Abstract
The personality-descriptive adjectives used in a famous Chinese novel of the 18th century, A Dream of Red Mansions, which is thought to broadly reflect Chinese culture, might help depict personality structure. Four hundred ninety-three personality-descriptive adjectives from the first 80 chapters of the novel were administered to 732 Chinese university students. After factor analyses, the one- to seven-factor solutions were extracted, and the five-factor one was relatively clearer. The five factors of personality titled Wicked, Intelligent, Amiable, Conscientious, and Frank, were intercorrelated. Men scored higher on Wicked and Conscientious but lower on Amiable compared with women. As a preliminary trial, our study demonstrates that personality-descriptive adjectives in a famous Chinese novel characterize the personality structure.
Introduction
A novel is an extended work in prose, either fictitious or partly so, dealing with character, action, thought, and so on, particularly in the form of a story, which offers a rich source for cultural studies. Therefore, not only arts and literature specialists but also social scientists, psychologists, and psychiatrists all take an interest in this cultural vector. Personality psychologists have noted the terms used in novels, which are trait-specific (Passakos & De Raad, 2009). For many years, lexical approaches have been used in studies of personality structure (De Raad et al., 2014; Galton, 1884; Goldberg, 1993). Factor analyses of self- and peer-ratings on the personality-descriptive terms of a given language (e.g., adjectives, nouns, and verbs) have revealed that personality structure involves two (Saucier et al., 2014), three (De Raad et al., 2010), five (Saucier & Goldberg, 1996), six (Ashton et al., 2004), or seven (Saucier, 2003) factors. For instance, using English adjectives, some scholars have demonstrated a five-factor personality structure (Goldberg, 1990): Extraversion, Agreeableness, Conscientiousness, Emotional Stability, and Intellect, or Imagination. A six-factor personality model has also been proposed, which covers an additional factor Honesty-Humility, besides the aforementioned five or their variants (Lee & Ashton, 2008; Saucier, 2009). Meanwhile, some investigators claim that there might be only two (Social Self-Regulation and Dynamism; Saucier et al., 2014) or three factors (Extraversion, Agreeableness, and Conscientiousness; De Raad et al., 2010, 2014), which are cross-culturally consistent.
Psycholexical studies have also been conducted in contemporary Chinese society but have yielded less convergent findings. K. Yang and Bond (1990) derived five factors: Social Orientation-Self-Centeredness, Competence-Impotence, Expressiveness-Conservation, Self-Control-Impulsiveness, and Optimism-Neuroticism. S. Yu, Wei, et al. (2009) also disclosed a five-factor structure: Intelligent, Emotional, Conscientious, Unsocial, and Agreeable. However, X. Zhou, Saucier, Gao, and Liu (2009) identified a seven-factor structure in mainland China: Extraversion, Conscientiousness/Diligence, Un-selfishness, Negative Valence, Emotional Volatility, Intellect/Positive Valence, and Dependency/Fragility. One reason for the discrepancy might be the different term pools used in different studies, which cover different aspects of Chinese culture.
Interpreting literary works was thought helpful to achieve a full understanding of personality traits and their facets, and many literature texts that are handed down across time could provide rich, trait-relevant information of different eras (Passakos & De Raad, 2009). Scholars have also studied ancient Chinese documents to learn more about personality descriptors. For instance, B. Yang (2005) examined the ancient Chinese personality terms from ShiJi (Records of the Grand Historian) and found four factors: Ren (benevolence, mercifulness or graciousness), Zhi (intelligence or wisdom), Yong (courage), and Yin (being a hermit). These results are similar to the contemporary personality descriptions to some extent (e.g., Ren likely represents Agreeableness, Zhi represents Intellect, Yong represents Emotional stability, and Yin represents Introversion). However, ShiJi is a biography of heroes or personages, rather than of ordinary people. One might wonder what the personality structures of normal people look like in the past. As the character, action, and thought of normal people are often narrated in a literary story, we would like to identify the possible personality structure in a Chinese novel.
