Age-related differences in brain electrical activity during extended continuous face recognition in younger children, older children and adults.
ABSTRACT To examine the development of recognition memory in primary-school children, 36 healthy younger children (8-9 years old) and 36 healthy older children (11-12 years old) participated in an ERP study with an extended continuous face recognition task (Study 1). Each face of a series of 30 faces was shown randomly six times interspersed with distracter faces. The children were required to make old vs. new decisions. Older children responded faster than younger children, but younger children exhibited a steeper decrease in latencies across the five repetitions. Older children exhibited better accuracy for new faces, but there were no age differences in recognition accuracy for repeated faces. For the N2, N400 and late positive complex (LPC), we analyzed the old/new effects (repetition 1 vs. new presentation) and the extended repetition effects (repetitions 1 through 5). Compared to older children, younger children exhibited larger frontocentral N2 and N400 old/new effects. For extended face repetitions, negativity of the N2 and N400 decreased in a linear fashion in both age groups. For the LPC, an ERP component thought to reflect recollection, no significant old/new or extended repetition effects were found. Employing the same face recognition paradigm in 20 adults (Study 2), we found a significant N400 old/new effect at lateral frontal sites and a significant LPC repetition effect at parietal sites, with LPC amplitudes increasing linearly with the number of repetitions. This study clearly demonstrates differential developmental courses for the N400 and LPC pertaining to recognition memory for faces. It is concluded that face recognition in children is mediated by early and probably more automatic than conscious recognition processes. In adults, the LPC extended repetition effect indicates that adult face recognition memory is related to a conscious and graded recollection process rather than to an automatic recognition process.
- [Show abstract] [Hide abstract]
ABSTRACT: The neuropeptides oxytocin and vasopressin are evolutionarily conserved regulators of social perception and behavior. Evidence is building that they are critically involved in the development of social recognition skills within rodent species, primates, and humans. We investigated whether common polymorphisms in the genes encoding the oxytocin and vasopressin 1a receptors influence social memory for faces. Our sample comprised 198 families, from the United Kingdom and Finland, in whom a single child had been diagnosed with high-functioning autism. Previous research has shown that impaired social perception, characteristic of autism, extends to the first-degree relatives of autistic individuals, implying heritable risk. Assessments of face recognition memory, discrimination of facial emotions, and direction of gaze detection were standardized for age (7-60 y) and sex. A common SNP in the oxytocin receptor (rs237887) was strongly associated with recognition memory in combined probands, parents, and siblings after correction for multiple comparisons. Homozygotes for the ancestral A allele had impairments in the range -0.6 to -1.15 SD scores, irrespective of their diagnostic status. Our findings imply that a critical role for the oxytocin system in social recognition has been conserved across perceptual boundaries through evolution, from olfaction in rodents to visual memory in humans.Proceedings of the National Academy of Sciences 12/2013; · 9.81 Impact Factor
- [Show abstract] [Hide abstract]
ABSTRACT: Although, in everyday life, patients with attention deficit hyperactivity disorder (ADHD) are frequently distracted by goal-irrelevant affective stimuli, little is known about the neural and behavioral substrates underlying this emotional distractibility. Because some of the most important brain responses associated with the sudden onset of an emotional distracter are characterized by their early latency onset and short duration, we addressed this issue by using a temporally agile neural signal capable of detecting and distinguishing them. Specifically, scalp event-related potentials (ERPs) were recorded while 20 boys with ADHD combined type and 20 healthy comparison subjects performed a digit categorization task during the presentation of three types of irrelevant, distracting stimuli: arousing negative (A-), neutral (N) and arousing positive (A+). Behavioral data showed that emotional distracters (both A- and A+) were associated with longer reaction times than neutral ones in the ADHD group, whereas no differences were found in the control group. ERP data revealed that, compared with control subjects, boys with ADHD showed larger anterior N2 amplitudes for emotional than for neutral distracters. Furthermore, regression analyses between ERP data and subjects' emotional ratings of distracting stimuli showed that only in the ADHD group, emotional arousal (ranging from calming to arousing) was associated with anterior N2: its amplitude increased as the arousal content of the visual distracter increased. These results suggest that boys with ADHD are more vulnerable to the distracting effects of irrelevant emotional stimuli than control subjects. The present study provides first data on the neural substrates underlying emotional distractibility in ADHD.Brain and Cognition 07/2013; 83(1):10-20. · 2.82 Impact Factor
