The authors have declared that no competing interests exist.
Conceived and designed the experiments: TG MC DP. Performed the experiments: MC. Analyzed the data: TG MC. Contributed reagents/materials/analysis tools: TG MC. Wrote the paper: TG MC DP.
This study examined emotional modulation of word processing, showing that the recognition potential (RP), an ERP index of word recognition, could be modulated by different emotional states. In the experiment, participants were instructed to compete with pseudo-competitors, and via manipulation of the outcome of this competition, they were situated in neutral, highly positive, slightly positive, highly negative or slightly negative emotional states. They were subsequently asked to judge whether the referent of a word following a series of meaningless character segmentations was an animal or not. The emotional induction task and the word recognition task were alternated. Results showed that 1) compared with the neutral emotion condition, the peak latency of the RP under different emotional states was earlier and its mean amplitude was smaller, 2) there was no significant difference between RPs elicited under positive and negative emotional states in either the mean amplitude or latency, and 3) the RP was not affected by different degrees of positive emotional states. However, compared to slightly negative emotional states, the mean amplitude of the RP was smaller and its latency was shorter in highly negative emotional states over the left hemisphere but not over the right hemisphere. The results suggest that emotional states influence word processing.
Emotion plays an important role in our everyday lives and interacts with our cognition
Emotional words contain both conceptual and emotional meanings, making these words particularly useful for research on emotional modulation of language processing. Previous studies using behavioral measures have shown that participants more accurately recognize negative words compared to positive and neutral words in a word recognition task, indicating a facilitation effect of negative emotional information on word processing
While these studies imply an effect of emotional words on word processing itself, other studies have also found that emotional states or backgrounds influence language processing at large. These studies first induced emotional states in participants with non-linguistic methods and subsequently employed a linguistic task to examine the effects of emotion on processing. For instance, Olafson and Richard (2001)
According to the literature, evidence for emotional modulation on language processing has been found in studies employing both linguistic and non-linguistic emotional stimuli. ERP studies with emotional words have consistently found a facilitation effect of emotional words, especially negative words, compared to neutral words
In addition to the well-known N400 ERP component related to semantic integration
To induce different emotional states, a method commonly used in previous studies is the presentation of emotional pictures
This study was approved by the Institutional Review Board of the Imaging Center for Brain Research of Beijing Normal University. All participants gave informed, written consent prior to the experiment, and were paid for their participation.
Twenty-five native Chinese speakers (12 females) aged from 17 to 25 years participated in the experiment. All participants were right-handed, with normal or corrected-to-normal vision. No one reported neurological disorders. The data of three participants were excluded because of artifacts. The data of an additional 3 participants were excluded because they failed to monitor the results of their pseudo-competitor in the emotion induction task. The data of 19 participants were included for final analyses.
There were three types of stimuli in the word recognition task: non-animal words, animal words, and word segmentations. All stimuli consisted of two Chinese characters. The non-animal words were the test stimuli whose ERPs were of interest, but to which participants were not required to respond. There were in total 66 non-animal words. The animal words (target stimuli) were used to provide subjects with an active word processing task (i.e., to maintain their attention). As such, accuracy rates for these words were interpreted as a measure of attention. We were not interested in brain responses to the animal words. There were 42 animal words in total, and participants were required to press a button when an animal word was presented. The background stimuli were composed of the fragments of another 132 non-animal words, which were matched in visual attributes with Chinese characters.
All stimuli were randomly presented in the center of a 19-inch screen with 24-point font size. Participants were seated in a quiet room at a distance of approximately 80 cm from the computer screen. Characters were 1.43° high by 1.43° wide. Participants were instructed to avoid eye blinks and body movements during the experiment. The order of stimulus presentation was programmed and controlled by the E-Prime Software (Psychology Software Tools). The refresh rate of the computer screen was 60 Hz.
Prior to beginning the experiment, participants were asked to read the experimental instructions carefully. The practice session started after participants indicated that they completely understood the experimental procedure. As illustrated in
The emotional induction task and the word recognition task were alternated. In this example, the first 4 stimuli after the fixation in Task 2 are word fragments, and the fifth is a word.
This task was used to induce 5 distinct emotional states in participants. Participants were told to compete with another participant, and that they would be paid 50–100 yuan (about 7–15 USDs) if they outperformed their competitor.
