported by a Natural Sciences and Engineering Research Council of Canada (NSERC) USRA award (LY), an NSERC Discovery Grant RGPIN-2019-05554 (IO) and an Accelerator Supplement RGPAS-2019-00026 (IO).