Emotion words affect eye fixations during reading.
ABSTRACT Emotion words are generally characterized as possessing high arousal and extreme valence and have typically been investigated in paradigms in which they are presented and measured as single words. This study examined whether a word's emotional qualities influenced the time spent viewing that word in the context of normal reading. Eye movements were monitored as participants read sentences containing an emotionally positive (e.g., lucky), negative (e.g., angry), or neutral (e.g., plain) word. Target word frequency (high or low) was additionally varied to help determine the temporal locus of emotion effects, with interactive results suggesting an early lexical locus of emotion processing. In general, measures of target fixation time demonstrated significant effects of emotion and frequency as well as an interaction. The interaction arose from differential effects with negative words that were dependent on word frequency. Fixation times on emotion words (positive or negative) were consistently faster than those on neutral words with one exception-high-frequency negative words were read no faster than their neutral counterparts. These effects emerged in the earliest eye movement measures, namely, first and single fixation duration, suggesting that emotionality, as defined by arousal and valence, modulates lexical processing. Possible mechanisms involved in processing emotion words are discussed, including automatic vigilance and desensitization, both of which imply a key role for word frequency. Finally, it is important that early lexical effects of emotion processing can be established within the ecologically valid context of fluent reading.
Emotion Words Affect Eye Fixations During Reading
Graham G. Scott
University of Aberdeen
Patrick J. O’Donnell and Sara C. Sereno
University of Glasgow
Emotion words are generally characterized as possessing high arousal and extreme valence and have typically
been investigated in paradigms in which they are presented and measured as single words. This study
examined whether a word’s emotional qualities influenced the time spent viewing that word in the context of
normal reading. Eye movements were monitored as participants read sentences containing an emotionally
positive (e.g., lucky), negative (e.g., angry), or neutral (e.g., plain) word. Target word frequency (high or low)
was additionally varied to help determine the temporal locus of emotion effects, with interactive results
suggesting an early lexical locus of emotion processing. In general, measures of target fixation time
differential effects with negative words that were dependent on word frequency. Fixation times on emotion
words (positive or negative) were consistently faster than those on neutral words with one exception—high-
frequency negative words were read no faster than their neutral counterparts. These effects emerged in the
earliest eye movement measures, namely, first and single fixation duration, suggesting that emotionality, as
defined by arousal and valence, modulates lexical processing. Possible mechanisms involved in processing
emotion words are discussed, including automatic vigilance and desensitization, both of which imply a key
role for word frequency. Finally, it is important that early lexical effects of emotion processing can be
established within the ecologically valid context of fluent reading.
Keywords: emotion words, word frequency, reading, eye movements, lexical access
The time course of processing written emotion words has been
the focus of much recent research (e.g., Hofmann, Kuchinke,
Tamm, Vo ˜, & Jacobs, 2009; Kissler, Herbert, Peyk, & Jungho ¨fer,
2007; Kousta, Vinson, & Vigliocco, 2009; Schacht & Sommer,
2009; Scott, O’Donnell, Leuthold, & Sereno, 2009). Emotion
words are typically characterized by the semantic dimensions of
arousal, a measure of internal activation, and valence, a measure of
value or worth (e.g., Osgood, Suci, & Tannenbaum, 1957). Emo-
tion words, in comparison to neutral words, have higher arousal
values that correlate with extreme (high/positive or low/negative)
values of valence (e.g., Bradley & Lang, 1999). Although there is
a general consensus that emotion words are differentially pro-
cessed, there remain several areas of uncertainty concerning their
time course of activation, the precise brain mechanisms involved,
and the features of such words that drive effects. Emotion word
processing has been evaluated with a variety of experimental
designs and measures, including reaction time, electrophysiologi-
cal recordings, and neuroimaging. One methodology that, to our
knowledge, has not been explored in this domain, despite its online
nature and ecological validity, is eye tracking during normal read-
ing. The current study investigates the time course of emotion
word recognition within the framework of fluent reading.
It is difficult to generalize across emotion word studies because of
differences in stimuli and method. Many studies, for example, do not
compare all emotion conditions (positive, negative, and neutral) but
rather a subset (e.g., positive vs. negative, positive vs. neutral, or
negative vs. neutral). In terms of method, many studies use a lexical
decision or emotional decision (categorization) task, but others use
self-referential, oddball, or recollection paradigms. The pervasive use
of additional manipulations such as masking, priming, mood induc-
tion, lateralized presentation, stimulus repetition, and/or blocked pre-
sentation of each condition—although essential to investigate specific
research questions—can further complicate the interpretation of re-
sults. Under such varied circumstances, it is difficult to determine the
relative contribution to emotion effects from lexical versus task-
related processes (e.g., Estes & Verges, 2008). Nevertheless, behav-
ioral studies, regardless of methodology, generally have demonstrated
a processing advantage for positive over neutral words (e.g.,
Kakolewski, Crowson, Sewell, & Cromwell, 1999; Kanske & Kotz,
2007; Kousta et al., 2009; Kuchinke, Vo ˜, Hofmann, & Jacobs, 2007;
Schacht & Sommer, 2009; Scott et al., 2009). Some studies have
shown an advantage for negative over neutral words (e.g., Kanske &
Kotz, 2007; Kousta et al., 2009; Nakic, Smith, Busis, Vythilingam, &
Blair, 2006; Schacht & Sommer, 2009; Tabert et al., 2001; Wind-
This article was published Online First February 13, 2012.
Graham G. Scott, Department of Psychology, University of Aberdeen,
Aberdeen, Scotland; Patrick J. O’Donnell, School of Psychology, Univer-
sity of Glasgow, Glasgow, Scotland; Sara C. Sereno, Institute of Neuro-
science and Psychology, University of Glasgow.
Portions of this research were presented at the Third China International
Conference on Eye Movement Studies in Zhuhai, China, June 2008. This
research was supported by an Economic and Social Research Council
(ESRC) postgraduate fellowship to Graham G. Scott.
Correspondence concerning this article should be addressed to Sara
C. Sereno, Institute of Neuroscience and Psychology, 58 Hillhead
Street, University of Glasgow, Glasgow G12 8QB, Scotland, United
Kingdom. E-mail: Sara.Sereno@glasgow.ac.uk
Journal of Experimental Psychology:
Learning, Memory, and Cognition
2012, Vol. 38, No. 3, 783–792
© 2012 American Psychological Association
mann, Daum, & Gu ¨ntu ¨rku ¨n, 2002). Others have shown an advantage
for positive over negative words (e.g., Atchley, Ilardi, & Enloe, 2003;
Dahl, 2001; Estes & Verges, 2008; Kiehl, Hare, McDonald, & Brink,
1999; Wentura, Rothermund, & Bak, 2000).
