Neuroscience Letters 442 (2008) 186–189
Contents lists available at ScienceDirect
Neuroscience Letters
journal homepage: www.elsevier.com/locate/neulet
Error-related negativity varies with the activation of gender stereotypes
Qingguo Ma a,b,∗ , Liangchao Shu a,b , Xiaoyi Wang a,b , Shenyi Dai a,b , Hongmin Che a,b
a
b
School of Management, Zhejiang University, PR China
Neuromanagement Lab., Zhejiang University, PR China
a r t i c l e
i n f o
Article history:
Received 24 February 2008
Received in revised form 23 June 2008
Accepted 30 June 2008
Keywords:
Gender stereotypes
Event-related potential (ERP)
Error-related negativity (ERN)
Conflict
a b s t r a c t
The error-related negativity (ERN) was suggested to reflect the response-performance monitoring process.
The purpose of this study is to investigate how the activation of gender stereotypes influences the ERN.
Twenty-eight male participants were asked to complete a tool or kitchenware identification task. The
prime stimulus is a picture of a male or female face and the target stimulus is either a kitchen utensil or
a hand tool. The ERN amplitude on male-kitchenware trials is significantly larger than that on femalekitchenware trials, which reveals the low-level, automatic activation of gender stereotypes. The ERN that
was elicited in this task has two sources—operation errors and the conflict between the gender stereotype
activation and the non-prejudice beliefs. And the gender stereotype activation may be the key factor
leading to this difference of ERN. In other words, the stereotype activation in this experimental paradigm
may be indexed by the ERN.
© 2008 Elsevier Ireland Ltd. All rights reserved.
A gender stereotype can be described as culturally specific, publicly
shared characteristics that are associated with one particular gender [15]. In Chinese traditional culture, the stereotypes of gender
are rather pervasive: the role of female is always associated with
housekeeping, while that of male is more associated with social
production. Although individuals may attempt to prevent gender
stereotypes from affecting their behavior by engaging in controlled
processes, the intentions to control gender bias are not always successful. Even for the self-avowed egalitarians, the prejudices of
gender have often slipped through in their behavior despite their
non-prejudiced intentions. Previous studies revealed that stereotyping is an automatic, implicit, cognitive process, and it can be
automatically activated by gender features [22]. The stimulus of
gender features can automatically start an activated gender stereotype [3]. When the activated stereotypic information conflicts with
one’s non-prejudiced beliefs, the following neural processes will
result.
In cognitive neuroscience, there are two separate neural systems that work in concert to arrive at an intended behavior in
the face of conflicting behavioral tendencies. The first one is a
conflict-detection system, which monitors ongoing responses and
is sensitive to competition between different response tendencies
[4,19]. This system is constantly active, requiring few cognitive
∗ Corresponding author at: School of Management, Zhejiang University, 38 Zheda
Road, Hangzhou 310058, PR China. Tel.: +86 571 87991338; fax: +86 571 87995372.
E-mail addresses: [email protected],
[email protected] (Q. Ma).
0304-3940/$ – see front matter © 2008 Elsevier Ireland Ltd. All rights reserved.
doi:10.1016/j.neulet.2008.06.080
resources, and has been shown to operate below conscious awareness. When a conflict is detected by this system, it alerts the second,
resource-dependent regulatory system designed to implement the
intended response while inhibiting the unintended response [9].
The error-related negativity (ERN) wave, a component of the
event-related potential (ERP), reflects the response-performance
monitoring process [23]. The onset of the ERN coincides with
response initiation, as determined by the onset of the electromyogram (EMG) associated with the responding hand, and peaks
roughly 80 ms thereafter. Its spatial distribution lies over the
frontal–central regions of the scalp [14], and is symmetrical to the
midline [23]. The anterior cingulate cortex, which is specifically
involved in the brain’s error processing system, seems to be the
generator site of the ERN [13].
