What’s special about personally familiar faces? A ...bctill/papers/facerec/Herzmann2005.pdf · What’s special about personally familiar faces? A multimodal approach GRIT HERZMANN,a
Post on 21-Feb-2019
219 Views
Preview:
Transcript
What’s special about personally familiar faces?
A multimodal approach
GRIT HERZMANN,a STEFAN R. SCHWEINBERGER,b WERNER SOMMER,a and INESJENTZSCHc
aDepartment of Psychology, Humboldt-University at Berlin, Berlin, GermanybDepartment of Psychology,University of Glasgow, Glasgow, UKcSchool of Psychology,University of St. Andrews, St. Andrews, UK
Abstract
Dual-route models of face recognition suggest separate cognitive and affective routes. The predictions of these models
were assessed in recognition tasks with unfamiliar, famous, and personally familiar faces. Whereas larger autonomic
responses were only triggered for personally familiar faces, priming effects in reaction times to these faces, presumably
reflecting cognitive recognition processes, were equal to those of famous faces. Activation of stored structural rep-
resentations of familiar faces (face recognition units) was assessed by recording the N250r component in event-related
brain potentials. Face recognition unit activation increased from unfamiliar over famous to personally familiar faces,
suggesting that there are stronger representations for personally familiar than for famous faces. Because the topog-
raphies of theN250r for personally and famous faces were indistinguishable, a similar network of face recognition units
can be assumed for both types of faces.
Descriptors:Face recognitionmodels, Skin conductance responses, Event-related brain potentials, Priming, Personally
known faces
Human faces are outstandingly rich sources of information for
social interaction, providing detailed information about famil-
iarity, identity, mood, gender, age, or focus of attention. It is
therefore hardly surprising that the issues of how the cognitive
system accomplishes and how the brain implements these aspects
of face perception have enjoyed a great deal of scientific interest.
Traditional models of face recognition (Bruce & Young, 1986;
Hay&Young, 1982) have focusedmainly on cognitive processes.
However, more recent models (Breen, Caine, & Coltheart, 2000;
Ellis & Lewis, 2001) have included affective aspects of face rec-
ognition as well. These models suggest that a so-called cognitive
route analyzes the identity of faces and provides access to se-
mantic knowledge and name of familiar persons. In addition, a
second route is thought to be involved in the production of af-
fective responses to familiar faces. The assumption of these dis-
tinct routes in face processing was made in order to explain
impairments of face recognition in prosopagnosia and its sup-
posed counterpart, Capgras delusion.
Patients with prosopagnosia are unable to identify the faces of
familiar persons and to learn new faces. Typically, prosopagnosia
is a consequence of acquired brain damage, involving inferior
occipito-temporal lesions of the right or both hemispheres. These
patients often remain able to recognize familiar persons by voice
or gait, and may also show preserved semantic memory for peo-
ple, for example, when confronted with their names. Neverthe-
less, prosopagnosic patients may be unable to recognize faces of
even highly familiar people. However, such faces, even though
overtly unrecognized, may elicit signs of covert recognition, such
as skin conductance responses (SCR). For example, Bauer
(1984) presented the prosopagnosic patient LF with familiar
faces, paired with spoken names, which could or could not cor-
respond to the face. Although LF could not identify the correct
names, his SCRs were larger to correct than incorrect face/name
pairs. In other studies, prosopagnosic patients showed larger
SCRs to familiar as compared to unfamiliar faces in the absence
of overt recognition (Tranel & Damasio, 1985). In terms of dual-
route models, preserved differential SCRs may indicate that
some patients with prosopagnosia, although impaired in overt
recognition along the cognitive route of face recognition, still
have a relatively intact affective route.
Patients with Capgras delusion show a pattern of impairment
that appears to be almost the mirror image of prosopagnosia
(Ellis&Young, 1990). Capgras delusionmay occur in the context
This research was supported by a Socrates-Erasmus exchange stu-
dentship to G.H. while she was visiting Glasgow, by a grant by the
Deutsche Forschungsgemeinschaft (So 177/14-1) to W.S., and by grants
by the Biotechnology and Biological Sciences Research Council (17/
S14233) and the Royal Society to S.R.S.Address reprint requests to: Grit Herzmann, Department of Psy-
chology, Humboldt-University at Berlin, Rudower Chaussee 18, D-10099 Berlin, Germany. E-mail: grit.herzmann@cms.hu-berlin.de, or toStefan R. Schweinberger, Department of Psychology, University ofGlasgow, Glasgow G12 8QQ, Scotland. E-mail: s.schweinberger@psy.gla.ac.uk.
Psychophysiology, 41 (2004), 688–701. Blackwell Publishing Inc. Printed in the USA.Copyright r 2004 Society for Psychophysiological ResearchDOI: 10.1111/j.1469-8986.2004.00196.x
688
of psychiatric conditions or as a result of structural or toxic brain
damage. Although these patients are still able to identify familiar
faces, they lack a sense of familiarity to these faces and believe
that impostors, doubles, or aliens have replaced these people
(e.g., spouses or children). Ellis and Young proposed that Cap-
gras patients, though unimpaired in their cognitive route for face
recognition, might have a damaged affective route. Their pre-
diction that these patients would fail to produce a differential
SCR response to overtly recognized familiar faces was confirmed
by subsequent research (Ellis, Young, Quayle, & DePauw, 1997;
Hirstein & Ramachandran, 1997).
The dissociation of autonomic responses and overt face rec-
ognition in prosopagnosia and Capgras delusion was taken as
support for the existence of two routes to face recognition. Bauer
(1984) postulated two routes for face recognition in both neuro-
anatomical and functional terms. A ventral visual-limbic route in
inferior temporal cortex was suggested to mediate overt identi-
fication and to be impaired in prosopagnosia. A dorsal route
projecting from primary visual cortex to limbic structures via the
superior temporal sulcus and inferior parietal lobe was consid-
ered to be involved in the detection of emotional significance and
to mediate the preserved SCR responses in prosopagnosic pa-
tients. Ellis and Young (1990) adopted Bauer’s dual-route model
to accommodate Capgras delusion, which was considered to be
the consequence of damage in the affective route, as indicated by
the absence of SCRs to familiar faces (Ellis et al., 1997).
The theories of both Bauer (1984) and Ellis andYoung (1990)
claim that a face has to be identified to some degree before its
relevance for affective responses can be noticed. In terms of cur-
rent concepts about face recognition (e.g., Bruce &Young, 1986)
these models therefore make the implicit assumption of two in-
dependent sets of stored representations of familiar faces (face
recognition units) feeding into the affective and the cognitive
routes. Recently, Breen et al. (2000) pointed out that Bauer’s
model does not specify possible mechanisms for face recognition
in the dorsal route and noted that there is little evidence for object
or face recognition in this route (Ungerleider & Mishkin, 1982).
Therefore, Breen et al. made the more parsimonious suggestion
of one common pool of FRUs residing within the ventral visual
stream and feeding into both the cognitive and affective routes.
Breen et al. adapted the face recognition model of Bruce and
Young, which contains only a singleFcognitiveFroute, by
adding a module triggering affective responses to familiar faces.
In the modified model, face recognition units feed information
simultaneously and independently into both person identity
nodes and the affective module. Face recognition is thought to
take place along an anatomical route in ventral temporal lobe
structures, with the amygdala triggering affective responses to
familiar faces. Within this framework, Breen et al. explained the
dissociation between patients with prosopagnosia and Capgras
delusion. Prosopagnosia can be caused by an impaired connec-
tion between the face recognition units and the person identity
nodes, while the connection between the face recognition units
and the affective response module may be intact. In contrast, the
locus of impairment in Capgras delusion is either within the af-
fective module or in the connection between the face recognition
units and the affective module (see Figure 1). Subsequently, Ellis
and Lewis (2001) slightlymodified the dual-routemodel of Breen
et al. by adding an integrative device, which compares the out-
puts of the affective and cognitive routes. This was done to ex-
plain the delusion inCapgras patients when a familiar face fails to
elicit a corresponding affective response.
