International Journal of Psychophysiology 79 (2011) 236–243

Contents lists available at ScienceDirect

International Journal of Psychophysiology

j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / i j p s yc h o

Influence of affective significance on different levels of processing using pupil

dilation in an analogical reasoning task
Kristin Prehn a,⁎, Hauke R. Heekeren a,b, Elke van der Meer c
a
Cluster of Excellence “Languages of Emotion”, Freie Universität, Berlin, Germany
b
Dept. of Education & Psychology, Freie Universität, Berlin, Germany
c
Dept. of Psychology, Humboldt University, Berlin, Germany

a r t i c l e i n f o a b s t r a c t

Article history: The present study investigates the interaction of cognition and emotion in decision making, using an
Received 3 September 2010 analogical reasoning task. In this task, two word pairs were presented simultaneously. Each word pair could
Received in revised form 22 October 2010 be characterized by an associative conceptual relation (object, actor, or location relation) as well as an
Accepted 26 October 2010
emotional relation (negative, neutral, or positive valence). Both types of relations were equally task-relevant:
Available online 31 October 2010
Participants had to identify both types of relations, to compare them, and to decide whether or not the word
Keywords:
pairs were analogous, i.e., corresponding in both conceptual and emotional relations. Behavioral data showed
Pupillary responses that emotional relations were identified preferentially and faster than conceptual relations. Pupil dilations
Analogical reasoning reflected the descending difficulty of the conditions and were greatest in amplitude when both conceptual
Decision making and emotional correspondence was shown, intermediate when only one type of relation (either the emotional
Resource allocation or the conceptual) corresponded, and least when neither correspondence existed. Additionally, a negative
Emotional and cognitive processing valence of the word material slowed down response times and increased pupil dilation relative to positive and
neutral items. In summary, pupil and response time data together support recent (neurobiological) models
concerning the interaction of emotion and cognition by showing that affective significance leads to a
processing advantage at a cognitively lower level of information processing (here, identification or retrieval of
relations from long-term memory) but can also distract people from higher level cognitive processes (here,
from the controlled comparison of retrieved relations).
© 2010 Elsevier B.V. All rights reserved.

1. Introduction stimuli compete for representation, processing, and control over

behavior. Sometimes, this competition results in an interference of
Everyday decision making includes both cognitive and affective more salient but task-irrelevant emotional stimuli on cognitive
aspects. Affective aspects of a situation are mostly vital for the processing (for an investigation of emotional interference on cognitive
immediate survival of an organism, because they can signify a potential control, see Hartikainen et al., 2000; McKenna, 1986; Mitchell et al.,
threat or a possible reward. This processing advantage of affective 2008).
aspects might be due to the fact that the brain can filter out relevant While many studies have investigated the selective advantage of
information and directs information processing automatically toward affective processing in general or an interference effect of task-irrelevant
aspects of a situation that are important (e.g., for immediate survival). affective stimuli on executive functioning (for reviews, see Compton,
The so-called biased competition model of attention (Desimone and 2003; Ochsner and Feldman Barrett, 2001; Pessoa and Ungerleider,
Duncan, 1995), for instance, suggests that more important aspects of a 2003; Phelps, 2006), it is still unclear how cognitive and affective
situation are selectively attended and processed faster with increased processes interact when both aspects are equally task-relevant and how
allocation of mental resources (for an extension of the biased this interaction of cognitive and affective aspects is reflected at different
competition model for conceptual representations, see Kan and (lower and higher) levels of information processing.
Thompson-Schill, 2004). Moreover, this model suggests that attending To approach this question, we developed an analogical reasoning
to one aspect or class of stimuli reduces the cognitive resources available task based on current models of long-term memory, according to
for others. Thus, different aspects of a situation or different classes of which semantic knowledge can be depicted as a network consisting of
concepts and semantic relations (for a review on the organization of
long-term memory in humans, see Martin and Chao, 2001).
⁎ Corresponding author. Freie Universität Berlin, Cluster of Excellence “Languages of
Emotion”, Habelschwerdter Allee 45, 14195 Berlin, Germany. Tel.: + 49 30 83851378;
In this task, two word pairs were presented simultaneously. To
fax: + 49 30 83851370. begin with, each word pair could be characterized by a conceptual
E-mail address: [email protected] (K. Prehn). relation, which was always an associative conceptual relation. That is,

0167-8760/$ – see front matter © 2010 Elsevier B.V. All rights reserved.
doi:10.1016/j.ijpsycho.2010.10.014
 K. Prehn et al. / International Journal of Psychophysiology 79 (2011) 236–243 237