- [Show abstract] [Hide abstract]
ABSTRACT: Previous studies in adults have revealed that attentional distraction modulates the late positive potential (LPP) during emotion regulation. To determine whether early visual components reflect developmental changes in attentional distraction during adolescence, we collected event-related potentials from 20 young adolescents, 18 older adolescents, and 18 young adults as they performed a distraction task (counting) while viewing affective images. Consistent with previous findings obtained in distraction studies, the distraction task (counting) reduced emotional modulation of the LPP. At an early stage of processing, counting reduced emotional modulation of P1 and increased the negativity bias of early frontal negativity (eFN) for negatively valenced pictures compared to simple viewing with no distraction. sLORETA analyses further revealed eFN indexing of rostral prefrontal cortical activation, a cortical area that has been shown in recent fMRI studies to be activated by distraction. Moreover, P1 amplitudes in young and older adolescents did not differ but were both larger than the P1s in young adults. In addition, eFN amplitudes significantly decreased with age. The dissociable distraction patterns between the posterior P1 and eFN provide evidence not only for the timing hypothesis of emotion regulation but also for different developmental trajectories of visual processing areas and the prefrontal cortex during affective processing in adolescence.Brain and Cognition 06/2014; 87:30–38. · 2.82 Impact Factor
Age-related differences in brain electrical activity during extended continuous
face recognition in younger children, older children, and adults
Jan W. Van Strien1
Johanna C. Glimmerveen2
Ingmar H.A. Franken1
Vanessa E.G. Martens3
Eveline A. de Bruin3
1Institute of Psychology, Faculty of Social Sciences, Erasmus University Rotterdam, The
2School of Social and Behavioral Sciences, Tilburg University, The Netherlands
3Sensation, Perception and Behaviour, Unilever R&D Vlaardingen, The Netherlands
To examine the development of recognition memory in primary-school children, 36 healthy
younger children (8-9 years old) and 36 healthy older children (11-12 years old) participated in
an ERP study with an extended continuous face recognition task (study 1). Each face of a series
of 30 faces was shown randomly for six times interspersed with distracter faces. The children
were required to make old vs. new decisions. Older children responded faster than younger
children, but younger children exhibited a steeper decrease in latencies across the five repetitions.
Older children exhibited better accuracy for new faces, but there were no age differences in
recognition accuracy for repeated faces. For the N2, N400 and late positive complex (LPC), we
analyzed the old/new effects (repetition 1 vs. new presentation) and the extended repetition
effects (repetitions 1 through 5). Compared to older children, younger children exhibited larger
frontocentral N2 and N400 old/new effects. For extended face repetitions, negativity of the N2
and N400 decreased in a linear fashion in both age groups. For the LPC, an ERP component
thought to reflect recollection, no significant old/new or extended repetition effects were found.
Employing the same face recognition paradigm in 20 adults (study 2), we found a significant
N400 old/new effect at lateral frontal sites and a significant LPC repetition effect at parietal sites,
with LPC amplitudes increasing linearly with the number of repetitions. This study clearly
demonstrates differential developmental courses for the N400 and LPC pertaining to recognition
memory for faces. It is concluded that face recognition in children is mediated by early and
probably more automatic than conscious recognition processes. In adults, the LPC extended
repetition effect indicates that adult face recognition memory is related to a conscious and graded
recollection process rather than to an automatic recognition process.
Key words: Event-related potentials; Old/new effect; P2; N400; Late positive complex (LPC);
Face recognition; Memory; Child development; extended continuous repetition
Behavioral research has demonstrated that recognition memory improves from infancy
into young adulthood (e.g., Billingsley, Smith, & McAndrews, 2002; Cycowicz, Friedman,
Snodgrass, & Duff, 2001). It has been suggested that the increase of recognition memory
proficiency with age is associated with differential developmental trajectories of two types of
retrieval processes: a fast familiarity process and a slower recollection process. With recollection,
the context in which information is presented is also remembered, while with familiarity it is not
(e.g., Curran, Tepe, & Piatt, 2006). Better recollection is associated with age-related increases of
activations in the dorsolateral prefrontal cortex (Ofen et al., 2007). As a consequence of the
longer maturational course of the frontal cortex, recollection may display a longer developmental
trajectory than familiarity (Cycowicz, 2000). Familiarity is usually thought to be a form of
explicit memory (see Rugg et al., 1998). People ‘know’ they have seen a stimulus before but can
not retrieve the details of the context of the first stimulus presentation. In research, the
dissociation of familiarity and recollection processes is typically observed with specific source
memory or ‘remember/know’ paradigms.
