Documentos de Académico
Documentos de Profesional
Documentos de Cultura
Original Communication
Abstract. Configural processing has been considered the major contributor to the face inversion effect (FIE) in face recognition. However,
most researchers have only obtained the FIE with one specific ratio of configural alteration. It remains unclear whether the ratio of
configural alteration itself can mediate the occurrence of the FIE. We aimed to clarify this issue by manipulating the configural information
parametrically using six different ratios, ranging from 4% to 24%. Participants were asked to judge whether a pair of faces were entirely
identical or different. The paired faces that were to be compared were presented either simultaneously (Experiment 1) or sequentially
(Experiment 2). Both experiments revealed that the FIE was observed only when the ratio of configural alteration was in the intermediate
range. These results indicate that even though the FIE has been frequently adopted as an index to examine the underlying mechanism of
face processing, the emergence of the FIE is not robust with any configural alteration but dependent on the ratio of configural alteration.
Keywords: face inversion effect, configural information, quantitative difference, qualitative difference, face processing
As social animals, human beings are surrounded in their daily given that all faces share the same first-order properties. A
lives by innumerable faces and are able to recognize faces considerable body of research supports the idea that the FIE
extremely quickly, even without paying specific attention to arises mainly from an impaired extraction of configural in-
them (Reddy, Wilken, & Koch, 2004; Reinitz, Morrissey, & formation in inverted faces (Diamond & Carey, 1986;
Demb, 1994). Face recognition is probably the most sophis- Freire, Lee, & Symons, 2000; Schwaninger & Mast, 2005;
ticated ability that human beings exhibit, especially consid- Searcy & Bartlett, 1996; Valentine, 1988; Yin, 1969).
ering the fact that all faces are composed of relatively few Empirically, the FIE can easily be replicated in face re-
constituent elements in an identical configuration. However, search and has been widely adopted as an index to examine
despite the extreme human capability for face processing, many important issues in face recognition research. For ex-
accumulated research has found that face recognition is dis- ample, it has been found that inversion costs are not equiva-
proportionately impaired by inversion when compared to the lent when different facial information is processed, for exam-
recognition of other objects. The impairment of inverted face ple, configural versus component information (Freire et al.,
recognition is referred to as the face inversion effect (FIE), 2000; Goffaux & Rossion, 2007; Murray, Yong, & Rhodes,
which implies that upright and inverted faces are processed 2000; Searcy & Bartlett, 1996; Sergent, 1984); when faces of
differently. The FIE serves as important evidence to support different races are to be recognized (Rhodes, Tan, Brake, &
different arguments in face research. Taylor, 1989); when objects versus faces are to be distin-
Generally, configural information in faces is divided into guished (Cooper & Brooks, 2004; Farah, Wilson, Drain, &
two different kinds of information: first-order and second- Tanaka, 1998; Tanaka & Farah, 1993; Yin, 1969); or when
order relational properties (Diamond & Carey, 1986; participants of different ages are to identify faces (Carey &
Rhodes, 1988). First-order properties refer to the overall Diamond, 1977; Chung & Thomson, 1995; Itier & Taylor,
spatial configuration of facial features, for example, the 2004a, 2004b; Mondloch, Grand, & Maurer, 2002).
eyes are above the nose, which in turn is above the mouth, In others words, many researchers have based their con-
etc. Second-order properties refer to the specific and quan- clusions on the presence or absence of the FIE. However,
titative relations among the facial features, such as the dis- almost all previous research has obtained the FIE with only
tance between the eyes, the distance between the eyes and one specific ratio of configural alteration. Therefore, it re-
the nose, etc. Although the first-order properties allow hu- mains unclear whether the ratio of configural alteration itself
mans to judge whether or not a particular spatial configu- can mediate the occurrence of the FIE. The present research
ration of arbitrary elements resembles a face, it is the sec- aims to clarify whether the ratio of configural alteration itself
ond-order properties that allow humans to tell faces apart, can mediate the occurrence of the FIE. In the experiment,
DOI 10.1024/1421-0185/a000018 Swiss J. Psychol. 69 (3) © 2010 by Verlag Hans Huber, Hogrefe AG, Bern
162 J. Yang & A. Schwaninger: Face Inversion Effect Emerges Under Critical Configural Discrepancy
Figure 1. Examples of the different ratios of configural alteration in one female Caucasian and one female Asian face.