concepts were related to each other, because they were sharing a mental resource allocation that reflects processing of both emotional
thematic category (e.g., CAR and GARAGE), in contrast to taxonomical and conceptual relations.
conceptual relations, in which concepts are from the same (hierarchi- Combining behavioral and psychophysiological measurements (in
cally structured) semantic category (e.g., CAR, TRAIN, BUS; cf. Lin and particular, using the measurement of allocated mental resources), in
Murphy, 2001; Sachs et al., 2008a). Types of associative conceptual the present study, we aimed to investigate how cognition and
relations (Collins and Quillian, 1969; Herrmann and Chaffin, 1986; Klix, emotion interact during decision making and whether one aspect
1992) used in the word material were actor relations (for instance, the dominates the other. Based on the literature reviewed above, we
concepts BIRD and CHIRP are characterized by an actor relation, i.e., “The hypothesized that emotional relations are processed preferentially
bird chirps.”), object relations (e.g., FEED–BIRD), and location relations (i.e., faster and with lower error rates), relative to associative
(e.g., BIRD–CAGE). Second, concepts and their relations were emotion- conceptual relations. Regarding the cognitively higher-level compar-
ally biased and could additionally be characterized by an emotional ison process (in contrast to the lower-level retrieval process), we
relation signifying either a pleasant (e.g., BIRD–SING), neutral (e.g., hypothesized that affective significance would lead to a processing
BIRD–FLY), or unpleasant (e.g., BIRD–ROT) evaluation (van der Meer, disadvantage leading to longer response times and increased mental
1989). An analogy was defined by correspondence of both conceptual resource allocation during the processing of items with negative
and emotional relations of the two word pairs presented. Participants emotional valence compared to neutral items.
had to identify, that is, retrieve conceptual and emotional relations in
each of the two word pairs from long-term memory, compare them, and 2. Methods
decide whether or not the word pairs were identical in both.
As the long-term memory model suggests, associative conceptual 2.1. Participants
and emotional relations should be retrieved automatically (lower level of
information processing; Sachs et al., 2008a,b; van der Meer, 1989). The 30 healthy subjects (n = 30, 11 males) with the mean age of
cognitively higher-level comparison process, however, which takes place 23.93 years (S.D. = 4.34) participated in this study. All participants
to check whether the analogy condition is fulfilled or not, requires some were native German speakers, had no history of neurological or
executive control function (Bunge et al., 2005; Krawczyk, 2002; psychiatric diseases, and did not take any medication that could
Krawczyk et al., 2008). influence the pupillary response. Participants reported no ophthal-
We measured response times and error rates as behavioral mologic problems (other than correctable eyesight). Participants gave
indicators for task performance and recorded pupillary responses as written consent prior to investigation, according to the Declaration
a reliable indicator for mental resource allocation. of Helsinki (1964), and received either course credit or payment
Pupil diameter is controlled by two muscles innervated by the (20 Euro) for their participation.
sympathetic and parasympathetic branches of the autonomic nervous
system, which get their input from the central nervous system and 2.2. Task and stimulus material
structures essential to both emotional and cognitive information
processing (Aston-Jones and Cohen, 2005; Granholm and Steinhauer, As described above, we developed an analogical reasoning task to
2004; for a detailed description of the neural basis of pupillary investigate the processing of cognitive and affective aspects during
responses, see Hoeks and Ellenbroek, 1993 or Steinhauer and decision making (see Prehn et al., 2008; van der Meer, 1989).
Hakerem, 1992). While tonic changes in pupil diameter reflect the In this task, participants had to identify both the conceptual and the
sensitivity of the cognitive system to all kinds of stimuli (“exploration emotional relations in two simultaneously presented word pairs,
mode”, cf. Aston-Jones and Cohen, 2005), phasic changes have been compare them, and decide whether or not the word pairs were an
proven as a sensitive and reliable psychophysiological measure of the analogy, i.e., corresponding in both conceptual and emotional relations.
task-related processing load, with larger pupil dilation reflecting To indicate whether the item presented on a computer screen was an
greater processing demands (Beatty, 1982; Granholm and Steinhauer, analogy or not, participants had to press one of two buttons (labeled
2004; Loewenfeld, 1993). The phasic “exploitation mode” (cf. Aston- “yes“ or “no“) of a response device as quickly and correctly as possible,
Jones and Cohen, 2005; van der Meer et al., 2010) typically occurs using the middle and index fingers of the dominant hand. The
when an individual is engaged in a particular task and focusing on assignment of “yes” and “no” to the response fingers was counter-
task-relevant stimuli while ignoring distracting or task-irrelevant balanced across participants. Half of the 216 items were identical in
ones. Using a digit span recall task, Kahneman and Beatty (1966), for both conceptual and emotional relations (analogy items). In the other
example, demonstrated that pupil diameter proportionally increases half (non-analogy items), either the conceptual or the emotional
as a function of the number of digits that have to be maintained in relations or both did not correspond (see Table 1 for a translated
short-term memory. Pupil diameter increases until individuals reach example for each of the four experimental conditions). Each word pair
their limit of available cognitive resources (i.e., until their memory was used twice.
capacity of 7 ± 2 digits; Granholm et al., 1996). Notably, Just et al. Thus, we had four different conditions (Con = Emo=: conceptual
(2003) have demonstrated that pupillary responses reflect an overall and emotional relations between the word pairs were identical or
aggregate of mental resource allocation that is not limited to a specific corresponding, n = 108; Con = Emo≠: conceptual relations cor-
part of the cognitive system. In studies using multiple tasks (for a responding, but emotional not corresponding, n = 36; Con ≠ Emo=:
review, see Beatty and Lucero-Wagoner, 2000), peak dilation has been conceptual relations not corresponding, but emotional corresponding,
found to increase systematically with enhanced processing demand, n = 36; Con ≠ Emo≠: conceptual and emotional relations both not
comprising language comprehension (e.g., Hyona et al., 1995; Just and corresponding, n = 36). The factors “conceptual correspondence”
Carpenter, 1993; Nuthmann and van der Meer, 2005; Raisig et al., (Con= vs. Con≠) and “emotional correspondence” (Emo= vs. Emo≠)
2007), attention allocation (e.g., Karatekin et al., 2004; Kim et al., were independently varied in a 2×2 factorial design.
2000), higher cognitive processes (van der Meer et al., 2010), memory The word material was matched across conditions for number of
maintenance and semantic elaboration (e.g., Granholm et al., 1996; letters [M = 28.57, S.D. = 3.94, F(3, 215) = 1.83, p = .14], number of
van der Meer et al., 2003), or valence identification (e.g., Siegle et al., syllables [M = 4.60, S.D. = 0.83, F(3, 215) = 0.25, p = .86] and word
2001; Steinhauer et al., 1983). In a previous study using a very similar frequencies [M = 18.23 per million words, S.D.= 17.82, F(3, 215) = 0.89,
task as in the present study, we showed the increase of pupil diameter p = .45; Baayen et al., 1993].
during emotional interference (Prehn et al., 2008). In summary, all these To further validate the word material, we conducted an additional
studies support the view that pupil size represents a general index of questionnaire-based investigation (n = 90) and collected rating data
238 K. Prehn et al. / International Journal of Psychophysiology 79 (2011) 236–243

Table 1
Examples for word material in the analogical reasoning task.