Numerous event-related potential (ERP) studies have supported a dual-process account of
recognition memory. When participants have to recognize studied items from a list of targets and
distracters, the correct identification of studied items is associated with an increased positivity of
the ERP waveform starting about 300 ms after stimulus onset. This ERP old/new effect typically
comprises an early mid-frontal component (N400 old/new effect, time window 300 to 500 ms)
that is thought to reflect familiarity and a later parietal component (late positive potential – LPC –
old/new effect, time window 500 to 800 ms) that is thought to reflect recollection (Curran et al.,
2006; Rugg & Curran, 2007).
With a continuous recognition paradigm, a similar dissociation of the early and late ERP
old/new effect has been established. Since in this paradigm old and new items are intermixed, the
early old/new effect may be akin to a more implicit recognition process that can still be
dissociated from a slower, explicit recollection process (Van Strien, Glimmerveen, Martens, &
De Bruin, 2009; Van Strien, Hagenbeek, Stam, Rombouts, & Barkhof, 2005). Using an extended
continuous-word-recognition paradigm in which each item was repeated nine times, Van Strien et
al. (2005) demonstrated that the positivity of the LPC increased linearly with the number of
repetitions. By contrast, the early N400 old/new effect was not affected by the number of
repetitions. These results suggested that the LPC old/new effect reflects a graded recollection
process that depends on the strength of the memory trace, whereas the early N400 old/new effect
reflects an automatic matching process that is not dependent on memory strength. The N400
old/new effect in the study by Van Strien et al. was observed at mid-parietal rather than mid-
frontal electrode positions. In another continuous word-recognition study, Van Strien et al. (2007)
found that this parietal N400 old/new effect was much larger with immediate than with delayed
repetitions. It is therefore likely that with continuous recognition, the N400 old/new effect
reflects implicit memory rather than familiarity. This is in concordance with Rugg and colleagues
(Rugg & Curran, 2007; Rugg et al., 1998), who also proposed that the early parietal old/new
effect is a neural correlate of implicit memory processes.
Earlier old/new effects than the N400 have also been observed in old/new paradigms.
With continuous face recognition, Guillem et al. (Guillem, Bicu, & Debruille, 2001) found a N2
old/new effect around 246 ms, with the N2 going more positive to old than new items. Van Strien
et al. (Van Strien, Langeslag, Strekalova, Gootjes, & Franken, 2009) found a similar N2 effect in
the 200-300 ms time window with continuous picture recognition. The early and later ERP
old/new effects may reflect the temporal and spatial course of consecutive recognition processes
such as perceptual matching/priming, familiarity, and recollection (see Ally & Budson, 2007, for
a proposed EEG model).
As the adult ERP studies demonstrated different neural correlates for an early implicit
recognition process and a later recollection process, developmental studies have provided some
evidence for a differential development of these correlates. Czernochowski et al. (2005) found
early frontal old/new effects for pictures in adults only, and later parietal old/new effects in
younger and older children, and adults. They concluded that children predominantly rely on
recollection, even though the prefrontal memory processes are not fully matured. Employing an
extended continuous word recognition task in which each word was repeated five times, Van
Strien et al. (Van Strien, Glimmerveen et al., 2009) found a much stronger N400 old/new effect
in 11- and 12-year-old children than in 8- and 9-year-old children. Both age groups exhibited
comparable LPC old/new and extended repetition effects. This LPC old/new effect suggested that
recollection plays a substantial role in word recognition memory of both younger and older
children, while the N400 old/new effect suggested that additional semantic representations may
be available for more automatic word recognition memory in older children. Several other ERP
studies found no developmental trends for the early and late old/new effects (Berman, Friedman,
& Cramer, 1990; Cycowicz, Friedman, & Duff, 2003).
Because recognition memory will depend on the development of a particular cognitive
domain, the developmental ERP trends could be affected by the type of stimuli that is used. For
instance, the early and late old/new effects for word recognition memory may be connected with
the development of other verbal skills such as reading or semantic processing (Van Strien,
Glimmerveen et al., 2009), while old/new effects for face recognition memory may be associated
with the development of facial processing abilities.
