Emotion
2008, Vol. 8, No. 3, 341–351
Copyright 2008 by the American Psychological Association
1528-3542/08/$12.00 DOI: 10.1037/1528-3542.8.3.341
Unpacking the Cognitive Architecture of Emotion Processes
Didier Grandjean and Klaus R. Scherer
Swiss Center for Affective Sciences, University of Geneva
The results of 2 electroencephalographic studies confirm Component Process Model (CPM) predictions
that different appraisal checks have specific brain state correlates, occur rapidly in a brief time window
after stimulation, and produce results that occur in sequential rather than parallel fashion. The data are
compatible with the assumption that early checks (novelty and intrinsic pleasantness) occur in an
automatic, unconscious mode of processing, whereas later checks, specifically goal conduciveness,
require more extensive, effortful, and controlled processing. Overall, this work, combined with growing
evidence for the CPM’s response patterning predictions concerning autonomic physiological signatures,
facial muscle movements, and vocalization changes, suggests that this model provides an appropriate
basis for the unpacking of the cognitive architecture of emotion and its computational modeling.
Keywords: appraisal, emotion, electroencephalographic timing, neurochronometry, brain waves
knitted brows, square mouth with teeth clenched, and loud, strident
vocal utterances).
Componential appraisal theory, especially the Component Process Model (CPM; Scherer, 1984, 2001), offers a number of
advantages as a guiding model for empirical research: (a) Emotions are defined as complex, multicomponential, dynamic processes that require sophisticated measurement of changes in the
different components; (b) highly specific predictions about the
determinants that elicit and differentiate emotions are made; and
(c) a concrete mechanism underlying emotional response patterning, allowing very specific hypotheses, is suggested (predicting
appraisal-driven responses based on functional considerations; see
Scherer, 2001). In consequence, the richness of emotion differentiation in humans is accounted for, allowing researchers to model
individual differences and emotional disorders (Scherer, 2004).
In this article, we report two experiments that were designed to
specifically address, using electroencephalographic recording of
electrical brain activity, three major postulates of the CPM as
summarized in Scherer (2001): (1) the decomposition of valence
into intrinsic pleasantness and goal conduciveness, (2) the assumption that appraisal criteria are evaluated on different levels of
processing, and (3) the prediction that, whereas the processing of
appraisal criteria takes place in a parallel fashion, results occur in
a fixed temporal sequence for different criteria.
1. Affect is often equated with valence evaluation: the general
positive or negative reaction toward a stimulus. In contrast, the
CPM considers the intrinsic pleasantness evaluation as separate
from, and prior to, a positive goal or need conduciveness check. In
particular, it is suggested that the pleasantness or unpleasantness
detected by the intrinsic pleasantness check is a feature of the
stimulus and, even though the preference may have been acquired,
it is independent of the momentary state of the organism. In
contrast, the positive evaluation of stimuli that help to reach goals
or satisfy needs depends on the relationship between the significance of the stimulus and the organism’s motivational state. In
fact, intrinsic pleasantness is orthogonal to goal conduciveness
because something intrinsically pleasant (like chocolate cake) can
be highly obstructive (if one is forced to eat three large pieces
Componential appraisal models have been developed to capture
the complexity of emotion as a dynamic episode that involves a
process of continuous change in all of its subsystems (e.g., cognition, motivation, physiological reactions, motor expressions—
the components of emotion) to adapt flexibly to events of high
relevance and potentially important consequences for an individual
(adopting a functional approach in the Darwinian tradition; Ellsworth & Scherer, 2003; Scherer, 1984, 2001). Importantly, feeling
is seen as a component of emotion, serving a monitoring function
to facilitate regulation (Scherer, 2004). The central emotion mechanism proposed by these theories is appraisal, based on the pioneering work of Arnold (1960) and Lazarus (1966, 1991): the
continuous, recursive evaluation of an event for criteria such as
novelty, intrinsic pleasantness, goal conduciveness, coping potential, and normative significance. The appraisal process consists of
the subjective evaluation of each individual, thus allowing for
differences between species, age groups, personal dispositions, and
cultural contexts.
The outcome of the appraisals based on the different criteria is
predicted to directly drive response patterning of physiological
reactions, motor expression, and action preparation. Thus, anger is
expected to be the result of an event being appraised as an
obstruction to reaching a goal or satisfying a need, produced by an
unfair, intentional act of another person, that could be removed by
powerful action (with a corresponding response pattern consisting
of aggressive action tendencies, involving sympathetic arousal,
Didier Grandjean and Klaus R. Scherer, Swiss Center for Affective
Sciences, University of Geneva, Switzerland.
This work was supported in part by an institutional grant of the Swiss
National Science Foundation to the Swiss Center for Affective Sciences.
We gratefully acknowledge technical advice provided by Christoph Michel
and Denis Brunet.
Correspondence concerning this article should be addressed to Didier
Grandjean, Swiss Center for Affective Sciences, University of Geneva,
7 rue des Battoirs, CH-1205, Geneva, Switzerland. E-mail:
[email protected]
341
342
GRANDJEAN AND SCHERER
while trying to loose weight). This work strives to demonstrate that
intrinsic pleasantness on the one hand and goal relevance and goal
conduciveness on the other can be separately manipulated in an
experimental design and shown to be processed independently of
each other.
2. Componential appraisal theories have been criticized for their
use of memory recall and verbal report during the early stages of
their development. In addition, it is often claimed that appraisal
requires an effortful, conscious processing of information that
cannot account for rapid automatic emotion responses. However,
very early on, Leventhal and Scherer (1987) suggested that appraisal of almost all important criteria can occur on different levels
of information processing—sensory–motor, schematic, and
conceptual–representational. In consequence, componential appraisal theories can model a wide variety of different forms of
emotion elicitation, ranging from extremely rapid, automatic, and
unconscious processing to extensive, effortful, controlled, and
conscious deliberation (see Sander, Grandjean, & Scherer, 2005).
Automatic, unconscious processing cannot be assessed via verbal
report and requires much more sophisticated measurement. Objections that appraisal processes are too slow and cumbersome to
account for emotion elicitation are difficult to maintain if the CPM
hypothesis of rapid multilevel processing of appraisal can be
empirically demonstrated. This work examines the plausibility of
this approach.
