Upright and inverted unfamiliar face-matching tasks - everything correlates everywhere all at once.

In a key study, Megreya and Burton (Memory & Cognition, 34, 865-876, 2006) argued that identity-matching tasks using unfamiliar faces may not effectively measure general 'real-world' face-processing ability - that is they are "not faces". They observed a high correlation in performance between upright and inverted unfamiliar face matching, a pattern not seen with familiar faces, which they interpreted as indicating unfamiliar face matching is qualitatively different and largely driven by image-specific factors. However, the authors cautioned that this limitation likely applies only to unfamiliar face-matching tasks for identity rather than other types of face judgements (e.g., emotion). The present study replicates and extends these findings by considering within-subject performance for upright/inverted unfamiliar face matching across various paradigms (sequential/simultaneous presentation or sorting) and face-judgement types (identity or emotion), whilst considering different types of measures (accuracy and reaction time). Our results illustrated high correlations for upright/inverted conditions were universally observed within tasks for both accuracy and reaction times. Subsequent factor analyses indicated that upright and inverted conditions loaded together into task-specific latent variables. These results concur with the conclusions of Megreya and Burton (2006) and extend to both identity and emotion matching tasks - that is such tasks exhibit low construct validity for testing hypotheses about much general 'everyday' face processing. We propose that researchers should carefully consider alignment between their test materials and the theoretical 'constructs' they aim to measure, ensuring more accurate and meaningful interpretations of their results.