Face recognition is an important type of recognition memory, especially in social
interactions. Although even very young children are capable of recognizing their own mother's
face (De Haan & Nelson, 1997; Ellis, 1992), face recognition in children is relatively poor when
compared to face recognition in adults. The gradual increase in face recognition performance
with age is traditionally thought to be due to increasing configural processing abilities (a view
dating back to Diamond & Carey, 1977), which allow adults to better grasp the relationships
between facial features. Recent research however, suggests that the configural processing abilities
already are mature in four- to seven-year-old children (Crookes & McKone, 2009; de Heering,
Houthuys, & Rossion, 2007) and that the psychophysiological correlates of face-sensitive
perceptual processes do not change from 4 years to adulthood (Kuefner, de Heering, Jacques,
Palmero-Soler, & Rossion, 2009).
Only few ERP studies investigated the development of face recognition memory in
children. Itier and Taylor (2004) presented upright, inverted and contrast-reversed faces to 8-16
year old children. One-third of the faces was repeated immediately or after one intervening face.
Old/new effects were found in the 250-500 ms time window and were not influenced by age or
face types. The old/new effect was larger for immediately repeated faces than for 1-lag repeated
faces (cf. Van Strien et al., 2007). The authors concluded that in all age groups a comparable
general working memory system was involved. The steady improvement in face recognition from
8 to 16 years for both upright and inverted faces led Itier and Taylor to conclude that the
increasing face recognition performance with increasing age was driven by general memory
improvements rather than by increased configural processing.
In another continuous-recognition ERP study with words and faces, Hepworth et al.
(2001) found a late parietal P3 old/new effect (peaking around 550 ms) in 11-14 year old
children, which was significantly larger for words than for faces over the left parietal region.
Further, they found increased latencies and decreased amplitudes for the early ERP components
to faces compared to words. The authors concluded that their findings demonstrated that for 11-
14 year olds, the early processing of faces is more difficult than the early processing of words, the
latter being comparable to adults.
As the results of Itier and Taylor (2004) imply a role of early automatic processes and the
results of Hepworth et al. (2001) imply a role of later recollection processes in face recognition
memory of school-aged children, the question arises whether the N400 old/new effect and the
LPC old/new effect will show different or similar developmental trajectories.
The present study
Here we examined developmental changes in face recognition memory by comparing
processing and performance in 8-9-year-old and 11-12-year-old children on an extended
continuous face recognition paradigm (Study 1). To provide a context for the developmental
findings, we also examined an adult sample in a separate study (Study 2). Because the research
settings and testing conditions between children and adults were different, the results for the adult
sample will be analyzed and reported separately.
This is the first ERP study to apply an extended continuous face recognition task in
school-aged children and adults. Previous studies with an extended continuous word recognition
paradigm have shown differential modulation of the N400 and LPC by multiple repetitions both
in children and adults (Van Strien, Glimmerveen et al., 2009; Van Strien et al., 2005). We
hypothesized that older children would recognize more faces correctly compared to younger
children, and that older children would rely more on conscious recollection (cf. Cycowicz, 2000).
We therefore expected larger early ERP (N2, N400) old/new and repetition effects in younger
children, and larger LPC old/new and repetition effects in older children.
Study 1 – Children
Healthy children from regular primary schools were screened for handedness. Handedness was
assessed with a 10-item handedness questionnaire (Van Strien, 1992). Only strong right-handed
children (right-handed for 9 out of 10 activities) were included. Exclusion criteria were
psychoactive medication and history of neurological disorder (as indicated by the parents). All
participants had normal or corrected to normal vision. The final sample included 36 younger
children (8 and 9 years old; 18 males; M= 107.6 months, SD = 6.6) and 36 older children (11 and
12 years old; 14 males, M=143.1 months, SD = 6.9). To avoid the other-race effect for face
recognition (e.g., Kelly et al., 2007), all children were Caucasian. The educational level of the
participants’ parents did not differ between the age groups. All parents provided written informed
consent. Preceding the EEG session, information about the procedure was given to the children.
Children received a small present, such as a key ring or a set of pencils, for their participation.
The study received approval from the Rotterdam Medical Ethics Review Committee.
For the continuous face recognition task, 90 different faces (45 female) were selected from four
different online databases: NimStim Set of Facial Expressions (Tottenham et al., 2009, 14 faces,
5 female), Nottingham Scans (http://pics.psych.stir.ac.uk, 6 faces, all female), Aberdeen Faces
(http://pics.psych.stir.ac.uk, 14 faces, 7 female), and AR Face Database (Martinez & Benavente,
1998, 56 faces, 27 female). The photographs depicted white Caucasian adults in a neutral, frontal
pose. All pictures were resized to 200 × 250 pixels and converted to grayscale. If necessary, the
pictures were edited to have a white background. Thirty different face stimuli (15 female) were
presented six times each, and were intermixed with 60 other faces (30 female) that were
presented only once to elicit supplementary ’new‘ responses. Each trial therefore contained either
a new face (New), or a first to fifth repetition (R1, R2, R3, R4, R5). The total number of trials
equaled 240. The face stimuli were presented in semi-random order, with at least three
intervening stimuli between the successive presentations of a particular face.
