Chapter 2 Literature Review
2.4 Summary
This chapter first discussed some studies on handedness and its interaction with hemispheric lateralization and then introduced the body-specificity hypothesis before reviewing literature on emotion word processing. To begin with, Section 2.1 discussed manifest handedness as well as familial handedness and their interaction with cerebral lateralization of language. In other words, studies reviewed in this section confirmed the modulation effect of handedness when individuals process language. Section 2.2 introduced the body-specificity hypothesis (Casasanto, 2009), which proposed that individuals with different dominant hands might form different mental representations in that they interact with the physical environment consistently in a different fashion. Specifically, the studies indicated that individuals not only preferred the objects shown at their dominant side of space but also associated positive ideas with their dominant side of space. Aside from the left-right space preference, this section also discussed the top-down associations with positive and negative valence common in the metaphorical expressions (Lakoff & Johnson, 1980, 1999).
Finally Section 2.3 reviewed research on emotion word processing, which exhibited a RH
14
advantage in right-handed subjects and also reported different behavioral results with depressed and non-depressed participants.
15
Chapter Three
Methods
This chapter illustrates the methods of the current experiment. First of all, the characteristics of the target participants and how they were divided into groups are described in Section 3.1. Experimental materials are introduced in Section 3.2. Then Section 3.3 illustrates how the experiment proceeded. Lastly, how the data were analyzed is depicted in Section 3.4.
3.1 Participants
This study recruited 37 left-handed and 68 right-handed participants, aged 20 to 40 (mean age = 27, 32 males). All of them were native speakers of Mandarin Chinese. All participants were required to have normal or corrected-to-normal vision for the experiment.
Participants with neurological or psychiatric disorders were not considered.
All of the subjects were verified by the Edinburgh handedness inventory (Oldfield, 1971, see Appendix A). Right-handers were further divided into either having familial sinistrals (FS+) or not (FS-) based on whether any of their blood relatives (including parents, siblings, grandparents, uncles, aunts or cousins) was left-handed: the participants who had at least one left-handed family member were categorized as the FS+ group, and the participants who reported no known left-handed family members were in the FS- group.
Aside from the Edinburgh handedness inventory (Oldfield, 1971), the participants also completed the Beck Depression Inventory-second edition in Chinese (BDI-II, see
16
Appendix B) to measure the level of depression in the participants since previous research has shown that people who suffered from depression responded to the negative verbal stimuli in the LVF differently from people without depression (Atchley et al, 2003; Walsh et al, 2010). BDI-II is a self-report questionnaire based on Beck et al. (1961). It has 21 questions of 4-point Likert scale (0 to 3), with the total score between 0 to 13 indicating minimal depression, 14 to 19 indicating mild depression, 20 to 28 indicating moderate depression, and 29 to 63 implying a severe depression. The questionnaire has the test-retest reliability at 0.93 (Beck et al., 1961). Its internal reliability (Cronbach’s alpha) for clinical patients is 0.92 and 0.93 for college students, which suggests good reliability and validity in patient samples as well as in college students.
Participants also filled out the adopted Chinese version (Zan, 1986) of Interpersonal Reactivity Index (IRI, see Appendix C) to measure both their cognitive and emotional components of empathy. This is because empathy, by its definition, is sharing the emotion of others. It is proposed in the previous studies that emotion embodiment occurs when one reads an emotional word. There has also been literature indicating the correlation of emotion and reading speed. Thus for better subject control, a questionnaire on empathy was also included in the study. IRI is a questionnaire designed by Davis (1980, 1983). It has 28 items on a 5-point Likert scale ranging from “Describes me very well” to “Not describing me at all.” IRI has good intra-scale and test-retest reliability, and convergent validity is indicated by correlations with other established empathy scales (Davis, 1980).
All subjects signed the consent forms prior to the experiment, and they were paid for the participation after the completion of the experiment.
