You are on page 1of 14

J Psycholinguist Res (2013) 42:379391

DOI 10.1007/s10936-012-9221-7

Evaluating Word in Phrase: The Modulation Effect


of Emotional Context on Word Comprehension
Hongyan Liu Zhiguo Hu Danling Peng

Published online: 12 June 2012


Springer Science+Business Media, LLC 2012

Abstract The present study aims to explore the influence of emotional context on word
evaluation. Participants were asked to read an adjective + noun phrase, where the adjective
could be a positive or negative word, and the noun could be a positive, neutral or negative
word, and then to make an emotional evaluation on the emotional tone of the target noun
based on a 9-point Likert scale. In a control condition, an isolated noun was presented with
no context. Results showed that positive context made the evaluation of target words bias
toward positive tone, while negative context shaped the evaluation of target words toward
negative tone. The modulatory effect of negative context was greater than that of positive
context in shaping evaluation of emotional words with opposite valence. Moreover, the modulatory effect of emotional context was constrained by the inherent meaning of target word.
The present study demonstrated the flexibility as well as the relative stability of emotional
meaning of word.
Keywords

Emotion Context Evaluation Modulation Word comprehension

Introduction
Words are the basic units of language, and understanding the meaning of words is crucial
for survival in our highly language-oriented society. For 1,000s of years, dictionaries have

H. Liu
Department of Psychology, Zhejiang Sci-Tech University, Hangzhou 310018, Peoples Republic of China
H. Liu D. Peng
State Key Laboratory of Cognitive Neuroscience and Learning, Beijing Normal University,
Beijing 100875, Peoples Republic of China
Z. Hu (B)
Center for Cognition and Brain Disorders, Hangzhou Normal University,
Hangzhou 310015, Peoples Republic of China
e-mail: maxnow_hu@hotmail.com; huzg@hznu.edu.cn

123

380

J Psycholinguist Res (2013) 42:379391

been used to record the meanings of words. For example, in the Chinese dictionary, the word
lie means a false statement or something meant to deceive or give a wrong impression.
Obviously, this was regarded as a negative word. Accordingly, it was often associated with
negative situations, such as an evil lie and a barefaced lie. However, in some contexts, the
emotional tone of the word was not as negative as in the dictionary. For example, a white lie,
which is told to avoid harmful implications of the truth, such as in a situation where a dying
mother was told that her son was delayed from visiting, rather than killed in an accident. In
this example, the lie was not regarded as negative, but rather as merciful and kindhearted,
giving it a positive connotation. In real-life conversation and written text, we could often
found that the emotional valence of a word is biased by certain context. However, to our
knowledge, the literature has been lacking in empirical evidence for the role of context in
emotional evaluation of word.
Extant studies have demonstrated that the meaning of a word could be affected by context. The most convincing evidence was provided by research on ambiguous words, which
have multiple meanings, such as homonymous words (with multiple unrelated meanings)
and polysemous words (with multiple related meanings). For example, the word bank is
homonymous, meaning a monetary institute, as well as the ground next to a river. A host of
studies have demonstrated that a supportive context could facilitate resolution of the lexical
ambiguity of a word (e.g., Binder and Morris 1995; Fischler and Bloom 1979; Onifer and
Swinney 1981; Rayner and Frazier 1989; Stanovich and West 1979; Zempleni et al. 2007;
for reviews on models and theories, c.f. Gorfein 2001). When an ambiguous word like bank
was embedded in context (e.g., He went to the bank to cash a check.), it was not detected
as ambiguous because the disambiguation processes rapidly and automatically selected one
particular meaning. Since the semantic meaning of a word can be selected by context, it seems
reasonable that evaluation of the emotional tone of a word could be shaped by an emotional
context, as emotional evaluation is more subjective and changeable than semantic meaning.
Though no study has directly explored the contextual influence on word evaluation, there
are some evidences to lend support for the hypothesis that the evaluation of a word could
be shaped by its emotional context. The most important evidences came from the studies
concerning the pairing effect of other words on a target word. A handful of affective priming
studies investigated the asynchronous pairing effect, in which a target emotional word was
preceded by a prime emotional word, and subjects were asked to perform an evaluative categorization task (e.g., De Houwer et al. 2002; Fazio et al. 1986; Hermans et al. 2001; Klinger
et al. 2000) on target word. Results have usually shown that the time needed to evaluate the
target word as either positive or negative is significantly shorter when the prime and target
are affectively congruent, meaning that they share the same valence (e.g., happy-flower),
compared to trials in which the prime and target are affectively incongruent (e.g., sad-flower)
(Hermans et al. 2001). Besides the asynchronous pairing effect, some other studies also
address the synchronous pairing effect. For example, in a study by Ochsner et al. (2008),
an affective Eriksen flanker paradigm was used, and participants were instructed to decide
whether a central target word was positive or negative while ignoring the other two words
that appeared above and below the central word. They found that participants responded
faster to the target when surrounded by emotional congruent flanking words than emotional
incongruent ones.
In both the affective priming studies and affective Eriksen Flanker studies, the pairing
context words and target words were either positive or negative, and had an emotional relationship (i.e., with the same or opposite valence), but not a semantic relationship. Thus
semantic dissociations from the target and context words prevented alterations in the emotional evaluation of target words by context words. In other words, a positive word such as