In past studies, it was demonstrated that the ERN can be elicited
by conflicts and errors. Response conflict monitoring theory of
ERN suggests that the amplitude of ERN reflects the degree of
response conflict. In higher-conflict incongruent trials, larger ERN
amplitudes are produced [23]. From reinforcement learning theory, the ERN is sensitive to the degree of error, as larger errors
correspond with larger negativity [14,18]. The ERN is elicited whenever it detects a mismatch between the produced response and
the correct, or the intended, response [8], such as the wrong hand,
the wrong finger, or both the wrong hand and the wrong finger
[5]. Recently, some studies correlate the ERN with some modulating factors. Confronted with the same error, the ERN peaks both
earlier and larger in the context of pleasant backgrounds than
neutral or unpleasant backgrounds [16]. The impulsivity of participant is correlated with the ERN amplitude, which suggests that
Q. Ma et al. / Neuroscience Letters 442 (2008) 186–189
the high-impulsive participants have smaller ERN amplitudes on
punishment trials than the low-impulsive participants [12]. Social
culture, such as racial stereotypes, can also induce lager ERNs [9].
Few studies however, have focused on how the gender stereotypes
affect the ERN.
In this study, the participants were asked to identify the target stimulus which was a tool or a piece of kitchenware, and the
prime stimulus was a male or female face. We speculated that the
prime stimulus would activate the gender stereotypes which would
influence the judgment of the target stimulus. The ERN would be
recorded on these trials that the participants’ identification of the
target stimulus was wrong, and larger ERNs would be observed on
the trials of gender stereotype-incongruent.
Twenty-eight right-handed male undergraduates aged between
20 and 35 years (mean = 27.5) were employed in this study. All
had normal or corrected-to-normal visual acuity. They did not
have any history of neurological or mental diseases. All the participants claimed holding predominantly non-prejudiced attitudes
of gender.1
The stimuli consisted of 144 prime-target pairs. The primes in
these pairs included 12 digital photographs of 6 Asian male and 6
Asian female faces, and the targets included 12 digital photographs
of 6 kitchen utensils (fork, knife, spatula, rolling pin, ladle, chopsticks) and 6 hand tools (drill, ratchet, wrench, pliers, brush, shovel).
These photographs were digitized at 228 × 172 pixels.
Stimuli were presented sequentially in the center of a computer screen, with a visual angle of 2.58◦ × 2.4◦ . Each trial began
with a pattern mask (1 s), followed by the prime (200 ms), and
then the target (200 ms). The next trial starts after the participant responded or after 1500 ms had elapsed since the onset of the
target. The prime-target pairs were divided into four conditions:
male-kitchenware, female-kitchenware, male-tool, and femaletool. The stimulus pairs of male-tool and female-kitchenware are
called stereotype-congruent pairs and those of male-kitchenware
and female-tool are called stereotype-incongruent pairs. The stimulus pairs were randomly presented in sequence and had the same
probability. Stimuli and recording triggers were presented using
STIM 2 software (Neurosoft Labs, Inc., Sterling, USA).
After participants provided informed consent, they were fitted
with an electrode cap for electroencephalographic (EEG) recording.
Then, the participants were instructed to classify the target picture
as a piece of kitchenware or a tool as quickly as possible, and were
told that an erroneous “kitchenware” response on a female-tool
trial or an erroneous “tool” response on a male-kitchenware trial
was indicative of gender bias.
Electroencephalogram was continuously recorded (band pass
0.05–100 Hz, sampling rate 500 Hz) with Neuroscan Synamp2
Amplifier (Scan 4.3.1, Neurosoft Labs, Inc., Sterling, USA), using
an electrode cap with 64 Ag/AgCl electrodes mounted according to the extended international 10–20 system and referenced to
linked mastoids. Vertical and horizontal electrooculograms were
recorded with two pairs of electrodes, one placed above and
below the right eye, and the other 10 mm from the lateral canthi.
Electrode impedance was maintained below 5 k throughout the
experiment. Following electrode application, participants sat in a
comfortable sofa located in a shielded room and were asked to fix a
point in the center of the computer display. Next, the experimenter
gave participants instructions to classify the second picture as a
kitchenware or a tool by pressing the corresponding key with their
right or left index finger. Responses were to be made within 500 ms
of the target presentation. If the responses exceeded this time limit,
1
All these participants were asked to fulfill a questionnaire that was created by
authors following the idea of Burt’s sex role stereotyping scales [7].