Taken together, all models of face recognition that attempt to
encompass both findings in prosopagnosia and Capgras delusion
and differences between overt and covert face recognition (Ba-
uer, 1984; Breen et al., 2000; Ellis & Lewis, 2001; Ellis & Young,
1990; Ellis et al., 1997) include two routes to face recognition,
termed affective and cognitive. As a result, dual-route models
provide a comprehensive framework for assumptions and inves-
tigations of face recognition. For intact processing systems, the
models postulate a similar relationship between facial familiarity
and response strength. For example, Breen et al. predict that the
affective response should increase with the degree of familiarity.
PreviousFmainly neuropsychologicalFresearch, which includ-
ed only one or two types of facial familiarity, confirmed this
relationship in the affective route. SCRs were more pronounced
for famous than for unfamiliar persons (Tranel & Damasio,
1985; Tranel, Fowles, & Damasio, 1985) or for personally famil-
iar than for famous faces (Tranel, Damasio, & Damasio, 1995).
From these observations and in line with dual-route models one
would expect SCRs to increase with facial familiarity when un-
familiar, famous, and personally familiar faces are compared.
Similarly, it can be thought that face familiarity should im-
prove cognitive processing. This has been demonstrated for the
comparison of famous with unfamiliar faces in several suggested
probes of processing within the cognitive route: priming, inter-
ference, matching effects, and the speed of face-name learning
(Bruce & Young 1986; Ellis, Lewis, Moselhy, & Young, 2000;
Pfutze, Sommer, & Schweinberger, 2002). Repetition priming
for faces reflects facilitation, for example, in reaction times to a
face that has been processed before, relative to initial encounter
(Bruce & Young, 1986). Current models of face recognition
(Burton, Bruce, & Hancock, 1999) assume that the decision
about facial familiarity is made at the person identity node level.
Whereas long-term repetition priming is thought to be mediated
mainly by the strengthening of links between face recognition
units and person identity nodes, immediate repetition priming
appears to involve facilitation at an earlier level of access to face
recognition units and/or structural encoding (Pfutze et al., 2002;
Schweinberger, Pfutze, & Sommer, 1995; Schweinberger, Picke-
ring, Jentzsch, Burton, & Kaufmann, 2002). Therefore, repeti-
tion priming in face identification may involve facilitation at
several loci along the cognitive route. Whereas previous studies
revealed stronger priming effects in RTs for famous than for
unfamiliar faces (Bruce & Valentine, 1985; Ellis et al., 2000;
Pfutze et al., 2002), personally familiar faces do not seem to have
been studied for priming effects. It is therefore an open question
whether there would also be an increase from famous to person-
ally familiar faces in RT priming.
Importantly, current research, mainly based on the model of
face recognition by Bruce and Young (1986), has identified at
least two consistent ERP correlates of face recognition within a
repetition priming task paradigm. First, a negativity for primed
relative to unprimed faces is seen over inferior temporal elec-
trodes at latencies of around 200–350 ms, an effect that has been
variously termed ‘‘N250r’’ or ‘‘early repetition effect’’ (ERE;
Pfutze et al., 2002; Schweinberger et al., 1995; Schweinberger,
Pickering, Jentzsch, et al., 2002). The N250r is typically larger
over the right than left hemisphere and is not present for asso-
ciative-semantic priming (Schweinberger, 1996). Moreover, the
N250r is reduced or even absent for unfamiliar as compared to
famous faces (Pfutze et al., 2002; Schweinberger et al., 1995),
suggesting that it depends on the contact with an existing mem-
ory representation. Although an N250r at similar latency as for
What’s special about personally familiar faces? 689
faces has been observed for written names, this effect is maximal
over the left hemisphere (Pfutze et al., 2002; Pickering &
Schweinberger, 2003). Thus, the N250r has a stimulus-depend-
ent topography, suggesting that it links to perceptual rather than
postperceptual representations. The face-elicited N250r has been
related to the transient activation of face recognition units, and is
therefore also of particular interest for dual-route models.
Another consistent ERP correlate of face priming is a late
ERP repetition effect (N400 or LRE), an increased centro-
parietal positivity (or reduced negativity) for repeated faces
at latencies around 300–600ms (Bentin & McCarthy, 1994;
Schweinberger et al., 1995; Schweinberger, Pickering, Burton, &
Kaufmann, 2002). This late repetition effect is likely related to
the N400, which was initially demonstrated to reflect the detec-
tion of semantic incongruency in sentences and was thought to
represent semantic integration (Kutas & Hillyard, 1980). An
N400-like pattern can also be elicited by faces, when these are
preceded by the same or by associated faces. The N400 may
reflect facilitation in assessing postperceptual or semantic mem-
ory codes for people (person identity node).
According to assumptions about cognitive networks for face
recognition (Burton et al., 1999; Mohr, Landgrebe, & Schwein-
berger, 2002), priming should especially facilitate recognition of
those faces for which there are stronger memory representations.
To the extent that a more frequent encounter of personally fa-
miliar as compared to famous faces might have induced stronger
neurocognitive networks, it might be expected that the N250r
increases from famous faces to personally familiar ones.
Importantly, the different versions of dual-route models dis-
agree with respect to the locus at which the two routes diverge.
Some models state that both routes diverge early in processing
and therefore imply two distinct sets of recognition mechanisms,
or face recognition units (Bauer, 1984; Ellis et al., 1997; Hirstein
& Ramachandran, 1997). Other models explicitly assert a com-
mon face recognition unit with the two routes diverging only
subsequently (Breen et al., 2000; Ellis & Lewis, 2001).
The N250r can provide functional evidence about face
processing, and the scalp topography of this component can,
to some extent, provide information about underlying sources. If
the N250r for two different types of facial familiarity showed
different scalp topographies, this would be evidence for the in-
volvement of at least partially different neural generators. To the
extent that N250r to personally familiar faces reflect the activa-
tion of a different set of face recognition units as compared to
famous faces, for example in a dorsal affective route (Bauer,
1984; Ellis et al., 1997; Hirstein & Ramachandran, 1997), it
690 G. Herzmann et al.
Figure 1. An adaptation of the dual-route models of face recognition proposed by Breen et al. (2000) and Ellis and Lewis (2001).
Face recognition is considered as matching a seen face to its stored representation at the level of the face recognition units. Face
recognition units activate simultaneously but independently the affective and the cognitive routes. The affective route leads to a
module triggering affective responses to familiar stimuli, of which skin conductance response is an indicator. The cognitive route
leads to person identity nodes, activates semantic information, and supports name retrieval. An abnormality at locationA will cause
a loss of overt face recognition and is therefore a possible explanation for prosopagnosia. An abnormality at location B will cause a
loss of differential SCR responses to familiar and unfamiliar stimuli. It will also cause a discrepancy in the integrative device and is
therefore a possible account for Capgras delusion. In the context of the present study, SCR responses are thought to be mainly
sensitive to activity along the affective route (B), RTpriming effects are thought to bemainly sensitive to activity along the cognitive
route (A), and the ERE should be sensitive to activity in the face recognition units.
should be reflected in distinguishable topographies of theN250r to
famous and personally familiar faces. In contrast, if affective and
cognitive routes of face recognition diverge only after the access to
a common pools of face recognition units (Breen et al., 2000; Ellis
& Lewis, 2001), the N250r topographies should not differ.
To our knowledge, the postulations of dual-route models for
both affective and cognitive aspects have not yet been tested for
more than two types of facial familiarity in the same healthy
population. For that reason itwas the first objective of the present
study to provide a more comprehensive test of the predictions of
dual-route models. We assessed several indices for affective and
cognitive aspects of face recognition, as a function of different
types of facial familiarity. More specifically, we investigated in
healthy participants how unfamiliar, famous, and personally fa-
miliar faces are processed at several levels of the face recognition
system. The affective route was assessed by recording skin con-
ductance responses, the cognitive route was investigated with a
short-term repetition priming task, and the N250r provided a
tool to investigate the effects of facial familiarity on the face
recognition units. Addressing the question of separate or com-
mon face recognition units was the second objective of the present
study. To answer this question we also used the N250r.
Because of the different methodological requirements for re-
cording SCRs and ERPs, the experiment was divided into two
parts where the same stimuli and the same participants were
employed. In Part 1 of the experiment, we explored the influence
of different types of facial familiarity on the affective route by
examining modulations in SCRs. In the second part of the ex-
periment, we investigated effects on the cognitive route as seen in
RT priming effects.