3. One of the central features of the CPM since its original
formulation is the prediction that the processing of major appraisal
criteria or stimulus evaluation checks (SECs), while proceeding in
a parallel fashion, yield outcomes or results in a fixed sequential
order: Novelty ⬎ intrinsic pleasantness ⬎ task– goal relevance ⬎
goal conduciveness ⬎ coping potential ⬎ compatibility with internal and external norms or standards. The essential criterion for
the sequence is the point in time at which a particular check
achieves preliminary closure, that is, yields a reasonably definitive
result, one that warrants efferent commands to response modalities. What are the arguments that justify this sequence assumption?
Relevance detection is considered to be a first selective filter a
stimulus or event needs to pass to merit further processing. It is
assumed that only objects or events that surpass a certain threshold
on novelty, or intrinsic pleasantness/unpleasantness, or goal/need
relevance will pass this filter. Attention will be focused on the
event and further processing will ensue. The next step is to assess
the causes and implications of the event for the organism. These
need to be established before the organism’s coping potential can
be determined because the latter is always evaluated with respect
to a specific demand. Finally, the normative significance of the
event, that is, its consequences for the self and its normative/moral
status, is evaluated. The general assumption is that, for logical and
economical reasons, the results of the earlier SECs need to be
processed before later SECs can operate successfully, that is, yield
a conclusive result. It can also be argued that the microgenetic
unfolding of the emotion-antecedent appraisal processes parallels
both phylogenetic and ontogenetic development in the differentiation of emotional states (see Scherer, 1999, 2001, for further
details on the sequence prediction).
The assumption of sequential processing of the SECs and, in
particular, the notion of a fixed order of the SECs has often been
challenged (e.g., Ellsworth, 1991; Lazarus, 1999). In particular, it
is claimed that the onset of an emotional reaction can occur
extremely rapidly, in a single step, especially if schematic processing is involved (Smith & Lazarus, 1990). However, the apparent
speed of an emotional reaction to an event does not rule out a
sequential model, if one assumes that at least some of the checks
can be performed extremely rapidly (particularly when lower brain
structures are involved). In consequence, we set out to empirically
test the sequence prediction for some of the early appraisal criteria
or SECs and to perform the first mental chronography of appraisal
to determine the plausibility of the assumption that sequential
checking can occur very early on after stimulus onset and at an
extremely rapid rate.
Given the need for high temporal resolution measurement of
brain activity required by these aims, we chose to use state-of-theart electroencephalography (EEG) and event-related potential
(ERP) methods in an appropriate experimental design. Two studies
were designed using EEG–ERP methods to measure the brain
electrical reactions to the visual presentation of affect-inducing
stimuli. These methods provide an excellent temporal resolution
and thus allow testing the fixed-sequence hypothesis and obtaining
precise estimates of the timing of the sequential phases of the
appraisal process.
The first experiment tested the sequence hypothesis by experimentally manipulating three different appraisal criteria: novelty,
intrinsic pleasantness, and goal relevance (or, specifically, task
relevance). On the basis of the CPM, early effects on electrical
signal patterns in the brain, specific to the manipulated criteria,
were predicted to occur in the following order: first, novelty;
second, intrinsic pleasantness; and third, task– goal relevance. The
second experiment consisted of a full factorial design in which the
intrinsic pleasantness and goal conduciveness/obstructiveness appraisals were manipulated (see the Method sections for further
details). The following order of occurrence of the traces for the
respective brain activity was predicted: first, intrinsic pleasantness;
second, goal conduciveness. These hypotheses are summarized in
Figure 1. The following dependent variables, based on three major
analysis techniques for EEG, were used to test these predictions:
(a) the timing of ERP topographical maps specifically related to
the different manipulated checks, (b) the timing of the amplitude
modifications of the global field power (GFP) related to the
manipulated appraisal checks, and (c) the timing of the appearance
of differences in amount of energy in the different frequency bands
related to the different manipulated checks.
Given the three central issues outlined above, we expected (a) to
obtain evidence for different patterns of brain activities involved in
the processing of intrinsic pleasantness and goal conduciveness,
(b) to use the results of mental chronography to obtain some
indirect evidence on the level of processing (related to automaticity and brain structures involved), and (c) to confirm the theoretical
predictions about the sequence of the results of specific checks.
Experiment 1
Method
Participants. Fifteen right-handed (mean of Edinburgh Handedness Inventory ⫽ 80.3, SD ⫽ 24.5) undergraduate students in
psychology participated in this experiment after they had given
their written consent (12 women; mean age ⫽ 25.1 years, SD ⫽
10.1). One participant did not perform the task correctly and was
excluded from the data analyses.
COGNITIVE ARCHITECTURE OF EMOTION
343
Figure 1. Theoretical predictions of the Component Process Model (CPM) concerning the sequence of
appraisals in emotion elicitation. The blue rectangle shows the three appraisals related to relevance detection,
which were all manipulated in Experiment 1 (marked ①). In Experiment 2, the intrinsic pleasantness (relevance)
and goal conduciveness (implication; green rectangle) appraisals were manipulated (marked ②). The solid red
circles show the theoretically predicted sequential effects of preliminary closure of the different appraisal results
on the different subsystems of the organism (e.g., autonomous nervous system, motor expression).