123

J Psycholinguist Res (2013) 42:379391

381

flower should be evaluated as positive, regardless of whether it is preceded or surrounded by


happy or sad, though the reaction time (RT) should be faster when preceded by happy.
In such situation, the pairing words can exert only an influence on the accuracy and speed of
target word evaluation.
In fact, to date some studies have suggested the necessity of examining the effect of semantic-related contexts on word evaluation. For example, in a study by Hertel and Parks (2002),
recall for neutral nouns improved when those nouns were paired with adjectives that changed
their emotional meaning (e.g., cruise ship, cargo ship, sinking ship). Similarly, Brierley et al.
(2007) found that memory for neutral words was enhanced when the words were presented in
a sentence with emotional context. Neutral words in an emotional context may be endowed
with a certain emotional tone, which might make them distinctive (Eich and Schooler 2000;
Heuer and Reisberg 1992), thus improving recall (Brierley et al. 2007; Hertel and Parks 2002;
Hunt and McDaniel 1993), consistent with the phenomenon of emotional enhancement of
memory (Winograd and Neisser 1992). These memory studies showed that neutral words
embedded in emotional context exhibited some features as emotional word (i.e., emotional
memory enhancement), indicating that contextual information that is semantically linked
to the target word may exhibit extra influence in the processing of a target word, i.e., the
inherent emotional tone of target word could be varied under certain emotional contexts. As
the neutral words could be processed similarly as emotional ones, a follow-up question is
that whether the evaluation of emotional words could also be biased by emotional context.
The present study aims to determine whether the evaluation of target words can be changed
by emotional contexts which are associated with the target words both emotionally and
semantically. As evaluation of a words emotional valence is often dependent on how it is
used linguistically (e.g., the word lie is negative in an evil lie, and positive in a white
lie), and the influence of context on word evaluation was proved to be limited by other
non-linguistical visual background (e.g., Erk et al. 2005), we used a phrase as context, just
as the situation we usually encounter a word in our everyday life. To simplify the issue, we
adopted short adjective + noun phrases as stimuli, in which a positive or negative adjective
word served as the context and a noun (positive, neutral or negative) as the target word.
Compared to priming and Flanker studies, such kind of phrases are more naturalistic stimuli
in our everyday language use, which would reflect the best integration procedure in word
comprehension.
Based on the everyday experiences, we hypothesized that, the phrase context would change
the evaluation of a words emotional meaning, in addition to emotional congruency effect
typically found in prior studies. Specially, words embedded in a positive context might be
biased toward a positive direction, and words embedded in a negative context might be biased
toward a negative direction. The second aim of the present study was to investigate whether
the bias effect (if present) was different in negative and positive contexts. As negative emotion
is more salient for survival and generally associates with attentional bias (Eastwood et al.
2003; Fox et al. 2001, 2002; hman et al. 2001), we speculated that the modulatory effect
of a negative context should be stronger than that of a positive context.
To answer the aforementioned questions, we utilized a dimensional rating task rather
than the more traditional categorization judgment task. In the present study, target words
were evaluated using a 9-point Likert rating scale ranging from 1 (extremely negative) to
9 (extremely positive). It is different from the categorization task adopted in prior studies,
during which participants were obligated to decide a target stimuli as either positive or negative (e.g., De Houwer et al. 2002; Fazio et al. 1986; Hermans et al. 2001; Klinger et al.
2000). The dimensional rating method allows to obtain a rating of degree of emotionality
(from strong negative to neutral and strong positive), thus is more sensitive to measure the

123

382

J Psycholinguist Res (2013) 42:379391

influence of context on evaluation of target word. Adopting this dimensional rating task, we
could measure the influence effect not only by the traditional index of processing speed (i.e.,
RT), but also by means of the precise degree to which the word portrays an emotion. In fact,
some recent studies have identified the necessity of using the dimensional rating method,
in examining the effect of emotional body gesture and background scene context on faces
(Aviezer et al. 2008; Koji and Fernandes 2010). To investigate the contextual influence of
context on word in a more ecological way, the adjective contexts were not instructed to be
ignored in the present study.