EXPERIMENT: PART 1FSCR
In Part 1 of the experiment we investigated the affective aspects in
face recognition by means of recording SCRs as a function of
facial familiarity.
Method
Participants
Twenty-four participants (21 women, 3 men) with ages ranging
from 19 to 33 years (M5 21.9 years, SD5 3.2) were paid to
contribute data to the experiment. According to an adapted ver-
sion of the Edinburgh Handedness Inventory (Oldfield, 1971)
participants were strongly right-handed (M5186.2, range5
50–100). They reported normal or corrected-to-normal visual
acuity. All participants were psychology students at the Univer-
sity of Glasgow, and had been residents of the United Kingdom
or Ireland for a minimum of 5 years.
Stimuli
Sixty portraits were used, consisting of 15 personally familiar, 15
famous, and 30 unfamiliar faces. Photographs of famous and
unfamiliar persons were made available from earlier research.
The celebrities had been selected on the basis of previous high
ratings for ease of face recognition (Schweinberger, Pickering,
Burton, et al., 2002). Portraits of personally familiar persons
were taken from the lecturing staff in the Department of Psy-
chology at theUniversity ofGlasgow.All portraits were edited to
a unitary format. They were converted to gray scale, framed
within an area of 170 � 216 pixels, corresponding to 6.0 �7.6 cm, and all background was removed. An attempt was made
to homogenize the pictures with respect to contrast and average
luminance across the sets of famous, personally familiar, and
unfamiliar faces.
Prior to the experiment, the 15 best known personally familiar
persons and 15 best known celebrities were selected for each
participant out of two pools of pictures of 15 men and 7 women
each. To avoid any priming or habituation for the face stimuli,
these selections were based on the person’s names. Each person
was rated for ease of possible face recognition on a 5-point scale.
The selected unfamiliar faces were matched to familiar counter-
parts with respect to gender, approximate age, general portrait
style, and expression.
Procedure
After the rating participants were seated in a dimly lit, sound-
attenuated, and electrically shielded chamber. A fixed chin rest
was used tomaintain a constant viewing distance of 1m. A white
fixation cross in the center of the monitor changed into bold font
5 s before the onset of a face. The 60 faceswere presented once for
2,000ms in randomized order with interstimulus intervals be-
tween 12 and 18 s. The familiarity of each face was indicated by
key presses with the index and middle finger of the right hand.
Half the participants pressed the left key (index finger) for a
familiar face and the right key (middle finger) for an unfamiliar
face; this assignment of key to response category was reversed for
the other half. Responses were scored as correct if the appropri-
ate keywas pressedwithin a timewindowof 200 to 2,000ms after
face onset. Errors of omission (no key press) and commission
(wrong key) were recorded separately.MeanRTswere calculated
for correct responses only.
The experiment started with three practice trials, to be dis-
carded later. After completion of both parts of the experiment
participants rated the portraits shown, using the same rating
scale as in the beginning. As expected, the face sets differed, with
both personally familiar and famous faces being both rated as
well known (M5 3.8 and 3.9, respectively), and unfamiliar faces
as unknown (M5 0.2).
Recording
For SCR measurement, sintered Beckman Ag/AgCl electrodes,
1 cm in diameter, were affixed to the thenar and hypothenar
eminences of the left hand; a ground electrode was placed at the
left forearm. Isotonic electrolyte gel (K-Y Jelly, Johnson 1
Johnsont) was used. Skin conductance was recorded with a
Coulbournt Isolated Skin Conductance Coupler (Model V71-
23), and digitized at a sampling rate of 200Hz. Off-line, 10-s
epochs, starting 1 s before stimulus onset, were analyzed. A total
of 2.1% of all trials were excluded from analysis because of ar-
tefacts.
In addition to RTs and error rates, several response param-
eters were derived from the SCR (see below). Results were eval-
uated by means of Huynh–Feldt (Huynh & Feldt, 1976)
corrected repeated-measures ANOVAs; pairwise comparisons
between levels of familiarity were Bonferroni corrected. Within-
subject repeated measures are reported with (a) uncorrected de-
grees of freedom, (b) the corrected p value, and (c) the epsilon
value of the correction factor.
What’s special about personally familiar faces? 691
Results and Discussion
Performance
Means and standard deviations of RTs and error rates are shown
in Table 1. In RTs, a significant main effect of familiarity was
observed, F(2,46)5 18.5, po.001, e5 .57. Responses to unfa-
miliar faces were slower when compared to both personally fa-
miliar, F(1,23)5 18.1, po.001, and famous faces,
F(1,23)5 20.4, po.001, which did not differ from each other,
F(1,23)5 1.0. No effect of familiarity was observed in the error
rates, F(2,46)o1.
SCR
Trials without artefacts were analyzed by a computerized algo-
rithm. The criterion for a valid response was an increase in skin
conductance of at least 0.01mS between 1 and 7 s after stimulus
onset. In 55.1% of the trials these requirements were met. The
other trials were defined as nonvalid and amplitude was scored as
zero.
SCR magnitudewas quantified as mean peak amplitude with-
in the response window of all single trials including valid and
nonvalid (zero-amplitude) SCRs. SCR amplitude was calculated
in the same way but for valid responses only. SCR frequencies
were defined as the proportion of stimuli eliciting valid SCR
responses. The means and standard deviations of these param-
eters are shown in Table 2.
There were significant effects on SCR magnitude,
F(2,46)5 6.5, po.01, e5 .91, SCR frequency, F(2,46)5 4.1,
po.05, e5 .83, as well as a strong trend for SCR amplitude,
F(2,46)5 3.5, p5 .057, e5 .68. In response to personally famil-
iar faces, mean SCR magnitudes and frequencies were higher
when compared with famous faces, F(1,23)5 9.6, po.01 and
F(1,23)5 5.3, p5 .06 (uncorrected po.05), respectively, and al-
so when compared with unfamiliar faces, F(1,23)5 7.4, po.05
and F(1,23)5 7.7, po.05, respectively. Although only as a trend,
the effect on SCR amplitude tended to be the same as for the
other parameters (cf. Table 2). In contrast, unfamiliar and fa-
mous faces did not significantly differ in any of these measures,
Fs(1,23)o1.1.
Partly conformable to the predictions of dual-route models,
this experiment shows that autonomic responses to personally
familiar target faces were both more frequent and larger than
those to famous and unfamiliar target faces. Therefore, the fa-
miliarity-induced modulation of SCR magnitude and amplitude
suggests a stronger involvement of affective components of face
processing for personally familiar faces. Unexpectedly and in
contrast to previous findings (Tranel et al., 1985) as well as to
predictions of dual-route models, no significant differences in
SCRs were seen between famous and unfamiliar faces. One pos-
sible explanation for this discrepancy may be that the present
inclusion of an additional category with strong significance (i.e.,
personally familiar faces) may have dominated the smaller dif-
ferences between famous and unfamiliar faces. It is worth noting
that the results are not explainable simply by an orienting re-
sponse due to signal value (Ohmann, 1979). Because all faces
were targets, they bore the same relevance for the task. Further-
more, SCR differences occurred especially within the same re-
sponse category (familiar faces), which would be hard to explain
in terms of signal value with respect to the task.
Interestingly, RTs showed a different pattern from the SCR
findings, with similar RTs to personally familiar and famous
faces and slower responses to unfamiliar faces. Although this
pattern may be explained by the task, which placed both types of
familiar faces into the same category, the SCRs nevertheless dis-
criminated between these face classes.
EXPERIMENT: PART 2FPRIMING
In Part 2 of the experiment we investigated priming effects in
reaction times and aimed at exploring face recognition units with
repetition effects in event-related brain potentials as a function of
facial familiarity.
Method
Participants
Sixteen participants (15 women, 1 man), who had also taken part
in the first part of the experiment, contributed data to this study.1
All were strongly right-handed (M5185.9, range5 50–100);
ages ranged from 19 to 27 (M5 20.9) years.
Stimuli
The same stimuli as in Part 1 of the experiment were used.
Procedure
Directly following the SCR experiment, EEG electrodes were
applied and participants received written task instructions.