Emotional relations Emotional relations

=

Conceptual relations TUMOR – BRAIN/ CANCER – BREAST/

= RAT – CELLAR SHELL – BEACH
n = 108 n = 36

Conceptual relations COCKROACH – KITCHEN/ MURDERER – PARK/

BODY – DECAY BIRD – CHIRP
n = 36 n = 36

Emotional relations =, emotional relations corresponding; Emotional relations ≠,

emotional relations non-corresponding; Conceptual relations =, conceptual relations
corresponding; Conceptual relations ≠, conceptual relations non-corresponding. The gray
colored field marks the analogy condition in which participants had to answer “yes”. That
is, both the conceptual and the emotional relations in the first word pair completely match
the conceptual and the emotional relations in the second word pair. All other conditions
are non-analogy conditions.

on how each word pair was judged regarding imagery, typicality and Fig. 1. Schematic illustration of an experimental trial. First, a fixation cross appeared
for 1 s (= baseline phase). Then, the item was presented (= stimulus presentation
personal familiarity. Rating data confirmed that the four conditions
phase). As soon as a response button was pressed by the participant, the item
were matched regarding these variables.1 disappeared from the screen, followed by another fixation cross for 5 s (= relaxation
phase). After relaxation phase, a smiley appeared on the screen indicating that
participants were now allowed to blink and could start the next trial by pressing one of
2.3. Experimental procedure the response buttons (= blinking phase).

The experiment took place in a quiet and moderately illuminated

room (about 500 lux). The participants were seated comfortably in
participants, first, rated all word pairs from the experiment regarding
front of a computer screen with a distance of approximately 0.7 m.
emotional arousal and, second, regarding emotional valence on a
Each trial consisted of four phases (see Fig. 1). First, a fixation cross
seven point rating scale from 0 (very unpleasant or low arousal) to 6
appeared for 1 s (= baseline phase). Then, the item (consisting of the
(very pleasant or high arousal).
two word pairs) was presented (= stimulus presentation phase). As
soon as a response button was pressed by the participant, the item
disappeared from the screen, followed by another fixation cross for 5 s 2.4. Apparatus
(= pupil relaxation phase). After the relaxation phase, a smiley
appeared on the screen, indicating that participants were now Stimuli were presented using the experimental control software
allowed to blink. Participants then could start the next trial (by Presentation (Neurobehavioral Systems Inc, Albany, CA), running on a
themselves), by pressing one of the response buttons (= blinking Microsoft Windows XP operating system. The computer used for
phase). During a trial, the participants were asked not to move their stimulus presentation collected the behavioral data (response times
head, to maintain fixation, and to restrict eye blinks if possible until and errors) and was connected with another computer for registra-
the blinking phase at the end of the trial. tion and storage of the pupil data for offline analyses, allowing the
The experiment itself contained 216 trials in total (presented in a transmission of trigger signals to mark the beginning of every trial.
randomized order) and consisted of two blocks with a break in Pupillary responses were continuously recorded using an iView
between. system (SensoMotoric Instruments, Teltow, Germany). The pupill-
Response times (measured as the time between appearance of the ometer (i.e., an infrared light source and a video camera sensitive to
item and the button press), error rates, and pupillary responses, as infrared light) was mounted on a headset. The light source (λ = 700–
indicator for resource allocation, were recorded. Prior to the 1049 nm) and the camera were pointed at the participants' right eye.
experiment, participants received written instructions and completed Pupil size was recorded at 50 Hz (i.e., every 20 ms).
a practice session (20 trials) with stimulus material from a different For minimizing possible pupillary responses as result of changes in
set (but similar to the material used in the experiment) to become screen illumination (associated with stimulus on- and offset) and thus
familiar with task and procedure. Immediately after the experiment, to control for the confounding effect of the initial light reflex reaction
(Loewenfeld, 1993; Steinhauer and Hakerem, 1992), stimuli were
1
Median ratings for each item (averaged across participants) were compared using drawn in black on a gray background. The mean luminance of the
nonparametric U-tests. There were no significant differences regarding these variables
between the four conditions (typicality: Con=Emo= vs. Con=Emo≠: U =1690.5,
stimuli was 24.64 cd/m2 (S.D. = 0.66). As mentioned above, the
p =.24; Con= Emo= vs. Con≠Emo=: U= 1878.5, p =.76; Con= Emo= vs. Con≠ number of syllables and number of letters was balanced for the
Emo≠: U= 1541, p = .06; Con= Emo≠ vs. Con≠Emo=: U =583.5, p =.46; Con= experimental conditions. This was done to keep reading times
Emo≠ vs. Con≠Emo≠: U =602.5, p= .61; Con≠ Emo= vs. Con≠Emo≠: U= 538, constant as well as to assure that the luminance of the stimuli did
p =.21; familiarity: Con= Emo= vs. Con= Emo≠: U= 1682, p = .23; Con= Emo= vs.
not systematically differ between the conditions.
Con≠ Emo=: U =1913, p = .89; Con= Emo = vs. Con ≠ Emo≠: U = 1623, p =.14;
Con = Emo≠ vs. Con ≠ Emo=: U = 579.5, p = .44; Con = Emo ≠ vs. Con ≠ Emo≠: The iView system samples pupil diameter in terms of pixels. To
U =620, p =.75; Con≠Emo= vs. Con≠Emo≠: U= 552.5, p= .28; imagery: Con= relate this measure to absolute pupil size, we recorded the size of a
Emo= vs. Con=Emo≠: U =1408, p =.01, not significant because of adjusted αb 0.0083 black dot of exactly 5 mm in diameter placed on the closed lid of the
for 6 comparisons; Con= Emo= vs. Con≠ Emo=: U =1848, p= .65; Con=Emo= vs. participants right eye before each of the two experimental blocks. This
Con≠Emo≠: U= 1792.5, p =.47; Con=Emo≠ vs. Con≠Emo=: U= 495, p =.08;
Con = Emo≠ vs. Con ≠ Emo≠: U = 504.5, p = .10; Con ≠ Emo = vs. Con ≠ Emo≠:
calibration procedure made it possible to convert pupil diameter from
U =631, p =.84). Thus, a confounding influence of these variables on task performance pixels to millimeters for each participant. Pupil diameter was assessed
can be ruled out. with an accuracy of 0.05 mm.
 K. Prehn et al. / International Journal of Psychophysiology 79 (2011) 236–243 239