A suitable candidate is A Dream of Red Mansions, which was written by Xueqin Cao (a pen name? See Cao, 2006, and an interesting web link) in the 18th century during the Qing Dynasty. The novel, also translated as Story of the Stone, Dream of the Red Mansion, or Dream of the Red Chamber, is one of China’s Four Great Classical Novels. “Xueqin Cao” was thought to have written it in 1740, and had completed its first 80 chapters (in some versions, Chapters 79 and 80 were not clearly separated but rather compiled together as chapter 79; in our study, we split the contents into Chapters 79 and 80) and possibly left behind a draft of the remaining chapters before his death in 1763 or 1764 (Chang & Owen, 2008). A Dream of Red Mansions is valued as a compendium of Chinese culture and is even regarded as a historical document (A. C. Yu, 1997). It is also used for studying the politics (Zhao, 2011), economics (Chan, Lew, & Tong, 2001), cultures (Levy, 1999; Liao, 2007), religions (A. C. Yu, 1989; Z. Zhou, 2001), social customs (Cooper & Zhang, 1993), mental disorders (Levy, 1994), and other details of the 18th-century China. Stone (2005) claimed that if one wants to understand modern-day China, one must read A Dream of Red Mansions. The novel touches on the lives of over 400 characters from virtually every class and profession, including maids-in-waiting, stewards, gardeners, cooks, nuns, actors, officials, members of the imperial family, gamblers, thieves, and so on (Y. Gao, 2009). Since its emergence, the novel has become very popular (Stone, 2005), suggesting that the descriptions inside the novel are wildly held and concerned. Indeed, the language styles (or terms) used in it are rich in number, and in depictions of the distinct and vivid characters and the realistic life-scenes. Moreover, Wu (2008) has shown that the narrations in the novel contain a rich source of adjectives, verbs, and motion-event descriptions throughout.
One might ask whether there is a clear picture of personality structure, which is depicted by the adjectives used in A Dream of Red Mansions. With this question in mind, we specifically designed the present study to decipher the possible implicit structures of personality by examining the trait-descriptive adjectives in the novel. Inspired by contemporary efforts to understand the ancient personality description (reviewed in Millon, 2012), we employed a group of well-educated matriculating university students to evaluate the personality-related adjectives used in the novel.
Method
Seven of our co-authors (all native Chinese speakers; 4 women and 3 men; 2 PhD holders, 4 PhD and 1 MSc candidates in clinical psychology) served as judges. They independently examined the 80-chapter version of A Dream of Red Mansions to search for the person-descriptive adjectives. The latter chapters of other versions were neglected for their distinctions from the preceding 80 chapters in writing style. In principle, the judges used questions such as, “What kind of person is he/she? He/she is [adjective]” and “What do you think of this person? He/she is [adjective]” to help determine whether an adjective referred to states or traits. The resulting 557 adjectives were compiled and approved by three authors of this article. An adjective would be removed from the list if only one author voted that it was personality descriptive. Four hundred ninety-three adjectives were retained to the word pool. Because A Dream of Red Mansions was written about 300 years ago, it contains certain adjectives that are difficult for contemporary people to comprehend. For these adjectives, a short explanation was attached immediately after the word.
Seven hundred thirty-two Chinese university students (474 women, mean age 20.42 years with 1.46 SD, range = 17-39 years; 258 men, 20.71 ± 1.48 SD, range = 18-32 years) majoring in the Arts, Education, Foreign Languages, Mechanics, and Medicine, were invited to participate in the present study. The 493 adjectives were used to construct a 5-point rating scale: 1 = very unlike me, 2 = moderately unlike me, 3 = somewhat like and unlike me, 4 = moderately like me, 5 = very like me. All participants were found to be free of and had no history of somatic or psychiatric illnesses. Participants completed the study anonymously, which was approved by a local Ethics Committee. All participants provided their written informed consent.