Almost without exception, these studies have presented and
measured emotion words in isolation. Outside the laboratory,
words seldom occur alone accompanied by tasks such as lexical
decision or categorization; words typically occur within a text and
are read for comprehension. A few studies have examined event-
related potentials (ERPs) to the second of two successively pre-
sented emotion words (e.g., Fischler & Bradley, 2006; Schacht &
Sommer, 2009). Another study presented short passages of text
followed by a target sentence containing a single emotion word
that matched or mismatched the emotional state implied by the
prior story context (Gernsbacher, Goldsmith, & Robertson, 1992).
Although this study involved reading an emotion word within a
sentence, it only reported total sentence reading time across con-
ditions. Measuring eye movements during fluent reading is an
established technique that has reliably demonstrated sensitivity to
online perceptual and cognitive aspects of lexical processing
(Rayner, 1998, 2009; Sereno & Rayner, 2000b, 2003). As a re-
sponse measure, fixation time possesses certain advantages over
traditional behavioral measures: Response latencies are faster, and
there is no secondary task involving overt decisions.
Our focus was to determine whether the emotionality of a word
affects early lexical processes within the context of normal reading.
Behaviorally, the earliest and most robust marker of lexical access
recognized is a word frequency effect (e.g., Sereno & Rayner, 2003),
in which commonly used high-frequency (HF) words elicit faster
responses than low-frequency (LF) words that occur much less often.
Such effects have been reliably demonstrated in numerous eye move-
ment studies (for the earliest demonstration of effects, see Sereno &
Rayner, 2000a; for recent reviews, see Hand, Miellet, O’Donnell, &
Sereno, 2010; Rayner, 1998; see also Miellet, O’Donnell, & Sereno,
2009; Miellet, Sparrow, & Sereno, 2007; Sereno, 1992; Sereno,
O’Donnell, & Rayner, 2006). The approach we adopted was to gauge
the time course of emotion effects by additionally manipulating word
frequency in order to examine the nature of the interaction of these
factors (e.g., Sternberg, 1969). An observed interaction between emo-
tion and frequency would suggest that these variables share the same
processing stage, supporting an early, lexical locus of emotion pro-
cessing. Alternatively, additive effects would suggest that the tempo-
ral locus of emotion processing is relatively delayed. An Emotion ?
Frequency design was used in two recent studies that demonstrated a
significant interaction in response times (Kuchinke et al., 2007; Scott
et al., 2009) as well as ERP voltages (Scott et al., 2009). However,
both studies presented words as stimuli in isolation for lexical deci-
sion. In this study, LF and HF positive, negative, and neutral words
were embedded as targets in single-line sentences. Participants’ eye
movements were recorded as they read, and their fixation times on
target words were analyzed.
Forty-eight members of the University of Glasgow community
(25 women; mean age ? 21.5 years) were paid £6 for their
participation. All were native English speakers and had normal,
uncorrected vision. None had been diagnosed as dyslexic.
Participants’ eye movements were monitored via a Fourward
Technologies Dual Purkinje Eyetracker (Gen V), which has a
resolution of 10 min of arc. The signal from the eyetracker was
sampled every millisecond by a 386 computer. Sentences were
displayed on a ViewSonic 17GS CRT in a nonproportional font
and were limited to the central 60 characters of an 80-character
line. Participants were seated 86 cm from the monitor, and 4
characters subtended 1oof visual angle. Although viewing was
binocular, eye movements were recorded from the right eye.
Design and Materials
A 3 (Emotion: Positive, Negative, Neutral) ? 2 (Frequency: LF,
HF) design was used. Emotion words were defined by their arousal
and valence values, which were acquired from the Affective Norms
for English Words (ANEW), a database of 1,000 words (Bradley &
Lang, 1999). Each word in ANEW has associated ratings for arousal,
from 1 (low) to 9 (high), and for valence, from 1 (low, having a
negative meaning) to 9 (high, having a positive meaning). The fol-
lowing criteria for word selection were employed. Arousal values
ranged from 6 to 9 for Positive and Negative words and from 1 to 5.5
for Neutral words. Valence values ranged from 6 to 9 for Positive, 1
to 4 for Negative, and 4 to 6 for Neutral words. Word frequencies
were acquired from the British National Corpus (http://www.natcorp
.ox.ac.uk/), a database of 90 million written word tokens. The average
arousal, valence, and frequency values of target words across condi-
tions are presented in Table 1.
There were 24 sets of word triples (Positive, Negative, and
Neutral), of which 12 were LF and 12 were HF. A set of three
neutral sentence frames corresponded to each set of three targets,
such that each target word could appear in any of the three possible
sentence frames. Target words were always positioned near the
middle of a line of text. Although LF and HF targets appeared in
different sentence frames, the pretarget text was equally long in
terms of the number of words and characters (LF vs. HF: 4.4 vs.
4.5 words and 23 vs. 24 characters, respectively). All experimental
sentences and corresponding targets are listed in the Appendix. For
counterbalancing, three participant groups read three versions of
the materials, differing on the target (Positive, Negative, or Neu-
tral) used in each sentence. In this way, all combinations of targets
and corresponding sentence frames were presented, but any given
target or sentence frame was only presented once to each partici-
pant. A total of 72 items across six conditions (Emotion ? Fre-
quency) yielded 12 items per participant per condition.
While we directly manipulated target word frequency and emo-
tionality (arousal and valence), we attempted to match our stimuli
across several other lexical dimensions that have been shown to
affect fixation times in reading. These include word length, im-
ageability, and age of acquisition (AoA; the age at which a word
is first acquired). When other variables are controlled, a fixation
time advantage emerges for words that are shorter in length, more
imageable, or acquired earlier (e.g., Juhasz & Rayner, 2003;
Rayner, Sereno, & Raney, 1996). Imageability norms (for 88% of
the words) were obtained from five sources: the Bristol Norms
SCOTT, O’DONNELL, AND SERENO
(Stadthagen-Gonzalez & Davis, 2006), the MRC Psycholinguistic
Database (Wilson, 1988), and norms of Bird, Franklin, and How-
ard (2001), Clark and Pavio (2004), and Cortese and Fugett (2004).