2.5. Data cleaning, reduction, and analyses Post hoc ratings of the word pairs were compared regarding
emotional arousal and valence using non-parametric Wilcoxon tests.
Response times and pupil data were only analyzed for correctly Neutral, positive, and negative word pairs differed regarding the rated
answered trials (analogy as well as non-analogy items). Trials with emotional valence [neutral vs. positive: Z = −4.993, p b .001; neutral vs.
response times 2 S.D. below or 2 S.D. above the mean were also negative: Z = −4.941, p b .001; positive vs. negative: Z = −4.891,
excluded from all analyses. Data outside this range were considered to p b .001; n = 30]. That is, participants were clearly able to discriminate
have been generated by processes unrelated to the task. Because the valence of emotional relations. Neutral, positive, and negative word
response times differed significantly between the four conditions (see pairs also differed regarding the rated emotional arousal [neutral vs.
Results), the boundaries for outliers were calculated for each positive: Z = −4.668, p b .001; neutral vs. negative: Z = −4.641, p b .001;
condition separately. Outliers (5.1% in total) were not systematically positive vs. negative: Z = −3.467, p b .001]. Arousal ratings were higher
distributed across experimental conditions [F(3, 119) = 0.11, p = .96]. for negative compared with positive and neutral word pairs indicating
In total, 12.8% of the trials were discarded from the outset due to their higher affective significance.
errors and outliers (for all participants and conditions).
Pupillary responses were analyzed using Matlab 7.0.4. (The Math- 3.2. Behavioral data
Works, Inc., MA, USA) following standard procedures (cf. Beatty and
Lucero-Wagoner, 2000; Granholm et al., 1996; Verney et al., 2001). First, Mean response times and error rates were computed for each of
we removed artifacts due to excessive blinking or eye-movements. This the four experimental conditions and averaged across participants
procedure eliminated another 27.5% of the trials in total (for all (n = 30, see Table 2).
participants and conditions). Artifacts were also not systematically A 2 (correspondence of conceptual relations: Con = vs. Con≠) × 2
distributed across experimental conditions [F(3, 119) = 2.22, p = .09]. (correspondence of emotional relations: Emo = vs. Emo≠) repeated
Very small blinks (defined as large changes in pupil diameter occurring measure analyses of variance (ANOVA) for response times was
too rapidly to signify actual pupil dilation or constriction) were replaced performed. There was a significant main effect for emotional
by linear interpolation. Second, we computed the baseline pupil correspondence [F(1, 29) = 67.36, M.S.E. = 141,378.91, p b .001,
diameter for each trial by averaging pupil diameter 200 ms (i.e., four η2 = .70], that is, participants responded faster when emotional
data points) before presentation of the item. This baseline was relations between the word pairs did not correspond. There was no
subtracted from the respective trial (baseline correction). Mean baseline effect for conceptual correspondence [F(1, 29) = 1.961, M.S.E. =
pupil diameter was 2.92 mm (S.D. = 1.18 mm, minimum: 1.38 mm, 66,992.93, p = .172, η2 = .06] and no interaction between emotional
maximum: 5.57 mm). We then computed stimulus-locked pupillary and conceptual correspondence [F(1, 29) = 1.6, M.S.E. = 33,104.76,
responses for each trial and averaged them for each condition and p = .216, η2 = .05]. The significant main effect for emotional corre-
participant. Data were smoothed using an unweighted seven point spondence was further analyzed by paired t-tests. For non-analogy
moving average filter. Finally, peak dilation and latency to peak of the items, we found a difference in response times between the
pupillary responses was determined as the maximum pupil diameter conditions Con = Emo≠ and Con≠Emo = [t(29) = −5.41, p b .001, s.
obtained in the interval of interest between 0.5 and 3.5 s after stimulus Fig. 3] but no difference between conditions Con = Emo≠ and
presentation for each participant and condition. Con≠Emo≠ [t(29) = 2.58, p = .015; only trend because of adjusted
The determination of the peak dilation was not based on individual α b 0.0083 for 6 comparisons].
trials, but on averaged pupillary response for each participant and For error rates, there was a significant main effect for emotional
each condition (see Beatty and Lucero-Wagoner, 2000; Granholm correspondence [F(1, 29) = 13.539, MSE = 0.015, p = .001, η2 = .32],
et al., 1996; Nuthmann and van der Meer, 2005; Prehn et al., 2008; no effect for conceptual correspondence [F(1, 29) = 0.842, MSE = 0.01,
Verney et al., 2004). Averaging across a certain number of trials is p = .366, η2 = .03], but an interaction between emotional and
necessary because pupil response is prone to spontaneous fluctua- conceptual correspondence [F(1, 29) = 29.22, M.S.E. = 0.223, p b .001,
tions. In the present study, average pupillary response was based on η2 = .50]. Error rate especially increased in the condition in
59.87 trials (55,44%) in condition Con = Emo = that included 108 which emotional relations corresponded while conceptual did not
trials in total (minimum: 39 trials, maximum: 97 trials) and on 21.99 (Con≠Emo=, see Table 2). Because response times were also
trials on average (61,08%) in the other three conditions that included increased in this condition, there is no evidence for a speed-accuracy
36 trials each in total (minimum: 11 trials, maximum: 36 trials). trade-off in the data.
All analyses [repeated measure analyses of variance (ANOVA) and
paired t-tests] on response times and pupil data were run with means Table 2
obtained for each participant and condition. A rejection criterion of Means and standard errors (S.E.) for response times (n = 30 participants; in milliseconds,
p b .05 was chosen in all analyses and the Greenehouse-Geisser ms), error rates (n = 30), and pupillary responses (n= 30; in millimeter, mm) in the four
correction was applied when necessary. Significant main effects conditions.
were further analyzed by separate t-tests (for multiple comparisons Con = Emo= Con = Emo≠ Con ≠ Emo= Con ≠ Emo≠
Bonferroni correction was applied).
Mean response 3459.38 2937.97 3435.21 2829.78
times (ms)
(S.E.) (146.86) (107.37) (124.78) (100.68)
3. Results Mean error rates 0.08 0.09 0.19 0.02
(S.E.) (0.01) (0.02) (0.03) (0.01)
Pupillary responses
3.1. Post hoc ratings
Mean baseline 2.92 2.92 2.92 2.93
pupil diameter
As expected, word pairs with a neutral valence were rated as (mm)
neutral (M = 3.05, S.D. = 0.20) and low-arousing (M = 1.18, S.D. = (S.E.) (0.22) (0.22) (0.22) (0.22)
Mean peak 0.29 0.27 0.27 0.24
1.28) on the seven point rating scale from 0 (very unpleasant, low
dilations (mm)
arousal) to 6 (very pleasant, high arousal). Word pairs with a positive (S.E.) (0.02) (0.02) (0.02) (0.02)
valence were rated as more pleasant (M = 4.35, S.D. = 0.60) and more Mean latencies 2872.67 2731.33 2888.00 2799.33
arousing (M = 3.02, S.D. = 1.25); word pairs with a negative valence to peak (ms)
were rated as unpleasant (M = 0.85, S.D. = 0.51) and highly arousing (S.E.) (66.66) (65.56) (67.83) (77.61)