Study 1 was part of a larger study concerning the developmental aspects of recognition
memory in school-age children. The EEG session took place in a separate room at school and
started with a word recognition task, which took approximately 17 minutes (not reported here,
see Van Strien, Glimmerveen et al., 2009). After a break, the face recognition task was
The face stimuli were displayed in the center of a black background using a Dell XPS
M170 laptop with a 17 inch TFT active matrix screen. The children were seated in a comfortable
office chair at a distance of approximately 50 cm from the screen, with the face stimuli
subtending approximately 7.8 × 9.8 ° of visual angle.
The sequence for each trial was: (1) the presentation of a fixation cross in the center of the
computer screen with a variable duration of 400 to 600 ms to reduce time-locked EEG phase or
expectancy effects, (2) the 500 ms presentation of the face in the center of the screen, (3) the
1200 ms presentation of the fixation cross, and (4) a 1500 ms inter-trial interval (a black screen).
The maximum response time was set at 2500 ms.
The participants were instructed to focus on the fixation cross and to give an ‘old’ or
‘new’ response as soon as they recognized the old and new faces that appeared on the screen. It
was explained that a ‘new’ response was correct when a face was presented for the first time,
while an ‘old’ response was correct when a face was presented for the second through sixth time.
Participants responded by pressing with their index fingers one of two buttons located at the left
and the right side of the screen. The assignment of ‘old’ and ‘new’ responses to the left and right
response button was counterbalanced across participants. Response latencies were collected using
two small response-button boxes connected to a Serial Response Box (Psychology Software
Tools, Pittsburgh, Pennsylvania, USA).
Preceding the experimental run, the participants were presented with a series of 21
practice trials with feedback on their performance at the end of each trial (‘correct’, ‘incorrect’,
‘too late’). If a participant had less than 80% correct in the first practice series, a second practice
series with 21 trials was started. After the second practice series, the experimental run was started
irrespective of the number of correct practice trials (six younger children and six older children
did not reach the 80% criterion after the second practice series). In the practice trials, faces were
presented that were not used in the experimental run. No feedback was given during the
EEG activity was recorded with a BioSemi Active-Two system from 64 pin-type active Ag/AgCl
electrodes mounted in an elastic cap according to the international 10–20 system. Flat-type active
electrodes were attached to the left and right mastoids. To measure eye movements, the electro-
oculogram (EOG) was recorded from four flat-type active electrodes positioned above and
beneath the left eye and at the outer canthi of the eyes. An additional active pin-type electrode
(CMS - common mode sense) and a passive pin-type electrode (DRL - driven right leg) were
used to comprise a feedback loop for amplifier reference. The EEG and EOG signals were
digitized with a 512 Hz sampling rate and 24-bit A/D conversion. Response latencies were
recorded online along with the EEG data.
Offline, the EEG signals were referenced to the averaged mastoids1 and phase-shift-free
filtered with a band pass of .15 Hz to 30 Hz. ERP epochs with an 1100-ms duration were
extracted, starting 100 ms before stimulus onset. Correction for ocular artifacts was done using
the Gratton, Coles, and Donchin (1983) algorithm. The ERPs were baseline corrected relative to
the mean amplitude of the prestimulus period and were averaged for each participant and each of
the six consecutive presentations (new, first repetition to fifth repetition). Epochs with an
incorrect response and epochs with a baseline-to-peak amplitude difference larger than +/- 150
µV on any channel were excluded. The mean number of valid epochs per condition ranged from
20.04 (new faces) to 24.92 (fifth repetition) with a mean across conditions of 22.97.
Based on inspection of the ERP waveforms and in accordance with previous research, the
anterior N2 was quantified by mean amplitude measures in the 200-275 ms time window (see
Guillem et al., 2001), the N400 by mean amplitude measures in the 350-450 ms time window
1 Although ERP studies concerning face perception often employ an average reference to obtain lateral N170 face
potentials, the present study primarily concerned recognition memory. For comparison with previous extended
continuous recognition studies, and because we expected within- and between-group differences in vertically
oriented ERP components like the N400 and LPC, we chose an averaged mastoids reference.