1 For example, in the trials with faces with configural alterations of 4%, two original faces will be presented in the same-configural-alteration
trial, while one original face and one configurally altered face will be presented in the different-configural-alteration trial. As a result, there
are three original faces and only one configurally altered face. The same situation also occurs in the configurally altered trials at 8%, 12%,
16%, 20%, and 24%. In the end, the original face will have been presented 18 times, whereas each face with a different extent of configural
alteration will have been presented only once.
Swiss J. Psychol. 69 (3) © 2010 by Verlag Hans Huber, Hogrefe AG, Bern
J. Yang & A. Schwaninger: Face Inversion Effect Emerges Under Critical Configural Discrepancy 163
2 In all analyses in this study, when Mauchly’s (1940) test of sphericity showed a significant deviance from equicorrection for a repeated
factor or for a combination of factors including at least one repeat factor, Greenhouse and Geisser’s (1959) epsilon was used to adjust the
degrees of freedom for the averaged tests of significance.
Swiss J. Psychol. 69 (3) © 2010 by Verlag Hans Huber, Hogrefe AG, Bern
164 J. Yang & A. Schwaninger: Face Inversion Effect Emerges Under Critical Configural Discrepancy
the performance of upright versus inverted conditions with response instructions were presented on the screen instruct-
different ratios. As can be seen from Figure 3, the FIE was ing participants to press the predesignated key correspond-
observed only with configural alterations of 16% and 20% ing to “same” or “different.” There was no time limit. After
(p values < .01). The FIE did not emerge with configural participants had responded, the response instructions dis-
alterations of 4%, 8%, 12%, or 24% (p values > .051). appeared and participants had to press any key to initiate
the next trial. Of the 384 trials, half were same-configural-
alteration and the other half different-configural-alteration
Discussion trials. For different-configural-alteration trials, the se-
quence of presentation (original face vs. configurally al-
The results show that configural processing systematically tered face) was balanced. Thus, on half of the trials, the
improved with increasing ratio of configural alteration, and original faces were presented first and on the other half the
for both upright and inverted faces. The face inversion ef- configurally altered faces were presented first.
fect (FIE) only emerged with intermediate ratios of config-
ural alteration, but not with the extremes of either subtle or
considerable configural alteration. In other words, there is Results
a critical configural discrepancy for the emergence of the
inversion effect in face processing. These results suggest d’ Score
that the ratio of configural alteration is an important factor
The method of analysis used in Experiment 2 was identical
for mediating the FIE. Accordingly, the FIE is not robust
to that used in Experiment 1. The data were converted to
with any configural alteration, but dependent on the ratio
d’ scores and subjected to a three-factor ANOVA with ori-
of configural alteration.
entation, race, and ratio as within-subject factors. Bonfer-
roni-adjusted posteriori pairwise comparisons were con-
ducted when the ANOVA revealed significant differences
Experiment 2 between conditions. Figure 4 shows the mean d’ score and
standard errors as a function of orientation and configural
Method alteration ratio in Experiment 2.
Participants
Swiss J. Psychol. 69 (3) © 2010 by Verlag Hans Huber, Hogrefe AG, Bern
J. Yang & A. Schwaninger: Face Inversion Effect Emerges Under Critical Configural Discrepancy 165
Face Inversion Effect (FIE) fact that subtle configural alterations make discrimination
difficult in both upright and inverted faces. The FIE also
The FIE was apparent only with configural alterations of failed to emerge with substantial configural alterations,
12%, 16% and 20% (p values < .05). No FIE emerged with which might reflect the fact that configural alterations can
configural alterations of 4%, 8%, or 24% (p values > .095). reach the perceptual criteria for detecting a configural dis-
crepancy in both upright and inverted faces. However,
with intermediate ratios of configural alteration, the con-
Discussion figural alteration can reach the perceptual criteria for de-
tecting a configural discrepancy in upright faces more eas-
Although overall performance declined when the paired ily than in inverted faces. As a result, the FIE emerges
faces were presented sequentially, the pattern revealed in under the inequivalent sensitivity of configural processing
Experiment 1 was replicated in Experiment 2. Performance in different orientations. The varying FIE with different
improved with an increasing ratio of configural alteration. ratios of configural alteration also rules out the possibility
This trend was evident in both upright and inverted faces. that participants match the faces by local distance. Schwa-
In addition, similar to the results in Experiment 1, the in- ninger, Ryf, and Hofer (2003) found that the FIE disap-
version effect emerges only in the intermediate range of pears when participants are required to compare the dis-
configural alteration, but not in the substantial or subtle tance between facial features in face recognition experi-
configural alterations. ments. Moreover, Farah, Drain, and Tanaka (1995) also
found that the FIE disappeared when participants were
asked to decompose a holistic dot pattern into several
parts based on color.