Method
Participants
A total of 31 college students (16 females, mean age of 18.6 years with a range of 1820 years)
from Beijing Normal University (Beijing, China) participated in this experiment for monetary
compensation. All were right-handed native Mandarin Chinese speakers. Written informed
consent was obtained from each participant following a research protocol approved by the
local institutional review board (IRB).
Materials and Design
Each phrase consisted of an adjective and a noun, and all adjectives and nouns were Chinese
two-character words. A total of 93 positive and 93 negative adjectives were used as context
words, and 93 nouns, including 31 positive, 31 neutral and 31 negative words, were used as
target words. The emotional valence of these words was rated by 20 additional participants
who were not participants of the formal experiment. Ratings were based on a 9-point Likert
scale where 1 represented the maximum negative valence and 9 represented the maximum
positive valence. The mean rating values were 7.03, 5.25, and 2.90 for the positive, neutral
and negative target words, respectively, and all were significantly different from each other
(all ps < 0.001). The context words were divided into three separate lists (Lists A, B and C),
corresponding to the positive, neutral and negative target words, respectively. Each list contained 31 positive and 31 negative adjectives. The context words were matched for valence
across lists (positive context: 7.44 for List A, 7.39 for List B, and 7.34 for List C; negative
context: 2.53 for List A, 2.65 for List B, and 2.45 for List C; all ps > 0.05).
There was no significant difference in the frequency of words or the number of strokes
of the characters constituting the words across positive, neutral and negative targets (all
ps > 0.05). The word frequency and number of strokes between the positive and negative
context words did not differ significantly among the three lists (all ps > 0.05).
The experiment had a 3 (context type: none, positive, negative) 3 (target valence:
positive, neutral, negative) within-subject repeated design. The no context condition, in
which words were presented individually without any context, was used as a baseline for the
evaluation of the words inherent emotional meanings. Under the positive/negative context
condition, a positive/negative adjective was followed by a target word with positive, neutral
or negative emotional tone. To form a syntactically valid phrase in Chinese, the auxiliary
word (de4) was added between an adjective and target word. Thus, all phrases consisted of
five Chinese characters. Furthermore, all phrases were semantically legal and contextually
appropriate, meaning that they could describe real objects or situations.

123

J Psycholinguist Res (2013) 42:379391

383

Table 1 Examples of stimuli


Target valence

Context
None

Positive

Negative

(Swan)

(Beautiful swan)

(Ugly swan)

(Environment)

(Graceful environment)

(Rugged environment)

(Lie)

(White lie)

(Barefaced lie)

Positive
Neutral
Negative
Target words were underlined

Phrases or single words were shown in black on a white background with the target word
underlined. Table 1 showed examples of the stimuli.
The experiment consisted of three sessions. The first session consisted of no context trials,
while the other two sessions included both positive and negative context trials. Sessions two
and three were counterbalanced across subjects, ensuring that a target word would not appear
twice in a session. All types of trials in each session were randomized.
Procedure
The experiment was conducted in a lightproof and soundproof room. A Windows PC controlled stimulus presentation and response recording using the E-prime software with millisecond timing accuracy. Participants were seated in front of the computer at a distance of
about 50 cm. All prompts and words were shown on the center of the screen, and participants
responded using the keyboard. For each trial, a + was presented for 1,000 ms as a focal
point, followed by a word or phrase. Participants were instructed to silently read the phrase
or single word firstly, and then evaluate how negative, neutral, or positive the underlined
target word appeared to them, using a 9 point Likert scale (1: strongly negative; 5: neutral; 9:
strongly positive). Ratings were made by pressing one of the 19 keys on a standard keyboard. Participants were asked to make their ratings as quickly and accurately as possible.
The time window for responding was 5,000 ms in the no context condition and 10,000 ms
in the positive/negative context conditions. The next trial started immediately following the
response or after the given time elapsed. Participants were exposed to a short practice session
before the formal sessions, and there was a 2- min short break between two sessions.
Results
Data from those trials where no response was made within the given time window, accounting
for less than 1 % of the total trials, were excluded from the analyses. Rating scores and RTs
in all nine conditions were shown in Tables 2 and 3, respectively.
A 3 (context type: none, positive, negative) 3 (target valence: positive, neutral, negative)
two-way repeated ANOVA on rating scores revealed significant main effects of context type
(F(2, 60) = 86.42, p < 0.001) and target valence (F(2, 60) = 174.30, p < 0.001), and
a significant interaction between context type and target valence (F(4, 120) = 59.09, p <
0.001). Post hoc analyses showed that there were significant context type effects for positive
(F(2, 60) = 85.16, p < 0.001), neutral (F(2, 60) = 78.19, p < 0.001) and negative
(F(2, 60) = 79.54, p < 0.001) target conditions. For all the three target valence, words in

123

384

J Psycholinguist Res (2013) 42:379391

Table 2 Mean (SD) rating scores for each condition


Target valence

Context
None

Positive

Negative

Positive

7.2 (0.9)

7.6 (1.0)

4.0 (1.6)

Neutral

5.8 (0.6)

7.3 (1.1)

4.0 (1.1)

Negative

3.3 (0.8)

5.6 (1.3)

3.0 (0.8)

Table 3 Mean (SD) RTs (in ms) for each condition


Target valence

Context
None

Positive

Negative

Positive

1514 (438)