AoA norms (for 82% of the words) were also obtained from five
sources: the Bristol Norms, the MRC Psycholinguistic Database,
and norms of Bird et al. (2001), Clark and Pavio (2004), and
Morrison, Chappell, and Ellis (1997). In our sample, the variables
of length, imageability, and AoA were closely matched across
conditions, and these values are presented in Table 1.
We had designed our materials so that targets would, at the same
time, be plausible but not be predictable from the prior context.
The predictability or anomaly of a word in a given context has
been shown to speed or slow associated fixation times, respec-
tively (e.g., Hand et al., 2010; Rayner, Warren, Juhasz, & Liv-
ersedge, 2004). In order to verify our intuitions regarding our
materials, we collected three sets of norms.
The first set of norms was obtained from a plausibility rating
task. The three versions of the materials (as described above) were
rated by 18 participants (three groups of six participants), none of
whom participated in the reading experiment. They were asked to
rate how plausible they considered each target word (presented in
bold font) to be within the context of its sentence on a scale of 1
(highly implausible) to 7 (highly plausible). Although the plausi-
bility ratings across conditions were tightly clustered around scale
point 6, we performed a two-way (Emotion ? Frequency) analysis
of variance (ANOVA) on the ratings both by participants (F1) and
by items (F2). There was a significant main effect of Frequency,
with HF targets (6.1) rated as more plausible than LF targets (5.7),
F1(1, 17) ? 10.13, p ? .01; F2(1, 11) ? 6.56, p ? .05. There was
no effect of Emotion, nor was there an interaction (Fs ? 1). The
effect of Frequency was not surprising in that HF words, by
definition, occur more often than LF words and, in this way, are
more likely to be encountered within any given context (e.g., Hand
et al., 2010). The average plausibility values across conditions are
presented in Table 1.
The second set of norms was a modified version of the plausi-
bility rating task described above. Experimental materials were
presented as sentence fragments up to and including the target
word, with each fragment followed by an ellipsis (i.e., “. . .”).
Unlike in the initial plausibility rating task, target words were not
presented in bold font; participants were simply asked to rate the
plausibility of the entire sentence fragment. The truncated versions
of the materials were rated by 12 participants (three groups of four
participants), none of whom participated in either the reading
experiment or the first plausibility task. The plausibility rating of
each of the six conditions was 6.5 (with standard deviations
ranging from 0.9 to 1.1). Neither main effects nor the interaction
was significant (Fs ? 1).
The third set of norms was obtained from a Cloze probability
task. Fourteen participants, none of whom participated in either the
reading experiment or the plausibility tasks, were given each
experimental item up to, but not including, the target word. Only
one set of materials was administered, because the target word was
absent. Participants were asked to generate the next word in the
sentence (i.e., the missing target). Responses were scored as 1 if
the target was correctly identified and 0 for all other guesses.
Results showed that in each condition the Cloze probability was
less than 0.01. Of a total of 3,024 possible responses (14 partici-
pants, 36 items per condition, six conditions), only five words
(arising from four different conditions) were correctly guessed. In
sum, these three sets of norms verified that our materials were
highly plausible without being predictable.1
A bite bar (to minimize head movements) was first prepared.
Participants were instructed to read each sentence on the monitor
1As mentioned earlier, different sentence frames were used for HF and
LF targets. It was not feasible to make sets of six sentence frames that
could accommodate triples (Positive, Negative, and Neutral) of both HF
and LF targets. Our pretarget contexts, however, were remarkably similar
across HF and LF conditions in terms of length (in both characters and
number of words) as well as judged plausibility and predictability. Nev-
ertheless, the issue remains that, although the three versions of any given
sentence set were read by different participants, the absence of having
identical pretarget contexts across frequency conditions may serve to limit
the conclusions that can be drawn.
Specifications of Experimental Words
PositiveNegativeNeutral Positive NegativeNeutral
Frequency in occurrences per million and Word Length in number of letters. The remaining variables are
expressed in units on the following scales: Arousal from 1 (low) to 9 (high); Valence from 1 (low, having a
negative meaning) to 9 (high, having a positive meaning); Imageability from 1 (low) to 7 (high); AoA (age of
acquisition) from 1 (early) to 7 (late); and Plausibility from 1 (highly implausible) to 7 (highly plausible). LF ?
low frequency; HF ? high frequency.
Mean values are shown with standard deviations in parentheses. Units of measurement are as follows:
READING EMOTION WORDS
and told that yes–no comprehension questions would follow half of
the sentences to ensure they were paying attention.
The experiment involved initial calibration of the eye-tracking
system, reading of six practice sentences, recalibration, and read-
ing of the 72 experimental sentences. A calibration display ap-
peared before every trial and involved a series of five calibration
points extending over the maximal horizontal range in which
sentences were presented. During this display, the calculated po-
sition of the eye was visible, allowing the experimenter to check
the accuracy of the calibration and recalibrate if necessary.
Each trial began with the calibration display. When participants
were fixating the leftmost calibration point (corresponding to the
first character of text), a sentence was presented. After reading
each sentence, participants fixated on a small box below and to the
right of the last word and pressed a key to clear the screen. The
calibration screen reappeared either immediately or after they had
answered a yes–no comprehension question by pressing corre-
sponding response keys (on average, 94% correct).
The target region comprised the target word and the space
preceding it. Standard lower (100 ms) and upper (750 ms) cutoff
values for individual fixations were used. Data were additionally
excluded if there was a track loss or blink on the target or if a
first-pass fixation on the target was either the first or last fixation
of the sentence. Overall, 7% of the data were excluded.
In reading, most content words are typically fixated once; some-
times they are immediately refixated or skipped. In this study, the
probabilities for target word single fixations, refixations, and skips
were .71, .11, and .11, respectively. Standard eye movement mea-
sures for target word processing include first fixation duration
(FFD) and gaze duration (GD). FFD is the average duration of the
first fixation on a word, whether it is a single fixation or one of two
or more consecutive fixations on that word. GD is the average sum
of all consecutive fixations on a word before the reader moves to
another word. The single fixation duration (SFD) measure (Rayner
et al., 1996; Sereno, 1992) has been used more recently in eye
movement data analysis and designates those cases when the target
was fixated exactly once. That is, SFD represents the proportion of
trials in which FFD and GD are identical and, in this study,
accounts for the majority (86%) of fixation time data. A final
measure is total fixation time (TT), which incorporates GD plus
any returning fixations to the target. That is, sometimes the reader
returns to the target after having earlier fixated (or skipped) it.