187
Table 1
Means (S.D.s) of RTs on correct responses and error rates in different conditions
Experimental condition
RT for correct response (ms)
Response error rate
Male-T
Male-K
Female-T
Female-K
344.88 (72.93)
363.03 (77.66)
379.72 (72.22)
345.92 (76.01)
0.3084 (0.1365)
0.4361 (0.1688)
0.4353 (0.1593)
0.3373 (0.1590)
Male-K, male-kitchenware trials; female-K, female-kitchenware trials; male-T,
male-tool trials; female-T, female-tool trials.
the participants would receive a red warning symbol “×” to make
them respond more quickly. After 30 practice trials, 576 stimulus
trials were presented.
An 800 ms epoch of EEG signal, centered on key press, was
selected for each artifact-free trial with the first 400 ms of the epoch
as a baseline. Electrooculogram artifacts were corrected using the
method proposed by Semlitsch et al. [21]. Trails contaminated by
amplifier clipping, bursts of electromyographic activity, or peak-topeak deflection exceeding ±80 ␮V were excluded from averaging.
The remaining trails were baseline corrected. ERPs for correct and
incorrect trails were averaged within their respective trial types
and the averaged ERPs were digitally filtered with a low pass filter
at 30 Hz (24 dB/Octave). The ERNs were scored, in this study, when
the peak negative deflection occurred between 50 ms before key
press and 150 ms after key press at the frontocentral site.
In each analysis, only participants with valid response on all
measures were adopted. To ensure reliable and stable ERP component with relative higher signal-to-noise ratio extracting from
digitized EEG signal, it was necessary to exclude the participants
with insufficient sample of valid trials. Fabiani et al. [10] suggested
that more than 20 valid trials for a condition are needed to obtain
the stable ERP. In this study, we excluded the participants with
fewer than 20 artifact-free error trials in any of the four conditions
(7 participants). Thus, the primary analyses were conducted on the
data from the rest (21 participants).
The data of means (Ms) and corresponding standard deviations (S.D.s) of reaction times (RTs) on correct responses and
of response error rates in different conditions are given in
Table 1. To examine whether the reaction times on stereotypecongruent trials are faster than on stereotype-incongruent trials,
we conducted a 2 (gender: male vs. female) × 2 (object: tool vs.
kitchenware) within-subjects analysis of variance (ANOVA) for RTs
on correct responses. This analysis produced a main effect for
gender of face, F(1,20) = 9.135, p = 0.007; responses to male faces
(M = 353.956, S.D. = 74.974) were faster than responses to female
faces (M = 362.819, S.D. = 75.198). However, there was not a main
effect for objects, F(1,20) = 2.778, p = 0.111. The interaction between
gender and object was significant, F(1,20) = 21.252, p = 0.000.
Simple effect analyses showed that participants were quicker
to identify kitchenware following female faces (M = 345.920,
S.D. = 76.008) than following male faces (M = 363.033, S.D. = 77.659),
F(1,20) = 10.766, p = 0.004, and were slower to identify tools following female faces (M = 379.718, S.D. = 72.217) compared with male
faces (M = 344.879, S.D. = 72.934), F(1,20) = 22.688, p = 0.000. These
results demonstrated an automatic association between the male
face and tools and between the female face and kitchenware. Additionally, the slower responses to the pairings of male-kitchenware
and female-tool suggested that participants adopted a more controlled response strategy on these trials in an effort to avoid gender
bias.
In addition, to examine the effect of stereotypic associations on
the response error rates, we conducted a 2 (gender) × 2 (object)
within-subjects ANOVA for error rates. There was no main effect
for gender, F(1,20) = 1.433, p = 0.245, nor object, F(1,20) = 0.234,
188
Q. Ma et al. / Neuroscience Letters 442 (2008) 186–189
Fig. 1. Response-locked event-related potential waveforms from site FZ for correct and erroneous kitchenware (A) and tool (B) trials as a function of gender of face. Zero
indicates the time of the key press. Errors on male-kitchenware or female-tool trials represent gender-biased responses.