1708 (553)

2045 (716)

Neutral

1555 (548)

1766 (582)

2071 (730)

Negative

1807 (553)

2195 (809)

2094 (753)

Fig. 1 The rating scores (a) and RTs (b) of positive, neutral and negative targets in different contexts. p values
indicate statistical significance for the comparison between different conditions. Bars represent one standard
error of the mean (SEM)

the positive context were rated significantly higher than those in the no context and negative
context (all ps < 0.01), and words in the no context were rated significantly higher than
those in the negative context (all ps < 0.01, except that there was only a trend for negative
words in the no context to be rated higher than in the negative context ( p = 0.097)), as shown
in Fig. 1a. Post hoc analyses also showed that there were significant target valence effects
in each of the three contexts: no context (F(2, 60) = 269.39, p < 0.001), positive context
(F(2, 60) = 83.95, p < 0.001) and negative context (F(2, 60) = 24.12, p < 0.001). For
all the three context type, the positive words were rated significantly higher than the neutral
and negative words (all ps < 0.01, except that in the negative context condition, the positive
and neutral words were rated similarly ( p > 0.05)), and the neutral words were rated higher
than the negative words (all ps < 0.05), as shown in Fig. 2.

123

J Psycholinguist Res (2013) 42:379391

385

Fig. 2 The rating scores of


different targets under no context,
positive context and negative
context. p values indicate
statistical significance for the
comparison between different
conditions. Bars represent one
standard error of the mean (SEM)

A 3 (context type: none, positive, negative) 3 (target valence: positive, neutral, negative) two-way repeated ANOVA on RT revealed significant main effects of context type
(F(2, 60) = 12.35, p < 0.001) and target valence (F(2, 60) = 33.48, p < 0.001), and
a significant interaction between context type and target valence (F(4, 120) = 8.47, p <
0.001). Post hoc analyses showed that there were significant context type effects for positive
(F(2, 60) = 23.79, p < 0.001), neutral (F(2, 60) = 23.13, p < 0.001) and negative
(F(2, 60) = 8.44, p < 0.001) targets. For all the three target valence, RTs in the negative
and positive context were significantly longer than those in the no context (all ps < 0.05),
and RTs in the negative context were significantly longer than those in the positive context
[all ps < 0.05, except that in the negative target condition, there was no significant difference
between RTs of negative and positive context ( p = 0.792)], as shown in Fig. 1b.
To explore the influence of positive and negative context on word rating scores, we calculated the modulatory effect of context by comparing scores in the positive/negative contexts
with scores in the no context condition. Specifically, the modulatory effect of positive context was measured by subtracting the score in the no context condition from the score in the
positive context condition. While the modulatory effect of negative context was measured by
subtracting the score in the negative context condition from that in the no context condition,
due to a decrease of rating score by the negative context.
A 2 (context type: positive, negative) 3 (target valence: positive, neutral, negative) twoway repeated ANOVA on modulatory effects of context revealed a significant main effect
of target valence (F(2, 60) = 24.02, p < 0.001), and a trend for a main effect of context
type (F(1, 30) = 3.88, p = 0.058), as well as a significant interaction between context type
and target valence (F(2, 60) = 74.87, p < 0.001). Post hoc analyses showed that there
were significant context effects for positive (F(1, 30) = 62.93, p < 0.001) and negative
(F(1, 30) = 44.17, p < 0.001) words, but not for neutral (F(1, 30) = 0.99, p = 0.33)
words. For positive words, the modulatory effect of negative context was greater than that
of positive context. While for negative words, the modulatory effect of positive context was
greater than that of negative context. Direct comparison showed that the modulatory effect
of negative context on positive words was greater than that of positive context on negative
words ( p < 0.01), while the modulatory effect of positive context on positive words was
similar to that of negative context on negative words ( p > 0.05). See Fig. 3a for details.

123

386

J Psycholinguist Res (2013) 42:379391

Fig. 3 The modulatory effect of positive and negative context on positive, neutral and negative target words
was shown, as measured by rating scores (a) or RTs (b). Bars represent 1 SEM ** p < 0.01