Such fixations (occurring on 13% of all trials in this study) are
added to first-pass durations. Participant means across all condi-
tions in each measure are presented in Table 2.
For each measure, a two-way analysis of variance (ANOVA)
was performed both by participants (F1) and by items (F2). As the
majority of target fixations were single fixations, we focus on the
SFD measure, although the results for all measures are presented.
In general, the other early target measures of FFD and GD pro-
duced a pattern of results highly similar to that for SFD. The later
TT measure usually produced a similar but sometimes weaker
version of results and is not reported in any further detail. ANOVA
results for all first-pass measures are presented in Table 3. The
SFD means (with standard error bars) are graphically depicted in
For SFD, as well as FFD and GD measures, the main effects of
Emotion and Frequency as well as the Emotion ? Frequency
interaction were significant both by participants and by items. We
therefore pursued follow-up contrasts, presented in Table 4, ex-
amining emotion word differences at each level of frequency as
well as frequency differences for each type of emotion word.
For LF words, SFDs on Neutral words (307 ms) were sig-
nificantly longer than those on either Positive (283 ms) or
Negative (286 ms) words, which did not differ from each other.
The pattern of effects was similar for FFD and GD. For HF
words, a different pattern emerged. SFDs on Neutral (282 ms)
and Negative (281 ms) words did not differ from each other,
and both were significantly longer than those on Positive words
(263 ms). Again, the pattern of effects was similar for FFD and
GD, although the Positive–Negative and Positive–Neutral con-
trasts were marginal by items in GD.
The frequency contrast for each type of emotion word in SFD
demonstrated significant effects for Positive (LF ? 283 vs. HF ?
263 ms) and Neutral (LF ? 307 vs. HF ? 282 ms) words but no
effect for Negative words (LF ? 286 vs. HF ? 281 ms). This
pattern was maintained in FFD. In GD, the frequency contrast for
Negative words additionally reached significance but only by
participants (trend by items).
We investigated the immediate effects of reading emotion words
within natural contexts. To our knowledge, no study to date has
examined the processing of single emotion words, as reflected in
Average Fixation Duration (Ms) Across Target Measures for LF and HF Positive, Negative, and
Positive Negative Neutral Positive NegativeNeutral
duration; SFD ? single fixation duration; GD ? gaze duration; TT ? total fixation time.
Standard deviations in parentheses. LF ? low frequency; HF ? high frequency; FFD ? first fixation
SCOTT, O’DONNELL, AND SERENO
fixation duration, in the natural context of fluent reading. Partici-
pants read sentences containing Positive, Negative, and Neutral
words. Critically, we additionally manipulated target word fre-
quency (LF, HF) to more precisely determine the temporal locus of
emotional processing. First-pass fixation time analyses demon-
strated significant effects of Emotion and Frequency as well as an
interaction, suggesting that lexical access (as indexed by word
frequency) is modulated by arousal and valence. Follow-up con-
trasts revealed that, for LF words, emotionality conferred a pro-
cessing benefit, with shorter fixation times on Positive and Neg-
ative than on Neutral words. For HF words, however, only Positive
words demonstrated a processing advantage. The dual nature of
our findings—a consistent advantage for LF and HF Positive
words versus a selective advantage for LF Negative words—may
help account for the mixed pattern of behavioral results in the
literature where frequency is not used as an experimental factor
(see also Kuchinke et al., 2007; Scott et al., 2009). Although
stimuli in past studies are often matched for frequency across
emotion conditions, the relative sampling of LF and HF words
varies within and between experiments. Nevertheless, our current
focus will be to address the theoretical implications of the pattern
Superficially, our results suggest that the expression of arousal
and valence effects is determined by how often a word is used.
That is, for LF words, word recognition during reading is facili-
tated by high arousal (Positive or Negative words) relative to low
arousal (Neutral words). For HF words, recognition is facilitated
only by high valence (Positive words) relative to an intermediate
valence (Neutral words), whereas low valence (Negative words)
confers no such processing advantage. Regarding arousal and
valence aspects of emotion independently, however, is somewhat
spurious, as extreme valence is strongly correlated with high
arousal (Bradley & Lang, 1999).
More substantially, emotional biases have been proposed to
explain how surrounding events are differentially processed,
with attentional resources modulated more so by events of one
type than another. Such biases, however, include a positivity
bias (e.g., Kakolewski et al., 1999; Mezulis, Abramson, Hyde,
high-frequency (LF, HF) Positive, Negative, and Neutral words. SFD ?
single fixation duration.
Single fixation duration (with standard error bars) on low- and
Analyses of Variance (ANOVAs) by Participants (F1) and Items
(F2) on Target Measures
Emotion ? Frequency
fixation duration; SFD ? single fixation duration; GD ? gaze duration.
df ? degrees of freedom; MSE ? mean squared error; FFD ? first
Follow-Up Emotion and Frequency Contrasts by Participants
(F1) and Items (F2) on Target Measures
LF vs. HF
LF vs. HF
LF vs. HF
duration; SFD ? single fixation duration; GD ? gaze duration.
LF ? low frequency; HF ? high frequency; FFD ? first fixation
READING EMOTION WORDS
& Hankin, 2004), a negativity bias (e.g., Cacioppo & Gardner,
1999; Rozin & Royzman, 2001), as well as a generalized
emotion bias (e.g., Ortigue et al., 2004; Zeelenberg, Wagen-
makers, & Rotteveel, 2006). Although a positivity bias can
account for the relative advantage of both LF and HF Positive
words (compared to Neutral words), it cannot explain why LF
Negative words are likewise advantaged. A negativity bias, on
the other hand, can work in both facilitative and inhibitory
ways. For example, in the automatic vigilance model (Pratto &
John, 1991), an initial, rapid and automatic evaluation (appet-
itive approach vs. aversive withdrawal) of a stimulus is fol-
lowed by prolonged attentional monitoring selectively for neg-
ative stimuli (cf. the mobilization–minimization hypothesis;
Taylor, 1991). This slower disengagement reflects additional
evaluation of potential threat and serves an adaptive purpose.