17
3.2 Materials
The materials were 180 two-character Chinese emotion words varied with valence:
either positive, neutral or negative, with each valence category containing 60 words. Some examples are provided in Table 1 below. No fillers were added because of the employment of neutral words, which served as a cushion against fatigue effect from viewing all the emotion words.
Table 1. Example words of the experimental materials Valence Example Word
Negative 貪心、邪惡、骯髒、世仇、小偷
Neutral 翻閱、遷徙、列舉、電池、人民
Positive 漂亮、激勵、珍藏、冠軍、恩惠
The wordlist of the present study was created by the following steps. First, 160 words of each valence condition (i.e., positive, neutral, negative) were chosen (480 in total) from the 750-word list of a former study on Chinese emotion word processing by Ku and Chan (2014).
In their study, the materials were selected from the research by Zhuo et al. (2013) as well as from the Chinese Linguistic Inquiry and Word Count Dictionary by Huang et al. (2012).
Based on these two resources, Ku and Chan (2014) preliminarily had 1040 words. After screening for repetition, word frequency, lexical categories, the frequency distribution of their lexical categories and also orthographic neighborhood sizes based on Chinese Word Sketch Engine1, the words were then manually inspected of homographs, in order to delete word candidates with multiple senses. Subsequent to the deletion of homographs, the remaining 750 words were rated online against their valence, arousal and imageability.
1 Chinese Word Sketch Engine (http://wordsketch.ling.sinica.edu.tw/) is a Corpus Query System incorporating concordance, word sketches, grammatical relations, and a distributional thesaurus based on Chinese Gigaword Corpus and Sinica Corpus 5.0 and maintained by Lexical Computing Ltd. in Academia Sinica.
18
Based on the rating data, the present study selected 480 words with the following two criteria: (1) Valence rating (on a scale of -3 to 3): high ratings of positive valence for positive words; ratings between -1 to 1 for neutral words; high negative ratings for negative words, and (2) Arousal rating: high arousal for positive and negative words, low arousal for neutral words. The first criterion was to maximize the valence difference between positive, neutral, and negative words. The second criterion was to control the arousal level among positive and negative words, so that the arousal of words would not be a confounding factor for the participants to interpret the valence of emotion words.
After the 480 words were selected, the second step of the creation of the wordlist was to control the valence and arousal differences among the three conditions. To achieve this, the words underwent an online rating with a 7-point Likert scale on two variables -- valence (-3 to 3 from negative to positive) and arousal (0 to 7 from low to high arousing). Four questionnaires were created, each containing 240 words (80 words on each valence condition). Two questionnaires asked the raters to evaluate the words by valence, the other two questionnaires, by arousal. The four questionnaires were respectively rated by 85, 39, 35, 37 native speakers of Chinese, aged 20 to 40, who did not participate in the formal experiment. The subject number of the first questionnaire was much larger than the latter three because the 240 words in the first questionnaire did not serve as a pool large enough for word selection. Thus one more questionnaire with 240 new words was then provided. To control the arousal levels of the experimental materials, two more questionnaires were also added online for arousal ratings.
Finally, 180 words (118 verbs, 62 nouns) were selected as the experimental materials.
To be more specific, 60 words rated higher than 1.5 (from -3 to 3) were chosen for the positive words, 60 words rated between -0.4 to 0.4 for the neutral words, and 60 words rated lower than -1.5 for the negative words. The arousal levels of the words were also carefully
19
matched during the word selection process. A three-way repeated measures ANOVA ensured the difference of the valence ratings between all the three valence conditions, F(2,118)=65307.285, p<.001 (mean score: 1.84 for positive words, 0.03 for neutral words, -1.79 for negative words). As for the arousal levels, statistical results of a three-way repeated measures ANOVA showed significant main effect of arousal ratings, F(2,118)=238.555, p<.001 (mean score: 4.22 for positive words, 2.59 for neutral words, 4.26 for negative words).