The modulatory effect of context could also be measured using differences in RT in the
positive/negative contexts versus no context. A 2 (context type: positive, negative) 3 (target
valence: positive, neutral, negative) two-way repeated ANOVA on modulatory effect measured by RT revealed a significant main effect of context type (F(1, 30) = 10.70, p < 0.01),
no significant main effect of target valence (F(2, 60) = 0.09, p = 0.92), and a significant
interaction between context type and target valence (F(2, 60) = 16.34, p < 0.001). Post hoc
analyses revealed significant context type effects for positive (F(1, 30) = 22.14, p < 0.001)
and neutral (F(1, 30) = 26.32, p < 0.0001) target words, but not for negative (F(1, 30) =
1.30, p = 0.26) target words. For both positive and neutral words, the modulatory effect of
negative context on RT was greater than that of positive context. But for negative words, the
modulatory effect of positive context was similar to that of negative context. See Fig. 3b for
details.
To argue against the possibility that the participants made their ratings based on the preceding adjectives rather than the target nouns in the context condition, we examined the
relationship between evaluation scores within an emotional context and evaluation scores of
the word alone under no context. A regression analysis was performed between the rating
scores in positive/negative contexts and the scores under no context. If the participants made
their ratings based on the preceding adjectives but not the target nouns, the rating scores in the
phrase conditions should bear no relationship with that under no context, since in both of the
positive and negative context conditions there are three different types of target words, i.e.,
positive, neutral and negative. However, the global linear fit of the analysis was significant
for both positive context (R 2 = 0.50, F(1, 91) = 90.95, p < 0.001) and negative context
(R 2 = 0.11, F(1, 91) = 11.17, p < 0.001), as shown in Fig. 4.

Discussion
We investigated how semantic-related emotional phrase context influenced the emotional
assessment of a word. Our results showed that the dimensional evaluation of a word was
biased by the emotional context, and the modulatory effect of negative context was greater

123

J Psycholinguist Res (2013) 42:379391

387

Fig. 4 Regression analysis between rating scores of target words in positive (a) and negative (b) contexts
with scores in no context. The global linear fit of the analysis was significant for both positive and negative
contexts (all ps < 0.01). Dashed red lines indicated linear fits with slope = 1

than that of positive context in shaping evaluation of emotional words with opposite valence.
Taken together, these data provided the first empirical evidence that the emotionality of a
phrase can change the way a word is evaluated.
The RT results in the current study were consistent with semantic context effects introduced
by Schuberth and Eimas (1977), who found that compared to a neutral condition, subjects
responded faster to a word preceded by a semantically congruent context and slower to a word
preceded by an semantically incongruent context. The present results were also consistent
with the affective priming effect (De Houwer et al. 2002, 2001; Fazio et al. 1986; Hermans
et al. 2001; Klinger et al. 2000; Wentura 2000) and emotional Flanker effect (Ochsner et al.
2008). In addition, the present results were also consistent with findings of the influence of
emotional prosody on emotional words (e.g., Schirmer and Kotz 2003), in which participants
responded faster to emotional words when expressed by an emotionally congruent prosody
than incongruent ones.
Besides the RT results as in the studies using traditional categorization task, our study
found more extensive results by adopting a dimensional rating task. Results on rating scores
showed that, positive context made the evaluation of target words bias toward positive tone,
while negative context shaped the evaluation of target words toward negative tone. This
pattern held true for words within all the three emotional categories (i.e., positive, neutral
and negative). Specially, a negative word viewed in a positive context were rated as appearing
significantly less negative or even positive to some extent than when viewed in a negative
context, a positive word viewed in a negative context were rated as appearing significantly
less positive or even negative to some extent than when viewed in a positive context (see
Fig. 1a). These results showed that despite the emotion a word is expressing, the emotional
information from the phrase context influenced the evaluation of that word. These findings
supported our first hypothesis, that emotional evaluation of a word can be changed by certain
phrase context. Many studies showed that the emotional context exerted crucial influence on
the evaluation of face expression (e.g., Aviezer et al. 2008; Koji and Fernandes 2010; Righart
and de Gelder 2008a,b; Russell and Fehr 1987). The present study extends the previous
findings in face perception by identifying that the emotional context could also influence the
emotional evaluation of word.

