Effect of inversion on the recognition of external and internal facial features
Introduction
One of the controversial issues regarding face recognition, which has important theoretical implications, concerns the question of whether or not inversion affects recognition of facial features. As a rule, when faces are inverted, recognition accuracy decreases (for a review, see Valentine, 1988). This decrement, called the “inversion effect”, is significantly bigger than a similar effect found for recognition of other inverted objects (Carey & Diamond, 1977; Diamond & Carey, 1986; Scapinello & Yarmey, 1970; Valentine & Bruce, 1986; Yarmey, 1971; Yin, 1969; but see Bruce, Doyle, Dench, & Burton, 1991). Whether or not this phenomenon demonstrates specificity of face perception is another controversial issue which has been dealt with elsewhere (de Gelder, Bachoud-Levi, & Degos, 1998; de Gelder & Rouw, 2000; Farah, Wilson, Drain, & Tanaka, 1995; Nachson, 1995).
One explanation of the inversion effect in face recognition is based on the assumption that upright faces are normally perceived configurationally (by considering the facial features as well as their spatial relationships), whereas inverted faces are recognized by componential analysis of the individual features (Bartlett & Searcy, 1993; Carey, 1978, Carey, 1981; Carey & Diamond, 1977; Endo, 1986; Frith, Stevens, Johnson, Owens, & Crow, 1983; Phillips & Rawls, 1979; Rhodes, Brake, & Atkinson, 1993; Sergent, 1984, Sergent, 1988; Tanaka & Farah, 1993; Vermeire & Hamilton, 1998; Yin, 1969). It was argued, that since inversion severely impairs configurational (but not componential) perception, recognition of inverted faces (but not of facial features) is impaired (Bartlett & Searcy, 1993; Bruyer & Coget, 1987; Carey, 1978, Carey, 1981; Leder & Bruce, 1998; Rhodes, Brake, Taylor, & Tan, 1989; Rhodes, Tan, Brake, & Taylor, 1989; Rhodes et al., 1993; Searcy & Bartlett, 1996; Tanaka & Farah, 1993; Valentine, 1988; Young, Hellawell, & Hay, 1987). Concerning configurational processing, Leder and Bruce (2000) recently argued that it should be defined in terms of spatial relationships (relational coding), rather than in terms of holistic processing (template-like coding), since only disruption of the former produces inversion effects. This notion is in line with findings showing that the qualitative changes in face processing which are involved in inversion of upright faces are due to encoding differences between the two orientations of the spatial–relational interactions among the internal features (Endo, 1986; Murray, Yong, & Rhodes, 2000; Rock, 1974; Sergent, 1984). However, Valentine, 1988, Valentine, 1991 argued that recognition of upright faces does not differ qualitatively from that of inverted faces, since accuracy differences between the two orientations is only a matter of degree.
The notion that recognition of facial features is unaffected by inversion was recently challenged by Rakover and Teucher (1997) who found that recognition of the forehead, eyes, nose, mouth and chin is less accurate when the features are inverted than when they are upright. According to the authors, this finding implies that configurational information extracted from a whole face is not necessary for obtaining inversion effects. This hypothesis was recently corroborated by Moscovitch and Moscovitch (2000) who found that inverted faces are not processed by piecemeal analysis of their features, since the spatial relationships of these features, as well as their orientation, are necessary for face recognition. If that is true, then the notion that upright and inverted faces are differentially processed by configural and featural analysis, respectively, should be revised.