Figure 2 shows the trial sequence. At the beginning of each tri-
al, a white fixation cross appeared for 500ms, being replaced by a
prime face, presented for 500 ms and followed by a green fixation
circle. After 1,300 ms the circle was replaced by a target face,
presented for 2,000 ms. The interval between prime onsets was
6,300 ms.
Participants were asked to indicate by key presses with their
left and right index fingers whether the target face was familiar or
unfamiliar. No motor response was required to prime faces. For
a given participant, the assignment of stimulus category to the
left or right response key was the same as in Part 1 of the ex-
periment. Both speed and accuracy were emphasized. Two dif-
ferent feedback tones of 300 ms duration were presented after
incorrect (500Hz) or missing responses (650Hz). Responses
were scored as correct if the appropriate key was pressed between
200 and 2,000 ms after target onset. Mean reaction times were
calculated for correct responses only. Throughout the experi-
ment, short breaks were allowed after every 45 trials.
The prime for each of the 60 target faces (15 personally fa-
miliar, 15 famous, and 30 unfamiliar) could be either the same
face (primed condition) or a different face (unprimed condition).
692 G. Herzmann et al.
Table 1. Mean Correct Reaction Times (RT, inMilliseconds) and
Percentage of Errors (PE) for Personally Familiar, Famous, and
Unfamiliar Faces in Part 1 of the Experiment
Familiarity type
RT PE
M SD M SD
Personally familiar 875.3 248.6 10.6 21.1Famous 855.5 270.4 6.9 20.0Unfamiliar 1163.0 299.8 6.6 11.8
1Although not all participants from the SCR study were available forthe ERP study, the results in Part 1 of the experiment were essentiallyunchanged even when only these 16 participants were considered.
In the unprimed condition, personally familiar and famous target
faces were preceded by unfamiliar primes, and unfamiliar target
faces were preceded by personally familiar or famous primes.
This allowed us to test repetition priming for both familiar and
unfamiliar targets without an impractically large amount of filler
trials and appeared appropriate because previous research
(Schweinberger et al., 1995) had indicated that responses to un-
primed target faces are independent of whether or not the pre-
ceding prime face is familiar.
The design involved the variables priming (primed vs. un-
primed) and familiarity (personally familiar, famous, and unfa-
miliar). Each of the 60 target faces appeared three times in the
primed condition and three times in the unprimed condition,
yielding a total of 360 experimental trials. In the unprimed con-
dition a given pair of faces was used only once in order to avoid
episodic priming. All trials were shown in randomized order.
Recording
The electroencephalogram (EEG) was recorded with sintered
Ag/AgCl electrodes mounted in an electrode cap (Easy-Capt) at
the scalp positions Fz, Cz, Pz, Iz, FP1, FP2, F3, F4, C3, C4, P3,
P4, O1, O2, F7, F8, T7, T8, P7, P8, FT9, FT10, P9, P10, PO9,
PO10, F90, F100, TP9, and TP10 (Pivik et al., 1993). The F90 andF100 electrodes were positioned 2 cm anterior to F9 and F10 at
the outer canthi of the left and right eyes, respectively. TP9 and
TP10 refer to inferior temporal locations over the left and right
mastoids, respectively. The TP10 electrode served as initial com-
mon reference and a forehead electrode (AFz) served as ground.
Impedances were typically kept below 5 kO. The horizontal
electrooculogram (EOG) was recorded from F90 and F100. Thevertical EOG was monitored from two additional electrodes
above and below the right eye. All signals were recorded with a
band-pass of 0.05Hz to 40Hz (� 6dB attenuation, 12 dB/oc-
tave), and a sampling rate of 200Hz.
Off-line, epochs of 1,700 ms starting 215 ms before target
onset were generated from the continuous record. Trials with
nonocular artefacts, saccades, and incorrect behavioral respons-
es were discarded. Trials with ocular blink contributions to the
EEG were corrected (Elbert, Lutzenberger, Rockstroh, &
Birbaumer, 1985) or excluded from data analysis if not enough
sample blinks could be obtained, as was the case in three par-
ticipants. ERPs were aligned to a 215-ms baseline before target
onset, averaged separately for each channel and experimental
condition, digitally low-pass filtered at 10Hz with zero phase
shift, and recalculated to average reference, excluding the vertical
EOG channel.
All dependent variables were analyzed bymeans of ANOVAs
with repeated measures on familiarity and priming, epsilon cor-
rected for heterogeneity of covariance, wherever appropriate.
Within-subject repeated measures are reported with (a) uncor-
rected degrees of freedom, (b) the corrected p value, and (c) the
epsilon value of the correction factor.
Results
Performance
Reaction times are shown in Table 3. Whereas RTs to unprimed
faces appear to be similar in the three familiarity conditions,
priming reduced RTs more for both types of familiar faces than
for unfamiliar faces. This pattern was confirmed by an interac-
tion between priming and familiarity, F(2,30)5 21.2, po.001,
What’s special about personally familiar faces? 693
Figure 2. Trial sequence for Part 2 of the experiment. Shown is an unprimed trial with a target face that was personally familiar to
the participants of this study, preceded by an unfamiliar prime face.
Table 2. Mean Magnitude (in Microsiemens), Amplitude (in Microsiemens), and Frequency
(in Percent) for Skin Conductance Responses to Personally Familiar, Famous, and Unfamiliar
Faces in Part 1 of the Experiment
Familiarity type
Magnitude Amplitude Frequency
M SD M SD M SD
Personally familiar 0.78 0.92 1.27 1.47 63 26Famous 0.54 0.62 0.88 1.06 58 25Unfamiliar 0.53 0.63 0.86 0.86 54 24
e5 .86, in addition to main effects of priming, F(1,15)5 67.6,
po.001, and familiarity, F(2,30)5 23.8, po.001, e5 .86. The
interaction is caused by the differential priming effects; there was
no familiarity effect in unprimed faces, F(2,30)o1.7, but a strong
effect of familiarity for primed faces, F(2,30)5 41.1, po.001,
e5 .72. Priming effects between personally familiar and famous
faces were indistinguishable, F(1,15)o1.7, p4.10, but priming
was smaller in unfamiliar faces relative to both personally fa-
miliar faces, F(1,15)5 35.1, po.001, and famous faces,
F(1,15)5 18.3, po.001, respectively. Although relatively small,
priming was still significant when tested for unfamiliar faces
alone, F(1,15)5 13.9, po.01.
Mean percentage of errors (Table 3) were generally low, ex-
cept that personally familiar faces were relatively often judged as
unfamiliar. This was reflected in a main effect of familiarity,
F(2,30)5 9.7, po.01, e5 .72, with higher error rates for per-
sonally familiar faces than for both famous, F(1,15)5 11.6,
po.01, and unfamiliar faces, F(1,15)5 9.4, po.05, which did
not differ from each other.2 However, effects of priming, Fo1,
and the interaction between priming and familiarity,
F(2,30)5 2.3, p4.10, e5 1.12, were not significant. Thus, error
rates do not suggest a speed–accuracy trade-off with respect to
the priming effects seen in RTs.
Event-Related Potentials
Figure 3 shows ERPs to personally familiar target faces as typical
examples of the observed waveforms. Figure 4 compares ERPs
to primed and unprimed faces for all three familiarities at the
most important electrode sites. ERPs were quantified with mean
amplitude measures, relative to a 215-ms prestimulus baseline, in
the time segments 120–140, 170–180, 230–270, 270–330, 330–
400, and 400–500 ms, relative to target onset. The first and sec-
ond segments correspond to the occipital P1 and the occipito-
temporal N170 peaks in the ERP waveforms, respectively. The
P1 is sensitive to variations in relatively early domain-general
visual processes, for example, those caused by differences in
contrast, brightness, or size (cf. Pfutze et al., 2002). TheN170 is a
face-specific component supposed to reflect structural encoding
(Bentin, Allison, Puce, Perez, & McCarthy, 1996; Eimer, 1998).