General Discussion The present results seem to contradict arguments claim-
ing that configural information in inverted faces is inac-
In face recognition research, FIE serves as an important cessible to people (Diamond & Carey, 1986; Sergent,
index which has frequently been used to examine various 1984; Yin, 1969). It shows that configural information in
issues regarding face processing. Many arguments are inverted faces is still extractable. Moreover, configural
based on the presence or absence of the FIE. However, processing of inverted faces also improves with increasing
in previous research, the FIE was often only obtained amounts of configural alteration. It seems that the sensi-
with one specific ratio of configural alteration. The ratio tivity of configural processing of inverted faces is im-
of configural alteration was rarely treated as a factor and paired, but not the capability of extracting face configural
manipulated systematically to examine its potential inter- information. The difference in the configural processing
action with the FIE. Therefore, it remained unclear of upright versus inverted faces might not be an “all-or-
whether the ratio of configural alteration can mediate the nothing” issue, but rather one concerning the extent of the
FIE. efficiency of configural processing. As proposed by Se-
In the present research, we altered the ratio of config- kuler, Gaspar, God, and Bennett (2004), the primary dif-
ural alteration parametrically, using six different ratios, ference between the processing of upright and inverted
and examined whether the FIE could be obtained with any faces might be quantitative rather than qualitative. Infor-
ratio or whether it was limited to specific ratios. Both ex- mation is extracted more efficiently in upright than in in-
periments revealed that the FIE emerged only with a crit- verted faces. The advantage in processing upright faces
ical discrepancy of face configural information. In addi- may simply be a by-product of relative expertise levels
tion, performance systematically improved with an in- because humans have more experience in recognizing up-
crease in ratio of configural alteration in both upright and right faces than inverted faces.
inverted faces. These results suggest that the ratio of con- Another issue worth addressing is the disappearance of
figural alteration plays an important role in mediating the the race effect in both experiments. It has been proposed
occurrence of the FIE. The FIE is not robust with any con- that it is easier for people to distinguish between faces of
figural alteration but dependent on the ratio of configural their own race than between those of another (see Meiss-
alteration. These results imply that the arguments based ner & Brigham, 2001, for a review). However, other re-
on the FIE should be interpreted cautiously because the search indicates that the race effect in face recognition
FIE may have been contaminated by the ratio of config- might be the result of an impairment of extracting the
ural alteration in previous experiments. This idea is in component information, holistic information, or configur-
agreement with Ellis (1975) and Rhodes, Hayward, and al information in other-race faces (Hayward, Rhodes, &
Winkler (2006) who cast doubts on the widespread inter- Schwaninger, 2008; Michel, Caldara, & Rossion, 2006;
pretation of FIE as diagnostic of configural coding. Rhodes et al., 2006). However, in those experiments, par-
The FIE only emerged with a critical configural dis- ticipants were required to assess whether the faces dif-
crepancy, namely, with an intermediate ratio of configural fered in either holistic, component, or configural informa-
alteration of around 12% to 20%. The FIE did not emerge tion throughout the entire task. In other words, the com-
with subtle configural alterations, which might reflect the ponent or configural processing was not enhanced since
Swiss J. Psychol. 69 (3) © 2010 by Verlag Hans Huber, Hogrefe AG, Bern
166 J. Yang & A. Schwaninger: Face Inversion Effect Emerges Under Critical Configural Discrepancy
participants were to pay attention to a variety of informa- Carey, S., & Diamond, R. (1977). From piecemeal to configura-
tion. However, in the present study, participants were en- tional representation of faces. Science, 195, 312–315.