When referring to the data distribution, a bipolar-scattered construct may easily (or misleadingly) emerge as two unipolar factors due to the influence of individual differences in the overall elevation of responses. Therefore, the ipsatization, which is a partialing procedure, is often used to avoid the erroneous interpretation (ten Berge, 1999) and has been used in several personality studies (e.g., Ashton et al., 2004; Boies, Lee, Ashton, Pascal, & Nicol, 2001; Saucier & Goldberg, 1996). In our study, the self-rating data from all participants were ipsatized first. Data were standardized within each participant across all terms to control for response style. The method also fits into our present study where some adjectives are of negative valence. Afterward, the data were submitted to the principal component analysis, and factor loadings were rotated orthogonally using the varimax normalized method. Once factor solutions were obtained, top adjectives with highest loadings on each target factor and without significant cross-loadings on other factors were selected for further analyses. Based on the top adjectives for each factor (scale), the internal reliability (Cronbach’s alpha coefficient) of each scale was calculated. Mean scale scores in women and men were also calculated, and their differences were submitted to a two-way ANOVA plus a post hoc analysis by Duncan’s multiple new range test. Pearson’s correlation was used to search for possible relations among these scale scores.
Results
The principal component analysis of the answers to the 493 items disclosed 20 factors with eigenvalues greater than 3.0. The first seven factors accounted for 8.9%, 4.5%, 2.9%, 2.2%, 1.9%, 1.4%, and 1.3% of the variance, respectively (Figure 1).
Scree plot of eigenvalues obtained from factor analysis of 493 Chinese terms.
The one-factor structure contrasted graceful and elegant (and other favorable attributes) with detestable and stupid (and other unfavorable attributes). In the two-factor solution, one factor contrasted kind, polite, and law-abiding with flippant, impudent, and black-hearted, which were related to propriety or impropriety. The other factor contrasted clever and resourceful, sensible and clever with cowardly, stupid-looking, incapable and useless, which were intelligence-related. With the top 10 items of each factor for calculating factor scores, the correlations between the only factor in one-factor solution and the two factors in two-factor solution were .87 and .40, respectively.
In the three-factor solution, the first two factors remained similar to those in the two-factor solution; the second factor contrasted clever and resourceful, sensible and clever with cowardly, stupid-looking, and incapable and useless, which was comparable to the extravert and intelligent traits. The third factor covered officious and officious and spiteful, and adulterous and fierce, which could be labeled as a negative valence. With the top 10 items of each factor for calculating factor scores, the correlations between the two factors in the two-factor solution and the three factors in the three-factor solution were between .87 (the factor one in the two-factor solution and the factor one in the three-factor solution) and .01 (the factor one in the two-factor solution and the factor two in the three-factor solution).
In the four-factor solution, the negative valence factor and the intelligence-related factor were retained. The third factor was defined most strongly by tactless, uncultivated and refractory, careless, refined and elegant, cautious, and careful, which characterized a Conscientious trait. The fourth factor covered suspicious, insolent, and haughty versus honest, magnanimous and broad-minded, and lenient and gentle, which were related to the Agreeable or Sociable traits. With the top 10 items of each factor for calculating factor scores, the correlations between the three factors in the three-factor solution and the four factors in the four-factor solution were between −.85 (the factor two in the three-factor solution and the factor one in the four-factor solution) and −.11 (the factor three in the three-factor solution and the factor four in the four-factor solution).
In the five-factor solution, factor structures were clearer (term examples are shown in Table 1). The first factor was defined most strongly by terms such as officious and spiteful, adulterous and fierce, shameless and wantonly, black-hearted, and sinister and vicious versus propriety-abiding (factor loading, −.36), devout (−.34), and prudent (−.34). The last three terms loaded less heavily on the opposite pole and were not shown in Table 1. The factor was named as Wicked after terms with high loadings. When scrutinizing the individual answers, we found a low-mean-response pattern for the factor. Most participants (about 77.12%) chose to define themselves as 1 (“very unlike me”), whereas far fewer defined themselves as 2 (about 13.83%, “moderately unlike me”), 3 (about 3.43%, “somewhat like and unlike me”), 4 (about 1.47%, “moderately like me”), or 5 (about 0.32%, “very like me”).