123

388

J Psycholinguist Res (2013) 42:379391

Researchers have suggested that top-down knowledge is used to guide visual stimuli
recognition (e.g., Bar 2004). The top-down factors consists of world knowledge, expectations, selective attention and et al. (Koji and Fernandes 2010), which can change bottom-up
processes based on information derived from the stimulus itself. Such effects may be responsible for the modulatory effect of context on word evaluation in the present study. When
we engage in reading, we do not process words or sentences in isolation. Instead, we determined the meaning of a particular word or sentence, in part, by the context in which it occurs
(Swinney and Hakes 1976). To get a contextually appropriate meaning, participants have to
integrate the target word into the context. Meanwhile, it is also necessary to invoke general
world knowledge in order to establish word meaning (Hagoort et al. 2004). As a particular
word was encountered, its inherent meaning was retrieved almost automatically. Then combinatorial processing was needed to converge on an appropriate interpretation of meaning,
in which different sources of information such as context and prior representations of the
world in memory were used. This process involves retrieving not only the meanings of target
words but also the meanings of all the words embedded within a phrase. Thus, evaluating the
emotional meaning of a word is modulated by its phrase context.
Our results also showed that, the modulatory effect was smaller when the context was
emotionally congruent with the target word (i.e., a positive word in a positive context or a
negative word in a negative context) than that when the context was emotionally incongruent
with the target word (i.e., a positive word in a negative context or a negative word in a positive
context) (see Fig. 3a). When a word occurred in an ordinary context (such as the word lie in
the phrase an evil lie), the inherent meaning of the word was consistent with the meaning of
the context, and fitted well with our general knowledge about the word, thus leading to rapid
response and less change of emotional tone (i.e., less modulatory effect). However, When a
word occurred in an unusual and emotionally incongruent context (such as the word lie in
the phrase a white lie), the original emotional tone of the word must be shifted to adapt to the
meaning and emotional tone of the phrase in which it occurs. Furthermore, our knowledge in
memory about the given objects or situations would facilitate the changing and lend the realistic basis for the adaption (e.g., we know that a white lie is not to be blamed and even worth
respect), though such objects or situations occurred rarely in the real world. In such cases,
extra efforts were required to overcome the inherent meaning of the target word, leading to
longer RTs and greater modulatory effects. In short, in positive/negative context conditions,
more or less dynamic integration or unification process was needed to form a coherent representation (Marslen-Wilson 1987), which was also demonstrated by the significantly longer
RTs in the positive/negative context conditions than in the no context condition.
An alternative mechanism that could explain the modulatory effect of emotional context
on word evaluation was expectancy. According to this interpretation, biased effects may arise
due to expectations that participants develop in response to the emotional context. Since a
positive adjective was often used to depict positive objects or situations, and a negative adjective was often used to describe negative objects or situations, when participants experienced
an emotional context, they might anticipate a congruent emotional target to follow (e.g.,
positive words should follow positive context and vice versa). When the valence of a target
word was as expected, the evaluation of that word was facilitated and the need to modulate
was decreased. Conversely, when the target word violated the expectation, evaluation of that
word would be disturbed and would be modulated to be close to the expected emotion. In
other words, the inconsistent target word would be shaped to accord to the original expectancy, thus leading to emotional bias to the neutral or even opposite valence from the inherent
emotional tone of the target word. This explanation was consistent with object recognition
studies, which found facilitated processing of visual objects that were surrounded by a context

123

J Psycholinguist Res (2013) 42:379391

389

with a high-probability occurrence with the objects, whereas processing was inhibited in an
unexpected context (Davenport and Potter 2004; Palmer 1975; see Bar 2004 for a review).
The finding that the responses to incongruent trials were different from congruent trials gave
further support for our claim that the inherent emotion of a word and context phrase both
influence the emotional evaluation of a word.
It is worth noting that positive and negative context differentially influenced word evaluation in the present study. In general, positive context biased words toward positive tone,
while negative context biased words toward negative tone. However, the modulatory effect of
a negative context on a positive word was greater than that of a positive context on a negative
word (see Fig. 3a, red line), supporting our second hypothesis. It is possible that a positive
evaluation tendency was more easily inhibited and modulated (Gross 2002; Mak et al. 2009).
On the other hand, negative information may be prone to capture attention and hold it from
disengagement (Eastwood et al. 2003; Fox et al. 2001, 2002; hman et al. 2001), making it
more immune to preceding contexts and harder to be shaped. In the current study, the asymmetry between the modulatory effect of a negative context upon a positive word and that
of a positive context upon a negative word was consistent with findings from studies about
emotional contextual influence on facial processing (e.g., Righart and de Gelder 2008a,b)
and affective flanker studies (e.g., Fenske and Eastwood 2003; Horstmann et al. 2006). For
example, Righart and de Gelder (2008a,b) found that categorization of happy facial expressions was significantly slower in fearful than happy scenes, while the interference effect did
not appear for fearful expression. In flanker studies, it was found that a positive face flanked
by negative faces induced greater interference than a negative face flanked by positive faces
(Fenske and Eastwood 2003; Horstmann et al. 2006). In short, these findings were in line with
the mechanism of negative bias (Eastwood et al. 2003; Fox et al. 2001, 2002; hman et al.
2001), reflecting the evolutionary value of negative information to humans. The negative bias
mechanism could also explain the more powerful ability of negative context to shape a word
relative to positive context, as illustrated by the greater deviation of the fitted line for negative
context from the diagonal line with slope = 1 (corresponding to evaluation of a word without
context), relative to the fitted line for positive context (see Fig. 4).
Some might argue that, because the participants saw both of the nouns and adjectives
in phrases, they may base their evaluations on the adjectives other than the target nouns.
For example, when presented with ugly swan, subjects might have rated ugly rather than
swan. Although participants may choose some strategy in response to the target, the above
mentioned situation could not be the case in the present study. Firstly, participants were
instructed to base their response on the target word explicitly in the present experiment. Secondly, supposing the participants rating the stimuli based on the preceding adjectives, there
should not be significant difference between the rating scores and RTs to different emotional
target in the same emotional type of context, which is not the case in the present results. In
fact, the evaluation scores of positive words were always higher than those of neutral words,
and the evaluation scores of neutral words were always higher than those of negative words,
under both positive and negative context, consistent with the results under no context (see
Fig. 2). One exception was that positive and neutral words received similar rating scores in
negative context, likely due to the strong modulatory effect of negative context. The much
decreased scores of positive words in a negative context might also be related to subjects
having a more negative outlook when a previously positive stimulus became negative (e.g.,
untruthful blessing). Strikingly, the regression analyses in our study showed that the rating
scores of words under positive and negative contexts significantly correlated with those of
isolated words without context, which strongly refuted the possibility that the subjects ratings
may base on the preceding adjectives rather than the target nouns. Such results also indicated