Experimental procedure. Given the requirements of mental
chronography, we needed a clear and stable time position for
stimulus onset to investigate the micromomentary unfolding processes of appraisal checks using EEG methodology (which is the
only option with respect to temporal resolution and direct mapping
of brain processes). In similar studies in the affective neurosciences, a picture-viewing paradigm is a method of choice as it
allows examination of emotional processes at different functional
levels, including behavioral (time reaction), central nervous system, and autonomic levels under well-controlled experimental
conditions. In many of these studies, the International Affective
Picture System (IAPS; Lang, Bradley, & Cuthbert, 1999) is used,
as it is allows choice of pictures on validated dimensions and,
consequently, direct comparison between different studies. We
have validated the norms for the French language and have empirically shown that the well-established affective reactions to
IAPS pictures are based on appraisal processes (Scherer, Dan, &
Flykt, 2006). In consequence, we decided to use the IAPS set in
Experiment 1 to systematically manipulate the novelty, intrinsic
pleasantness, and task– goal relevance appraisals. Novelty was
manipulated by the frequency of presentation of specific stimuli
(single exposure for novelty, repeated exposure for familiar). Intrinsic pleasantness (IP) was manipulated by selecting pictures
according to the IAPS norms: positive valence (80 pictures, mean
IAPS valence norm ⫽ 7.98), negative valence (80 pictures, mean
IAPS valence norm ⫽ 1.92), and neutral pictures (80 pictures,
mean IAPS valence norm ⫽ 5.15). Goal relevance refers to the
pertinence of a stimulus or event for the current concerns of a
person (see Scherer, 2001). In experimental situations, these concerns are generally manipulated by specific tasks the participants
have to perform and for which they generally show a fairly high
degree of achievement motivation. Thus, we manipulated the goal
relevance check specifically in the form of task– goal relevance by
asking participants to identify target pictures out of the total set of
pictures shown (for a subset of these targets, participants were
asked to press a button on identification). A previous fMRI study
showed that emotional brain activation, particularly the amygdala,
is differentially affected by a similar matching task (Wright & Liu,
2006).
The experimental paradigm was an adapted oddball paradigm,
often used in the literature to investigate novelty and task– goal
relevance (for a review, see Ranganath & Rainer, 2003). Three
types of stimuli were used: (a) new stimuli (15%, novel), (b)
identification of task/goal relevant stimuli (15%, task/goal relevant), and (c) background stimuli (70%, familiar/nonrelevant pictures). The new and task/goal relevant stimuli could be positive,
negative, or neutral, whereas the familiar and nonrelevant background stimuli were always neutral. In consequence, the analysis
design included a Pertinence factor with three levels—novel, task/
goal relevant, and familiar/nonrelevant— crossed by an IP factor
(with the levels positive, negative, and neutral). During EEG
recordings, the button press responses were requested only when
an interrogation mark appeared on the screen to avoid EEG motor
artifacts during the presentation of the stimuli themselves (the
344
GRANDJEAN AND SCHERER
response buttons were counterbalanced across participants). Ten
percent of the stimuli were followed by a question mark; these
trials were excluded from further EEG analyses. Each stimulus
was preceded by a fixation cross (with variable duration of 500 to
1,000 ms) to avoid expectancy effects, presented for 500 ms, and
followed by a black screen (500 ms).
EEG acquisition, analyses, and ERPs. The EEG was recorded
using Quickcap Neuroscan with 64 electrodes, a nose reference,
and a sampling rate of 500 Hz, with a notch at 50 Hz and filters
(high pass of 0.15 Hz and low pass at 70 Hz). The participants
were seated comfortably in a Faraday cage in front of a 17-in.
screen where the stimuli were presented, subtending 6.8° H 5.7° of
visual angle at 100 cm viewing distance. The electro-oculogram
was measured with two bipolar recordings to measure the vertical
as well as the horizontal ocular movements. To analyze the EEG
signals, we used topographical clustering based on ERPs and
time–frequency analyses using the discrete wavelet transform
(DWT) method (Başar, Demiralp, Schürmann, Başar-Eroglu, &
Ademoglu, 1999; Samar, Bopardikar, Rao, & Swartz, 1999) on
ERPs and raw EEG data (see the next section for further details).
The ERPs were computed by averaging the trials for each experimental condition using a high-pass filter at 0.5 Hz and a low-pass
filter at 30 Hz, with a baseline of 100 ms before stimuli onset. The
ocular artifacts were corrected using established procedures (see
Gratton, Coles, & Donchin, 1983). To extract and test the different
electrical brain activation maps obtained, we used the procedure
described by Michel and collaborators (Michel, Murray, Lantz,
Gonzalez, & Grave de Peralta, 2004) using the Cartool software
(http://brainmapping.unige.ch/cartool.php).
Wavelet analyses. Wavelet analysis is a well-known mathematical technique allowing disentanglement of the relative amplitude of the different frequencies in EEG and ERP signals. The
EEG signal obtained at the scalp surface is a complex sinusoidal
signal result of the averaging of different neuronal sources. The
goal of wavelet analysis is to decompose the complex signal into
simple sinusoids in different frequency bands. This technique
allows researchers to quantify the energy in the different frequency
bands contributing to the raw EEG signal or to the ERPs. Furthermore, the size of the time window used for the analysis is adapted
to the level of frequency band of interest, contrary to other decomposition algorithms like the Fast Fourier Transform. The algorithm used here consists of a pyramidal discrete type, allowing
us to decompose systematically, without redundancy between the
extracted frequencies, the energy in different frequency bands. The
advantage of performing frequency analyses on the raw EEG
signals consists of the possibility of quantifying the energy of
signal even when the EEG modulations are not time-locked to
stimulus onset. Thus, the EEG signal can be dissociated into an
“evoked” part, which is time-locked, and an “induced” part, which
is not time-locked to the stimulus onset (see Bertrand & TallonBaudry, 2000, for more details). In the analyses described below,
we tested the differences between the absolute values of the
wavelet coefficients for the different frequency bands. To quantify
the intensity of the EEG signal in these bands, we performed the
DWT using the algorithm implemented in the Wavelets toolbox,
which is part of the S⫹ software. We used the bio-orthogonal
B-spline wavelet, bs3.7, which is particularly adapted to EEG
analysis, allowing us to decompose the signal with a systematic
pyramidal algorithm (Başar et al., 1999; Samar et al., 1999). The
signal (512 time frames at 500 Hz) was decomposed into 11 levels;
we selected 5 levels corresponding to the delta band (⬃3.7 Hz),
theta band (⬃7.4 Hz), alpha band (⬃14.8 Hz), beta band (⬃29.6
Hz), and gamma band (⬃58.7 Hz). We performed these analyses
on the ERPs of each participant for all experimental factors.
Results
As explained above, the unique topographical maps related to
the manipulated checks were extracted from the averaged ERPs.