couraged to perform only configural processing through- Chung, M. S., & Thomson, D. M. (1995). Development of face
out the entire task. In addition, the configural information recognition. British Journal of Psychology, 86, 55–87.
was systematically altered to different extents. This might Cooper, E. E., & Brooks, B. E. (2004). Qualitative differences in
the representation of spatial relations for different object class-
also have enhanced participants’ configural processing
es. Journal of Experimental Psychology, 30, 243–256.
during the task. The results of Experiment 2 show that
Diamond, R., & Carey, S. (1986). Why faces are and are not spe-
participants’ performance in discriminating between cial: An effect of expertise. Journal of Experimental Psychol-
Asian and Caucasian faces was comparable under this par- ogy: General, 115, 107–117.
adigm. These results were not surprising because the race Ellis, H. D. (1975). Recognizing faces. British Journal of Psy-
effect has not always been replicated (Bothwell, Brigham, chology, 66, 409–426.
& Malpass, 1989; Tanaka, Kiefer, & Bukach, 2004). How- Farah, M. J., Drain, H. M., & Tanaka, J. W. (1995). What causes
ever, Experiment 1 produced an unexpected result, name- the face inversion effect? Journal of Experimental Psycholo-
ly, that Caucasian participants performed better in recog- gy: Human Perception and Performance, 21, 628–634.
nizing Asian than Caucasian faces with configural altera- Farah, M. J., Wilson, K. D., Drain, H. M., & Tanaka, J. (1998).
tions of 20% and 24%. Although the opposite race effect What is “special” about face perception? Psychological Re-
seems surprising, these results might be explained by ra- view, 105, 482–498.
cial differences in facial characteristics. A recent anthro- Freire, A., Lee, K., & Symons, L. A. (2000). The face-inversion
pometric analysis reported by Kunjur, Sabesan, and Ilan- effect as a deficit in the encoding of configural information:
kovan (2006) found that Chinese men and women have Direct evidence. Perception, 29, 159–170.
Goffaux, V., & Rossion, B. (2007). Face inversion disproportion-
wider intercanthal distances than Indian and Caucasian
ately impairs the perception of vertical but not horizontal re-
men and women. As explained by the authors, the inter- lations between features. Journal of Experimental Psychology:
canthal distance refers to the most medial part of the Human Perception and Performance, 33, 995–1002.
palpebral fissure, that is, the closest distance between the Greenhouse, S. W., & Geisser, S. (1959). On methods in the anal-
two eyes. In other words, the separation of the eyes is ysis of profile data. Psychometrika, 32, 95–112.
greater in Asian than in Caucasian faces. This racial dif- Hayward, W. G., Rhodes, G., & Schwaninger, A. (2008). An own-
ference in eye separation also occurs in the faces used in race advantage for components as well as configurations in
the present research, the average eye separation being face recognition. Cognition, 106, 1017–1027.
greater in Asian faces (140.75 pixels, or about 4.10 cm on Itier, R. J., & Taylor, M. J. (2004a). Face inversion and contrast-
a 17-inch monitor) than in Caucasian faces (135.06 pixels, reversal effects across development: In contrast to expertise
about 3.95 cm on a 17-inch monitor). Although ratio was theory. Developmental Science, 7, 246–260.
controlled, the greater eye separation in Asian faces Itier, R. J., & Taylor, M. J. (2004b). Face recognition memory and
makes the actual movement of the eyes larger, especially configural processing: A developmental ERP study using up-
when the ratio was considerable. This might explain why right, inverted, and contrast-reversed faces. Journal of Cogni-
tive Neuroscience, 16, 487–502.
the configural discrepancy was easier to detect in Asian
Kunjur, J., Sabesan, T., & Ilankovan, V. (2006). Anthropometric
faces when the ratio of configural alteration was 20% and
analysis of eyebrows and eyelids: An interracial study. British
24%. Journal of Oral and Maxillofacial Surgery, 44, 89–93.