Follow-up pairwise comparisons further revealed that the arousal level of the neutral word category was significantly lower than the other two categories (p’s<.001) while the positive and negative categories did not differ significantly (p=.124). Table 2 summarizes the rating results of the experimental materials. Table 3 summarizes their statistical results, while a complete list of the words is detailed in Appendix D.
Table 2. Summary of the valence and arousal rating means and SDs (in parenthesis) for the positive, neutral and negative conditions of the experimental materials
Positive Words Neutral Words Negative Words Valence (from -3 to 3) 1.84 (0.03) 0.03 (0.02) -1.79 (0.02) Arousal (from 0 to 7) 4.22 (0.31) 2.59 (0.74) 4.26 (0.41)
Table 3. Summary of the statistical results for valence and arousal ratings among the positive, neutral and negative conditions of the experimental materials
Variable Pair t value Pairwise p value
Valence Positive, Neutral t(59)= -172.53 .000**
Positive, Negative t(59)= -348.194 .000**
Neutral, Negative t(59)= -198.91 .000**
Arousal Positive, Neutral t(59)= -1.43 .000**
Positive, Negative t(59)= .531 1.79 Neutral, Negative t(59)= 15.93 .000**
Note: p-values were Bonferroni-corrected. **= <.001
20
3.3 Procedure
The experiment was carried out in a sound attenuated room in the Neurolinguistics Lab at National Taiwan Normal University. After signing the written informed consent, participants were seated in a chair facing a computer screen. They were instructed to rest their chin on a chinrest positioned 68 cm from the computer screen, so the distance to the screen was maintained throughout the experiment. To familiarize themselves with the procedure, the participants were required to experience 24 practice trials before they started the 180 target trials.
When the experiment started, a white fixation cross (a plus sign “+”) appeared in the center of the screen with black background. The subjects were instructed to fixate their eye gaze on the central cross throughout the experiment. After 1000ms, a target two-character word appeared in either the LVF or the RVF for 185ms. The degree of visual angle to the inner edge of the laterally displayed stimulus was 2°. After 185ms, the target word was replaced by three pound signs (###) at the same location to mask the word to prevent the participant from moving his/her eyes away from the fixation point, while at the same time the participants had to make a response. They were asked to indicate whether the valence of the word was positive, neutral, or negative by key-pressing on the response box. The participants were instructed to put his/her index finger, middle finger and ring finger of the right/left hand on the respective buttons of the response box (fingers and hands counterbalanced across participants and groups). Response buttons were also counterbalanced: the index finger and the ring finger were switched for indicating “positive” or “negative”. Participants were required to respond within 2500 ms after target onset. If there was no response within the time frame, the trial was recorded as incorrect, and the screen automatically moved onto the next trial. The participants were instructed to respond as quickly and accurately as possible.
Please see Figure 2 for the experimental procedure of stimulus presentation. The total time
21
for the experiment lasted 10 to 15 minutes with two short breaks in between.
Figure 2. Procedure of stimulus presentation
Each trial began with a central fixation (+) and the subjects were randomly presented with a two-character word either in the RVF or LVF. The stimulus was presented for 185ms then covered by a pattern of three pound signs (###). Participants pressed the key to indicate if they thought the word was positive, neutral or negative within 2500ms (from stimulus onset).
3.4 Data Acquisition and Analysis
The E-Prime 2.0 software from Psychology Software Tools, Inc. was used in the Windows 7 environment on a PC to present the stimuli and to record the behavioral data of the participants (i.e., reaction time and response accuracy). The SPSS software version 21 (IBM Corporation) was used to analyze the collected data.
Data with accuracy rate lower than two standard deviations were excluded, and only trials with correct responses were analyzed. The collected RTs and accuracy rates were respectively analyzed with a three-way repeated measures ANOVA with three independent
22
variables: group (left-handed, FS+, FS-), valence (positive, neutral, negative), and visual field (LVF, RVF). Post-hoc pairwise comparisons were carried out and Bonferroni corrected when needed. In addition, the p-values were adjusted with the Greenhouse-Geisser correction (Greenhouse & Geisser, 1959) when the Mauchly's test of Sphericity was violated.
