The Vanderbilt Holistic Face Processing Test (VHPT-F): A Short and Reliable Measure of Holistic Processing

摘要

Studying individual differences requires measures with sufficient reliability, often ignored in high-level vision. Holistic processing is central to research on face recognition and, more recently, to the study of individual differences in this area. A popular measure of holistic processing, the composite task, is highly sensitive in group studies (Richler & Gauthier, 2014), but shows low reliability (~.2; DeGutis et al., 2013; Ross et al., 2014). We present a measure of holistic face processing, the Vanderbilt Holistic Face Processing Test (VHPT-F), specifically designed for individual differences research. As in the composite task, subjects try to match a target face segment while ignoring the rest of the face, and holistic processing is measured as a failure of selective attention. The VHPT-F adopts a 3-AFC design, with variability among trials to increase discriminability at various levels of holistic ability. The VHPT-F 1.0 was more reliable than the composite task (.41) and produced the expected interaction between holistic processing and alignment (?·2p=.61, p .001). While misaligned trials are critical to establishing the measurea??s validity, we found little shared variance between congruency effects on aligned and misaligned trials (r=.14); regressing out misaligned performance does not affect the measure of holistic processing and was deemed unnecessary. The VHPT-F 2.0 with only aligned trials had a large effect size (?·2p=.75) and higher reliability (.56). The VHPT-F appears to measure a stable trait, with test-retest reliability of .52 after 3 weeks. Holistic processing in the VHPT-F and composite task were nearly as correlated (r=.28; n=136) as might be expected by the reliability of the composite task, suggesting they measure the same ability. Surprisingly, the VHPT-F did not correlate with the Cambridge Face Memory Test (CFMT) (r=-.17; n=97), suggesting that either holistic processing does not contribute to face recognition, or that the CFMT promotes the use of other strategies.