123

390

J Psycholinguist Res (2013) 42:379391

that the modulatory effect of emotional context on word evaluation was not unconditional
and was constrained by its inherent emotional meaning, identifying the relative stability of a
words emotional meaning, though it could be shaped by emotional context.
To summarize, the present study investigated how integration of a word within a phrase
influenced their ratings of emotionality. The results suggested that emotional evaluation of
a word could not be independent from emotional context. Instead, it could be modulated by
an emotional context. Moreover, the modulatory effects of positive and negative contexts
differed from each other. Furthermore, we found that the modulatory effect of context on
a word was constrained by its inherent emotional meaning. This study was one of the first
attempts to investigate the influence of emotional context on word evaluation. The findings
demonstrated the flexibility of word meaning, as well as the relative stability of meaning.
Acknowledgments This work was supported by the Youth Foundation of Humanities and Social Sciences
of Ministry of Education of China (12YJC190022 to HL), the Startup Scientific Research Foundation of the
Zhejiang Sci-Tech University (1113832-Y to HL), and the National Natural Science Foundation of China
(Grant Nos. 30700234 to ZH and 30670705 to DP).

References
Aviezer, H., Hassin, R. R., Ryan, J., Grady, C., Susskind, J., Anderson, A., & Bentin, S. (2008). Angry,
disgusted, or afraid? Studies on the malleability of emotion perception. Psychological Science, 19, 724
732.
Bar, M. (2004). Visual objects in context. Nature Review Neuroscience, 5, 617629.
Binder, K. S., & Morris, R. K. (1995). Eye movements and lexical ambiguity resolution: Effects of
prior encounter and discourse topic. Journal of Experimental Psychology: Learning, Memory, &
Cognition, 21, 11861196.
Brierley, B., Medford, N., Shaw, P., & David, A. S. (2007). Emotional memory for words: Separating
content and context. Cognition and Emotion, 21, 495521.
Davenport, J. L., & Potter, M. C. (2004). Scene consistency in object and background perception. Psychological Science, 15, 559564.
De Houwer, J., Hermans, D., & Spruyt, A. (2001). Affective priming of pronunciation responses: Effects
of target degradation. Journal of Experimental Social Psychology, 37, 8591.
De Houwer, J., Hermans, D., Rothermund, K., & Wentura, D. (2002). Affective priming of semantic
categorization responses. Cognition and Emotion, 16, 643666.
Eastwood, J. D., Smilek, D., & Merikle, P. M. (2003). Negative facial expression captures attention and
disrupts performance. Perception & Psychophysics, 65, 352358.
Eich, E. E., & Schooler, J. W. (2000). Cognitionemotion interactions. In E. E. Eich, J. F. Kihlstrom,
G. H. Bower, J. P. Forgas, & P. Niedenthal (Eds.), Cognition and emotion (pp. 329). New York: Oxford
University Press.
Erk, S., Martin, S., & Walter, H. (2005). Emotional context during encoding of neutral items modulates
brain activation not only during encoding but also during recognition. NeuroImage, 26, 829838.
Fazio, R. H., Sanbonmatsu, D. M., Powell, M. C., & Kardes, F. R. (1986). On the automatic activation
of attitudes. Journal of Personality and Social Psychology, 50, 229238.
Fenske, M. J., & Eastwood, J. D. (2003). Modulation of focused attention by faces expressing emotion:
Evidence from flanker tasks. Emotion, 3, 327343.
Fischler, I., & Bloom, P. A. (1979). Automatic and attentional processes in the effects of sentence contexts
on word recognition. Journal of Verbal Learning and Verbal Behavior, 18, l20.
Fox, E., Russo, R., & Dutton, K. (2002). Attentional bias for threat: Evidence for delayed disengagement
from emotional faces. Cognition and Emotion, 16, 355379.
Fox, E., Russo, R., Bowles, R., & Dutton, K. (2001).Do threatening stimuli draw or hold attention in
subclinical anxity? Journal of Experimental Psychology: General, 130, 681700.
Gorfein, D. S. (2001). On the consequences of meaning selection: Perspectives on resolving lexical
ambiguity. Washington: American Psychological Association.
Gross, J. J. (2002). Emotion regulation: Affective, cognitive, and social consequences. Psychophysiology, 39, 281291.