(see Van Strien, Langeslag et al., 2009), and the LPC by mean amplitude measures in the 650-
850 ms time window (see Czernochowski et al., 2005).
The behavioral and EEG data were subjected to analyses of variance (ANOVAs). The
factors that were included in each individual ANOVA are indicated in the various results sections
below. Note that in this paper the factor 'old/new' refers to the first vs. second presentation (New,
R1), the factor ‘repetition’ refers to the first through fifth repetition (R1 through R5), and the
factor ‘presentation’ refers to the six consecutive presentations of a face (New, R1 through R5).
In case of significant effects for the repetition factor, linear and quadratic trends were tested.
Where appropriate, F-ratios were tested with Greenhouse-Geisser corrected degrees of freedom.
The F-values, uncorrected degrees of freedom, epsilon values and corrected p-values are
Preliminary statistical analyses on both the behavioral and EEG data revealed no readable
main or interaction effects for gender. Therefore the age groups were collapsed across boys and
For each level of presentation (New, R1 through R5), we determined the participant's accuracy
and mean reaction time across correct trials. Table 1 presents the mean reaction times and mean
accuracies (percentages correct responses) for new faces and repetitions, as a function of age
group. The reaction-time and accuracy data were analyzed by means of ANOVAs with age group
as a factor between subjects and presentation (New, R1 through R5) as a factor within subjects.
For the reaction-time data, we found significant main effects of age group, with older
children showing faster responses than younger children, F(1,70) = 5.44, p = .023 (younger: M =
900 ms, SD = 201; older: M = 795 ms, SD = 182), and presentation, F(5,350) = 82.89, epsilon =
.588, p < .001. These main effects were qualified by the significant interaction of age group and
presentation, F(5,350) = 5.04, epsilon = .588, p = .002, with younger children showing a
relatively larger increase in response speed across repetitions than older children (see Table 1).
Single comparisons revealed that older children responded faster than younger children to new
faces (p = .002), first repetitions (p = .009), second repetitions (p = .012), and third repetitions (p
= .053), but not to fourth and fifth repetitions (both p-values > .205) .
For the accuracy data, we found a significant main effect for presentation, F(5,350) =
52.08, epsilon = .447, p < .001, with accuracy increasing across the number of repetitions (see
Table 1). When analyzed separately, the accuracy for new faces was significantly larger in older
compared to younger children, F(1,70) = 4.80, p = .032.
To examine whether the age groups differed in response bias, the bias measure Br was
computed as the false alarm rate (across all new words, including distracter words) divided by 1
minus the difference of hit rate (across R1 through R5) and false alarm rate (Snodgrass &
Corwin, 1988). Br scores range from 0 to 1, with scores above .50 indicating a liberal response
strategy, that is, a tendency towards “old” decisions. The mean response bias measures indicated
that both younger (Br = .75, SD = .15) and older (Br = .72, SD = .15) children exhibited a
moderate response tendency towards “old" decisions. No significant age difference was found for
Br (F < 1).
ERP old/new effects (R1 vs. new)
For each age group, the grand-average ERPs at selected electrodes for ‘new’ vs. ‘old’ (= first
repetition) faces are depicted in Figure 1. To analyze the N2, N400, and LPC old/new effects,
individual electrodes were clustered in six regional averages arranged in a three-by-two layout
with a left frontal (AF3, F5, F3, F1), a right frontal (AF4, F6, F4, F2), a left temporal (FT7, FC5,
T7, C5), a right temporal (FT8, FC6, T8, C6), a left central (FC3, FC1, C3, C1), and a right
central cluster (FC3, FC1, C3, C1). Individual ANOVAs were conducted for each ERP
component, with age group as between-subjects factor and old/new (R1 vs. New), location
(frontal, temporal, central), and laterality (left, right) as within-subjects factors.