As predicted, we found an early brain microstate, specifically
related to the novelty appraisal manipulation. This microstate was
followed by another specific microstate related to the task/goal
relevance manipulation. Topographical clustering of the ERP data
in the first experiment revealed 13 different topographical maps
based on the cross-validation index (Michel et al., 2004). To test
the sequence hypothesis for the early checks that were manipulated
in this experiment, we focused the following analyses on the first
250 ms after onset of the stimulus. To statistically test the specificity and the timing of the topographical maps of the initial
appraisal checks, we performed a repeated measures analysis of
variance (ANOVA) on the explained variance for each participant’s ERPs. This analysis revealed an interaction effect between
the Topographical Maps factor with levels consisting of four
different maps (Figure 2, A–C) and the Pertinence factor with the
levels novel, task/goal relevant, and familiar/nonrelevant, indicating that some maps are specific to particular conditions, F(6, 78) ⫽
13.6, p ⬍ .001. The relevant contrasts showed that two maps are
common to all three experimental conditions, F(1, 13) ⫽ 3.1, p ⬎
.10, whereas the other two are specific for the novel and task/goal
relevant conditions, F(1, 13) ⫽ 5.3, p ⬍ .05. The two contrasts for
the topographical maps in the novel and task/goal relevant conditions compared with the familiar/nonrelevant condition yielded
significant effects, F(1, 13) ⫽ 6.8, p ⬍ .02, and F(1, 13) ⫽ 4.3,
p ⬍ .05. To test the timing of appearance of these two specific
maps, we tested their onset in the respective novel and task/goal
relevant experimental conditions. This analysis revealed that (a)
the novel map occurred at about 95 ms (mean of 96.00 ms) and (b)
the task/goal relevant map occurred at about 160 ms (mean of
156.36 ms) after the onset of the stimuli (Figure 2D). The difference between these two onset times is significant, F(1, 10) ⫽ 9.74,
p ⬍ .02, indicating that the occurrence of the novel map precedes
the task/goal relevant map by about 50 ms. These topographical
analyses did not reveal any specific maps related to the IP factor.
To investigate effects of manipulated appraisals not revealed by
the topographical analyses, we performed further analyses on the
GFP of the ERPs, mainly representing the intensity of signal
changes without corresponding modifications of the topography of
the electrical fields. A repeated measures multivariate analysis of
variance (MANOVA) yielded a significant effect of the Pertinence
factor, F(40, 16) ⫽ 3.83, p ⬍ .005. The contrasts for the first time
window (0 –50 ms) revealed a significant effect of novel compared
with the familiar/nonrelevant, F(1, 13) ⫽ 14.49, p ⬍ .005, and
task/goal relevant, F(1, 13) ⫽ 7.23, p ⬍ .02, conditions, confirming the sequence shown by the topographical analysis. The contrast
between task/goal relevant and familiar/nonrelevant was not significant, F(1, 13) ⫽ 2.82, p ⬎ .10. To investigate the effects of the
IP manipulation, we performed a repeated measures MANOVA
with two factors, the Pertinence factor (with the levels novel and
COGNITIVE ARCHITECTURE OF EMOTION
345
Figure 2. Experiment 1: Results of the topographical analyses and the global field power (GFP) analyses. (A)
Topographical results are shown for the experimental condition “familiar/nonrelevant” (the different colors in the
graph for the time course correspond to the different topographical maps observed); the different maps (red
corresponding to the positive, blue corresponding to the negative part of the electrical field, and the upper part of the
topographical maps corresponding to the anterior part of the head) are not specific for this experimental condition. (B)
Specific novelty map for the second section. (C) Specific task/goal relevance map for the third section. (D) Timing
of the two specific event-related potential topographical maps: the novelty map and the task/goal relevance map after
the onset of the stimuli. (E) Timing of the GFP for the negative, positive, and neutral levels of intrinsic pleasantness
after onset of the stimuli. The superimposed red rectangle indicates the time windows for which the positive and
negative levels are significantly different from the neutral level ( p ⬍ .05).
task/goal relevant) and the IP factor (with the positive, negative,
and neutral levels), on the GFP. The contrasts for the first time
window showed a significant effect of the Pertinence factor, F(1,
13) ⫽ 7.24, p ⬍ .02, with an increase of GFP for novel compared with
task/goal relevant, but not for IP (F ⬍ 1). The contrasts on the second
time window (50 –100 ms) revealed a significant effect of Pertinence,
F(1, 13) ⫽ 11.95, p ⬍ .005, and of IP, F(2, 26) ⫽ 9.52, p ⬍ .005, with
the negative level showing a significant effect compared with the
neutral level, F(1, 13) ⫽ 28.40, p ⬍ .0005, and compared with the
positive level, F(1, 13) ⫽ 5.57, p ⬍ .05 (Figure 2E).
The decomposition of the ERPs by wavelet analyses (DWT)
allows us to quantify the energy of the different frequency bands
(namely, delta, theta, alpha, beta, and gamma bands) to track
potential effects of experimentally manipulated appraisals on these
specific frequency bands. Systematic analysis revealed the following sequence of events: In the first time window in the delta low
frequency band (at about 0 –130 ms), novelty induced a significant
increase of energy compared with both task/goal relevant and
familiar/nonrelevant conditions (Figure 3A). No early effect of IP
could be demonstrated for this frequency band (Figure 4A). In the
GRANDJEAN AND SCHERER
346
theta band, we confirmed a specific effect for task/goal relevant
condition (compared with familiarity/nonrelevant and novel) in the
second time window (at about 130 –260 ms; Figure 3B). Figure 4A
shows other significant effects of the contrasts for the levels of the
two experimental factors separately for the five frequency bands
and the different time windows, confirming results on the sequential process presented above.
Overall, the patterns of results from all three types of analysis
confirm the theoretically predicted sequence novelty ⬎ IP ⬎
task/goal relevance.
Experiment 2
Method
Participants. Seventeen right-handed (mean of Edinburgh
Handedness Inventory ⫽ 71.9, SD ⫽ 31.2) undergraduate students
in psychology participated in this experiment after they had given
their written consent (10 women; mean age ⫽ 23 years, SD ⫽ 4.1).