p = 0.634. The interaction between gender and object was significant, F(1,20) = 13.313, p = 0.002. Simple effect analyses expressed
that participants made a higher error rate on tool trials when the
target was preceded by a female face (M = 0.435, S.D. = 0.159) compared with a male face (M = 0.308, S.D. = 0.136), F(1,20) = 14.379,
p = 0.001, and on kitchenware trials they made higher error
rates when the target was preceded by a male face (M = 0.436,
S.D. = 0.169) compared with a female face (M = 0.337, S.D. = 0.159),
F(1,20) = 9.120, p = 0.007. These results indicated a prepotent tendency to favor a “kitchenware” response after viewing a female face
and a “tool” response after viewing a male face.
On the basis of previous findings that the ERN is relatively focal
over frontocentral locations [11] and the scalp distribution maps of
the present data, the electrode site Fz was chosen for analysis. The
data of ERN amplitudes at the scalp site of Fz in different conditions
are given in Fig. 1.
We conducted 2 (gender: male vs. female) × 2 (object: tool vs.
kitchenware) × 2 (response: correct vs. erroneous) within-subject
repeated measure ANOVA for the ERN peak amplitude to examine the effect of the ERN evoked by the activation of gender
stereotype. The ANOVA produced a marginally significant main
effect for gender of face, F(1,20) = 3.655, p = 0.070, the mean difference of amplitudes between cues of male and female faces was
−0.581 ␮V. A main effect emerged for response, F(1,20) = 16.735,
p = 0.001; the mean difference of amplitudes between erroneous
and correct responses was −3.112 ␮V. There was no significant main
effect for object, F(1,20) = 1.030, p = 0.322. These effects were qualified by a marginally significant interaction of gender × response,
F(1,20) = 4.217, p = 0.053, and a significant interaction of gender × object × response, F(1,20) = 5.374, p = 0.031, whereas both
the interaction of gender × object and of object × response were
not significant, F(1,20) = 0.006, p = 0.937, F(1,20) = 1.307, p = 0.266,
respectively.
To further understand the significant interaction of gender × response, a simple effect analysis was conducted, showing
that the ERN amplitudes on male error trials (M = −1.601,
S.D. = 3.417) were significantly larger (negative-polarity) than that
on female error trials (M = −0.508, S.D. = 4.156), t(41) = −2.555,
p = 0.014, while the ERN amplitudes on male correct trials were
not significantly different from that on female correct trials,
t(41) = −0.193, p = 0.848.
To further understand the significant interaction of gender × object × response, we conducted the analyses of pairwise
comparisons, and obtained the following results. It was not significant that the mean difference of the ERN amplitudes of the
correct responses on tool trails when the target was preceded by
a male face compared with a female face, t(20) = −1.659, p = 0.113,
and on kitchenware trials when the target was preceded by a male
face, compared with a female face, t(20) = 0.982, p = 0.338. The ERN
amplitudes were significantly larger when kitchen utensils were
erroneously classified as tools following male faces (M = −2.226,
S.D. = 3.163) than when the same errors were made following
female faces (M = −0.467, S.D. = 3.677), t(20) = −3.475, p = 0.002,
whereas the ERNs associated with erroneously classifying tools
as kitchen utensils following male and female faces did not differ statistically, t(20) = −0.637, p = 0.532. The significantly larger
ERN observed for errors on male-kitchenware trials suggests that
greater conflict is detected when a response tendency is gender
biased than when it is not biased.
This study demonstrates that ERNs are obviously elicited
in erroneous responses to four types of prime-target conditions: male-tool, female-tool, male-kitchenware, and femalekitchenware. Statistical test shows that the mean amplitude of
ERNs of male-kitchenware errors is significantly larger than that of
female-kitchenware errors, whereas the mean amplitude of ERNs
of male-tool errors does not differ significantly from that of femaletool errors.
The ERN has been linked with response conflict, reinforcement
learning or response monitoring (see [14], for a review). As the ERN
reflects a response-conflict monitoring process, it can be elicited by
negative feedback and by error commission [14], and modulated by
modulating factors, such as affectiveness or impulsiveness [12,16].