Loadings of the Top 20 Adjectives (Chinese Plus English Translation) on Each Factor in 732 Participants.
Note. A word might appear more than once in a chapter; adjectives with a cross-loading (absolute value) higher than .30 were omitted.
Loadings larger than .35 are in bold for clarity.
The second factor in the five-factor solution was defined by terms such as clever and resourceful, smart, intelligent, sharp-witted, and talented versus stupid-looking, inept and tongue-tied, inarticulate, incapable and useless, stereotyped and pedantic, and stupid and clumsy. This factor was then named Intelligent. The third factor was highly loaded by terms such as amiable, kind, lenient and gentle, easy-going, obedient and considerate, and virtuous and educated, and on the other pole by haughty, highly ambitious, insolent, eccentric, extremely frivolous, wayward, and highly strung. It was named Amiable. The fourth factor clearly represented Conscientious, was highly connected with careful, cautious, prudent, painstaking, meticulous, steady, and elegant versus tactless, mischievous, naughty, careless, crazy and lunatic, and stubborn and roguish. Some terms, such as refined and elegant, civilized and well-bred, and uncultivated and refractory, implied that this factor was education-related. The fifth factor, named Frank, was loaded with bold, plucky, stern, firm, frank, straightforward, quick and efficient, and absolutely fearless versus old-womanish, sticky, fussy, timid and weak, cowardly, and oversensitive. Adjectives loaded on this factor (Frank) were loaded on the intellect-related factor in the two- or three-factor solution. With the top 10 items of each factor for calculating factor scores, the correlations between the four factors in the four-factor solution and the five factors in the five-factor solution were between .97 (the Factor 2 in the four-factor solution and Factor 1 [Wicked] in the five-factor solution) and .09 (Factor 3 in the four-factor solution and Factor 5 [Frank] in the five-factor solution).
In the six-factor solution, the three factors—Wicked, Intelligent, and Conscientious—identified in the five-factor solution were retained. The fourth factor was loaded with suspicious, insolent, and haughty versus honest. The fifth factor was loaded with wayward, eccentric, and unsociable. The sixth factor was loaded with softly and flatteringly speaking, spoiled, crying, and other terms related to the perceived femininity, which might be considered a dependent or fragile trait. With the top 10 items of each factor for calculating factor scores, correlations between the five factors in the five-factor solution and the six factors in the six-factor solution were between .97 (Factor 1 [Wicked] in the five-factor solution and Factor 2 in the six-factor solution) and .05 (Factor 2 [Intelligent] in the five-factor solution and Factor 6 in the six-factor solution).
In the seven-factor solution, factors identified in the five-factor solution were all retained. The sixth factor was loaded with reticent, solitary and uncompromising, and demure versus talkative, sweet-spoken, and smooth-spoken, which might be related to an Introvert trait. The seventh factor was loaded with wayward, bossy, and tyrannical versus broadly caring, which might be related to the opposite side of the Agreeable trait. With the top nine items (some factors had only nine highly loaded items left in the round) of each factor for calculating factor scores, the correlations between the six factors in the six-factor solution and the seven factors in the seven-factor solution were between .98 (Factor 2 in the six-factor solution and Factor 1 in the seven-factor solution) and .09 (Factor 6 in the six-factor solution and Factor 5 in the seven-factor solution).
When scrutinizing the scree plot of the eigenvalues, a clear level-off after the fifth factor appeared (see Figure 1). We selected the five-factor model for further analyses as it was much clearer and understandable than either of the six or the seven factor solutions. Table 1 illustrates the top 20 adjectives with highest loadings (absolute values) on the target factor (scale) and cross-loadings (absolute values) below .30 on other non-target factors.