However, it is unclear how this translates into the temporal
parameters of behavioral response measures. Recent debates
regarding automatic vigilance have focused on whether the
nature of the relationship between valence and recognition is
categorical (Estes & Adelman, 2008a, 2008b) or linear (Larsen,
Mercer, & Balota, 2006; Larsen, Mercer, Balota, & Strube,
2008). In any case, although automatic vigilance can explain the
lack of facilitation for HF Negative words (compared to Neutral
words), it cannot explain the relative facilitation observed for
LF Negative words. Finally, a generalized bias to any emotional
stimulus cannot explain a lack of facilitation for HF Negative
It may be possible to account for our results via a modified
version of automatic vigilance. Such a model would need to be
additionally sensitive to word frequency in order to account for
the presence of facilitation without subsequent inhibition in the
case of LF Negative words. Given that there is a rough corre-
spondence between the frequency of a word and the frequency
of what it represents, it should follow that a seldom-occurring
threat is less threatening. However, this discounts the fact that
a threat, no matter how rare, can still be fatal. Another possible
explanation for our results derives from the psychotherapy
literature on desensitization of negative stimuli, where repeated
exposure to a cue perceived as dangerous or threatening neu-
tralizes its negativity (cf. “the boy who cried wolf”; Aesop,
1919). For example, it has been demonstrated that affect label-
ing decreases amygdala activation in response to negative im-
ages (e.g., Lieberman et al., 2007). On this view, as others have
suggested, all high arousal words, regardless of valence, are
facilitated. HF Negative words, however, despite their designa-
tion (e.g., emotional ratings), should not be classified as being
high arousal. By definition, HF words are used often and in
many contexts. Repeated exposure leading to affect diminution
selectively affects Negative words because there is no equiva-
lent mechanism for Positive words. As a corollary, this view
would need to claim that emotion ratings themselves are valid
but only for the conditions under which they are obtained,
namely in a task in which explicit arousal and valence judg-
ments are made offline on words presented in isolation. At
present, both explanations seem able to account for the pattern
In terms of the neural substrates underpinning the processing of
emotion words, hemodynamic studies have typically shown
amygdala involvement (e.g., Hamann & Mao, 2002; Herbert et al.,
2009; Lewis, Critchley, Rotshtein, & Dolan, 2007; Nakic et al.,
2006; Tabert et al., 2001). The relative timing of these activations,
however, cannot be specified because they are metabolic conse-
quences of neural activity. For example, in a study using intracra-
nial recording, Naccache et al. (2005) found amygdala involve-
ment during subliminal presentation of emotion words but only
800 ms after stimulus onset. This activation is quite delayed with
respect to word recognition, which is estimated to take place
within 200 ms (Sereno & Rayner, 2003; Sereno, Rayner, & Posner,
As the perceived emotional nature of a word, defined by the
characteristics of arousal and valence, is modulated by frequency,
it presumably affects the early stages of lexical access, and this has
implications for the range of information stored about words in the
lexicon. The significant early FFD and SFD effects found in our
study indicate that the degree of lexical access required by eye
movement control models such as E-Z Reader (Reichle, Rayner, &
Pollatsek, 2003) is indeed influenced by complex semantic infor-
mation. Our results seem to be explicable only in terms of the
word’s emotional quality and not by confounding features such as
its length, imageability, AoA, or plausibility. Although our find-
ings identify emotion as a feature of the lexicon, nonetheless, there
may be more parsimonious explanations. The first of these would
appeal to a mechanism other than some form of lexical access that
drives eye movements in fluent reading. Consequently, it is pos-
sible, although unlikely, that a superficial aspect of emotion words
per se, such as orthographic distinctiveness, may give rise to an
early advantage. The second might argue that, if lexical access is
the main trigger for eye movements in reading, the form of access
may be only partial.
Traditional theories of lexical access argued that what became
available at the “magic moment” of recognition would be the
meaning of the word, its orthography and phonology, and its
syntactic class (Balota, 1990). However, even as such a definition
was suggested, it was recognized that lexical access was likely to
be incremental and would activate broader sources of information.
Recently, it has been argued that more extensive semantic knowl-
edge is acquired early on in word recognition (Dambacher, Rolfs,
Go ¨llner, Kliegl, & Jacobs, 2009; Hauk, Davis, Ford, Pulvermu ¨ller,
& Marslen-Wilson, 2006; Norris, 2006; Pulvermu ¨ller, Assadollahi,
& Elbert, 2001; Sereno, Brewer, & O’Donnell, 2003). Ultimately,
determining temporal activation functions of the different kinds of
information that drive word recognition is an empirical issue.
Nevertheless, a growing body of electrophysiological evidence
suggests that the emotional quality of a word is accessed relatively
early (Bernat, Bunce, & Shevrin, 2001; Schacht & Sommer, 2009;
Scott et al., 2009; Skrandies, 1998; see also Kissler, Assadollahi,
& Herbert, 2006). Our current results not only are consistent with
such findings but, notably, extend them within the context of fluent
In sum, despite the amount of research on affective word
processing, a clear picture is only beginning to emerge regard-
ing when and how emotion-specific lexical mechanisms oper-
ate. It is important to recognize the limitations of emotion word
studies in which words are presented in isolation and response
time is relatively delayed. In fluent reading, eye movement data
demonstrate that individual word meanings are rapidly acti-
vated and integrated online into a developing discourse context.
Establishing reliable emotion word effects in early reading
SCOTT, O’DONNELL, AND SERENO
measures provides positive evidence for a central role of emo-
tion in language processing.
Aesop. (trans. 1919). The Aesop for children with pictures by Milo Winter.
Retrieved from http://www.gutenberg.org/files/19994/19994-h/19994-
Atchley, R. A., Ilardi, S. S., & Enloe, A. (2003). Hemispheric asymmetry
in the processing of emotional content in word meanings: The effect of
current and past depression. Brain and Language, 84, 105–119. doi:
Balota, D. A. (1990). The role of meaning in word recognition. In D. A.
Balota, G. B. Flores d’Arcais, & K. Rayner (Eds.), Comprehension
processes in reading (pp. 9–32). Hillsdale, NJ: Erlbaum.