The subsequent segments were chosen in order to evaluate the
N250r (230–330 ms) and the N400 (330–500 ms). For each time
segment, ANOVAs were performed analogous to those for the
performance data, except for the inclusion of an additional re-
peated measurement factor electrode (30 levels). Note that, be-
cause the average reference sets the mean activity across all
electrodes to zero, condition effects in these ANOVAs are only
meaningful in interaction with electrode. Therefore, we report
only such interactions butFfor brevity’s sakeFwithout men-
tioning the electrode factor. If significant effects of priming, fa-
miliarity, or of their interaction showed up in the ANOVAs,
additional analyses were performed on specific regions of interest
(ROIs) in order further to locate these effects. These ROIs were
chosen at regions for which the components in question are most
clearly visible, and were, for N250r (a) frontal (Fz, F3, F4), (b)
temporal (T7, T8, P7, P8, TP9, TP10, P9, P10), and for N400 (c)
central-parietal (Cz, Pz, C3, C4, P3, P4).
P1. The P1 component can be seen at the occipital electrodes
around 125 ms after target onset. No significant effects of the
experimental variables on the P1 amplitude were observed,
Fso1.8, p4.10.
N170. The N170 is most pronounced at occipito-temporal
electrodes. TheANOVAof theN170 amplitude yielded effects of
priming, F(29,435)5 2.6, po.05, e5 .18, and familiarity,
F(58,870)5 3.1, po.01, e5 .13, but no interaction between
these factors, F(58,870)5 1.4, p4.10, e5 .16. The effects of
priming and familiarity onN170 amplitudewere rather weak and
inconsistent and will not be elaborated upon further.3
N250r (early repetition effect). The N250r is to be seen best
at inferior temporal electrodes (e.g., TP10). Moreover, a polarity
inversion of this effect is prominent at frontal electrodes (e.g.,
Fz). Figure 5 depicts the repetition effects by showing the dif-
ference waveforms between ERPs to primed and unprimed faces
for the three conditions; the topographies of the repetition effects
for the different time segments are shown in Figure 6. Inspection
of these figures not only shows that (a) the repetition effects
reflect an electrical negativity for primed faces over temporal
electrodes and a positivity over fronto-central electrodes, and but
also suggests that (b) the early repetition effect increased in am-
plitude from unfamiliar over famous to personally familiar faces.
From around 230 ms onward priming caused increased negativ-
ity at inferior temporal electrodes.
The amplitude measures in both the 230–270- and 270–330-
ms segments yielded strong effects of priming, Fs(29,435)5 12.4
and 26.4, respectively, pso.001, es5 .16. There were also effects
694 G. Herzmann et al.
Table 3. Mean Correct Reaction Times (RT, inMilliseconds) and Percentages of Errors (PE)
for Personally Familiar, Famous, and Unfamiliar Target Faces in Part 2 of the Experiment
Familiarity type
Primed condition Unprimed condition
RT PE RT PE
M SD M SD M SD M SD
Personally familiar 530.2 102.6 6.0 5.2 663.6 81.9 6.3 6.6Famous 549.7 107.4 1.2 1.9 669.1 89.6 1.8 1.8Unfamiliar 642.7 108.6 3.1 3.6 685.5 102.1 1.7 2.2
2It is interesting that the observation of higher percentages of errors isin contrast with the finding of larger SCRs to personally familiar thanfamous faces. This would seem to exclude an explanation of the SCReffect in terms of overall better knowledge of personally familiar faces.
3Several studies (Bentin &Deouell, 2000; Eimer, 2000) have indicatedthat the N170 is insensitive to the familiarity of faces. Similarly, althoughthere are several studies that report face repetition effects for the N170 oreven earlier components (e.g., Braeutigam, Bailey, & Swithenby, 2001),one difficulty with these effects is that they appear to be rather incon-sistent across studies. Although we would not completely reject the pos-sibility that there may be small repetition effects on the N170, the presenteffects in this time range might also reflect temporal overlap with therising slope of the N250r, rather than differences in the N170 itself.
of familiarity, Fs(58,870)5 6.1 and 7.3, pso.001, es5 .20 and
.17, and significant interactions between priming and familiarity,
Fs(58,870)5 3.5 and 6.6, pso.001, es5 .16. Pairwise compari-
sons revealed significant priming effects at each level of famil-
iarity, all Fs(29,435)44.5, all ps o.01, all eso.19. Moreover,
priming effects in both time segments differed between personally
familiar and famous faces, Fs(29,435)5 5.0 and 4.3, po.001 and
po.01, es5 .28 and .18, respectively, and also between person-
ally familiar and unfamiliar faces, Fs(29,435)5 4.9 and 12.4,
pso.001, es5 .18 and .20, respectively. In contrast, priming ef-
fects differed between famous and unfamiliar faces in the 270–
330 ms segment, F(29,435)5 3.2, po.05, e5 .14, but not in the
preceding 230–270-ms segment, Fo1. It may be noted that in
both segments there were also significant familiarity-induced
modulations of amplitude for unprimed faces, Fs(29,435)5 5.2
and 4.0, pso.001, es5 .19 and .20. Bonferoni-corrected pairwise
comparison within the unprimed condition revealed significant
differences in amplitude between all familiarity conditions at the
230–270-ms segment, Fs(29,435)44.4, pso.01, eso.23. In the
270–330-ms time segment amplitudes differed significantly only
between personally familiar and unfamiliar, F(29,435)5 4.8,
po.001, eo.24, as well as between famous and unfamiliar faces,
F(29,435)5 5.8, po.001, eo.25, whereas personally familiar
faces did not differ from famous faces, Fo2.0.4
In the time segment between 230 and 270 ms the ROI analysis
at frontal electrodes revealed significant main effects of priming,
F(1,15)5 17.8, po.001, familiarity, F(2,30)5 6.7, po.01,
e5 1.13, and a trend for an interaction of priming and famili-
arity, F(2,30)5 2.6, p5 .10, e5 .81. This interaction was due to
priming effect differences between personally familiar and fa-
mous faces, F(1,15)5 7.5, po.05, but not for the other famil-
iarity comparisons, Fs(1,15)o1.7. In the same time segment,
temporal electrodes only revealed significant main effects of
priming, F(1,15)5 68.2, po.001, and familiarity, F(2,30)5
15.5, po.001, e5 .91, but no differences in priming effects
across familiarity conditions, F(2,30)5 1.6, p5 .22, e5 .95.
In the 270–330-ms segment the ROI analysis revealed similar
effects. At both frontal and temporal electrode sites the main
effects of priming and familiarity reached significance, pso.001.
Moreover, there was a significant interaction of familiarity and
priming at frontal electrodes, F(2,30)5 3.9, p5 .05, e5 .93, this
interaction again arising from priming effect differences between
personally familiar and famous faces, F(1,15)5 7.1, po.05.
Again, no significant interaction showed up at temporal elec-
trode sites, F(2,30)5 1.8, p5 .19, e5 .97.
To address the question of whether or not these data provide
evidence for separate face recognition units for the affective and
cognitive routes, the topographies of the N250r for each category
of facial familiarity were analyzed. If scalp topographies of the
N250r differ, one may conclude that at least some of the gen-
erator sources of this effect are different, lending support for the
idea of separate N250r. Interactions in ERP amplitudes of ex-
perimental variables with electrode site may derive from differ-
ences in the underlying neuronal source configuration only when
differences in source strength are ruled out. Therefore, ANOVAs
were calculated with factors familiarity and electrode site for the
N250r after scaling them to the same overall amplitude within
each condition with the average distance of the mean, derived
from the grand mean ERPs, as the divisor (McCarthy & Wood,
What’s special about personally familiar faces? 695
Figure 3. ERPs recorded for primed (solid lines) and unprimed (dashed lines) personally familiar target faces. Recordings are shown
for all 30 channels. Arrows indicate the P1, N170, the early repetition effect (N250r), and the late repetition effect (N400).
4The observed familiarity effects within the unprimed condition weresignificant but weaker than the N250r effect and showed a different to-pography. They appear to originate from a different source, and willtherefore not be analyzed any further.