Based on the top 20 adjectives, the internal reliability of each scale was calculated (Table 2). Two-way ANOVA detected a significant gender effect, F(1, 730) = 4.15, MSE = 182.8, p < .05, on these scale scores. Men scored higher on Wicked (p < .01, 95% confidence interval [2.20, 4.90]) and Conscientious (p < .05, 95% confidence interval [0.50, 3.70]), and lower on Amiable (p < .01, 95% confidence interval [−5.16, −2.24]) than women did (Table 2). Moreover, in all participants (n = 732), some scales were significantly intercorrelated. For example, Wicked was negatively correlated with Amiable (r = −.45), whereas Intelligent was positively correlated with Frank (r = .53; Table 3).
Internal Alphas of the Five Factors and Their Respective Gender Scores (M ± SD) Based on the Top 20 Adjectives for Each Factor.
Note. *p < .05 vs. men.
Intercorrelations Between the Five Factors Based on the Top 20 Adjectives for Each Factor (n = 732).
Note. All factors were significantly correlated with each other except Frank and Conscientious (p < .05).
Discussion
We selected 493 personality-descriptive adjectives from the novel A Dream of Red Mansions, which was written 300 years ago, and identified different factor-solution models. The one- and two-factor solutions were similar to the ones reported earlier (De Raad et al., 2010; X. Zhou et al., 2009), and the three- and four-factor solutions raised factors that were similar to the ones reported previously (Saucier, 2009; X. Zhou et al., 2009). In the six-factor solution, the factors Wicked, Intelligent, and Conscientious were similar to those reported in previous studies (Saucier, 2009; X. Zhou et al., 2009). The sixth factor was comparable with the dependency/fragility trait as reported (X. Zhou et al., 2009), and the other two factors in the solution reflected the agreeableness trait. However, our six-factor solution did not fit well into the six-factor model proposed by Saucier (2009). Our seven-factor solution also did not fit well into the previously reported seven-factor model, although the results of our seven-factor solution were comparable with those of Saucier (2003).
The results we obtained show that the factor development was progressive, and that the five-factor solution presents a relatively clearer picture of personality structure. The first factor, Wicked, which suffered from the response style, was most likely an artificial structure. Nevertheless, it characterized a small number of our participants, which might be its representation of social unacceptability, corresponding to the negative pole of agreeableness (Peabody & De Raad, 2002; Saucier & Goldberg, 1996) and of Ren, as noted in ShiJi (B. Yang, 2005). It also appears as the Negative Valence, (Dis) Honesty (Ashton, Lee, & Son, 2000; Saucier, 2009; X. Zhou et al., 2009), or Integrity Values (Szirmák & De Raad, 1994). Highly evaluative terms, such as wicked and black-hearted, manifest patterns discernible in the behavior, thinking, or emotional expression (Saucier, 2009; X. Zhou et al., 2009). The negative attributes were important because they provide the threshold between acceptable and unacceptable, and between socially tolerable and intolerable behavior (Saucier, 2003). As East Asians are believed to show more avoidance motivation (Heine & Buchtel, 2009), the factor Wicked might reflect people’s concern for their own safety by keeping them away from dangerous people. The negative correlation between Wicked and Amiable found in our study was also in line with this reasoning. Men scored significantly higher on the Wicked factor than women did in our present study, as was the case in a previous study of anti-social personality traits in China (R. Yu, Yu, et al., 2009).
The second factor, Intelligent, which represents talent, clever, and reflective, corresponds to the Intellect (Saucier & Goldberg, 1996; X. Zhou et al., 2009), Intelligent (S. Yu, Wei, et al., 2009), and Zhi (B. Yang, 2005) reported previously. Intelligence is suggested as a trait dimension that affects nearly every aspect of an individual’s life (Costa & McCrae, 1994). This factor corresponds to the wisdom, which is a highly valued attribute, and might be the foundation of other virtues (Thompson, 2007). Interestingly, terms relating to communication skills also loaded on this factor. Sharing feelings with each other through language is one of the primary functions of the Chinese community; it helps to maintain existing personal relationships, to reinforce role and status differences, and to preserve harmony within the group (G. Gao, 1998). Previous results have shown that Intelligent is a highly treasured trait among Chinese people (S. Yu, Wei, et al., 2009; X. Zhou et al., 2009).