Importantly, planned comparisons, including paired t-tests on valence under two visual field conditions and paired t-tests on visual field under three valence conditions for each of the subject groups, were carried out to test the body-specificity hypothesis.
23
Chapter Four
Results
The current study recruited 104 participants. Five subjects’ data were excluded due to technical problems or unexpected health issues of the participants, such as unilateral near-sightedness. Then the data of the remaining 99 participants (33 FS-, 35 FS+, 34 Left-handed) were further subjected to elimination based on the following criteria -- (1) mean accuracy rates lower than 2 SDs, (2) for the left-handed group only, the score (0-10) of the Edinburgh left-handedness inventory lower than five, and (3) the BDI-II depression score equal to or higher than 20 (indicating a moderate depression or above). After the screening, the data of 84 subjects ( 27 FS-, 30 FS+, 27 Left-handed) remained. In order to keep the number consistent across groups, three subjects’ data in the FS+ group were randomly excluded, which left 81 subjects in total, with 27 subjects in each group.
4.1 Accuracy Rates
Table 4 summarizes the mean accuracy rates under the six conditions (3 valence levels x 2 visual field levels) for each group. A three-way repeated measures ANOVA with the factors of group (FS-, FS+, Left-handed), valence (positive, neutral, negative) and visual field (LVF, RVF) was conducted. The results revealed a main effect of valence, F(1.268, 98.934) = 46.940, p<.001. Follow-up pairwise comparisons showed that neutral words had a significantly higher accuracy rate (91.79 ± 6.81%) than the other two categories of words (positive: 85.92 ± 11.04%; negative: 85.86 ± 11.83%), while the other two categories did not
24
significantly differ from each other (p = 1.00). In addition, there was also a main effect of visual field (F(1, 78)=42.055, p<.001), with higher accuracy in the RVF. The average RVF accuracy rate was 91.33 ± 7.24%, while the LVF accuracy rate was 84.39 ± 12.00% (mean difference: 6.94%). As for the between-subject factor, the test results suggested no significant effect of group. On average, the left-handed group had a slightly higher accuracy (88.21%) than the FS- group (87.73%), and the FS- group performed slightly better than the FS+ group (87.65%), though the difference was not significant, F(2,78)=.097, p=.908.
Table 4. Summary of the accuracy rates (Mean and SD) in each experimental condition across groups
Variables
Valence x Visual field
Negative Words Neutral Words Positive Words
LVF RVF LVF RVF LVF RVF
Group FS- 0.81 observed, F(2,156)=21.342, p<.001. To better understand the simple effects of valence and visual field respectively, six follow-up paired t-tests were conducted to further investigate the pairwise relations, the results of which are illustrated in Table 5. As shown in Table 5, only the pair of positive/neutral stimuli and the pair of neutral/negative words in the LVF showed significant accuracy differences. Finally, except for the valence x visual field interaction, no other interactions reached significance: valence x group, F(4,156)=.140, p=.911; visual field x group, F(2, 78)=.052, p= .949; valence x visual field x group, F(4,156)=.138, p=.968.
25
Table 5. The summary of the follow-up paired t-tests on accuracy rates after a valence x visual field interaction
Visual Field Valence t p-value
LVF Positive/Neutral t(80)= -6.67 .000**
Positive/Negative t(80)= -0.71 2.89
Neutral/Negative t(80)= 6.26 .000**
RVF Positive/Neutral t(80)= -0.34 4.39
Positive/Negative t(80)= 1.04 1.80
Neutral/Negative t(80)= 0.82 2.49
Note: p-values were Bonferroni-corrected.
**= <.001
4.2 Reaction Times
Table 6 summarizes the RTs under the six conditions (3 valence levels x 2 visual field levels) for each of the group. A three-way repeated measures ANOVA with the factors of group, valence and visual field was conducted. The results revealed a main effect of valence, F(2,156)=149.019, p<.001. Pairwise comparisons indicated that the three levels of valence all distinguished themselves from one another on a significant level (p’s <.001). To be specific, the RTs of positive words were the shortest while those of neutral words were the longest.
