1 Introduction
Faces convey a wealth of social cues informing viewers about a person's gender, age,
emotional state, speech production, and identity. The relationships between these
aspects of face processing have been the subject of considerable debate. According to
the influential model of Bruce and Young (1986), the perception of both identity and
expression is based on a process of structural encoding. The identification of familiar
faces and the recognition of variant aspects of faces, such as emotional expressions
and lip reading, are carried out in parallel. Although high-level information is involved
in all these processes, they have been thought to operate along separate routes and be
functionally independent (Calder et al 2000).
Evidence for the putative independence of identity and expression recognition has
been obtained in intact as well as in neurologically impaired subjects. For example,
Etcoff (1984) reported lack of interference from the irrelevant stimulus dimension when
subjects were sorting faces according to identity or to expression. Other studies showed
that, when smiling faces were used in an expression-matching task, no reaction time
(RT) advantage was found for familiar over unfamiliar faces, although familiarity did
speed up identity matching (Bruce 1986). Similar findings were also reported in other
studies (eg Young et al 1986; Calder et al 2001, and many more).
The most striking evidence in support of the independence of face recognition and
perception of facial expressions came from double dissociations reported in neurologi-
cal patients. Patients with brain lesions who exhibited relatively selective impairments
in either face identification or expression recognition have been described, for example,
by Tranel et al (1988), Humphreys et al (1993), Young et al (1993), and Keane et al (2002)
among others.
Recent neuroimaging research on patients suggested specific brain areas that
subserve the recognition of different emotions (Adolphs 2002). Those brain areas
were unique to emotions and were not thought to be involved in identity recognition.
Nevertheless, such anatomical facts do not necessarily imply functional indepen-
dence, since locating brain activities more specifically in particular areas does not
preclude interaction among components, which will have behavioural manifestations.
Interactive processes in matching identity and expressions
of unfamiliar faces: Evidence for mutual facilitation effects
Perception, 2008, volume 37, pages 915 ^ 930
Yonata Levyô, Shlomo Bentin½
Department of Psychology (ô and Hadassah-Hebrew University Medical School; ½ and Center for
Neural Computation), Hebrew University of Jerusalem, Mount Scopus, 91905 Jerusalem, Israel;
e-mail: msyonata@huji.ac.il
Received 16 April 2006, in revised form 24 July 2007; published online 1 May 2008
Abstract. We investigated the interactions between matching identity and expressions of unfamiliar
faces. In experiment 1, participants matched expressions in frontal and in oblique views, while
we manipulated facial identity. In experiment 2, participants matched identity in frontal and in
oblique views, while facial expressions were manipulated. Labeling of expressions was not
required. Results showed mutual facilitation between matching facial identity and facial expres-
sions, in accuracy as well as in reaction times. Thus, matching expressions was better and faster
for same-identity images in oblique as well as in frontal views (experiment 1), and matching
identity was better and faster for same-expression images in oblique as well as in frontal views
(experiment 2). The discussion focuses on the implications of these results for the structural
encoding of facial identity and facial expressions.
doi:10.1068/p5925

Similarly, evidence of involvement of the same brain areas in processing identity and
expression does not in itself suffice as an argument against the possible dissociation
of these cognitive processes. Behavioural data along with appropriate neural functional
modeling are, indeed, crucial in order to clarify the relationships between these systems.
Recent behavioural studies reported mutual effects between identity recognition
and recognition of facial expressions. However, the interpretation of these results is not
unequivocal. There were studies showing that expressions affected the identification of
famous faces, but not the identification of unfamiliar faces (Young et al 1986; Endo
et al 1992; Sansone and Tiberghien 1994, reported in Posamentier and Abdi 2003).
Campbell et al (1996) extended this familiarity effect to personally known faces. Their
study went beyond happy faces, testing RT in decisions about sad and disgusted faces
as well, with similar results across different emotions.
In a study of healthy participants, Schweinberger and Soukup (1998) found that
RTs for identity were not influenced by expressions or by facial speech, whereas expres-
sion recognition and recognition of facial speech were affected by variation in identity.
These authors therefore argued that there might be asymmetries between different
components of face processing: whereas identity was independent of other face varia-
tions, expression and facial speech recognition were affected by identity. In a later
study, Schweinberger et al (1999) did not find an effect on identity classification of
unfamiliar faces through morphing across emotions. However, much like previous
work, expressions clearly influenced the identification of familiar faces (Kaufman and
Schweinberger 2004).
Note that, while exposure to famous faces is notoriously biased in favour of partic-
ular expressions, extending the familiarity effects to personally known faces (cf Campbell
et al 1996) still does not rule out the possibility that familiarity with multiple examples
of conjunctive identity and expression may account for the observed effects of identity
on recognising expression (cf Tarr and Bu
«
lthoff 1995). Thus, the crucial question remains
with the processing of unfamiliar faces, where there is no previous familiarity with expres-
sive variants of the face, and performance depends on decoding the information that is
inherent in the stimulus. Studying identity and expression recognition of unfamiliar faces
will capitalise on the inherent difficulties of these tasks, as reviewed in Hancock et al (2000).
These authors point out factors such as differences in lighting and viewpoints, as well as
configuration and distinctness that affect performance.
To the best of our knowledge, the only apparent evidence of an interaction involving
the identity and expressions of unfamiliar faces came from a study by Ganel and Goshen-
Gottstein (2004) who found interference effects of expression on identity as well as of
identity on expression recognition, not only for familiar faces but for unfamiliar faces
as well, albeit the latter effect was only partial and less conspicuous. In a recent
fMRI study by the same group the same sets of stimuli were used showing active
involvement of the fusiform face area (FFA) in processing facial expressions when
attention was directed to expression, as well as when expression was an irrelevant
variable (Ganel et al 2005b). These data extend and elaborate previous findings that
showed stronger FFA involvement in processing fearful faces compared with neutral
faces (eg Pessoa et al 2002).
With respect to these data, one wonders to what extent was familiarity indeed
controlled for in the studies reported above. For example, in Schweinberger et al (1999)
two unfamiliar faces were given namesöPeter and Simonöand were introduced to
the subjects during 32 trial presentations (16 with name tags and 16 without name
tags). Following these familiarisation trials, there were 280 presentations of these two
individuals, varying between two expressions. Similarly, in Ganel and Goshen-Gottstein
(2004) there were 16 trial presentations and 84 test presentations, varying between
just two individuals with two different expressions. Furthermore, in the latter study
916 Y Levy, S Bentin