Validation of the Curiosity and Exploration Inventory–II (CEI–II) Among Chinese University Students in Hong Kong

This study aimed at validating the Curiosity and Exploration Inventory–II (CEI–II; Kashdan et al., 2009) in a Chinese context. A total of 294 Chinese first-year undergraduate students in Hong Kong completed the CEI–II and measures of satisfaction with university life, the Big Five personality traits, and human values. The results of exploratory structural equation modeling, parallel analysis, and confirmatory factor analysis supported a 1-factor solution and did not replicate the original 2-factor structure. Time invariance of the 1-factor structure was obtained among 242 participants who completed the questionnaires again after 4 months. The latent means and correlation indicated that curiosity as measured by the CEI–II was quite stable over the period of investigation. The CEI–II was found to be positively correlated with satisfaction with university life, extraversion, agreeableness, conscientiousness, openness to experience, and openness to change values, but negatively with neuroticism and conservation values. The results of hierarchical multiple regression analyses showed that the CEI–II score had incremental validity above and beyond the Big Five personality traits in predicting human values and satisfaction with university life.

Curiosity is an essential element and a key motive for both animal and human behaviors, "essential to the survival not only of the individual but of the species" (Bruner, [7], p. 115). It also plays an important role in helping individuals adapt to an ever-changing environment (Loewenstein, [36]) and maintain optimal psychological well-being (Jovanovic & Brdaric, [26]). Throughout an individual's life span, curiosity can facilitate the development of secure attachments and self-identity (Kashdan et al., [27]) and contribute to desirable outcomes including global life satisfaction, work satisfaction, living a pleasurable life, living an engaging life, and living a meaningful life (Silvia & Kashdan, [46]).

Based on previous empirical studies, researchers have developed various tools for measuring curiosity (e.g., Collins, Litman, & Spielberger, [13]; Kashdan et al., [28]; Kashdan, Rose, & Fincham, [30]; Litman & Silvia, [33]; Litman & Spielberger, [34]), among which is the Curiosity and Exploration Inventory (CEI) designed by Kashdan et al. [30]. The CEI was designed to measure two dimensions: the exploration (diversive) and absorption (specific) aspects of curiosity. Exploration refers to curiosity toward novel and challenging stimuli (e.g., new people, objects, events, or ideas), with an aim to integrate experience and knowledge. Absorption refers to the ability to regulate one's attention and become immersed in novel and challenging situations. Initial data provided support for the psychometric properties of the CEI by showing satisfactory internal consistency, test–retest reliability, interpretable factor structure, consistency between informant and self-ratings, and both convergent and discriminant validity (Kashdan et al., [30]). Evidence has also shown that social desirability has less influence on the CEI scores than it does on other measures of positive qualities (Kashdan et al., [30]). Subsequent research has further revealed that the CEI's construct validity is largely maintained regardless of the settings under investigation (Kashdan & Steger, [31]; Silvia, [44], [45]).

Despite the aforementioned merits, several limitations of the CEI have been found (Kashdan et al., [28]). First, partly due to a negatively worded item, the absorption subscale did not perform as well as the exploration subscale in terms of internal consistency and the correlations with related variables. Second, the CEI failed to capture the breadth of the construct. In particular, the scale did not measure the willingness to manage or embrace the tension arising from the contact with novelty and uncertainty. To address these limitations, Kashdan and colleagues ([28]) modified the original CEI and developed the CEI–II, which includes two subscales: stretching and embracing. Stretching refers to individual motivation to seek knowledge and new experiences, whereas embracing focuses on the willingness to enjoy new, uncertain, and unpredictable events in everyday life. The original absorption subscale was dropped, as it contributed little to the psychometric properties of the scale and could be adequately represented by the stretching and embracing subscales. Empirical research among college students demonstrated satisfactory psychometric properties of the CEI–II (Kashdan et al., [28]). In particular, the CEI–II and its subscales were found to be reliable (αs =.75–.85), and sufficiently brief to reduce respondent burden. Moreover, the validity of the stretching and embracing dimensions was established by their correlations with various measures of emotion, personality, and well-being. The significance and valence of the correlations were generally comparable, but the stretching subscale was found to have higher correlations than the embracing subscale, except for two variables: mindful awareness and extraversion.