The amplitude of ERN reflects the degree of response conflict or
response error [14,18,23]. In this study, the ERNs that were elicited
in the task of forced-alternative identification response have two
sources—“pure” operation errors2 and the conflict between the
gender stereotype activation and the non-prejudice beliefs. The
numbers of “pure” hand manipulation errors on male-kitchenware
and female-kitchenware trials are random variables which should
have the same mean for any participant. Therefore, the finding that
the mean amplitude of ERNs on erroneous response trials of malekitchenware is significantly larger than that of female-kitchenware
is mainly a result of the activated stereotypes of gender.
Previous studies demonstrated that stereotypes and unconscious prejudices could be automatically activated when one
identifies another person’s skin color, facial features, gender characteristics, and the like [17]. Borgida et al. suggested that gender
is a fundamental dimension of categorization. Once an individual is categorized as belonging to one gender, the stereotypes of
that gender may quickly come to the perceiver’s mind [6]. From
the view of dual system theory, there may be two explanations for
2
A “pure” operation error by hand means the participant wants to press the correct key but pressed the incorrect key by using a wrong finger. This error is different
from the error that participant pressed a key he wants to press.
Q. Ma et al. / Neuroscience Letters 442 (2008) 186–189
stereotypic affect. One is that the first system does not appraise
stereotype-biased response, and therefore it does not signal the
second system. The alternative explanation is that the second system does not implement the control after it received the signal
of need of control [9]. Some race bias studies suggested that controlled, belief-based processes are more effectively implemented
in deliberate responses [1]. While deliberate responses, like selfreports, do not invariably reflect attitudes and beliefs, particularly
when the expression of the belief is perceived to be socially inappropriate (see [20], for a review). In this study, we infer that the
second system has no time to implement the control, the unconscious, erroneous response had submitted already when the signal
of need of control being delivered from the first system to the second.
The unintended stereotypic activation may be the source of the
difference of ERN amplitude. In this identification task, the participants’ gender stereotype is activated, and the activated stereotypes
have a strong conflict with the participants’ intended response (the
gender-unbiased response). The results supported our hypothesis.
Some researches suggest that the ERN represents a conflict monitoring function in regulating race-biased behavior [2,9]. Therefore,
gender stereotypes in the task may be associated with responseperformance monitoring. It may imply that the gender stereotypes,
as a subjective factor, may modulate the ERN. The ERPs are sensitive to violations of gender-based occupational stereotypes even
when subjects judged a stereotype violation to be acceptable [20].
The present findings prove that the ERPs might be advantageous
tools for studying automatic activation of gender stereotypes. The
present study focuses on the gender-role stereotype of males. It will
be extended to females and do some comparisons between them
in the future.
Taken together, the ERN that is elicited in this experiment may
come from an error commission and modulated strongly by a gender stereotypic factor, which suggest that the activation of gender
stereotype may be indexed by the ERN.
Conflict of interest
None.
Acknowledgements
We would like to appreciate the advice of two anonymous
reviewers. This work was supported by grant No. 08BJL054 from
the National Social Science Foundation of China.
Appendix A. Supplementary data
Supplementary data associated with this article can be found,
in the online version, at doi:10.1016/j.neulet.2008.06.080.
189
References
[1] D.M. Amodio, E. Harmon-Jones, P.G. Devine, Individual differences in the activation and control of affective race bias as assessed by startle eye blink response
and self-report, J. Pers. Soc. Psychol. 84 (2003) 738–753.
[2] D.M. Amodio, E. Harmon-Jones, P.G. Devine, Individual differences in the regulating of intergroup bias: the role of conflict monitoring and neural signals for
control, J. Pers. Soc. Psychol. 94 (2008) 60–74.
[3] J.A. Bargh, E.L. Williams, The automaticity of social life, Curr. Dir. Psychol. Sci.
15 (2006) 1–4.
[4] G.S. Berns, J.D. Cohen, M.A. Mintun, Brain regions responsive to novelty in the
absence of awareness, Science 276 (1997) 1272–1275.
[5] P. Bernstein, M.K. Scheffers, M.G.H. Coles, “Where did I go wrong?”: a psychophysiological analysis of error detection, J. Exp. Psychol. Hum. Percept. 6
(1995) 1312–1322.