N2. The topographic distribution of the N2 old/new effect is given in Figure 2A. From
this figure, it can be seen that younger children in particular showed a widespread N2 old/new
effect across bilateral central and frontal regions. We found a significant main effect of old/new,
F(1,70) = 12.83, p=.001, with smaller negative N2 peaks for ‘old’ (M = -3.6 µV, SD = 5.0) than
for ‘new’ faces (M = -5.4 µV, SD = 6.0). There was no main effect of age group (F < .5), but
there was a significant interaction of age group and old/new, F(1,70) = 7.21, p = .009. The
old/new effect appeared to be significant for younger children (new: M = -6.3 µV, SD = 6.4; old:
M = -3.1 µV, SD = 5.0; p < .001) but not for older children (new: M = -4.5 µV, SD = 5.7; old: M
= -4.1 µV, SD =5.1; p = .449). These effects were further qualified by a significant interaction of
age group, old/new, location, and laterality, F(2,140) = 6.47, epsilon=.901, p = .003. From Figure
2B, it can be seen that the N2 old/new effect was much larger at all regions in younger children,
and that in the younger children the old/new effect tended to be larger at right than at left
temporal electrodes (p = .058).
N400. From Figure 2C, it can be seen that younger children exhibited a larger N400
old/new effect than older children at bilateral central and fronto-central regions. There was a
significant main effect for age group, F(1,70) = 8.76, p = .004, with younger children (M = -15.6
µV, SD = 5.4) showing larger negative N400 amplitudes than older children (M = -11.6 µV, SD =
6.2). The main old/new effect was also significant, F(1,70) = 21.73, p<.001, with larger negative
N400 amplitudes for ‘new’ faces (M = -14.8 µV, SD = 6.6) than for ‘old’ faces (M = -12.3 µV,
SD = 5.9). In addition, the interaction of age group and old/new was significant2, F(1,70) = 7.16,
p = .009. Follow-up tests revealed that the N400 old/new effect was significant for younger
children (new: M = -17.6 µV, SD = 6.6; old: M = -13.6 µV, SD = 5.2; p <.001) but not for older
children (new: M = -12.1 µV, SD = 6.6; old: M = -11.0 µV, SD =6.5; p = .133).
LPC. The ANOVA on the LPC data yielded no significant main effects or interactions for
age group or old/new.
ERP repetition effects (R1 through R5)
Figure 3A displays the extended repetition effect (R1 to R5) at central sites for younger and older
children. To analyze the N2, N400, and LPC extended repetition effects, individual ANOVAs
were conducted for each ERP component at the central clusters with age group as between-
subjects factor, and repetition (R1 through R5) and laterality (left, right) as within-subjects
N2. For the N2, we found a significant main effect for repetition, F(4,280) = 13.34
epsilon=.932, p < .001. Both the linear contrast, F(1,70) = 27.45, p < .001, and the quadratic
contrast, F(1,70) = 16.07, p < .001, were significant. Fig 3B displays the N2 repetition effect,
2 When we controlled for possible age differences in skull thickness and bone conductivity by means of converting
the old/new amplitudes to within-age-group Z-scores, thus cancelling the main age group effect, the interactions of
age group and old/new remained significant for both N2, F(1,70) = 6.50, p = .013, and N400, F (1,70) = 7.73, p =
with diminishing N2 negativity across the first four repetitions. No interaction of age group and
repetition was found (p = .575).
N400. For the N400, the repetition effect was significant, F(4,280) = 13.35, epsilon =
.947, p < .001, as was the linear contrast, F(1,70) = 45.41, p < .001. The N400 negative amplitude
decreased with the increasing number of repetitions. Further, there was a significant interaction of
age group and repetition, F(4,280) = 2.97, epsilon = .947, p = .022. This interaction is depicted in
Figure 3C. From this picture, it can be seen that the younger children exhibited a steeper decline
in N400 negativity across the five repetitions than the older children did. Single comparisons
revealed larger N400 amplitudes for younger vs. older children at R1 (p = .030) and R4 (p =
.055). To explore a possible association between the N400 and the behavioral results, we
correlated the N400 repetition effect (R5 minus R1) with RT and with accuracy (R5 minus R1).
For neither younger nor older children, significant correlations were found.
LPC. For the LPC, no main or interaction effects for age group or repetition were found.
The aim of study 1 was to examine developmental changes in face recognition memory in 8- to 9-
year-old and 11- to 12-year-old children. We used an extended continuous face recognition
paradigm to assess developmental changes in behavioral performances and in early (N2, N400)
and late positive (LPC) components of the ERP old/new and extended repetition effects.
Childrens’ behavioral data
Across repetitions, both age groups displayed increasing response speed and accuracy, which
indicates better encoding and retrieval after multiple repetitions. Better initial encoding and
retrieval in older than in younger children was reflected by shorter reaction times for correct 'old'
or 'new' responses. At the outset, younger children were slower, but after two repetitions they
started to approach the response speed of older children. Apparently, after repeated encoding,
retrieval becomes faster in younger children and comparable to older children. With regard to
accuracy, older children better recognized new faces, but exhibited no differences from younger
children across the five repetitions. Older children might show superior recognition of new faces
because they generally have less difficulty recognizing unfamiliar faces than younger children
(e.g., Taylor, Batty, & Itier, 2004).