Stimuli. IAPS pictures for different levels of IP (positive,
negative, and neutral pictures) were chosen, matched for mean
arousal between the different categories and the IP levels. An
ANOVA on the valence norm values of the chosen IAPS pictures
was performed with two factors: (a) IP (with three levels: positive,
negative, and neutral) and (b) Categories (with three levels: animal, people, object–landscape). The analysis revealed an effect of
IP, F(2, 45) ⫽ 471.25, p ⬍ .0001, and no effect of categories (F ⬍
1). An ANOVA on the values (IAPS norms) reflecting the arousal
dimension with the same factors revealed no effect for IP, F(2,
45) ⫽ 1.09, ns, nor for Categories (F ⬍ 1). To avoid low-level
visual effects on EEG measures, we controlled the mean of the
luminance of the IAPS pictures; an ANOVA confirmed that there
were no effects for IP, F(2, 45) ⫽ 1.61, ns, nor for Categories (F ⬍
1). Moreover, we controlled for the low and high spatial frequencies of pictures to avoid specific effects related to the energy in
these different frequency bands, as shown by previous studies
(Vuilleumier, Armony, Driver, & Dolan, 2003). We performed
two-dimensional DWT on the 54 IAPS pictures to test whether the
energy was different between the high and low spatial frequencies
between the experimental conditions. A MANOVA on the first
three levels of high frequencies indicated no effect for IP, F(6,
86) ⫽ 1.67, ns, nor for Categories (F ⬍ 1). For the low frequencies, the MANOVA also indicated no effect between the three
levels of low frequencies (IP: F ⬎ 1; Categories: F ⬍ 1). To
manipulate the goal conduciveness/obstructiveness and IP appraisal, we selected 54 IAPS pictures that satisfied two criteria: (a)
showing one of three different content categories (animal, people,
and object–landscape) and (b) corresponding to one of the three
valence groups—positive, negative, and neutral—as defined
above.1
Experimental procedure. Again, the picture-viewing task using IAPS was chosen to investigate the micromomentary unfolding
processes of appraisal checks with the help of EEG methodology
(which is the only measurement option providing the necessary
temporal resolution and allowing direct mapping of electrical brain
activity). Goal conduciveness/obstructiveness was manipulated by
associating the three different content categories (animal, people,
and object–landscape) with gain or loss of money or having no
effect on the outcome (these contingencies were counterbalanced
between the participants). The experiment occurred at two times:
(a) on the first day, the participants were exposed to a broad range
of pictures (not taken from IAPS), including animals, people, and
object–landscape, training them on the specific contingencies concerning gain, loss, and no effect; and (b) on the second day, EEG
was recorded using the IAPS picture-viewing task.
The training session on the first day (⬃45 min) served to firmly
associate the occurrence of a content category in a picture with the
conduciveness/obstructiveness manipulation, represented by the
monetary outcome of the picture event. For this purpose, participants were shown picture series and had to identify the categories
and compute their respective gains and losses. Initially, feedback
was given to ensure that the contingencies (gain or loss of 1 Swiss
franc per contingent stimulus) were well understood. The data on
calculation errors were used to ascertain that the contingencies had
been well automatized. One person was excluded from further
participation because of a high level of errors.
In a session of 60 min on the following day, the EEG was
recorded during the IAPS picture presentation (500 ms), with each
picture being preceded by a fixation cross. The participant’s task
was to decide for each picture whether it represented a gain, a loss,
or no effect on the amount of money accrued. During EEG
recordings, the motor responses were required only when an interrogation mark appeared on the screen for the previously seen
stimulus (which was the case for only 10% of the stimuli) to avoid
EEG motor artifacts during the presentation of the stimuli themselves (the response buttons were counterbalanced across participants).
EEG acquisition and ERPs. We recorded the EEG using the
same procedure as in the first experiment.
Wavelet analyses. DWT analyses were performed on the ERPs
as well as on the raw EEG signals using the same bio-orthogonal
wavelet as in the first experiment. Each trial used to compute the
ERPs was analyzed using DWT to obtain the intensity of the signal
in the different frequency bands before the averaging process. This
method allowed us to investigate the energy in the high frequencies, which are strongly diminished by the usual averaging processing (Lachaux et al., 2005; Tallon-Baudry, Bertrand, Delpuech,
& Pernier, 1996).
As the results reported here are related exclusively to our
testing the sequence of the appraisal check hypothesis, we did
not correct for alpha error given that all contrasts were planned
and only the main hypotheses were tested. We have not reported
or interpreted any other patterns in the data found in ERP and
EEG analyses that would have yielded significance by conventional standards.
Results
During the training phase (first day), we recorded the categorization reaction time for picture category identification. An
ANOVA of these data with group contingencies (three groups,
e.g., animal ⫽ loss, people ⫽ gain, and object–landscape ⫽ no
effect) as the between-factors variable and blocks (three successive
blocks) and goal conduciveness (gain, loss, and no effect) as the
1
The list of the pictures may be obtained from the authors on request.
COGNITIVE ARCHITECTURE OF EMOTION
347
Figure 3. Experiment 1: Timing of the means and standard error of the mean (SEM) of different wavelet
coefficients (absolute values) for the event-related potential (ERP) after the onset of the stimuli for the novel, task/goal
relevant, and familiar/nonrelevant conditions: (A) delta band, (B) theta band. Experiment 2: Timing of the means and
SEM of different wavelet coefficients (absolute values) for the ERP and raw electroencephalography (EEG): (C) delta
band, (D) theta band, (E) alpha band. (F) Timing of the wavelet coefficients of the raw EEG in the alpha band.
within-factor variables revealed (a) no effect of group (F ⬍ 1), (b)
a significant effect of blocks, F(3, 39) ⫽ 44.5, p ⬍ .0001, and (c)
a significant effect of goal conduciveness, F(2, 26) ⫽ 7.9, p ⬍ .01.
The block effect was interpreted as a learning effect; the reaction
time decreased with a quadratic function: quadratic polynomial
contrast, F(1, 13) ⫽ 22.7, p ⬍ .001. The contrasts between loss
versus gain as well as loss versus no effect revealed significant
effects, F(1, 13) ⫽ 34.8, p ⬍ .0001, and F(1, 13) ⫽ 4.7, p ⬍ .05,
respectively. This goal conduciveness effect was interpreted as an
acquired negative bias.