In sum, the finding that the FIE emerges only with a Mauchly, J. W. (1940). Significance test for sphericity of a normal
critical configural discrepancy is important and enlighten- n-variate distribution. Annals of Mathematical Statistics, 11,
ing for research associated with face recognition. In the 204–210.
domain of face recognition, the inversion effect is widely Meissner, C. A., & Brigham, J. C. (2001). Thirty years of inves-
and frequently used as an index to examine the underlying tigating the own-race bias in memory for faces: A meta-ana-
mechanism of face processing. The FIE is always regard- lytic review. Psychology, Public Policy, and Law, 7, 3–35.
ed as an important indicator to justify or disprove different Michel, C., Caldara, R., & Rossion, B. (2006). Same-race faces
arguments. However, as revealed in the present research, are perceived more holistically than other-race faces. Visual
the emergence of the FIE does not only depend on the Cognition, 14, 55–73.
stimuli, task, participant age, etc., but is also a conse- Mondloch, C. J., Grand, R. L., & Maurer, D. (2002). Configura-
tion face processing develops more slowly than featural face
quence of a specific ratio of configural alteration.
processing. Perception, 31, 553–566.
Murray, J. E., Yong, E., & Rhodes, G. (2000). Revisiting the per-
ception of upside-down faces. Psychological Science, 11,
492–496.
Reddy, L., Wilken, P., & Koch, C. (2004). Face-gender discrimi-
References nation is possible in the near-absence of attention. Journal of
Vision, 4, 106–117.
Bothwell, R. K., Brigham, J. C., & Malpass, R. S. (1989). Cross- Reinitz, M. T., Morrissey, J., & Demb, J. (1994). Role of attention
racial identification. Personality and Social Psychology Bulle- in face encoding. Journal of Experimental Psychology: Learn-
tin, 15, 19–25. ing, Memory, and Cognition, 20, 161–168.
Swiss J. Psychol. 69 (3) © 2010 by Verlag Hans Huber, Hogrefe AG, Bern
J. Yang & A. Schwaninger: Face Inversion Effect Emerges Under Critical Configural Discrepancy 167
Rhodes, G. (1988). Looking at faces: First-order and second-order ural processes underlying face perception. British Journal of
features as determinants of facial appearance. Perception, 17, Psychology, 75, 221–242.
43–68. Tanaka, J. W., & Farah, M. J. (1993). Parts and wholes in face
Rhodes, G., Hayward, W., & Winkler, C. (2006). Expert face cod- recognition. The Quarterly Journal of Experimental Psychol-
ing: Configural and component coding of own and other race ogy, 46A, 225–245.
faces. Psychonomic Bulletin and Review, 13, 499–505. Tanaka, J. W., Kiefer, M., & Bukach, C. M. (2004). A holistic ac-
Rhodes, G., Tan, S., Brake, S., & Taylor, K. (1989). Expertise and count of the own-race effect in face recognition: Evidence
configural coding in face recognition. British Journal of Psy- from a cross-cultural study. Cognition, 93, B1–B9.
chology, 80, 313–331. Valentine, T. (1988). Upside-down faces: A review of the effect
Schwaninger, A., & Mast, F. W. (2005). The face-inversion effect of inversion upon face recognition. British Journal of Psychol-
can be explained by the capacity limitations of an orientation ogy, 79, 471–491.
normalization mechanism. Japanese Psychological Research, Yin, R. (1969). Looking at upside-down faces. Journal of Exper-
47, 216–222. imental Psychology, 81, 141–145.
Schwaninger, A., Ryf, S., & Hofer, F. (2003). Configural infor-
mation is processed differently in perception and recognition
of faces. Vision Research, 43, 1501–1505.
Searcy, J. H., & Bartlett, J. C. (1996). Inversion and processing of Jisien Yang
component and spatial-relational information in faces. Journal of
Experiment Psychology: Human Perception and Performance, Department of Psychology
22, 904–915. University of Zurich
Sekuler, A. B., Gaspar, C. M., God, J. M., & Bennett, P. (2004). CASRA / Thurgauerstrasse 39
Inversion leads to quantitative, not qualitative, changes in face CH - 8050 Zurich
processing. Current Biology, 14, 391–396. Switzerland
Sergent, J. (1984). An investigation into component and config- jisien.yang@casra.ch
Swiss J. Psychol. 69 (3) © 2010 by Verlag Hans Huber, Hogrefe AG, Bern