The third factor, Amiable, corresponded to Agreeableness (Saucier & Goldberg, 1996) and Agreeable (S. Yu, Wei, et al., 2009). McCrae and Costa (1989) showed that the interpersonal circumplex was defined by the two dimensions of Extraversion and Agreeableness. Marušić, Kamenov, and Jelić (2011) found that avoidance in friendships was also characterized by these two personality traits. In a collectivistic culture (Earley, 1989; Kwan, 2009; Tjosvold, Wu, & Chen, 2010), self is not an independent entity and not complete by itself, and is relational and defined by the surrounding personal encounters (G. Gao, 1998). It is thus understandable that an Amiable trait is connected with agreeable and extravert tendencies. Consistent with that, women scored higher than men on Extraversion (Feingold, 1994) and Agreeableness (Costa, Terracciano, & McCrae, 2001). In our study, women scored higher than men on Amiable.
The fourth factor, Conscientious, representing careful, steady, and solemn attributes, corresponds to Conscientiousness (Saucier & Goldberg, 1996; X. Zhou et al., 2009) and to Conscientious (S. Yu, Wei, et al., 2009). Terms such as refined and elegant, and civilized and well-bred were also loaded on this factor. This outcome is in line with the practicing Ren, which refers to self-restraint and self-discipline on a personal level (G. Chen & Chung, 1994). The related disciplines also help to explain the correlation between Amiable and Conscientious found in our study. Moreover, as breadwinners in the earlier, male-dominated society (Chia, Moore, Lam, Chuang, & Cheng, 1994; Marshall, 2008), Chinese men had to take more social and familial responsibilities, which required them to be more conscientious. The fifth factor, Frank, representing bold, stern, and fearless characteristics, corresponds generally to Yong as noted in ShiJi (B. Yang, 2005) and to an Emotional stability trait as previously illustrated (X. Zhou et al., 2009). The trait is closely related to courage, which guarantees the practicing Ren together with wisdom (L. Chen, 2010). These connections might also explain the correlation between Frank and Intelligent found in our study.
Nevertheless, as this was a preliminary study, one should bear in mind some limitations of our design. First, the adjective pool was taken only from A Dream of Red Mansions, a novel written about 300 years ago, rather than from a genuine lexical-design. It captures some terms but ignores others. Therefore, it remains unknown how much the writing style of “Xueqin Cao” influenced the personality description. The novel contains a great variety of characters and character descriptions that took place in everyday life during the 18th century, but the adjectives used might not be exhaustive of that era. A novel describes characters on stage only, without offering a full blown vocabulary of trait terms. Second, we did not include terms from other novels as controls, nor other independent measures. Studies on other literary works of that time period might serve as a wonderful comparison for future study. Third, our participants were not asked to rate a contemporarily used personality questionnaire. Therefore, it is unknown how many personality traits identified in the present study are cross-cultural. Furthermore, the factor Wicked, which displayed a low-mean-response property, might also suffer from this third response style limitation, despite having a satisfactory internal reliability (.93).
Overall, our results show that this design study was effective. Personality factors identified in A Dream of Red Mansions were comparable with those reported in the literature. This implies that an excellent novel such as this one might help us better understand the structure of personality and the worldview in our contemporary lives, and better understand the personality disorders that exist in society. Future study should elucidate the importance of this trait on modern life and mentality. Similarly, attention should be paid to the Intelligent trait, which might play a significant role in modern psychiatry. For instance, the concept of wisdom in ancient India has helped in the development of psychotherapies in modern Indian society (Jeste & Vahia, 2008). Amiable covers both Agreeableness and Extraversion, and focuses on interpersonal relationships. Frank is also a mixture of Extraversion and Emotional Stability. Based on these results, personality psychologists might pay more attention to the role of a different worldview and emphasize personal relationships in Chinese personality traits.
Footnotes
Authors’ Note
J. Zhu and W. Chen contributed equally to the work described in the study.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research and/or authorship of this article: The study was supported by a grant from the Natural Science Foundation of China (No. 91132715) to Dr. W. Wang.