Figure 4. Onset timing of the significant effects and statistical coefficients for the discrete wavelet transform (DWT) on event-related potentials (ERPs) and electroencephalography (EEG). (A) Onset
timing of the significant effects and statistical coefficients for the DWT analyses on ERPs in the first experiment for the different time windows (ms) after the onset of the stimulus. Blue: Comparison
between novel and other experimental conditions. Green: Comparison between the different levels of intrinsic pleasantness. Yellow: Comparison between task/goal relevant and the other experimental
conditions. FamNR ⫽ familiar/nonrelevant; TaskR ⫽ task/goal relevant; Neg ⫽ negative; Neut ⫽ neutral; Nov ⫽ novel; Pos ⫽ positive. (B) Onset timing of the significant effects and statistic
coefficients for the DWT analyses on ERPs in the second experiment for the different time windows (ms) after the onset of the stimulus. Green: Comparison between the different levels of intrinsic
pleasantness. Red: Comparison between the different levels of goal conduciveness. Neg ⫽ negative; Neut ⫽ neutral; Pos ⫽ positive. (C) Onset timing of the significant effects and statistic coefficients
for the DWT analyses on raw EEG in the second experiment for the different time windows (ms) after the onset of the stimulus. Green: Comparison between the different levels of intrinsic pleasantness.
Red: comparison between the different levels of goal conduciveness. Neg ⫽ negative; Neut ⫽ neutral; Pos ⫽ positive.
348
GRANDJEAN AND SCHERER
COGNITIVE ARCHITECTURE OF EMOTION
349
Figure 5. Experiment 2: (A) Timing of the wavelet coefficients (absolute values) for the raw electroencephalography in the gamma band. (B) Timing for the wavelet coefficients (z score of the absolute values) for each
participant (n ⫽ 16) for the loss level of the goal conduciveness factor (upper part) and for the negative level
of the intrinsic pleasantness factor (lower part). A significant interaction effect between these two levels occurs
from 560 ms to 640 ms after onset of the stimuli, F(1, 16) ⫽ 4.63, p ⬍ .05.
For the IP factor, the results of the second experiment are similar
to those reported for the first experiment; that is, we did not find
any specific topographic maps for this factor (nor for the goal
conduciveness factor). However, the GFP analyses did reveal
significant effects for these appraisal factors. For IP, effects started
at about 200 ms, F(2, 32) ⫽ 24.08, p ⬍ .0001, and continued to
750 ms, F(2, 32) ⫽ 7.83, p ⬍ .005. For goal conduciveness, a
significant effect appeared much later, at about 400 – 450 ms, F(2,
32) ⫽ 3.72, p ⬍ .05.
The wavelet analyses of the ERPs revealed a significantly
earlier effect for IP than for goal conduciveness in the delta band
(Figure 3C), as well as in higher frequency bands (Figures 3D and
3E; see Figure 4B).
The wavelet analyses of the raw EEG data revealed an early IP
effect in the alpha band at about 260 –320 ms (Figure 3F) but no
effect for goal conduciveness. However, a significant effect for
goal conduciveness was found in the higher frequency region,
specifically in the beta and gamma range from about 580 to 640
ms. The contrast revealed an effect of the gain/loss conditions
compared with the no effect level (see Figure 4C). These later
effects in the gamma band (not present in the DWT analyses of
unfiltered ERPs), indicating an effect of goal conduciveness on the
so-called induced gamma (Tallon-Baudry et al., 1996) at about 600
ms after the onset of the stimuli (see Figure 5), suggest that a high
level of cognitive processing is involved in this type of appraisal.
The results support the CPM predictions and suggest that novelty and intrinsic pleasantness may be appraised early, on an
unconscious, automatic, and possibly schematic level, whereas
goal conduciveness tends to be evaluated later in the sequence, on
a conscious, controlled, and propositional level.
General Discussion
These studies show that both the nature and the timing of the
emotion– constitutive–appraisal processes can be examined by objective measures of electrical activity in the brain. This approach
allows theory-guided investigation of the nature of appraisalgenerated emotion elicitation, both for automatic, unconscious
processing and for higher, controlled levels of processing, in much
greater detail than has previously been the case. The results show
350
GRANDJEAN AND SCHERER
that the different appraisal checks have specific brain state correlates, occur rapidly in a brief time window after stimulation, and
have effects that occur in sequential rather than parallel fashion.
Furthermore, the results show that the duration of processing of
different checks varies, as they do not achieve preliminary closure
at the same time, again supporting model predictions (Sander et al.,
2005; Scherer, 2001, 2004). From the results on the timing of the
different checks, we infer that early appraisal checks, including
novelty and intrinsic pleasantness detection, are likely to occur in
an automatic, unconscious mode of processing, whereas later
checks, specifically goal conduciveness, require more extensive,
effortful, and controlled processing.
Given the lack of prior work on the mental chronography of
appraisal, this first attempt to examine the dynamic nature of this
process is necessarily not without limitations. One type of limitation concerns the nature of the manipulation of the different
appraisal checks. Thus, the goal relevance check in the first experiment was manipulated by the task that the participant had to
achieve, a frequently used paradigm (i.e., Wright & Liu, 2006) that
corresponds well to a controlled laboratory setting. Obviously, the
goal–need relevance check can take many other forms depending
on the context. In everyday life, the current goals or needs of
individuals can be very different, sometimes related to the current
task, at other times related to long-term goals or needs. Further
studies are needed to systematically manipulate the different types
of goal–need relevance to examine potential differences.
Similarly, we operationalized goal conduciveness using monetary gains and losses on the basis of the literature on reward and
punishment. The purpose of this experimental design was to compare the notion that the reward value of intrinsic pleasantness (e.g.,
of a picture) can be clearly differentiated from the reward value of
reaching a goal of winning money, as well as to test the prediction
of differential timing of the respective effects of these processes on
ERPs and EEG measures. Again, appraisal in real life often concerns different kinds of goals, and future research needs to examine the potential effects of different needs, goals, or values.