[6] E. Borgida, C. Hunt, A. Kim, On the use of gender stereotyping research in sex
discrimination litigation, J. Law Policy 13 (2005) 613–628.
[7] M.R. Burt, Cultural myths and supports for rape, J. Pers. Soc. Psychol. 38 (1980)
217–230.
[8] M.G.H. Coles, M.K. Scheffers, C.B. Holroyd, Why is there an ERN/Ne on correct
trials? Response representations, stimulus-related components, and the theory
of error-processing, Biol. Psychol. 56 (2001) 173–189.
[9] M.A. David, H.J. Eddie, G.D. Patricia, J.C. John, L.H. Sigan, E.C. Alison, Neural
signals for the detection of unintentional race bias, Psychol. Sci. 15 (2004)
88–93.
[10] M. Fabiani, G. Gratton, M.G.H. Coles, Event-related brain potentials methods,
theory, and applications, in: J.T. Cacioppo, L.G. Tassinary, G.G. Berntson (Eds.),
Handbook of Psychophysiology, Cambridge University Press, New York, 2000,
p. 54.
[11] W.J. Gehring, B. Goss, M.G. Coles, D.E. Meyer, E. Donchin, A neural system for
error detection and compensation, Psychol. Sci. 4 (1993) 385–390.
[12] F.P. Geoffrey, R.M.G. Mary, E.M. Laura, S.B. Ernest, Reduced punishment sensitivity in neural systems of behavior monitoring in impulsive individuals, Neurosci.
Lett. 397 (2006) 130–134.
[13] E.S.M. Gillian, D. Jane, T.H. Grant, M. Shari, U. Ayse, J.S. Sidney, ERN varies with
degree of psychopathy in an emotion discrimination task, Biol. Psychol. 76
(2007) 31–42.
[14] C.B. Holroyd, M.G.H. Coles, The neural basis of human error processing: reinforcement learning, dopamine, and the error-related negativity, Psychol. Rev.
109 (2002) 679–709.
[15] E.A. Holroyd, M.H. Bond, H.Y. Chan, Perceptions of sex-role stereotypes, selfconcept, and nursing role ideal in Chinese nursing students, J. Adv. Nurs. 37
(2002) 294–303.
[16] M.J. Larson, W.M. Perlstein, D. Stigge-Kaufman, K.G. Kelly, V.M. Dotson, Affective
context-induced modulation of the error-related negativity, Neuroreport 17
(2006) 329–333.
[17] M.J. Monteith, L. Ashburn-Nardo, C.I. Voils, A.M. Czopp, Putting the brakes on
prejudice: on the development and operation of cues for control, J. Pers. Soc.
Psychol. 83 (2002) 1029–1050.
[18] S. Nieuwenhuis, C.B. Holroyd, N. Mol, M.G.H. Coles, Reinforcement-related brain
potentials from medial frontal cortex: origins and functional significance, Neurosci. Biobehav. Rev. 28 (2004) 411–418.
[19] S. Nieuwenhuis, K.R. Ridderinkhof, J. Blom, G.P.H. Band, A. Kok, Error-related
brain potentials are differently related to awareness of response errors: Evidence from an antisaccade task, Psychophysiology 38 (2001) 752–760.
[20] L. Osterhout, M. Bersick, J. Mclaughlin, Brain potentials reflect violations of
gender stereotypes, Mem. Cogn. 25 (1997) 273–285.
[21] H.V. Semlitsch, P. Anderer, P. Schuster, O. Presslich, A solution for reliable and
valid reduction of ocular artifacts applied to the P300 ERP, Psychophysiology
23 (1986) 695–703.
[22] D.M. Wegner, J.A. Bargh, Control and automaticity in social life, in: D.T. Gilbert,
S.T. Fiske, G. Lindzey (Eds.), Handbook of Social Psychology, McGraw-Hill Press,
Boston, 1998, pp. 446–496.
[23] N. Yeung, M.M. Botvinick, The neural basis of error detection: conflict monitoring and the error-related negativity, Psychol. Rev. 111 (2004) 931–959.