Bernat, E., Bunce, S., & Shevrin, H. (2001). Event-related brain potentials
differentiate positive and negative mood adjectives during both supra-
liminal and subliminal visual processing. International Journal of Psy-
chophysiology, 42, 11–34. doi:10.1016/S0167-8760(01)00133-7
Bird, H., Franklin, S., & Howard, D. (2001). Age of acquisition and
imageability ratings for a large set of words, including verbs and
function words. Behavior Research Methods, Instruments, & Comput-
ers, 33, 73–79. doi:10.3758/BF03195349
Bradley, M. M., & Lang, P. J. (1999). Affective norms for English words
(ANEW). Instruction manual and affective ratings. Gainesville: NIMH
Center for the Study of Emotion and Attention, University of Florida.
Cacioppo, J. T., & Gardner, W. L. (1999). Emotion. Annual Review of
Psychology, 50, 191–214. doi:10.1146/annurev.psych.50.1.191
Clark, J. M., & Paivio, A. (2004). Extensions of the Paivio, Yuille, and
Madigan (1968) norms. Behavior Research Methods, Instruments, &
Computers, 36, 371–383. doi:10.3758/BF03195584
Cortese, M. J., & Fugett, A. (2004). Imageability ratings for 3,000 mono-
syllabic words. Behavior Research Methods, Instruments, & Computers,
36, 384–387. doi:10.3758/BF03195585
Dahl, M. (2001). Asymmetries in the processing of emotionally valenced
words. Scandinavian Journal of Psychology, 42, 97–104. doi:10.1111/
Dambacher, M., Rolfs, M., Go ¨llner, K., Kliegl, R., & Jacobs, A. M. (2009).
Event-related potentials reveal rapid verification of predicted visual
input. PLoS ONE, 4, e5047. doi:10.1371/journal.pone.0005047
Estes, Z., & Adelman, J. S. (2008a). Automatic vigilance for negative
words in lexical decision and naming: Comment on Larsen, Mercer, and
Balota (2006). Emotion, 8, 441–444. doi:10.1037/1528-3518.104.22.1681
Estes, Z., & Adelman, J. S. (2008b). Automatic vigilance for negative
words is categorical and general. Emotion, 8, 453–457. doi:10.1037/
Estes, Z., & Verges, M. (2008). Freeze or flee? Negative stimuli elicit
selective responding. Cognition, 108, 557–565. doi:10.1016/j.cognition
Fischler, I., & Bradley, M. (2006). Event-related potential studies of
language and emotion: Words, phrases, and task effects. In S. Anders, G.
Ende, M. Jungho ¨fer, J. Kissler, & D. Wildgrubber (Eds.), Progress in
brain research: Understanding emotions (Vol. 156, pp. 185–203). Am-
sterdam, the Netherlands: Elsevier.
Gernsbacher, M. A., Goldsmith, H. H., & Robertson, R. R. W. (1992). Do
readers mentally represent characters’ emotional states? Cognition &
Emotion, 6, 89–111. doi:10.1080/02699939208411061
Hamann, S., & Mao, H. (2002). Positive and negative emotional verbal
stimuli elicit activity in the left amygdala. NeuroReport, 13, 15–19.
Hand, C. J., Miellet, S., O’Donnell, P. J., & Sereno, S. C. (2010). Word
frequency and contextual predictability effects in reading: It depends
where you’re coming from. Journal of Experimental Psychology: Hu-
man Perception and Performance, 36, 1294–1313. doi:10.1037/
Hauk, O., Davis, M. H., Ford, M., Pulvermu ¨ller, F., & Marslen-Wilson,
W. D. (2006). The time course of visual word recognition as revealed by
linear regression analysis of ERP data. NeuroImage, 30, 1383–1400.
Herbert, C., Ethofer, T., Anders, S., Jungho ¨fer, M., Wildgruber, D., Grodd,
W., & Kissler, J. (2009). Amygdala activation during reading of emo-
tional adjectives: An advantage for pleasant content. Social Cognitive
and Affective Neuroscience, 4, 35–49. doi:10.1093/scan/nsn027
Hofmann, M. J., Kuchinke, L., Tamm, S., Vo ˜, M. L.-H., & Jacobs, A. M.
(2009). Affective processing within 1/10th of a second: High arousal is
necessary for early facilitative processing of negative but not positive
words. Cognitive, Affective, & Behavioral Neuroscience, 9, 389–397.
Juhasz, B. J., & Rayner, K. (2003). Investigating the effects of a set of
intercorrelated variables on eye fixation durations in reading. Journal of
Experimental Psychology: Learning, Memory, and Cognition, 29, 1312–
Kakolewski, K. E., Crowson, J. J., Jr., Sewell, K. W., & Cromwell, R. L.
(1999). Laterality, word valence, and visual attention: A comparison of
depressed and non-depressed individuals. International Journal of Psy-
chophysiology, 34, 283–292. doi:10.1016/S0167-8760(99)00085-9
Kanske, P., & Kotz, S. A. (2007). Concreteness in emotional words: ERP
evidence from a hemifield study. Brain Research, 1148, 138–148.
Kiehl, K. A., Hare, R. D., McDonald, J. J., & Brink, J. (1999). Semantic
and affective processing in psychopaths: An event-related potential
(ERP) study. Psychophysiology, 36, 765–774. doi:10.1111/1469-
Kissler, J., Assadollahi, R., & Herbert, C. (2006). Emotional and semantic
networks in visual word processing: Insights from ERP studies. In S.
Anders, G. Ende, M. Jungho ¨fer, J. Kissler, & D. Wildgrubber (Eds.),
Progress in brain research: Understanding emotions (Vol. 156, pp.
147–183). Amsterdam, the Netherlands: Elsevier.
Kissler, J., Herbert, C., Peyk, P., & Jungho ¨fer, M. (2007). Buzzwords:
Early cortical responses to emotional words during reading. Psycholog-
ical Science, 18, 475–480. doi:10.1111/j.1467-9280.2007.01924.x
Kousta, S.-T., Vinson, D. P., & Vigliocco, G. (2009). Emotion words,
regardless of polarity, have a processing advantage over neutral words.
Cognition, 112, 473–481. doi:10.1016/j.cognition.2009.06.007
Kuchinke, L., Vo ˜, M. L.-H., Hofmann, M., & Jacobs, A. M. (2007).
Pupillary responses during lexical decisions vary with word frequency
but not emotional valence. International Journal of Psychophysiology,
65, 132–140. doi:10.1016/j.ijpsycho.2007.04.004
Larsen, R. J., Mercer, K. A., & Balota, D. A. (2006). Lexical characteristics
of words used in emotional Stroop experiments. Emotion, 6, 62–72.