Considering facial features, a distinction has been made between recognition of external (e.g., hair, ears, facial outline) and internal (e.g., eyes, nose, mouth) features. In a typical experiment on face matching, two facial stimuli, a target face and a test facial stimulus (consisting of either a full face, external features or internal features) are presented to the participants who are asked to tell whether they depict the same or different faces (de Haan & Hay, 1986; de Haan, Young, & Newcombe, 1987; Hines, Jordan-Brown, & Juzwin, 1987; Young, 1984; Young, Hay, McWeeny, Flude, & Ellis, 1985). Participants' responses are scored in terms of accuracy and reaction time. Differential results have been reported for matching familiar and unfamiliar faces (de Haan & Hay, 1986; Ellis, Shepherd, & Davies, 1979; Young et al., 1985). For the latter, matching of external features is usually faster and more accurate than matching of internal features (de Haan & Hay, 1986; Nachson, Moscovitch, & Umiltá, 1995; Young et al., 1985). Conflicting data exist regarding matches of isolated facial features with full faces. While some authors (Ellis et al., 1979; Hines et al., 1987; Phillips, 1979; Young et al., 1985, Experiment 1) have found no differences in accuracy rates and reaction times between the various features, other authors (Young et al., 1985, Experiment 2) reported shorter reaction times for matching external than internal features. Furthermore, when inverted, recognition of internal features is more impaired than that of external features (Phillips, 1979).
In the present study the interactive effects of feature and orientation on the recognition of facial stimuli were systematically investigated by matching upright target faces with three categories of facial test stimuli: full faces, external features and internal features which were presented in either upright or inverted orientations. It was predicted that face matching would be faster and more accurate under upright than under inverted orientations; mainly for full faces, less for external features, and least for internal features. These predictions were tested under two experimental conditions: pair matching (PM) and multiple choice (MC) matching.
Section snippets
Participants
24 female students at Bar-Ilan University were tested.
Stimuli
100 male faces were used in the two experiments; four target (full) faces, and 96 test faces which were divided into three series of 32 stimuli each: full faces, external features (hair, ears, and face outline) and internal features (eyes, nose and mouth) (see Fig. 1). Each series consisted of 16 facial stimuli in the upright orientation, and 16 identical stimuli in the inverted orientation. All stimuli were 3 cm long and 2.5 cm wide,
Results
For each participant, accuracy and reaction time scores were recorded for each task. Accuracy was measured in terms of percent correct responses. The data were analyzed by 2 Tasks (PM, MC) × 2 Orientations (upright, inverted) × 3 Modes (full face, external features, internal features) analysis of variance with repeated measurements. Additional analyses were performed where necessary.
Discussion
Overall, the data of the present study are consistent across tasks (PM/MC) and measures (accuracy/reaction time) regarding both, stimulus orientation (upright/inverted) and mode of presentation (full faces/external features/internal features).
Acknowledgements
The authors acknowledge the help given by Judith Abulafia in data analysis, and thank two anonymous reviewers for their helpful comments on an earlier draft of this paper.
References (53)
- et al.
Inversion and configuration of faces
Cognitive Psychology
(1993) - et al.
The influence of inversion on the judgment of facial and non-facial attributes
Acta Psychologica
(1997) - et al.
Remembering facial configurations
Cognition
(1991) - et al.
Features of laterally displayed faces: Saliency or top-down processing?
Acta Psychologica
(1987) - et al.
Inversion superiority in visual agnosia may be common to a variety of orientation polarised objects besides faces
Vision Research
(1998) - et al.
Paradoxical configuration effects for faces and objects in prosopagnosia
Neuropsychologia
(2000) - et al.
The inverted face inversion effect in prosopagnosia: Evidence for mandatory, face-specific perceptual mechanisms
Vision Research
(1995) - et al.
Hemispheric visual processing in face recognition
Brain and Cognition
(1987) Some explanatory experiments on memory for photographs of faces
Acta Psychologica
(1979)- et al.
What's lost in inverted faces?
Cognition
(1993)
The effect of race, inversion and encoding activity upon face recognition
Acta Psychologica
Inversion effects for faces in split-brain monkeys
Neuropsychologia
Upright versus upside-down faces: How interface attractiveness varies with orientation
Perception and Psychophysics
Models for the processing and identification of faces
Perception and Psychophysics
Lesions cérébrale unilaterale et troublé de la perception des visages: Spécificité du deficit?