1985). Recently, Urbach and Kutas (2002) criticized this proce-
dure as being potentially unreliable in its intended application of
identifying generator differences, particularly when overall base-
line differences exist between conditions, or whenmultiple sourc-
es are present simultaneously. In our experience, overall baseline
differences are not likely to be a strong concern in the present
study, which used average reference such that any overall dif-
ferences should be eliminated. In addition, dipole source local-
ization of theN250r to famous faces has suggested a single source
in the fusiform gyrus (Schweinberger, Pickering, Jentzsch, et al.,
2002). Nevertheless, when using scaling to investigate differences
in underlying neuronal source configuration, potential limita-
tions of this procedure should be kept in mind.
N250r topographies of priming effects differed significantly
across familiarity conditions in the 270–330-ms segment,
F(58,870)5 4.8, po.001, e5 .14. Figure 6 suggests that the pos-
itive aspect of the N250r peaked in midfrontal regions for un-
familiar faces, but for personally familiar and famous faces it was
more pronounced in more posterior central regions. Pairwise
comparisons confirmed differences in topography between un-
familiar and both personally familiar faces, F(29,435)5 9.0,
po.001, e5 .18, and famous faces, F(29,435)5 3.4, po.05,
e5 .17. However, N250r topographies for personally familiar
and famous faces were indistinguishable, F(29,435)5 2.5,
p4.10, e5 .14.
N400 (late repetition effect). In both the 330–400-ms and the
400–500-ms segment there were significant effects of priming,
Fs(29,435)5 22.6 and 5.0, po.001 and po.01, es5 .17 and .15,
of familiarity, Fs(58,870)5 4.1 and 2.6, pso.001, es5 .19 and
.17, and of their interactions, Fs(58,870)5 9.0 and 2.0, po.001
and po.05, es5 .18, respectively. These priming effects resemble
an N400-like modulation of the late positive complex, with more
negativity (or less positivity) for unprimed than primed faces at
central-parietal locations, and less negativity at prefrontal and
lateral frontal locations (see Figures 3–5). Furthermore, Figure 5
suggests that between 330 and 400 ms the N400 priming effect
(i.e., the difference between primed and unprimed faces) in-
creased in amplitude from unfamiliar over famous to personally
familiar faces. In the 330–400-ms segment, priming was signif-
icant at each level of familiarity, Fs(29,435)44.6, pso.001,
es5 .19. Post hoc comparisons showed differences in the N400
696 G. Herzmann et al.
Figure 4. ERPs for personally familiar (first column), famous (second column), and unfamiliar target faces (third column)
comparing primed (solid lines) and unprimed conditions (dashed lines) at themost important electrode sites (Pz, Fz, TP10, and P10).
Vertical timelines indicate the areas of the early repetition effect, which were used in the ANOVA (230–270 ms and 270–330 ms).
priming effect between personally familiar and famous,
F(29,435)5 3.7, p5 .01, e5 .27, personally familiar and unfa-
miliar, F(29,435)5 15.3, p5 .001, e5 .21, and between famous
and unfamiliar faces, F(29,435)5 6.2, p5 .001, e5 .16. In the
400–500-ms segment, priming was significant for personally
familiar, F(29,435)5 4.9, po.001, e5 .20, and famous faces,
F(29,435)5 3.5, po.05, e5 .17, but was reduced to insignifi-
cance for unfamiliar faces,F(29,435)5 2.7, p4.10, e5 .10. ROIs
analysis of the 330–400-ms time segment revealed significant
main effects of priming, F(1,15)5 58.2, p5 .001, and familiarity,
F(2,30)5 8.8, p5 .001, e5 1.15, as well as an interaction be-
tween priming and familiarity at central-parietal electrodes,
F(2,30)5 23.9, p5 .001, e5 .93. Pairwise comparison showed
differences in priming effects between personally familiar and
famous, F(1,15)5 9.5, p5 .01, personally familiar and unfamil-
iar, F(1,15)5 48.4, p5 .001, and between famous and unfamiliar
faces, F(1,15)5 13.6, p5 .01. However, in the 400–500-ms
ROI analysis at central-parietal electrodes revealed only signif-
icant main effects of priming, F(1,15)5 9.6, p5 .01, and famil-
iarity, F(2,30)5 5.7, p5 .01, e5 .96, but no significant
interaction of both, Fo1.8. Finally, N400 topographies of prim-
ing effects (applying the scaling procedure described above)
did not differ significantly across familiarity, Fs(58,870)o2.9,
ps4.05.
Discussion
At the level of RTwe found least priming for unfamiliar faces and
larger but equivalent priming for personally familiar and famous
faces. There appears to be no advantage in the cognitive aspects
of face processing for personally familiar compared with famous
faces.
The ERP data revealed an early repetition effect (N250r)
around 230–330 ms that was negative at posterior temporal
electrodes but positive at mid-frontal sites, replicating previous
reports (Pfutze et al., 2002; Schweinberger et al., 1995;
Schweinberger, Pickering, Jentzsch, et al., 2002). Most impor-
tantly, N250r amplitude was significantly modulated by facial
familiarity. It was smallest for unfamiliar faces, increased to fa-
mous faces (see Pfutze et al., 2002 for similar results), and was
largest for personally familiar faces. The sensitivity of the N250r
to familiarity indicates that it is not simply a correlate of the
repetition of the stimulus but reflects the contact with memory
representations for faces. Furthermore affective dimensions,
such as likeability or attractiveness, could be thought to have
modulated the observed N250r amplitude effects, especially for
personally familiar faces. However, the present interstimulus in-
terval between prime and target of 1,300 ms was much too long
for possible affective priming, which would require an interstim-
ulus interval below 300ms (Greenwald, Klinger, & Schuh, 1995).
Decisions about the familiarity of a face are supposed to be
made at the person identity node level (Burton et al., 1999).
Repetition priming has been proposed to involve processing fa-
cilitation already during the access to face recognition units or
even structural encoding (Pfutze et al., 2002; Schweinberger et
al., 1995; Schweinberger, Pickering, Jentzsch, et al., 2002). In this
respect, the observed amplitude differences of the N250r for
personally familiar compared with famous or unfamiliar faces
could be a result of more widespread and stronger neural net-
works coding these faces. These network differences might result,
for example, from a more extensive range of visual experience
with personally familiar faces, compared with famous faces.
Stronger networks can be assumed to expedite information
processing by reducing the threshold for face recognition (Mohr
et al., 2002). In contrast to the quantitative differences in N250r
amplitudes, the topographies of N250r for personally familiar
and famous faces were indistinguishable. This is consistent with
What’s special about personally familiar faces? 697
Figure 5. ERP difference waves (primedminus unprimed) for personally
familiar (solid lines), famous (dashed lines), and unfamiliar target faces
(dotted lines) at the most important electrode sites (Pz, Fz, TP10, and
P10). Vertical timelines indicate the areas of the early repetition effect,
which were used in the ANOVA (230–270 ms and 270–330 ms).
the notion of similar underlying neural sources. The only top-
ographical difference found was between the N250r for unfamil-
iar faces and both personally familiar and famous faces, confirm-
ing earlier suggestions of qualitatively different underlyingmech-
anism for unfamiliar faces (Hancock, Bruce, & Burton, 2000).
The present N400 priming effect broadly replicates the N400-
like effect observed before (Bentin & McCarthy, 1994; Schwein-
berger et al., 1995; Schweinberger, Pickering, Burton, et al.,
2002). For the present data, this interpretation must be seen with
some restrictions because the N400 may have been partially in-
fluenced by a latency shift in a late positive ERP component (see
Figures 3 and 4) especially when RTs differed between condi-
tions. In the absence of such RT differences, in the 330–400-ms
segment the N400 priming effect was more pronounced for per-
sonally familiar than for famous faces, as can be seen in Figure 5.
According to network theories of face recognition (Burton et al.,
1999), this could reflect more elaborated semantic processing for
personally familiar than for famous people.
One must be aware that because of the small number of crit-
ical stimuli and of the preceding SCR experiment, each stimulus
was repeated four times across the two experiments. Multiple
repetitions could be assumed to have made the unfamiliar faces
familiar for the participants and consequently may have influ-
ences the RT and ERP results. However, multiple repetitions
were used in all conditions and therefore should not have caused
differences between conditions. In addition, although an influ-
ence of repetitions over longer time intervals is usually found for
N400, N250r reflects a more transient effect and seems to be
insensitive to repetitions over longer intervals (Schweinberger,
Pickering, Burton, et al., 2002).