123

J Psycholinguist Res (2013) 42:379391

391

Hagoort, P., Hald, L., Bastiaansen, M., & Petersson, K. M. (2004). Integration of word meaning and world
knowledge in language comprehension. Science, 304, 438441.
Hermans, D., De Houwer, J., & Eelen, P. (2001). A time course analysis of the affective priming
effect. Cognition and Emotion, 15, 143165.
Hertel, P. T., & Parks, C. (2002). Emotional episodes facilitate word recall. Cognition and Emotion, 16, 685
694.
Heuer, F., & Reisberg, D. (1992). Emotion, arousal, and memory for detail. In S. Christianson (Ed.), The
handbook of emotion and memory: Research and theory (pp. 269288). Hillsdale, NJ: Erlbaum.
Horstmann, G., Borgstedt, K., & Heumann, M. (2006). Flanker effects with faces may depend on perceptual
as well as emotional differences. Emotion, 6, 2839.
Hunt, R. R., & McDaniel, M. A. (1993). The enigma of organization and distinctiveness. Journal of
Memory and Language, 32, 421445.
Klinger, M., Burton, P., & Pitts, S. (2000). Mechanisms of unconscious priming: I. Response competition, not spreading activation. Journal of Experimental Psychology: Learning, Memory, and
Cognition, 26, 441455.
Koji, S., & Fernandes, M. (2010). Does it matter where we meet? The role of emotional context in
evaluative first impressions. Canadian Journal of Experimental Psychology, 64, 107116.
Mak, A. K. Y., Hu, Z. G., Zhang, J. X. X., Xiao, Z., & Lee, T. M. C. (2009). Neural correlates of
regulation of positive and negative emotions: An fMRI study. Neuroscience Letters, 457, 101106.
Marslen-Wilson, W. D. (1987). Functional parallelism in spoken word-recognition. Cognition, 25, 71102.
Ochsner, K. N., Hughes, B., Robertson, E. R., Cooper, J. C., & Gabrieli, D. E. (2008). Neural systems
supporting the control of affective and cognitive conflicts. Journal of Cognitive Neuroscience, 21, 1841
1854.
hman, A., Flykt, A., & Esteves, F. (2001). Emotion drives attention: Detecting the snake in the
grass. Journal of Experimental Psychology: General, 130, 466478.
Onifer, W., & Swinney, D. A. (1981). Accessing lexical ambiguities during sentence comprehension:
Effects of frequency, meaning, and contextual bias. Memory and Cognition, 9, 225236.
Palmer, S. E. (1975). The effects of contextual scenes on the identification of objects. Memory and
Cognition, 3, 519526.
Rayner, K., & Frazier, L. (1989). Selection mechanisms in reading lexically ambiguous words. Journal
of Experimental Psychology: Learning, Memory, and Cognition, 15, 779790.
Righart, R., & de Gelder, B. (2008a). Rapid influence of emotional scenes on encoding of facial expressions:
An ERP study. Social Cognitive & Affective Neuroscience, 3, 270278.
Righart, R., & de Gelder, B. (2008b). Recognition of facial expressions is influenced by emotional scene
gist. Cognitive Affective Behavior Neuroscience, 8, 264272.
Russell, J. A., & Fehr, B. (1987). Relativity in the perception of emotion in facial expressions. Journal
of Experimental Psychology: General, 116, 223237.
Schirmer, A., & Kotz, S. A. (2003). ERP evidence for a sex-specific Stroop effect in emotional speech. Journal
of Cognitive Neuroscience, 15, 11351148.
Schuberth, R. E., & Eimas, P. D. (1977). Effects of context on the classification of words and nonwords. Journal of Experimental Psychology: Human Perception and Performance, 3, 2736.
Stanovich, K. E., & West, R. F. (1979). Mechanisms of sentence context effects in reading: Automatic
activation and conscious attention. Memory and Cognition, 7, 7785.
Swinney, D. A., & Hakes, D. T. (1976). Effects of prior context upon lexical access during sentence
comprehension. Journal of Verbal Learning and Verbal Behavior, 15, 681689.
Wentura, D. (2000). Dissociative affective and associative priming effects in the lexical decision task:
Responding with yes vs. no to word targets reveals evaluative judgment tendencies. Journal of
Experimental Psychology: Learning, Memory, and Cognition, 26, 456469.
Winograd, E., & Neisser, U. (1992). Affect and accuracy in recall: Studies of flashbulb memories.
Cambridge, MA: Harvard University Press.
Zempleni, M. Z., Renken, R., Hoeks, J. C., Hoogduin, J. M., & Stowe, L. A. (2007). Semantic ambiguity
processing in sentence context: Evidence from event-related fMRI. Neuroimage, 34, 12701279.

123

Copyright of Journal of Psycholinguistic Research is the property of Springer Science &


Business Media B.V. and its content may not be copied or emailed to multiple sites or posted
to a listserv without the copyright holder's express written permission. However, users may
print, download, or email articles for individual use.

You might also like