Acta Neurologica Belgica
From piecemeal to configurational representation of faces
Science
Matching of facial features: Continuous processing, improper filtering, and holistic comparison
Perception and Psychophysics
Face recognition without awareness
Cognitive Neuropsychology
Why faces are and are not special: An effect of expertise
Journal of Experimental Psychology: General
Identification of familiar and unfamiliar faces from internal and external features: Some implications for theories of face recognition
Perception
Integration of schematic faces and other complex objects in schizophrenia
Journal of Nervous and Mental Disease
Cited by (23)
Contributions of individual face features to face discrimination
2017, Vision ResearchThe role of inner and outer face parts in holistic processing: A developmental study
2014, Acta PsychologicaCitation Excerpt :Research with adults has demonstrated that familiar faces are recognised more accurately from inner rather than outer features, but vice versa for unfamiliar faces (Ellis, Shepherd, & Davies, 1979; Frowd, Bruce, McIntyre, & Hancock, 2007; Jarudi & Sinha, 2003; Veres-Injac & Persike, 2009; Veres-Injac & Schwaninger, 2009; Young, Hay, & Ellis, 1985; Young, Hay, McWeeny, Flude, & Ellis, 1985). With respect to holistic vision, outer features are often found to be less affected by face inversion than inner features (Barton, Keenan, & Bass, 2001; Malcolm, Leung, & Barton, 2005; Meinhardt-Injac, Meinhardt, & Schwaninger, 2009; Moscovitch & Moscovitch, 2000; Nachson & Shechory, 2002; Phillips, 1979; Rakover & Teucher, 1997; Rhodes et al., 1993). These differences have led some authors to suggest that inner and outer features may be implicated in different sensory routes — with the former stimulating the face recognition system and the latter stimulating the object recognition system.
The time course of face matching by internal and external features: Effects of context and inversion
2010, Vision ResearchCitation Excerpt :Therefore, studying the perceptual dependency of face parts across the time continuum could help to decide among both alternative concepts of face perception, and could help to further elaborate the holistic approach. Research on the differential effects of external or internal features in recognition and matchings tasks has shown that performance with external features is generally less affected by inversion (Barton et al., 2001; Malcolm et al., 2005; Meinhardt-Injac, Meinhardt, & Schwaninger, 2009; Moscovitch & Moscovitch, 2000; Nachson & Shechory, 2002; Phillips, 1979; Rakover & Teucher, 1997; Rhodes et al., 1993; Winocur, & Behrmann, 1997). Moreover, there are recent findings that face matching while attending internal features is strongly modulated by orientation and by viewpoint, but hardly for attending the external features of the same faces (Meinhardt-Injac et al., 2009).
Does matching of internal and external facial features depend on orientation and viewpoint?
2009, Acta PsychologicaCitation Excerpt :While it is generally accepted that configural information is strongly affected by inversion there is no consensus regarding the inversion effect for isolated facial features. Previous studies show a conflicting pattern of results, with alternating support of orientation-independency and orientation-dependency (Barton, Keenan, & Bass, 2001; Nachson & Shechory, 2002; Rakover & Teucher, 1997; Rhodes et al., 1993; Schwaninger, Carbon, & Leder, 2003; Valentine, 1988; Veres-Injac & Schwaninger, 2009). Although an inversion effect is reported in some of these studies, there are remarkable differences in its strength mediated by the type of facial features.
From theory to implementation: Building a multidimensional space for face recognition
2009, Acta PsychologicaCitation Excerpt :Nevertheless, Distinctiveness, Face Appearance and Eyes were significantly correlated with face recognition. Thus, although face recognition relies mainly on holistic characteristics, featural characteristics also play a substantial role in this process (Collishaw & Hole, 2000; Maurer et al., 2002; Nachson & Shechory, 2002; Rakover & Teucher, 1997; Rossion, 2008; Valentine, 1988, 1991). Valentine (1991) argues that any feature that discriminates among faces may be considered a dimension.
Three studies on configural face processing by chimpanzees
2006, Brain and Cognition