698 G. Herzmann et al.
Figure 6. Topographical voltage maps of ERP differences between primed and unprimed conditions showing priming effects for
personally familiar (left column), famous (middle), and unfamiliar faces (right column) at different latencies after target
presentation. All maps were obtained by using spherical spline interpolation. Equipotential lines are separated by 1 mV/line;negativity is shaded.
General Discussion
Within the framework of dual-route models (Breen et al., 2000;
Ellis & Lewis, 2001), the present study had two main objectives.
Objective 1
In the present study we explored the effects of three different
types of facial familiarity on affective as well as cognitive aspects
of face recognition, and on the face recognition units, using
SCRs, priming effects, and the N250r.
We found that face familiarity caused different effects in SCRs
and RT priming. Whereas stronger cognitive effects (RT prim-
ing) were elicited by famous than unfamiliar faces, this was not
the case for the autonomic responses (SCR). Strong autonomic
responses were only triggered for personally known faces for
which, however, cognitive activation did not exceed the level
attained for famous faces.
According to dual-route models, the SCR is assumed to rep-
resent the activation within an affective route. However, these
models are not very specific about what is meant with affect: Is it
an emotional response to the familiar face or is it a reflection of
the significance of the displayed person for the observer? It
should also be kept in mind that, although SCRs are a result of
activation of the sympathetic autonomic nervous system, they
are rather unspecific responses, which can be caused by a number
of affective as well as cognitive processes. Therefore, cognitive
components of face processing could, in principle, have contrib-
uted to the observed SCR results.
Electrodermal responses have been suggested to be triggered
by processes such as novelty, emotion, or significance of the
stimulus (e.g., Dawson, Schell, & Filion, 2000). In the setting of
the present study, it is unlikely that novelty has triggered the
obtained results of larger SCRs to personally familiar faces. It
can be assumed that the participants have seen their lecturers as
frequently as or even more often than the shown celebrities and,
of course, also more often than the unfamiliar faces. Because
personally familiar and famous faces were both used as targets
and placed into the same category, they had the same significance
with respect to the task, yet they differed in SCRs. Therefore the
‘‘affective’’ response reflected in the SCRs to personally familiar
faces might relate to emotional responses or to the ‘‘personal
significance’’ for the observer. In principle, it may be possible
that the participants experienced a specific emotional response to
their lecturers as their portraits were shown such as joy, fear, and
so forth. However, to us it seems to be more plausible that the
observed SCRs may reflect the importance or significance of the
portrayed lecturers for our participantsFtheir students. There-
fore, the SCRs may represent the significance of the stimuli as
defined by the personal background of the observer rather than
by the experimental task.
The amplitude of the N250r exhibited reliable quantitative
differences in the face recognition units between all three types of
familiarity. Showing a continuous increase from unfamiliar over
famous to personally known faces, this result could point to the
assumptions of increasing face recognition units strength with
facial familiarity.
It may be argued that lecturers may not be typical instances
for personally familiar people, as might be family members or
friends. Although it would be desirable to replicate the present
findings with these kinds of stimuli, the pictures employed here
did elicit significant effects, especially in the N250r amplitudes,
which are thought to be a sign of stored facial structures. In SCRs
and the N250r, personally familiar faces were found to cause the
largest effects. This can additionally be caused by their high (so-
cial) importance, which lecturers as authorities bear for their
students. These results confirm the notion that personally famil-
iar compared with famous faces encompass especially an advan-
tage in affective aspects of face processing and may have a
stronger network of face recognition units.
Objective 2
We also addressed the question of whether there are unitary or
multiple face recognition mechanisms, or face recognition units,
feeding into the affective and cognitive routes of face processing,
using both functional and neural information of the early rep-
etition effect (N250r).
We found that the N250r, taken to reflect face recognition
unit activation, did not differ in its neural underpinnings between
personally familiar and famous faces. Although a quantitatively
larger N250r was seen for personally familiar faces, the scalp
topographies of the N250r to personally familiar and famous
faces were indistinguishable. Recent brain electric source anal-
yses of the N250r are consistent with a generator in fusiform
gyrus (Schweinberger, Pickering, Jentzsch, et al., 2002), an area
that is strongly implicated in face recognition (Kanwisher,
McDermott, & Chun, 1997) and face repetition priming (Hen-
son, Shallice, & Dolan, 2000).
Before taking this result as evidence for a common face
recognition unit, one needs to be aware of certain limitations of
ERP research. In particular, whereas topographical differences
between two ERPs or ERP effects indicate that these were gen-
erated by at least partially different sources, there might beFat
least in principleFa number of reasons for the absence of such
differences. For an ERP to be recordable at the scalp, simulta-
neous activity of a vast number of suitably aligned neurons is
required. Although this is often the case for cortical sources, it
may not hold true in many subcortical sources (Wood & Allison,
1981). We believe that this problem is unlikely to have biased the
present interpretation, as scalp-recorded ERPs can likely be re-
corded from key structures in the dorsal route, specifically the
superior temporal sulcus (Puce, Smith, & Allison, 2000) and
cingulate gyrus (Badgaiyan & Posner, 1998). However, respons-
es to personally familiar faces along the affective route would
include activity in ventral limbic structures (and amygdala in
particular), and it is well possible that such activitymight bemore
difficult to pick up. Despite these caveats, we would like to em-
phasize that the present ERP data did show topographical dif-
ferences in N250r between unfamiliar faces and both famous and
personally familiar faces. This not only lends support to the idea
that the recognition of familiar and unfamiliar faces is governed
by a qualitatively different mechanism (Hancock et al., 2000), it
also suggests that the equivalence of the topographies for per-
sonally and famous faces is unlikely to be a result of insufficient
power to find such topographical differences in the present ex-
periment.
Using the framework of dual-route models to test the influ-
ence of facial familiarity on face processing reinforced the sug-
gestion that face recognition is the result of the interaction of
several components in the information processing system. In
addition, the results of the study strengthen the notion of one
single set of face recognition units, feeding into both the affective
and cognitive route. The current findings are consistent with the
views of Breen et al. (2000) and Ellis and Lewis (2001) that the
What’s special about personally familiar faces? 699
initial parts of face recognition take place along a single ana-
tomical routeFthe ventral route.
Finally, with the present study we provide insight into the class
of familiar faces by adding findings about personally familiar faces
to the existing results of familiarity-induced effects on SCR, RT
priming, and the N250r. For personally familiar faces, more
widespread and elaborative networks may be assumed for (a)
stored representations of facial structures (face recognition unit),
as indicated by larger effects of the N250r amplitude; and (b)
stored semantic codes, as indicated by a higher amplitude of the
N400. Additionally, these faces are associated with a stronger af-
fective response, as seen in larger SCRs. However, as the topog-
raphies of the N250r to personally familiar and famous faces did
not differ, recognition of both types of familiar faces can be
thought to be triggered essentially by the same neural generator,
most likely the fusiform gyrus (Schweinberger, Pickering,
Jentzsch, et al., 2002).
To conclude, by using a multidimensional methodological
approach we have been able to give a comprehensive insight into
modulations of face recognition by facial familiarity. In partic-
ular, we have provided new evidence that face recognition units
may serve as common modules feeding into both affective and
cognitive routes. The strength of the networks representing face
recognition units appears to increase as a function of the degree
of facial familiarity.
REFERENCES
Badgaiyan, R. D., & Posner, M. I. (1998). Mapping the cingulate cortexin response selection and monitoring. NeuroImage, 7, 255–260.
Bauer, R. M. (1984). Autonomic recognition of names and faces inprosopagnosics: a neuropsychological application of the GuiltyKnowledge Test. Neuropsychologia, 22, 457–469.
Bentin, S., Allison, T., Puce, A., Perez, E., & McCarthy, G. (1996).Electrophysiological studies of face perception in humans. Journal ofCognitive Neuroscience, 8, 551–565.
Bentin, S., & Deouell, L. Y. (2000). Structural encoding and identifica-tion in face processing: ERP evidence for separate mechanisms. Cog-nitive Neuropsychology, 17, 35–54.