Larsen, R. J., Mercer, K. A., Balota, D. A., & Strube, M. J. (2008). Not all
negative words slow down lexical decision and naming speed: Impor-
tance of word arousal. Emotion, 8, 445–452. doi:10.1037/1528-
Lewis, P. A., Critchley, H. D., Rotshtein, P., & Dolan, R. J. (2007). Neural
correlates of processing valence and arousal in affective words. Cerebral
Cortex, 17, 742–748. doi:10.1093/cercor/bhk024
Lieberman, M. D., Eisenberger, N. I., Crockett, M. J., Tom, S. M., Pfeifer,
J. H., & Way, B. M. (2007). Putting feelings into words: Affect labeling
disrupts amygdala activity to affective stimuli. Psychological Science,
18, 421–428. doi:10.1111/j.1467-9280.2007.01916.x
Mezulis, A. H., Abramson, L. Y., Hyde, J. S., & Hankin, B. L. (2004). Is
there a universal positivity bias in attributions? A meta-analytic review
of individual, developmental, and cultural differences in the self-serving
attributional bias. Psychological Bulletin, 130, 711–747. doi:10.1037/
Miellet, S., O’Donnell, P. J., & Sereno, S. C. (2009). Parafoveal magnifi-
cation: Visual acuity does not modulate the perceptual span in reading.
READING EMOTION WORDS
Psychological Science, 20, 721–728. doi:10.1111/j.1467-9280
Miellet, S., Sparrow, L., & Sereno, S. C. (2007). Word frequency and
predictability effects in reading French: An evaluation of the E-Z Reader
model. Psychonomic Bulletin & Review, 14, 762–769. doi:10.3758/
Morrison, C. M., Chappell, T. D., & Ellis, A. W. (1997). Age of acquisition
norms for a large set of object names and their relation to adult estimates
and other variables. Quarterly Journal of Experimental Psychology:
Human Experimental Psychology, 50(A), 528–559. doi:10.1080/
Naccache, L., Gaillard, R., Adam, C., Hasboun, D., Cle ´menceau, S.,
Baulac, M., . . . Cohen, L. (2005). A direct intracranial record of emo-
tions evoked by subliminal words. Proceedings of the National Academy
of Sciences, USA, 102, 7713–7717. doi:10.1073/pnas.0500542102
Nakic, M., Smith, B. W., Busis, S., Vythilingam, M., & Blair, R. J. R.
(2006). The impact of affect and frequency on lexical decision: The role
of the amygdala and inferior frontal cortex. NeuroImage, 31, 1752–
Norris, D. (2006). The Bayesian reader: Explaining word recognition as an
optimal Bayesian decision process. Psychological Review, 113, 327–
Ortigue, S., Michel, C. M., Murray, M. M., Mohr, C., Carbonnel, S., &
Landis, T. (2004). Electrical neuroimaging reveals early generator mod-
ulation to emotional words. NeuroImage, 21, 1242–1251. doi:10.1016/
Osgood, C. E., Suci, G. J., & Tannenbaum, P. H. (1957). The measurement
of meaning. Urbana: University of Illinois Press.
Pratto, F., & John, O. P. (1991). Automatic vigilance: The attention-
grabbing power of negative social information. Journal of Personality
and Social Psychology, 61, 380–391. doi:10.1037/0022-3522.214.171.1240
Pulvermu ¨ller, F., Assadollahi, R., & Elbert, T. (2001). Neuromagnetic
evidence for early semantic access in word recognition. European Jour-
nal of Neuroscience, 13, 201–205. doi:10.1046/j.0953-816X
Rayner, K. (1998). Eye movements in reading and information processing:
20 years of research. Psychological Bulletin, 124, 372–422. doi:
Rayner, K. (2009). Eye movements and attention in reading, scene per-
ception, and visual search. Quarterly Journal of Experimental Psychol-
ogy, 62, 1457–1506. doi:10.1080/17470210902816461
Rayner, K., Sereno, S. C., & Raney, G. E. (1996). Eye movement control
in reading: A comparison of two types of models. Journal of Experi-
mental Psychology: Human Perception and Performance, 22, 1188–
Rayner, K., Warren, T., Juhasz, B. J., & Liversedge, S. P. (2004). The
effect of plausibility on eye movements in reading. Journal of Experi-
mental Psychology: Learning, Memory, and Cognition, 30, 1290–1301.
Reichle, E. D., Rayner, K., & Pollatsek, A. (2003). The E-Z Reader model
of eye-movement control in reading: Comparisons to other models.
Behavioral and Brain Sciences, 26, 445–476. doi:10.1017/
Rozin, P., & Royzman, E. B. (2001). Negativity bias, negativity domi-
nance, and contagion. Personality and Social Psychology Review, 5,
Schacht, A., & Sommer, W. (2009). Time course and task dependence of
emotion effects in word processing. Cognitive, Affective, & Behavioral
Neuroscience, 9, 28–43. doi:10.3758/CABN.9.1.28
Scott, G. G., O’Donnell, P. J., Leuthold, H., & Sereno, S. C. (2009). Early
emotion word processing: Evidence from event-related potentials. Bio-
logical Psychology, 80, 95–104. doi:10.1016/j.biopsycho.2008.03.010
Sereno, S. C. (1992). Early lexical effects when fixating a word in reading.
In K. Rayner (Ed.), Eye movements and visual cognition: Scene percep-
tion and reading (pp. 304–316). New York, NY: Springer-Verlag.
Sereno, S. C., Brewer, C. C., & O’Donnell, P. J. (2003). Context effects in
word recognition: Evidence for early interactive processing. Psycholog-
ical Science, 14, 328–333. doi:10.1111/1467-9280.14471
Sereno, S. C., O’Donnell, P. J., & Rayner, K. (2006). Eye movements and
lexical ambiguity resolution: Investigating the subordinate bias effect.
Journal of Experimental Psychology: Human Perception and Perfor-
mance, 32, 335–350. doi:10.1037/0096-15126.96.36.1995
Sereno, S. C., & Rayner, K. (2000a). Spelling-sound regularity effects on
eye fixations in reading. Perception & Psychophysics, 62, 402–409.