(M = 3.97, S.D. = 1.01). ms = milliseconds; mm = millimeter.

240 K. Prehn et al. / International Journal of Psychophysiology 79 (2011) 236–243

We additionally compared error rates and response times for items

with neutral, positive and negative emotional valence in the analogy
condition (Con = Emo=, see Table 3). Responses were slower for items
with negative valence compared to items with positive [t(29) = 3.297,
p = .003] and neutral emotional valence [t(29)= −2.931, p = .007].
Error rates were also higher for items with negative emotional valence
compared to items with positive emotional valence [t(29) = 2.817,
p = .009].

3.3. Pupil data

Pupil diameter increased during all conditions, reflecting increased

resource allocation during information processing (see Fig. 2).
Mean peak dilations were computed for each of the four experimental
conditions and averaged across participants (n=30). A 2 (correspon-
dence of conceptual relations)×2 (correspondence of emotional rela-
tions) repeated measure ANOVA revealed a significant main effect for
emotional correspondence [F(1, 29) = 11.40, MSE = 0.001, p = .002, Fig. 2. Stimulus-locked averaged pupillary responses in the four experimental conditions
η2=.28] as well as for conceptual correspondence [F(1,29)=13.47, (Con = Emo=: conceptual and emotional relations between the word pairs cor-
responding; Con = Emo≠: conceptual relations corresponding, emotional relations non-
MSE=0.001, p=.001, η2=.32]. There was no significant interaction corresponding; Con≠Emo=: conceptual relations non-corresponding, emotional relations
between the two factors [F(1, 29) = 0.47, MSE = 0.001, p = .499, corresponding; Con≠Emo≠: conceptual and emotional relations non-corresponding).
η2=.02]. The greatest increase in pupil dilation was measured when
both the conceptual and the emotional relations corresponded (analogy
items). Paired t-tests showed that there was no significant difference
in maximum pupil diameter between condition Con = Emo≠ and
condition Con≠Emo= [t(29)=.30, p=.77; i.e., between non-analogy
items in which one, either the conceptual or the emotional relation, did
not correspond]. However, there was a significant difference between
condition Con = Emo≠ and condition Con≠Emo≠ [t(29) = 3.02,
p=0.005], that is, pupil dilations were smaller when both relations did
not correspond. Response time data and peak dilations are contrasted in
Fig. 3.
Latencies to peak showed the same pattern as response time data. A
repeated measure ANOVA revealed a significant main effect for emotional
correspondence [F(1,29)=7.43, M.S.E.=53,412.07, p=.011, η2 =.20]
but no effect for conceptual correspondence [F(1,29)=1.21, M.S.E.=
43,117.82, p=.281, η2 =.04]. Because latencies to peak were significantly
correlated with response times in all four conditions (Con=Emo=:
r=.56, p=.001; Con=Emo≠: r=.73, pb .001; Con≠Emo=: r=.67,
pb .001; Con≠Emo≠: r=.63, pb .001), latencies reflect the moment of
maximal resource consumption and, thus, provide another index of
processing speed (cf. Kruger et al., 2001; Prehn et al., 2008).
Further, we compared peak dilations for items with neutral,
positive, and negative emotional valence in the analogy condition
(Con = Emo=, see Fig. 4 and Table 3). Pupil diameter increased very
similar to response times during processing of word pairs with
negative valence compared to positive [t(29) = −2.37, p = .025; only
trend because of adjusted α b 0.017 for 3 comparisons] and neutral
ones [t(29)=−2.92, p=.007]. There was no difference between neutral
and positive items [t(29)=1.01, p=.322].

Table 3
Means and standard errors (S.E.) for response times (n = 30 participants; in milliseconds,
ms), error rates (n = 30), and pupillary responses (n= 30; in millimeter, mm) for items
with neutral, positive, and negative valence in the analogy condition (Con= Emo=).

Neutral valence Positive valence Negative valence

Mean response times (ms) 3597.08 3594.40 3824.35

(S.E.) (233.17) (196.77) (197.93)
Mean error rates 0.08 0.07 0.10
(S.E.) (0.02) (0.01) (0.02)
Mean peak dilations (mm) 0.28 0.29 0.31
Fig. 3. a) Response times [mean and standard error (S.E.) in seconds, s] and b) peak
(S.E.) (0.02) (0.02) (0.02)
dilations [mean and standard error (S.E.) in millimeters, mm] in the four experimental
ms = milliseconds; mm = millimeter. conditions.
 K. Prehn et al. / International Journal of Psychophysiology 79 (2011) 236–243 241

4.1. Indicators of task performance

Response times and error rates showed a main effect of emotional

correspondence but no such effect in regard to conceptual corre-
spondence. That is, participants preferentially compared emotional
relations between the two word pairs of an item and responded faster
and more accurately when emotional relations did not correspond.
There was no difference in response times between the two
experimental conditions (Con = Emo ≠ vs. Con ≠ Emo≠), in which
emotional relations were not corresponding (regardless of whether
the conceptual relations corresponded or not). Moreover, the high
error rate in the non-analogy condition in which the emotional
relations corresponded while the conceptual did not (Con ≠ Emo=)
indicates that participants primarily took into account the correspon-
dence of emotional relations while associative conceptual relations
were frequently neglected (i.e., participants accepted these non-
analogy items mistakenly as analogous).
The results on processing speed and accuracy are in line with
Fig. 4. Stimulus-locked averaged pupillary responses in the analogy condition several studies providing evidence that emotional processing dom-
(Con = Emo=) for word pairs with neutral, positive, and negative evaluation.
inates conceptual processing (for reviews, see Compton et al., 2003;
Ochsner and Feldman Barrett, 2001; Pessoa and Ungerleider, 2003).
Examination of baseline pupil diameter as a possible confound on
pupillary responses 4.2. Indicators of resource allocation