Another limitation concerns the fact that we inferred the level of
processing at which the respective appraisals might have occurred
on the basis of plausibility considerations, including rapidity of
onset. Further work is needed to increase our understanding of
these different levels of processing, attempting to determine the
respective processing level more directly, preferably using brain
imagery methods, and studying the interactions between different
checks and the levels on which they are processed.
Many of these limitations are imposed by the fact that our
measurement techniques, while representing current state of the
art, do not yet allow us to examine some of the issues outlined
above. Overall, this work, combined with growing evidence for the
CPM’s response patterning predictions concerning autonomic
physiological signatures (Aue, Flykt, & Scherer, 2007; Pecchinenda & Smith, 1996; van Reekum et al., 2004), facial muscle
movements (Lanctot & Hess, 2007), and vocalization changes
(Johnstone, van Reekum, Hird, Kirsner, & Scherer, 2005) encourage the use of this model to guide further work on the dynamics of
emotion– constituent appraisal. Because of its explicit specification of the dynamics and recursiveness of the afferent and efferent
processes involved, the CPM also provides an appropriate basis for
the unpacking of the cognitive architecture of emotion and its
computational modeling (Sander et al., 2005).
In this work, we focused on only one component of emotion.
Full-blown emotions are characterized not only by the underlying
appraisal process but also by a series of efferent phenomena like
action tendencies, motor expressions, motivational processes, and
subjective feeling that can, in turn, influence the appraisal process.
One of the tasks for future research will be the assessment of
efferent effects of the appraisal results occurring at a specific point
in time. As described in the beginning of this article, we assume
that the processing of different appraisal checks occurs in fact in a
parallel fashion but that the results of major appraisal checks (as
shown in unique brain activation patterns) occur in a sequence. We
expect that it is at these points that efferent effects are produced,
for example, somatovisceral activation and innervations of the
striate musculature involved in expression. However, with our
present approach we cannot demonstrate this as it requires synchronized recording of EEG signals on the one hand and somatovisceral variables on the other. We are currently working on the
technical realization of this complex measurement design.
With respect to research strategy, the neurochronometry of
appraisal, establishing precise timing for different checks, constitutes a royal road to the investigation of the fundamental properties
of the central nervous system, allowing organisms to appraise
events in their environment in a systematic, sequential, and recursive manner and then to react adaptively to the exigencies of the
situation. Further work using neurochronometry to test the predicted timing of the coping potential and normative significance
appraisal checks, to further disentangle the levels of processing,
and to examine the synchronization of emotion components by
sequential appraisal is currently in preparation.
References
Arnold, M. B. (1960). Emotion and personality: Vol. 1. Psychological
aspects. New York: Columbia University Press.
Aue, T., Flykt, A., & Scherer, K. R. (2007). First evidence for differential
and sequential efferent effects of stimulus relevance and goal conduciveness appraisal. Biological Psychology, 74, 347–357.
Başar, E., Demiralp, T., Schürmann, M., Başar-Eroglu, C., & Ademoglu,
A. (1999). Oscillatory brain dynamics, wavelet analysis, and cognition.
Brain and Language, 66, 146 –183.
Bertrand, O., & Tallon-Baudry, C. (2000). Oscillatory gamma activity in
humans: A possible role for object representation. International Journal
of Psychophysiology, 38, 211–223.
Ellsworth, P. C. (1991). Some implications of cognitive appraisal theories
of emotion. In K. T. Strongman (Ed.), International review of studies of
emotion (Vol. 1, pp. 143–161). New York: Wiley.
Ellsworth, P. C., & Scherer, K. R. (2003). Appraisal processes in emotion.
In R. J. Davidson, K. R. Scherer, & H. Goldsmith (Eds.), Handbook of
affective sciences (pp. 572–595). New York: Oxford University Press.
Gratton, G., Coles, M. G., & Donchin, E. (1983). A new method for
off-line removal of ocular artifact. Electroencephalography and Clinical
Neurophysiology, 55, 468 – 484.
Johnstone, T., van Reekum, C. M., Hird, K., Kirsner, K., & Scherer, K. R.
(2005). Affective speech elicited with a computer game. Emotion, 5,
513–518.
Lachaux, J. P., George, N., Tallon-Baudry, C., Martinerie, J., Hugueville,
L., Minotti, L., et al. (2005). The many faces of the gamma band
response to complex visual stimuli. NeuroImage, 1, 491–501.
Lanctot, N. G., & Hess, U. (2007). The timing of appraisals. Emotion, 7,
201–212.
Lang, P. J., Bradley, M. M., & Cuthbert, B. N. (1999). International
Affective Picture System (IAPS): Instruction manual and affective rat-
COGNITIVE ARCHITECTURE OF EMOTION
ings (Tech. Rep. No. A-4). Gainesville: The Center for Research in
Psychophysiology, University of Florida.
Lazarus, R. S. (1966). Psychological stress and coping process. New York:
McGraw-Hill.
Lazarus, R. S. (1991). Emotion and adaptation. New York: Oxford University Press.
Lazarus, R. S. (1999). The cognition-emotion debate: A bit of history. In
T. Dalgleish & M. J. Power (Eds.), Handbook of cognition and emotion
(pp. 3–19). New York: John Wiley & Sons Ltd.
Leventhal, H., & Scherer, K. R. (1987). The relationship of emotion and
cognition: A functional approach to a semantic controversy. Cognition
and Emotion, 1, 3–28.
Michel, C. M., Murray, M. M., Lantz, G., Gonzalez, S., & Grave de
Peralta, R. (2004). EEG source imaging. Clinical Neurophysiology, 115,
2195–2222.
Pecchinenda, A., & Smith, C. A. (1996). The motivational significance of
skin conductance activity during a difficult problem-solving task. Cognition and Emotion, 10, 481–503.
Ranganath, C., & Rainer, G. (2003). Neural mechanisms for detecting and
remembering novel events. Nature Reviews Neuroscience, 4, 193–202.
Samar, V. J., Bopardikar, A., Rao, R., & Swartz, K. (1999). Wavelet
analysis of neuroelectric waveforms: A conceptual tutorial. Brain and
Language, 66, 7– 60.