Bentin, S., & McCarthy, G. (1994). The effects of immediate stimulusrepetition on reaction time and event-related potentials in tasks ofdifferent complexity. Journal of Experimental Psychology: Learning,Memory, and Cognition, 20, 130–149.
Braeutigam, S., Bailey, A. J., & Swithenby, S. J. (2001). Task-dependentearly latency (30–60 ms) visual processing of human faces and otherobjects. NeuroReport, 12, 1531–1536.
Breen, N., Caine, D., & Coltheart, M. (2000).Models of face recognitionand delusional misidentification: A critical review. Cognitive Neuro-psychology, 17, 55–71.
Bruce, V., & Valentine, T. (1985). Identity priming in the recognition offamiliar faces. British Journal of Psychology, 76, 373–383.
Bruce, V., & Young, A. (1986). Understanding face recognition. BritishJournal of Psychology, 77, 305–327.
Burton, A. M., Bruce, V., & Hancock, P. J. B. (1999). From pixels topeople: Amodel of familiar face recognition.Cognitive Science, 23, 1–31.
Dawson, A. E., Schell, A. M., & Filion, D. L. (2000). The electrodermalsystem. In J. T. Cacioppo, L. G. Tassinary, & G. G. Berntson (Eds.),Handbook of psychophysiology (2nd ed, pp. 200–223). Cambridge,UK: Cambridge University Press.
Eimer, M. (1998). Does the face-specific N170 component reflect theactivity of a specialized eye processor? NeuroReport, 9, 2945–2948.
Eimer, M. (2000). Event-related brain potentials distinguish processingstages involved in face perception and recognition. Clinical Neuro-physiology, 111, 694–705.
Elbert, T., Lutzenberger, W., Rockstroh, B., & Birbaumer, N. (1985).Removal of ocular artifacts from the EEGFA biophysical approachto the EOG. Electroencephalography and Clinical Neurophysiology,60, 455–463.
Ellis, H. D., & Lewis, M. B. (2001). Capgras delusion: A window on facerecognition. Trends in Cognitive Sciences, 5, 149–156.
Ellis, H. D., Lewis, M. B., Moselhy, H. F., & Young, A. W. (2000).Automatic without autonomic responses to familiar faces: Differen-tial components of implicit face recognition in a case of Capgrasdelusion. Cognitive Neuropsychiatry, 5, 255–269.
Ellis, H. D., & Young, A. W. (1990). Accounting for delusional mis-identifications. British Journal Of Psychiatry, 157, 239–248.
Ellis, H. D., Young, A. W., Quayle, A. H., & DePauw, K. W. (1997).Reduced autonomic responses to faces in Capgras delusion. Proceed-ings of the Royal Society of London B, 264, 1085–1092.
Greenwald, A. G., Klinger, M. R., & Schuh, E. (1995). Activation bymarginally perceptible (‘‘subliminal’’) stimuli: Dissociation of uncon-scious from conscious cognition. Journal of Experimental Psychology:General, 124, 22–42.
Hancock, P. J. B., Bruce, V., & Burton, A. M. (2000). Recognition ofunfamiliar faces. Trends in Cognitive Sciences, 4, 330–337.
Hay, D. C., & Young, A. W. (1982). The human face. In H. D. Ellis(Ed.), Normality and pathology in cognitive functions (pp. 173–202).London, New York: Academic Press.
Henson, R., Shallice, T., &Dolan, R. (2000). Neuroimaging evidence fordissociable forms of repetition priming. Science, 287, 1269–1272.
Hirstein, W., & Ramachandran, V. S. (1997). Capgras syndrome: Anovel probe for understanding the neural representation of the iden-tity and familiarity of persons. Proceedings of the Royal Society ofLondon B, 264, 437–444.
Huynh, H., & Feldt, L. S. (1976). Estimation of the Box correction fordegrees of freedom from sample data in randomized block and split-plot designs. Journal of Educational Statistics, 1, 69–82.
Kanwisher, N., McDermott, J., & Chun, M. M. (1997). The fusiformface area: A module in human extrastriate cortex specialized for faceperception. The Journal of Neuroscience, 17, 4302–4311.
Kutas, M., & Hillyard, S. A. (1980). Reading senseless sentences: Brainpotentials reflect samantic incongruity. Science, 207, 203–205.
McCarthy, G., & Wood, C. C. (1985). Scalp distributions of event-re-lated potentials: An ambiguity associated with analysis of variancemodels. Electroencephalography and Clinical Neurophysiology, 62,203–208.
Mohr, B., Landgrebe, A., & Schweinberger, S. R. (2002). Interhemi-spheric cooperation for familiar but not unfamiliar face perception.Neuropsychologia, 40, 1841–1848.
Ohmann, A. (1979). The orienting response, attention, and learning: Aninformation processing perspective. In H. D. Kimmel, E. H. vanOlst,& J. F. Orlebeke (Eds.), The orienting reflex in humans (pp. 443–472).Hillsdale, NJ: Erlbaum.
Oldfield, R. C. (1971). The assessment and analysis of handedness: TheEdinburgh inventory. Neuropsychologia, 9, 97–113.
Pfutze, E.-M., Sommer, W., & Schweinberger, S. R. (2002). Age-relatedslowing in face and name recognition: Evidence from event-relatedbrain potentials. Psychology and Aging, 17, 140–160.
Pickering, E. C., & Schweinberger, S. R. (2003). N200, N250r and N400event-related brain potentials reveal three loci of repetition primingfor familiar names. Journal of Experimental Psychology: Learning,Memory, and Cognition, 29, 1298–1311.
Pivik, R. T., Broughton, R. J., Coppola, R., Davidson, R. J., Fox, N., &Nuwer, M. R. (1993). Guidelines for the recording and quantitativeanalysis of electroencephalographic activity in research contexts.Psychophysiology, 30, 547–558.
Puce, A., Smith, A., & Allison, T. (2000). ERPs evoked by viewing facialmovements. Cognitive Neuropsychology, 17, 221–239.
Schweinberger, S. R. (1996). How Gorbachev primed Yeltsin: Analysesof associative priming in person recognition by means of reactiontimes and event-related brain potentials. Journal of ExperimentalPsychology: Learning, Memory, and Cognition, 22, 1383–1407.
Schweinberger, S. R., Pfutze, E.-M., & Sommer, W. (1995). Repetitionpriming and associative priming of face recognition: Evidence fromevent-related potentials. Journal of Experimental Psychology: Learn-ing, Memory, and Cognition, 21, 722–736.
Schweinberger, S. R., Pickering, E. C., Burton, A. M., & Kaufmann, J.M. (2002). Human brain potential correlates of repetition priming inface and name recognition. Neuropsychologia, 40, 2057–2073.
700 G. Herzmann et al.
Schweinberger, S. R., Pickering, E. C., Jentzsch, I., Burton, A. M., &Kaufmann, J. M. (2002). Event-related brain potential evidence for aresponse of inferior temporal cortex to familiar face repetitions. Cog-nitive Brain Research, 14, 398–409.
Tranel, D., &Damasio, A. R. (1985). Knowledge without awareness: Anautonomic index of facial recognition by prosopagnosics. Science,228, 1453–1454.
Tranel, D., Damasio, H., & Damasio, A. R. (1995). Double dissociationbetween overt and covert face recognition. Journal of CognitiveNeuroscience, 7, 425–432.
Tranel, D., Fowles, D. C., & Damasio, A. R. (1985). Electrodermaldiscrimination of familiar and unfamiliar faces: A methodology. Psy-chophysiology, 22, 403–408.
Ungerleider, L., &Mishkin,M. (1982). Two cortical visual systems. InD.J. Ingle, M. A. Goodale, & R. J. W. Mansfield (Eds.), Analysis ofvisual behavior (pp. 549–586). Cambridge, MA: MIT Press.
Urbach, T. P., & Kutas, M. (2002). The intractability of scaling scalpdistributions to infer neuroelectric sources. Psychophysiology, 39,791–808.
Wood, C. C., & Allison, T. (1981). Interpretation of evoked potentials: Aneurophysiological perspective. Canadian Journal of Psychology, 35,113–135.
(Received July 31, 2003; Accepted January 5, 2004)
What’s special about personally familiar faces? 701
top related