To rule out an influence of baseline pupil diameter on pupillary Numerous studies have demonstrated that task-evoked pupillary
responses, baseline pupil diameter was also averaged for each responses reflect the processing load of a task (see Introduction; e.g.,
experimental condition and participant (see Table 2) and subjected Beatty and Lucero-Wagoner, 2000). Thus, pupillary responses “can be
to a 2 (correspondence of conceptual relations) × 2 (correspondence used to index the extent of central nervous system processing
of emotional relations) repeated measure ANOVA. There was neither a allocated to a task” (Granholm and Steinhauer, 2004, p. 2).
significant effect of conceptual [F(1, 29) = 2.297, M.S.E.= 0.001, p = .14, In general, pupil size increased in all conditions of the
η2 = .073] nor of emotional correspondence [F(1, 29) = 0.309, M.S.E. = analogical reasoning task. The increase, however, was greater
0.001, p = .582, η2 = .011] and thus no evidence for an influence of when more relations corresponded (Con = Emo= N Con = Emo≠ /
baseline pupil diameter on the task-evoked pupillary responses. Con ≠ Emo= N Con ≠ Emo≠) and the greatest increase in pupil
dilation was measured for analogy items (Con = Emo=), in which
4. Discussion both conceptual and emotional relations corresponded. The
increase of resource allocation during processing of corresponding
With the present study we aimed to investigate the interaction of relations might be due to the fact that an item could be categorized
cognition and emotion in decision making with special focus on the as a non-analogy as soon as at least one type of relations (conceptual
question whether one aspect (cognition or emotion) dominates the or emotional) could be identified as non-corresponding (i.e., proces-
other in terms of task performance (processing speed, accuracy) and sing could stop there and the second type of relations could be
mental resources allocated to the task. For this purpose we developed ignored). In a similar vein, if processing is self-terminating (and thus,
an analogical reasoning task, in which participants had to identify and highly effective) then the analogy condition (Con = Emo=) should be
compare associative conceptual and emotional relations. A remark- more difficult than the three non-analogy conditions (Con ≠ Emo=,
able feature of this task is that emotional and conceptual relations Con = Emo≠, and Con ≠ Emo≠) because most resource consuming
(i.e., cognitive and affective aspects of a situation) were equally task high-level cognitive comparison processes are needed in this condition
relevant. Moreover, this task requires on the one hand cognitively for making a decision.
lower-level information processing (here, identification, i.e., retrieval In addition to this effect, peak dilations, in contrast to response
of associative conceptual and emotional relations from long-term time data, showed a significant main effect for both conceptual and
memory) and on the other hand cognitively higher-level comparison emotional correspondence. Thus, our data show that emotional
processes. processing (here, the comparison of emotional relations) and
The experiment yielded three key results: First, response times conceptual processing (here, the comparison of conceptual relations)
and error rates showed that emotional relations were processed both consume a significant amount of mental resources.
preferentially and faster than conceptual relations. Second, pupil Latencies to peak covaried with response times and thus provide
dilations reflected the complexity of the comparison process, showing another index of processing speed reflecting the moment of maximal
that comparing both emotional and conceptual relations consumed a resource allocation (cf. Kruger et al., 2001; Prehn et al., 2008). In contrast
significant amount of mental resources. Third, in the analogy- to peak dilations, response times and latencies to peak were shorter
condition (Con = Emo=), items with a negative emotional valence when emotional relations did not correspond (Con = Emo ≠ and
were processed more slowly, with an increased error rate, and greater Con ≠ Emo≠) indicating that resource allocation (due to the compar-
pupil dilation than items with a positive or neutral emotional valence. ison process) in these conditions might be earlier. Differences in
In the following, we argue that indicators of task performance processing speed (response times and latencies to peak), in our study,
(response time and error rate) and indicators of resource allocation seem to reflect the cognitively lower-level relation retrieval processes
(pupil dilation) together provide evidence for the view that the that are assumed to be based on automatic activation, which is less
affective significance of stimuli leads to a processing advantage during resource consuming and faster for emotional relations, whereas phasic
the identification of emotional relations (lower-level information pupil dilations rather reflect the cognitive comparison processes that
processing) but can also be distracting and provokes interference on require a significant amount of mental resources for both emotional and
higher levels of information processing. associative conceptual relations.
242 K. Prehn et al. / International Journal of Psychophysiology 79 (2011) 236–243