Sereno, S. C., & Rayner, K. (2000b). The when and where of reading in the
brain. Brain and Cognition, 42, 78–81. doi:10.1006/brcg.1999.1167
Sereno, S. C., & Rayner, K. (2003). Measuring word recognition in
reading: Eye movements and event-related potentials. Trends in Cogni-
tive Sciences, 7, 489–493. doi:10.1016/j.tics.2003.09.010
Sereno, S. C., Rayner, K., & Posner, M. I. (1998). A time-line of word
recognition: Evidence from eye movements and event-related potentials.
NeuroReport, 9, 2195–2200. doi:10.1097/00001756-199807130-00009
Skrandies, W. (1998). Evoked potential correlates of semantic mean-
ing—A brain mapping study. Cognitive Brain Research, 6, 173–183.
Stadthagen-Gonzalez, H., & Davis, C. J. (2006). The Bristol norms for age
of acquisition, imageability, and familiarity. Behavior Research Meth-
ods, 38, 598–605. doi:10.3758/BF03193891
Sternberg, S. (1969). Memory-scanning: Mental processes revealed by
reaction-time experiments. American Scientist, 57, 421–457.
Tabert, M. H., Borod, J. C., Tang, C. Y., Lange, G., Wei, T. C., Johnson,
R., . . . Buchsbaum, M. S. (2001). Differential amygdala activation
during emotional decision and recognition memory tasks using unpleas-
ant words: An fMRI study. Neuropsychologia, 39, 556–573. doi:
Taylor, S. E. (1991). Asymmetrical effects of positive and negative events.
The mobilization–minimization hypothesis. Psychological Bulletin, 110,
Wentura, D., Rothermund, K., & Bak, P. (2000). Automatic vigilance: The
attention-grabbing power of approach- and avoidance-related social in-
formation. Journal of Personality and Social Psychology, 78, 1024–
Wilson, M. D. (1988). The MRC Psycholinguistic Database: Machine
readable dictionary, Version 2. Behavioral Research Methods, Instru-
ments, & Computers, 20, 6–10. doi:10.3758/BF03202594
Windmann, S., Daum, I., & Gu ¨ntu ¨rku ¨n, O. (2002). Dissociating prelexical
and postlexical processing of affective information in the two hemi-
spheres: Effects of the stimulus presentation format. Brain and Lan-
guage, 80, 269–286. doi:10.1006/brln.2001.2586
Zeelenberg, R., Wagenmakers, E., & Rotteveel, M. (2006). The impact of
emotion on perception: Bias or enhanced processing? Psychological
Science, 17, 287–291. doi:10.1111/j.1467-9280.2006.01700.x
SCOTT, O’DONNELL, AND SERENO
Sentence frames for low-frequency targets
Dave kept staring at the _____ performer on the stage.
Her movements were _____ and made Ryan feel uncomfortable.
Ed thought that Bea was _____ and that night she proved it.
Looking carefully, you can see the _____ behind the trees.
Located on a hill, the _____ could be seen for miles.
Everyone in the _____ was beginning to feel ill.
The lowly private thought he would _____ the major general.
Traditionally recruits would _____ their superior officers.
Diego knew he could only _____ some members of his company.
annoy salute cheer
The documentary on the _____ was very interesting.
Lisa read about the _____ in her animal book.
A sturdy creature, the _____ can survive in many habitats.
The young boy felt _____ about being in the school play.
Amanda was totally _____ about public speaking.
The referee was _____ as he ran onto the pitch.
The relatives thought Jane was _____ for a girl of her age.
Jason knew he was _____ and he didn’t want to change.
Heather felt very _____ after having a few drinks.
Hannah could only think about the _____ as she lay awake.
The appearance of the _____ alarmed the investigator.
Andrew read about the _____ in his magazine.
Agent Ross would have to _____ Mary during his assignment.
She speculated Dylan might _____ her at the party.
Fred knew he could never _____ her in a crowd.
The sailor wanted to _____ the passing jet-skier.
The young fishermen tried to _____ the Greenpeace boat.
The crowd wanted to _____ the champion with drinks.
The man who carried out the _____ had a long beard.
Rumours of the _____ spread from village to village.
The recipient of the _____ was a young refugee.
The court heard how the _____ had been treated.
The student wanted to dress as a _____ for Halloween.
Judith thought that Phil could be a _____ when he grew up.
The teenagers were _____ during their return journey.
Nigel was frequently _____ because he drank too much.
The tramp often became _____ due to slight distractions.
Sentence frames for high-frequency targets
The reporter described David’s _____ in great detail.
She spoke of a mysterious _____ that would soon follow.
Their inevitable _____ was the central theme of the story.
Janet and Sheena were lectured by the _____ professor.
Jane thought that the _____ actor was quite attractive.
The student was _____ in her response to the question.
The careless man dropped the _____ on the floor.
Oliver forgot to bring Kevin’s _____ to the table.
The vagrant had found the _____ in the dumpster.
READING EMOTION WORDS
Sentence frames for low-frequency targets
Negative Neutral Positive
The counsellor was quick to notice _____ in many children.
The coach nurtured _____ in some of the younger players.
There had been evidence of Robin’s _____ since childhood.
They were discussing the young _____ over dinner.
The first _____ was the oldest of them all.
The article described each _____ as tall and thin.
victim writer winner
The confetti landed on the _____ child in the brown jumper.
She wore glasses and was described as _____ by most people.
Steam rose as the _____ woman disembarked from the train.
Linda listened to the _____ through the wall.
When Nicole heard the _____ she thought of her childhood.
As Lucy thought about the _____, Scott poured her a drink.
As the jury watched, the _____ defendant began to cry.
The detective kept the _____ girl waiting for over an hour.
After talking to the _____ clerk, he saw her point of view.
guilty quiet pretty
The cat watched the _____ hurry down the street.
There was silence as the _____ walked into the theatre.
The shop was closing as the _____ rushed to the checkout.
criminal teacher friends
Stories of the group’s _____ travelled quickly.
The origin of their _____ was not well understood.
We discovered that Dr. Falkin’s _____ was a hoax.
As Kyle anticipated the _____, his heart began racing.
After the _____ everyone looked for someone to blame.
In the end, the _____ wasn’t as bad as people had feared.
Nobody realised the impact that the _____ would cause.
The journalist reported the _____ at the general’s estate.
Reginald still dreamt about the _____ months later.
Received June 3, 2010
Revision received October 31, 2011
Accepted November 14, 2011 ?
SCOTT, O’DONNELL, AND SERENO