Why does picture-plane inversion sometimes dissociate perception of features and spacing in faces, and sometimes not? Toward a new theory of holistic processing

Abstract

778Face recognition is critical for our daily social interactions. The ease with which we discriminate among the thousands of faces that we encounter throughout our life is remarkable, given the very similar structure and features that all faces share. An important contribution to this ability, which is found for upright but not inverted faces, is presumed to be a style of perceptual processing commonly referred to as holistic (Tanaka & Farah, 1993), configural ), or second-order relational (Diamond & Carey, 1986. This type of processing has classically been thought to include information about the spacing among the major face features but to exclude information about the shape and/ or color of these features. The goal of this review is to examine whether published data support this prevalent claim. We first will describe the theoretical question and present the basic terminology used in studies of holistic/ configural face processing. We then will critically examine the results of 22 studies that investigated the effect of picture-plane inversion on the processing of spacing and/ or features in faces. At the end, we will integrate these findings with the results of studies in which various other methodologies were used (e.g., childhood development, neuropsychology, neuroimaging, individual differences) and will suggest an alternative view about the nature of holistic/configural processing of faces. The Aspects of Facial Information Included in Holistic Processing for Upright FacesIt is well established that holistic/configural/secondorder relational processing operates only for upright faces and not for inverted (upside-down) faces. Classic evidence includes the composite effect (Young, Hellawell, & Hay, 1987), in which identifying a top half-face suffers interference from alignment of a competing-identity bottom half, as compared with a misaligned condition, and the part-whole effect (Tanaka & Farah, 1993), in which memory for a face part is much better in the context of the studied whole face than when presented alone. Both the composite and part-whole effects are obtained for upright faces but not inverted faces, and these and many other paradigms confirm a qualitative difference between the perceptual processing of faces in the upright and inverted orientations (for recent reviews, see McKone, in press;Rossion, 2008).An open theoretical question concerns the aspect or aspects of facial information that are included within the mental holistic/configural/second-order relational repre-© Classically, it has been presumed that picture-plane inversion primarily reduces sensitivity to spacing/ configural information in faces (distance between location of the major features) and has little effect on sensitivity to local feature information (e.g., eye shape or color). Here, we review 22 published studies relevant to this claim. Data show that the feature inversion effect varied substantially across studies as a function of the following factors: whether the feature change was shape only or included color/brightness...

Faces in Context: Does Face Perception Depend on the Orientation of the Visual Scene?

“…Importantly, the main effect of Face Orientation is largely consistent with the qualitative account of face inversion effect (Le Grand et al, 2001;McKone & Yovel, 2009;Rhodes et al, 1993;Rossion, 2009;Rossion & Boremanse, 2008), suggesting that upright faces are processed as such irrespective of the background scene orientation. However, when we took correct reaction time into account, the IE data suggested that congruency with the scene had a different impact on upright faces than it did on inverted faces; congruency facilitated performance when faces were upside down.…”

“…Moreover, by using a number of different photographs to represent each identity, we have eliminated any concern that participants were able to determine matching pairs based on image properties, rather than faces per se. Therefore, this method of using multiple photographs of highly familiar faces may prove useful in the future because it avoids both the problems normally associated with simultaneous matching tasks and those associated traditional recognition memory paradigms (for recent reviews, see McKone & Yovel, 2009;Rossion, 2009).…”

“…However, several other studies have cast some doubts on a strict qualitative interpretation of the face inversion effect (Collishaw & Hole, 2002;McKone & Yovel, 2009;Riesenhuber, Jarudi, Gilad, & Sinha, 2004;Riesenhuber & Wolff, 2009). A quantitative view of this phenomenon was first asserted by Valentine and Bruce (1988).…”

“…This classic observation (Hochberg & Galper, 1967;Yin, 1969) has been widely replicated (for a review, see McKone & Yovel, 2009) and has converged with many different neural markers (Kanwisher, Tong, & Nakayama, 1998;Rossion, Prieto, Boremanse, Kuefner, & Van Belle, 2012;Taubert, Van Belle, Vanduffel, Rossion, & Vogels, 2015;Yovel & Kanwisher, 2005), suggesting that faces are processed by orientation-dependent brain mechanisms in participants that have developed normally. Indeed, a number of studies have confirmed that equivocal performance with upright and inverted faces is associated with impaired or inefficient face recognition in people living with Prosopagnosia, a specific inability to recognize faces (de Gelder & Rouw, 2000;Farah, Wilson, Drain, & Tanaka, 1995;Righart & de Gelder, 2007).…”

Holistic processing of static and moving faces.

“…Answers to these questions will not only help reconcile the discrepancies between mixed findings about holistic processing of moving faces, but they will also help address one fundamental question about current theories of holistic processing-whether they can be generalized to moving stimuli or apply only to static faces and objects. Moreover, determining whether holistic and part-based processing are similarly involved between static and rigidly moving faces may also help clarify what is holistic representation of faces (McKone & Yovel, 2009;Piepers & Robbins, 2012;Rossion, 2013) and what type of facial information underlies holistic processing (Rossion, 2013;Zhao et al, 2016b). …”