In this study, a possible impact of confounding variables such as enhanced and sometimes even disturbed by affective significance, for
the baseline pupil diameter as well as typicality, personal familiarity, instance, when the decision cannot be made based on the correspon-
or imagery of the used word material could be excluded (even though dence of emotional relations.
there was a small but statistically insignificant difference in imagery). Taken together, our study on the impact of “affective significance”
Additionally, we controlled the amount of display light, using dark on different levels of information processing emphasizes the necessity
letters on gray background to eliminate any light reflex at display and the advantage of using both behavioral and psychophysiological
onset. However, screen luminance decreased minimally during measures to differentiate between cognitive and affective aspects
stimulus display (black letters replacing some gray background where when investigating complex decision making processes.
subjects were fixating), which might have led to a minimal increase in
dilation (after the participants' response, the words disappeared, Acknowledgements
returning the screen to a minimal brighter gray). This could lead to
greater dilation for conditions or subjects with longer response times, This study was financially supported by grants from the Gruter
because longer response times result in less light to the pupil. This Institute of Law and Behavioral Research, the Graduate Program Berlin
confound might explain the greater dilation found in the condition with (Scholarship Nachwuchsfoerderung), the BMBF (Berlin NeuroImaging
longer responses (Con = Emo=). However, this is improbable to Center, BNIC), and the DFG [Emmy-Noether-Program (HE 3347/2-1)].
explain the pattern of results found in this study for two reasons: We thank Susanne Raisig, Torsten Schiemann and Roman Purkhart for
First, responses generally occurred after peak dilation (see Table 2 and assistance during the acquisition and analysis of pupil data and an
Fig. 2). Thus display luminance was constant during processing that anonymous reviewer for helpful comments.
led to peak dilation. Second, response times were longer in the
Con ≠ Emo = condition than in other conditions yet pupil dilation References
was not greater. We raise this issue, because we think it is important to
bring this potential confound to the attention of readers who may want Aston-Jones, G., Cohen, J.D., 2005. An integrative theory of locus coeruleus–norepinephrine
function: adaptive gain and optimal performance. Annu. Rev. Neurosci. 28, 403–450.
to replicate or extend our work using pupillometry (see Prehn et al., Baayen, R.H., Piepenbrock, R., Rijn, H., 1993. The CELEX Lexical Database (Release 1)
2008; here, as soon as a response button was pressed by the participant, [CD-ROM], Linguistic Data Consortium. University of Pennsylvania, Philadelphia,
the item was replaced by a mask). PA.
Beatty, J., 1982. Task-evoked pupillary responses, processing load, and the structure of
processing resources. Psychol. Bull. 91, 276–292.
4.3. The impact of emotional valence on the controlled comparison process Beatty, J., Lucero-Wagoner, B., 2000. The pupillary system, In: Cacioppo, J.T., Tassinary,
L.G., Berntson, G.G. (Eds.), Handbook of psychophysiology, 2nd ed. Cambridge
University Press, New York, pp. 142–162.
To additionally examine the effect of emotional valence of the items
Bunge, S.A., Wendelken, C., Badre, D., Wagner, A.D., 2005. Analogical reasoning and
in the analogical reasoning task, we compared response times and pupil prefrontal cortex: evidence for separable retrieval and integration mechanisms.
dilations for items with neutral, positive, and negative emotional Cereb. Cortex 15, 239–249.
valence in the analogy condition (Con = Emo=). We found increased Collins, A.M., Quillian, M.R., 1969. Retrieval time from semantic memory. J. Verb. Learn.
Verb. Behav. 8, 240–247.
response times and error rates as well as increased pupil dilations in Compton, R.J., 2003. The interface between emotion and attention: a review of evidence
items with negative emotional valence compared to items with neutral from psychology and neuroscience. Behav. Cogn. Neurosci. Rev. 2, 115–129.
and positive emotional valence. Compton, R.J., Banich, M.T., Mohanty, A., Milham, M.P., Herrington, J., Miller, G.A., Scalf,
P.E., Webb, A., Heller, W., 2003. Paying attention to emotion: an fMRI investigation
As mentioned above, the analogy condition was the most difficult of cognitive and emotional stroop tasks. Cogn. Affect. Behav. Neurosci. 3, 81–96.
experimental condition and always required two comparison processes Desimone, R., Duncan, J., 1995. Neural Mechanisms of Selective Visual-Attention. Annu.
(checking the correspondence of emotional relations and of associative Rev. Neurosci. 18, 193–222.
Dresler, T., Mériau, K., Heekeren, H.R., van der Meer, E., 2009. Emotional Stroop task: effect
conceptual relations). In this condition, affective significance was of word arousal and subject anxiety on emotional interference. Psychol. Res. 73,
not as advantageous as in other conditions (e.g., Con = Emo≠ and 364–371.
Con ≠Emo≠, in which a non-analogy decision could be made as soon as Granholm, E., Steinhauer, S.R., 2004. Pupillometric measures of cognitive and emotional
processes. Int. J. Psychophysiol. 52, 1–6.
a non-correspondence in emotional relations was detected) or even
Granholm, E., Asarnow, R.F., Sarkin, A.J., Dykes, K.L., 1996. Pupillary responses index
distracting and detrimental for the controlled comparison process cognitive resource limitations. Psychophysiology 33, 457–461.
(e.g., Hartikainen et al., 2000; McKenna, 1986; Mitchell et al., 2008). Hartikainen, K.M., Ogawa, K.H., Knight, R.T., 2000. Transient interference of right
hemispheric function due to automatic emotional processing. Neuropsychologia
Moreover – and also in line with the literature –, the distracting
38, 1576–1580.
influence was greater in items with negative emotional valence Herrmann, D.J., Chaffin, R., 1986. Comprehension of semantic relations as a function of the
compared to items with positive emotional valence. A possible reason definitions of relations. In: Klix, F., Hagendorf, H. (Eds.), Human memory and cognitive
is the greater emotional significance as measured by rated arousal capabilities. Mechanisms and Performances Elsevier Science Publishers B.V, North-
Holland, pp. 311–319.
elicited by items with negative emotional valence (see Dresler et al., Hoeks, B., Ellenbroek, B.A., 1993. A neural basis for a quantitative pupillary model.
2009; Mitchell et al., 2008). Psychophysiology 7, 315–324.
Hyona, J., Tommola, J., Alaja, A.M., 1995. Pupil-dilation as a measure of processing load
in simultaneous interpretation and other language tasks. Q. J. Exp. Psychol. A 48,
5. Conclusion 598–612.
Just, M.A., Carpenter, P.A., 1993. The intensity dimension of thought - pupillometric
In conclusion, the results of the present study have implications for indexes of sentence processing. Can. J. Exp. Psychol. 47, 310–339.
Just, M.A., Carpenter, P.A., Miyake, A., 2003. Neuroindices of cognitive workload:
(neurobiological) models of the interaction of emotion and cognition. neuroimaging, pupillometric and event-related potential studies of brain work.
They are, for instance, consistent with the “dual competition” Theory Issues Ergonomie Sci. 4, 56–88.
framework (Pessoa, 2009). According to this model, the competition Kahneman, D., Beatty, J., 1966. Pupil diameter and load on memory. Science 154,
1583–1585.
for limited processing resources and behavioral control between Kan, I.P., Thompson-Schill, S.L., 2004. Selection from perceptual and conceptual
different situational aspects is differentially influenced by affective representations. Cogn. Affect. Behav. Neurosci. 4, 466–482.
significance at both lower-level (here, automatic identification or Karatekin, C., Couperus, J.W., Marcus, D.J., 2004. Attention allocation in the dual-task
paradigm as measured through behavioral and psychophysiological responses.
retrieval from long-term memory) and higher-level information
Psychophysiology 41, 175–185.
processing (here, controlled comparison of retrieved relations). The Kim, M., Beversdorf, D.Q., Heilman, K.M., 2000. Arousal response with aging:
study clearly reveals that at the cognitively lower level of processing pupillographic study. J. Int. Neuropsychol. Soc. 6, 348–350.
(i.e., relation retrieval) emotional relations are more salient. There- Klix, F., 1992. Higher-order learning-mechanisms in knowledge domain. Z. Psychol.
200, 91–103.
fore, they are processed preferentially and faster. The cognitively Krawczyk, D.C., 2002. Contributions of the prefrontal cortex to the neural basis of
higher level of processing (i.e., relation comparison), however, is not human decision making. Neurosci. Biobehav. Rev. 26, 631–664.
 K. Prehn et al. / International Journal of Psychophysiology 79 (2011) 236–243 243