Sander, D., Grandjean, D., & Scherer, K. R. (2005). A systems approach to
appraisal mechanisms in emotion. Neural Networks, 18, 317–352.
Scherer, K. R. (1984). On the nature and function of emotion: A component process approach. In K. R. Scherer & P. Ekman (Eds.), Approaches
to emotion (pp. 293–317). Hillsdale, NJ: Erlbaum.
Scherer, K. R. (1999). On the sequential nature of appraisal processes:
Indirect evidence from a recognition task. Cognition and Emotion, 13,
763–793.
351
Scherer, K. R. (2001). Appraisal considered as a process of multilevel
sequential checking. In K. Scherer, A. Schorr, & T. Johnstone (Eds.),
Appraisal processes in emotion: Theory, methods, research (pp. 92–
120). New York: Oxford University Press.
Scherer, K. R. (2004). Feelings integrate the central representation of
appraisal-driven response organization in emotion. In A. S. R. Manstead,
N. H. Frijda, & A. H. Fischer (Eds.), Feelings and emotions: The
Amsterdam Symposium (pp. 136 –157). Cambridge, England: Cambridge
University Press.
Scherer, K. R., Dan, E., & Flykt, A. (2006). What determines a feeling’s
position in three-dimensional affect space? A case for appraisal. Cognition and Emotion, 20, 92–113.
Smith, C. A., & Lazarus, R. S. (1990). Emotion and adaptation. In L. A.
Pervin (Ed.), Handbook of personality: Theory and research (pp. 609 –
637). New York: Guilford Press.
Tallon-Baudry, C., Bertrand, O., Delpuech, C., & Pernier, J. (1996).
Stimulus specificity of phase-locked and non-phase-locked 40 Hz visual
responses in human. Journal of Neuroscience, 16, 4240 – 4249.
van Reekum, C. M., Johnstone, T., Banse, R., Etter, A., Wehrle, T., &
Scherer, K. R. (2004). Psychophysiological responses to appraisal dimensions in a computer game. Cognition and Emotion, 18, 663– 688.
Vuilleumier, P., Armony, J., Driver, J., & Dolan, R. J. (2003). Distinct
spatial frequency sensitivities for processing faces and emotional expressions. Nature Neuroscience, 6, 624 – 631.
Wright, P., & Liu, Y. (2006). Neutral faces activate the amygdala during
identity matching. NeuroImage, 15, 628 – 636.
Received February 28, 2007
Revision received February 11, 2008
Accepted February 11, 2008 䡲
A
Time windows (ms)
Frequency bands
0-128
257-384
TaskR > FamNR
[F(1,13)=36.30, P<.0005]
129-256
Nov > FamNR
[F(1,13)=6.58, P<.05]
Nov > TaskR
[F(1,13)=5.77, P<.05]
Delta
TaskR > FamNR
[F(1,13)=8.82, P<.05]
TaskR > Nov
[F(1,13)=17.06, P<.005]
Theta
0-64
65-128
0-32
33-64 65-96
Beta
Neg > Neut
[F(1,13)=6.91, P<.05]
129-192
TaskR > FamNR
[F(1,13)=5.21, P<.05]
TaskR > Nov
[F(1,13)=12.53, P<.005]
Neut > Neg
[F(1,13)=23.76, P<.001]
Alpha
97-128
Nov < FamNR
[F(1,13)=6.15, P<.05]
Nov < TaskR
[F(1,13)=5.73, P<.05]
Neut > Neg
[F(1,13)=23.01, P<.0005]
Neut > Pos
[F(1,13)=5.92, P<.05]
129-160
161-192
513-640
385-512
Neg > Neut
[F(1,13)=12.99, P<.005]
193-256
Nov > FamNR
[F(1,16)=9.11, P<.01]
257-320
193-224 225-256
B
257-288
289-320
385-448
321-384
321-352
353-384
385-416
449-512
417-448 449-480 481-512
513-576
513-544
577-640
545-576
577-608
609-640
Time windows (ms)
Frequency bands
Delta
0-128
129-256
257-384
Neg > Neut
[F(1,16)=9.43, P<.005]
385-512
513-640
Loss > Gain
[F(1,16)=6.53, P<.05]
Neg > Pos
[F(1,16)=17.65, P<.0005]
Pos > Neg
[F(1,16)=10.53, P<.01]
Pos > Neut
[F(1,16)=25.61, P<.0005]
129-192
193-256
Theta
0-64
65-128
257-320
Alpha
0-32 33-64 65-96
Beta
97-128
Neg > Pos
[F(1,16)=7.35, P<.01]
129-160 161-192 193-224 225-256
257-288
321-384
385-448
Neg > Pos
[F(1,16)=4.57, P<.05]
289-320 321-352 353-384 385-416
417-448
C
449-512
449-480
Loss > Gain
[F(1,16)=8.37, P<.05]
513-576
Loss > Gain
[F(1,16)=5.53, P<.05]
481-512 513-544
545-576
577-640
577-608
609-640
Time windows (ms)
Frequency bands raw EEG
0-128
129-256
257-384
385-512
513-640
Neg > Pos
[F(1,16)=21.3, P<.01]
Neut > Pos
[F(1,16)=5.65, P<.05]
Delta
Theta
0-64
65-128
129-192
193-256
0-32
33-64
65-96
97-128 129-160
161-192
193-224
257-320
Neg < Pos
[F(1,16)=20.42, P<.05]
Neg < Neut
[F(1,16)=8.43, P<.05]
225-256
257-288
289-320
0-32
33-64
65-96
97-128 129-160
161-192
193-224
225-256
Alpha
321-384
385-448
321-352
353-384 385-416
417-448
321-352
353-384 385-416
417-448
Beta
Gamma
257-288
289-320
449-512
513-576
481-512
Loss > NoEff
[F(1,16)=5.60, P<.05]
Loss > Gain
[F(1,16)=4.86, P<.05]
449-480
481-512
513-544 545-576
449-480
513-544 545-576
577-640
577-608
609-640
577-608
609-640
Gain > NoEff
Gain > NoEff
[F(1,16)=6.93, P<.05] [F(1,16)=7.19, P<.05]
Loss > NoEff
[F(1,16)=5.34, P<.05]