At the beginning of this task, a left arrow and a right arrow were displayed on the screen. Participants were instructed to choose one of the arrows, which resulted in either the gain or loss of points. Subsequent to their choice, the results of the competition were displayed for 2000 ms; the participant's score appeared on the left and the pseudo-competitor's score appeared on the right. In the first part of the experiment, we aimed to induce a neutral emotional state (prior to the induction of positive or negative states so as to avoid possible contamination of the neutral state). Each participant was informed in advance that they would be unable to view their point totals and that the system was recording them automatically. Thus, the points presented on the screen after each trial were “0, 0” (i.e., 0 points for the participant and 0 for the pseudo-competitor), where a score of 0 did not mean 0 points were awarded. This was intended to introduce uncertainty as to their real scores, and may further have prevented participants from counting the total points they received. As such, we may exclude the effect of the total points and attribute the evoked emotional states to the points participants received on each trial. In the second part, the points given to the participant and pseudo-competitor were presented on the screen and were expected to induce 4 different degrees of emotional states: 1) +1000,−1000 (highly positive emotion): a participant received 1000 points while the pseudo-competitor lost 1000 points, which was expected to induce highly positive emotion; 2) +1000,+1000 (slightly positive emotion): both the participant and the pseudo-competitor won a 1000-points, which was expected to induce slightly positive emotion, as the participants increased their point total without expanding the difference between their score and that of the pseudo-competitor; 3) −1000,+1000 (highly negative emotion): The participant lost 1000 points while the pseudo-competitor won 1000 points in this condition, which was expected to induce highly negative emotion; 4) −1000,−1000 (slightly negative emotion): Both the participant and pseudo-competitor lost 1000 points in this condition.
The emotional induction task was conducted 36 times for each of the five conditions (180 times in total). The same results were never displayed more than 4 times in a row.
Stimuli were presented using the rapid stream stimulation procedure
Every non-animal word was repeated 5 times so that the same words were used for every type of emotional state. The numbers of trials for each experimental condition was equal, and presented in pseudorandom order.
Participants alternatively performed the emotional induction task and the word recognition task. They were given four short breaks during the experiment. After the entire experiment, the participants were asked to rate the degree of their emotional state when they saw their results. This rating was completed on a scale spanning from −5 to +5, with negative numbers indicating more negative emotion and positive numbers indicating more positive emotion. Absolute numerical values indicated the intensity of the emotional states. For example, −5 stands for very disappointed, 0 for neutral, +5 for very happy. At debriefing, all participants indicated that they believed they had been competing with another person during the emotional induction task.
Electroencephalographic (EEG) data were recorded using the Scan 4.3 package (NeuroScan, Inc.). Brain electrical activity was recorded from 64 scalp sites using Ag/Agcl electrodes with the left mastoid as the reference. The data were algebraically re-referenced off-line according to the common averaged reference method, which has been proved to be the best procedure to obtain the RP
The EEG data were segmented into 1000 ms epochs, including a 100 ms baseline before the onset of non-animal words under each emotional state. Epochs with artifacts exceeding ±65 μv were automatically rejected.
The RP peak latencies were measured during a 160∼417 ms interval following stimulus onset. A commonly used way to obtain the RP is to subtract ERPs elicited by background stimuli from those elicited by words (which reduces the presence of driving rhythms generated by rapid stream stimulation
Repeated measure ANOVAs with the aim of comparing the activity evoked by non-animal words under all 5 kinds of emotion states were performed on a selected sample of 10 electrodes (P1/P2, P5/P6, PO3/PO4, PO7/PO8, O1/O2), which showed a clear RP. These ANOVAs included three factors: emotional state (5 levels: neutral, highly positive, slightly positive, highly negative, slightly negative), hemisphere (2 levels: left, right), and electrode position (5 levels). The Greenhouse-Geisser correction was applied when appropriate. According to previous literature on the RP
Participants' self-rating scores of their emotional states are illustrated in
Points | +1000, −1000 | +1000, +1000 | 0, 0 | −1000, −1000 | −1000, +1000 |
Emotional State | Highly Positive | Slightly Positive | Neutral | Slightly Negative | Highly Negative |
Ratings | 4.29 (0.00) | 2.08 (0.71) | 0.05 (0.00) | −2.10 (0.71) | −3.87 (0.00) |
Performance on the animal words in the word recognition task showed that participants were able to perform the task successfully and intensively, with 98.3% correct responses under the neutral emotional state, 98.1% under the highly positive emotional state, 98.4% under the slightly positive emotional state, 97.9% under the highly negative emotional state, and 98.6% under the slightly negative emotional state.
The bottom panel shows topographic maps for difference waves between neutral and each of the other four emotional states during the time window of 250–280 ms after stimulus onset. A clear recognition potential (RP) over the occipital scalp can be indentified under the neutral (NEU), highly positive (HP), slightly positive (SP), highly negative (HN) and slightly negative (SN) emotional states, which is highest in the case of the neutral emotional state.