Krawczyk, D.C., Morrison, R.G., Viskontas, I., Holyoak, K.J., Chow, T.W., Mendez, M.F., Raisig, S., Welke, T., Hagendorf, H., van der Meer, E., 2007. Investigating dimensional
Miller, B.L., Knowlton, B.J., 2008. Distraction during relational reasoning: the role of organization in scripts using the pupillary response. Psychophysiology 44, 864–873.
prefrontal cortex in interference control. Neuropsychologia 46, 2020–2032. Sachs, O., Weis, S., Krings, T., Huber, W., Kircher, T., 2008a. Categorical and thematic
Kruger, F., Nuthmann, A., van der Meer, E., 2001. Pupillometric indices of temporal knowledge representation in the brain: neural correlates of taxonomic and
order representation in semantic memory. Z. Psychol. 209, 402–415. thematic conceptual relations. Neuropsychologia 46, 409–418.
Lin, E.L., Murphy, G.L., 2001. Thematic relations in adults' concepts. J. Exp. Psychol. Gen. Sachs, O., Weis, S., Zellagui, N., Huber, W., Zvyagintsev, M., Mathiak, K., Kircher, T.,
130, 3–28. 2008b. Automatic processing of semantic relations in fMRI: neural activation
Loewenfeld, I.E., 1993. The pupil. Iowa State University Press, Ames. during semantic priming of taxonomic and thematic categories. Brain Res. 1218,
Martin, A., Chao, L.L., 2001. Semantic memory and the brain: structure and processes. 194–205.
Curr. Opin. Neurobiol. 11, 194–201. Siegle, G.J., Granholm, E., Ingram, R.E., Matt, G.E., 2001. Pupillary and reaction time
McKenna, F.P., 1986. Effects of unattended emotional stimuli on color-naming measures of sustained processing of negative information in depression. Biol.
performance. Curr. Psychol. Res. Rev. 5, 3–9. Psychiatry 49, 624–636.
Mitchell, D.G.V., Luo, Q., Mondillo, K., Vythilingam, M., Finger, E.C., Blair, R.J.R., 2008. The Steinhauer, S.R., Hakerem, G., 1992. The Pupillary Response in Cognitive Psychophys-
interference of operant task performance by emotional distracters: an antagonistic iology and Schizophrenia. Ann. N.Y. Acad. Sci. 658, 182–204.
relationship between the amygdala and frontoparietal cortices. Neuroimage 40, Steinhauer, S.R., Boller, F., Zubin, J., Pearlman, S., 1983. Pupillary dilation to emotional
859–868. visual-stimuli revisited. Psychophysiology 20, 472.
Nuthmann, A., van der Meer, E., 2005. Time's arrow and pupillary response. Psychophys- van der Meer, E., 1989. Impacts of emotions on conceptual structures. In: Bennett, A.F.,
iology 42, 306–317. McConkey, K.M. (Eds.), Cognition in Individual and Social Contexts. Proceedings of
Ochsner, K.N., Feldman Barrett, L., 2001. A multiprocess perspective on the the XXIV International Congress of Psychology Elsevier, Amsterdam, pp. 349–356.
neuroscience of emotion. In: Mayne, T., Bonnano, G. (Eds.), Emotion: current van der Meer, E., Friedrich, M., Nuthmann, A., Stelzel, C., Kuchinke, L., 2003. Picture-
issues and future directions. The Guilford Press, New York, pp. 38–81. word matching: flexibility in conceptual memory and pupillary responses.
Pessoa, L., 2009. How do emotion and motivation direct executive control. Trends Cogn. Psychophysiology 40, 904–913.
Sci. 13, 160–166. van der Meer, E., Beyer, R., Horn, J., Foth, M., Bornemann, B., Ries, J., Kramer, J.,
Pessoa, L., Ungerleider, L.G., 2003. Neuroimaging studies of attention and the processing Warmuth, E., Heekeren, H.R., Wartenburger, I., 2010. Resource allocation and fluid
of emotion-laden stimuli. Prog. Brain Res. 144, 171–182. intelligence: insights from pupillometry. Psychophysiology 47, 158–169.
Phelps, E.A., 2006. Emotion and cognition: Insights from studies of the human Verney, S.P., Granholm, E., Dionisio, D.P., 2001. Pupillary responses and processing
amygdala. Annu. Rev. Psychol. 57, 27–53. resources on the visual backward masking task. Psychophysiology 38, 76–83.
Prehn, K., Heekeren, H.R., Blasek, K., Lapschies, K., Mews, I., van der Meer, E., 2008. Verney, S.P., Granholm, E., Marshall, S.P., 2004. Pupillary responses on the visual
Neuroticism influences pupillary responses during an emotional interference task. backward masking task reflect general cognitive ability. Int. J. Psychophysiol. 52,
Int. J. Psychophysiol. 70, 40–49. 23–36.