Younger and older children both showed comparable response biases (Br = .75 vs. .72)
and tended to respond in a moderately liberal direction (i.e., toward 'old' responses). Previous
recognition memory research in older adults has suggested that a more liberal response bias is
associated with diminished frontal functioning (Huh, Kramer, Gazzaley, & Delis, 2006). It could
be hypothesized that a liberal response bias in children is a similar consequence of immature
frontal functioning. However, it may also be the result of task-specific demands, because with
words both groups of children exhibited a much smaller tendency toward ‘old’ decisions (Br =
.61, see Van Strien, Glimmerveen et al., 2009) than with faces.
Childrens’ ERP old/new effects
In younger, but not in older children, the N2 was more negative going for new compared to old
(i.e., first repetition) presentations at frontocentral sites. Previous research has suggested that in
adults, the frontocentral N2 may reflect a mismatch between stimulus input and existing
representations (Folstein & Van Petten, 2008). In 18-month-old children, a parietal N2 peaking at
250 ms was larger (more negative) in response to unfamiliar than in response to familiar toys
(Carver, Meltzoff, & Dawson, 2006). In general, novel stimuli (i.e., unfamiliar faces) will elicit a
larger N2 than known stimuli. The larger N2 old/new effect in younger children may therefore be
related to their difficulty to process unfamiliar faces. In these younger children, the N2 old/new
effect tended to be larger at right than at left temporal regions, which might reflect right-
hemispheric specialization in face processing (e.g., Schweinberger, Pfutze, & Sommer, 1995).
In addition, younger children displayed larger N400 amplitudes and larger N400 old/new
effects at frontal and central regions than older children. The larger N400 amplitudes suggest the
allocation of more resources for face recognition in younger compared to older children.
Consistent with the present influence of age on the N400 amplitude, Itier and Taylor (2004)
found a general age effect in the 300-600 ms time window in children from 8 to 16 years of age,
with less negative amplitudes with increasing age. These authors further found frontal N400
old/new effects for faces in all age-groups but reported no interaction of age and repetition. This
latter outcome may be a consequence of the repetition lags chosen in the Itier and Taylor study.
The faces were repeated immediately after the first presentation (0-lag) or after one intervening
face (1-lag). With these short lags, the involvement of a general working-memory system may
have been larger than with longer lags as used in the present study.
Younger and older children did not demonstrate a LPC old/new effect. The LPC old/new
effect is thought to reflect more conscious recollection. Apparently the children based their
old/new decisions on more or less automatic priming processes (e.g., template matching) rather
than on conscious recollection. This would be in accordance with research showing that children
rely more on early familiarity-related processing than on recollection (e.g., Ofen et al., 2007).
This reliance on more automatic recognition appears to be specific for face recognition, because
with words, both age groups exhibited a clear LPC old/new effect suggesting that recollection
plays a substantial role in word recognition memory of both younger and older children. (Van
Strien, Glimmerveen et al., 2009). With line drawings, Czernochowski et al. (2005) also found
later parietal old/new effects in younger and older children.
Childrens’ repetition effects
At left and right central clusters, we found robust repetition effects (R1 through R5) for both the
N2 and N400. The decline of the N2 negativity upon repeated presentations may reflect the
diminishing novelty of the face stimuli. The N400 repetition effect differed between age groups.
Younger children showed a steeper decline in N400 negativity over multiple face repetitions than
older children, which seems to parallel the age differences in response speed. However, we found
no significant correlations between the N400 repetition effect and the behavioral data. The
repetition effects for the early N2 and N400 ERP components suggest that in children face
recognition is based on more automatic processing. The larger N400 repetition effect in younger
children when compared to older children indicates a gradual development of these automatic
memory processes, with less allocation of resources for face recognition with increasing age.
For the 650-850 ms time window (LPC) we found no repetition effects. This result
suggests that the later and more conscious recollection processes do not play a substantial role in
face recognition in children. It was expected that on repeated presentations, a face would be
recollected more consciously, which would be reflected by a greater LPC repetition effect,
especially in older children. With the present continuous face recognition task, the earlier
automatic recognition processes apparently suffice and no further memory updating takes place.