PO7 | PO8 | |||
Amplitude (μV) | Latency (ms) | Amplitude (μV) | Latency (ms) | |
Neutral | −6.64 (2.56) | 276 (20.56) | −6.35 (2.46) | 264 (13.45) |
Highly positive | −5.67 (2.45) | 266 (19.57) | −5.68 (2.52) | 257 (12.92) |
Slightly positive | −5.59 (2.47) | 269 (21.92) | −5.55 (2.19) | 257 (16.84) |
Highly negative | −5.35 (2.42) | 261 (20.24) | −5.53 (2.16) | 256 (12.66) |
Slightly negative | −5.74 (2.60) | 268 (16.76) | −5.42 (2.26) | 255 (18.90) |
According to a three-way repeated measure ANOVA performed on the mean amplitudes, there was a significant main effect of emotional state [
As for the peak latencies, results revealed a significant main effect of hemisphere [
Consistent with previous studies
For the mean amplitudes of the RP over the PO7 electrode, the orthogonal contrasts revealed significant differences between the neutral and other emotional states [
For the peak latencies, similar results were obtained. At the PO7 electrode, the orthogonal contrast revealed a significant difference between the peak latencies of the RP elicited under the neutral and other emotional states [
To investigate if and how word processing is modulated by emotional states, the present study first induced emotional states by making participants compete with a pseudo-participant, and then required them to complete a word recognition task. The results demonstrated that the RP, a good index of early word processing, is sensitive to an individual's emotional states. The results are discussed below.
In the present experiment, a clear-cut RP was found in the word recognition task. This further suggests that the RP can be elicited by recognizable stimuli. Consistent with the findings of Hinojosa et al. (2001)
More interestingly, compared with the neutral emotional state, word recognition under the positive and negative emotional states elicited a smaller RP over each hemisphere. The difference mainly distributed over the occipital scalp. A reduced RP reflects less cognitive resource is required to process concrete words relative to abstract words
In addition, a significant effect of emotion was observed on the RP peak latency. Specifically, a shorter peak latency of the RP was found under emotional states. The reduced latency reflects a faster word processing
In our daily experience, cognitive processing is relatively more efficient under a peaceful emotional state than under an intense emotional state. Indeed, emotional background with high arousal was found to impede word processing but that with low arousal was not
The results of the present study showed that, over the left hemisphere, significant differences were observed in the mean amplitude and peak latency of the RP elicited under different degrees of negative emotional states. Specifically, a smaller RP with earlier peak latency was elicited under the highly negative emotional state over the left hemisphere than that under the low negative emotional state. In other words, under the highly negative emotional state, word processing was faster and required fewer attentional resource over the left hemisphere. No such differences were found over the right hemisphere. As mentioned earlier, the right lingual gyrus and fusiform gyrus may be recruited earlier for sub-word level processing, followed by the left lingual gyrus and fusiform gyrus for word level processing. This finding of an asymmetry between the two hemispheres under different degrees of negative emotional states suggests that the early stage of word level processing is more sensitive to negative emotional states.
In terms of positive emotional states, no difference was found in either the mean amplitude or peak latency of the RP under different degrees of emotional intensity. That is, positive emotional states of different intensity had similar effects on word recognition. However, it is curious that over the left hemisphere, different degrees of negative emotion had different effects on word processing, while positive emotions did not. There are two possible explanations for this effect. First, even though in general there may be no difference between the effect of the positive and negative emotions on word processing, the early stage of word processing may be more sensitive to negative emotions. Some previous studies also showed that negative emotional information
Furthermore, no difference was observed between RPs under positive and negative emotional states, which indicates that positive and negative emotions have similar effects on word processing. This is inconsistent with the findings in several previous studies. Some of these studies found that negative emotion has a greater effect on word processing than positive emotion
To summarize, the influence of emotional states on word processing manifests as a smaller RP with an earlier peak latency. In addition, positive and negative emotional states have similar effects on word processing, but positive and negative emotions of differing degree have different effects. The effects of positive emotions of varying degree on word processing are similar, but compared to low negative emotion, highly negative emotion facilitates word processing over the left hemisphere.
The authors would like to thank Profs. Jinghan Wei and Yan Song for their helpful comments on the experimental design, Dr. Jingjing Guo for her assistance with data analysis. We also thank Ji Zhou, Brian Dillon, Tim Poepsel and Benjamin Zinszer for proofreading.