The RTs of positive words (608.35ms) were approximately 50ms shorter than those of negative words (658.00ms), and the RTs of negative words were roughly 120ms shorter than those of the neutral words (772.17ms). Moreover, although there was no significant valence x visual field interaction, planned analyses revealed that different valence levels distinguished from each other in both visual fields (p’s <.001), with positive words soliciting shortest RTs while neutral words longest RTs.
As for the visual field factor, similar to the results of the accuracy rates, a main effect of visual field was observed, F(1,78)=5.667, p<.05. Words appeared in the RVF/LH were
26
generally processed 12ms faster than words in the LVF/RH.
Table 6. Summary of the mean RTs and SDs (ms) in each experimental condition across groups
Variables
Valence x visual field
Mean RT (SD) Negative Words Neutral Words Positive Words
LVF RVF LVF RVF LVF RVF
group. Although the effect of group did not reach significance, F(2,156)=.902, p=.410, this
“left-handed < FS- < FS+ group” pattern could be observed across valence conditions and also across visual fields, as is shown in Tables 6 and 7. Table 7 shows the RT difference collapsed by valence or visual field. Please note that the arrangement of the groups from top to down is by the longest RTs to the shortest ones, for the ease of observation. Therefore, the FS+ group is placed at the top row and the left-handed group at the bottom row. As can be seen from Tables 6 and 7, the “left-handed < FS- < FS+ group” pattern on RTs held across every condition.
Although valence x visual field interaction was significant with the accuracy data, no such interaction was observed with RTs (F(2,156)=1.226, p= .296). Also, no other interactions reached significance: valence x group, F(4,156)=.204, p= .936; visual field x group, F(2,78)=.743, p= .479; valence x visual field x group, F(4,156)=.484, p= .748.
27
Table 7. Summary of the mean RTs and SDs(ms) collapsed by valence or visual field
Valence VF
4.3 Analysis of the Headedness of the Materials
It is not clear if the lack of the effect of the body-specificity hypothesis was due to the difference of the headedness distribution of valence in the bisyllabic materials. In the current study, the headedness of the materials was not controlled because the employed bisyllabic words were considered a unit. It is thus not clear if the preference for the dominant side, if any, was obscured by the uncontrolled headedness of valence within the words. Therefore, we analyzed the headedness distribution of the words in the three valence conditions and the results are summarized in Table 8 below.
Table 8. The distribution of headedness among the three valence categories
Valence\Headedness left-headed right-headed double-headed compositional
Positive 10 (樂觀) 15 (中獎) 23 (幸福) 12 (天堂)
Neutral 4 (發呆) 2 (打牌) 51 (填寫) 3 (傻笑)
Negative 24 (吵架) 8 (世仇) 26 (偷竊) 2 (自大)
Note: Examples are provided in the parenthesis.
As can be seen from Table 8, the valence of the word could be derived from the left (e.g. 樂觀) or the right (e.g. 中獎) character of the word, which is termed left- or
right-28
headed in the current study. Moreover, the valence could also come from both characters (e.g.
幸福), termed double-headed here. In addition, it could also be that neither of the two characters carried the word-level valence. Rather, the valence was derived from the composition of the two characters when they formed a word (e.g. 天堂). Although for both positive and negative conditions, valence was contributed by both characters in more than 20 instances, the rest of the instances showed an inconsistent pattern, with more right-headed words in the positive condition and more left-headed words in the negative one. Also, positive words exhibited a greater number of compositionality than neutral or negative words.
As for neutral condition, more than two-thirds of the instances were double-headed (e.g. 巷 道).
29
Chapter Five
Discussion
This study aimed to investigate the hemispheric lateralization of emotion word
This study aimed to investigate the hemispheric lateralization of emotion word