Although the distinction between stretching and embracing seems conceptually meaningful, several issues have raised the question of whether it is necessary to empirically differentiate these two dimensions in the CEI–II. First, the high correlations between the two subscales (rs =.79–.85; Kashdan et al., [28]) suggested a substantial overlap between the two concepts, limiting their usefulness and uniqueness in both research and practical settings. Second, as stretching and embracing are used to measure the motivation and the willingness to approach and enjoy novel experiences, it is unlikely they will be independent from each other. Psychologically well-adjusted individuals are typically consistent in their behaviors and attitudes so as to minimize the stress from cognitive dissonance (Festinger, [16]). People who are motivated to seek out new knowledge and experiences are usually willing to embrace novel and uncertain situations, and vice versa. Third, because of the conceptual similarity and intrinsic connection between the two concepts, it is difficult to operationalize them separately in the scale. As a result, an item intended for one subscale could look perfectly suitable for measuring the other subscale. This problem can contribute to undesirable cross-loadings of the factor structure. For instance, in recent validation work done by Acun, Kapıkıran, and Kabasakal [1], an embracing item was found to load on the stretching factor. Finally, past research has commonly used the total score of the CEI–II without differentiating the two subscales (e.g., Connelly, [14]; Kashdan, McKnight, Fincham, & Rose, [29]). Even among the few studies that calculated and analyzed the two subscale scores separately (e.g., Jovanovic & Brdaric, [26]; Kashdan et al., [28]), the correlations of these scores with other measures were quite similar in terms of their significance and valence.

To address the preceding issues, further research can be conducted to extend the current knowledge about curiosity in three ways. First, it would be helpful to further examine the structures and functions of trait curiosity in a different cultural context. As shown in the literature review, despite the significant development of theories and research on curiosity, most previous work has been done in Western cultures (e.g., Jovanovic & Brdaric, [26]; Kashdan et al., [28]; Litman & Silvia, [33]). Although various measurement models and associations of curiosity with other constructs have been reported in the literature, the extent to which these findings can be generalized to other cultures has not been well understood. For instance, Chinese culture has been characterized as collectivism, with an emphasis on obedience, conformity to social norms, interdependence with significant others, and respect for authority (Hsu, [22]). Whereas some researchers believe that being curious and autonomous are universally beneficial, and therefore the strict and controlling parenting styles and the great emphasis on academic success among the Chinese are likely to hinder children's development (Barber, Stolz, & Olsen, [3]), others have argued that being curious and autonomous is not an important developmental task in Chinese culture compared with conventional achievement goals, and hence curiosity and autonomy are not necessarily associated with positive outcomes (Greenfield, Keller, Fuligni, & Maynard, [19]). Given the unsettled debate and inconsistent findings, it would be worthwhile to examine how curiosity is related to other variables in a Chinese context using the new CEI–II.

Second, little is known about the stability of trait curiosity in the current literature. It would be interesting to examine whether the mean level and factor structure of the CEI–II remain invariant across time. An increasing number of longitudinal studies have been published in recent years because they can assess intraindividual change over time and provide stronger evidence for inferring causal relationships among two or more constructs (Little, [35]). In longitudinal research, the factorial invariance of multi-item measures is an important concern, because it can ensure that an instrument assesses the same underlying construct on each occasion and warrants the comparisons of latent factor means across different time points (Little, [35]). Therefore, it is necessary to evaluate whether the CEI–II has a stable factor structure over time and if this measure is suitable for investigating mean level changes in longitudinal studies.

Third, to further establish the validity of trait curiosity, it is important to expand the nomological network that demonstrates its associations with various related constructs. Aside from personality and well-being, key constructs such as human values have not been studied in the previous validation work. According to Schwartz [41], human values are a set of fundamental goals that are regarded as important in people's life, including self-direction, stimulation, hedonism, achievement, power, security, conformity, tradition, benevolence, and universalism. These values can be further grouped into four higher order values, arranged along two orthogonal dimensions: openness to change versus conservation and self-enhancement versus self-transcendence. Recently, Schwartz et al. [42] refined value theory and the circular motivational structure of human values. Along one dimension of the new framework, openness to change and self-enhancement pertain to personal focus values, whereas self-transcendence and conservation pertain to social focus values. Along the other dimension, openness to change and self-transcendence express anxiety-free motivations and emphasize self-expansion and growth, whereas self-enhancement and conservation are based on anxiety avoidance and emphasize self-protection against threats. These values are found to be widely applicable in different cultures and are associated with various attitudes, opinions, behaviors, and personality traits. As both curiosity and human values are psychological constructs that are related to motivation and goal setting, it is meaningful to examine how these two constructs are related to or different from each other.

In sum, this study has three major aims. First, the study was to test the factor structure and mean levels of the CEI–II using exploratory structural equation modeling (ESEM) and confirmatory factor analysis (CFA) among Chinese university students. Based on previous findings about the CEI (Kashdan et al., [30]; Kashdan & Yuen, [32]), it was hypothesized that the mean level of the CEI–II among Hong Kong students would be lower than that of Kashdan et al. [28]. Second, this study was to investigate the longitudinal factorial invariance of the CEI–II. Because the CEI–II measures trait curiosity, it was predicted that the factor structure would remain stable over time. Third, this study sought to validate the CEI–II by investigating its relationships with related constructs (i.e., life satisfaction, personality traits, and human values) in a Chinese context. In light of previous findings on the beneficial effects of curiosity (e.g., Jovanovic & Brdaric, [26]; Silvia & Kashdan, [46]), we hypothesized that curiosity would be positively correlated with satisfaction with university life, a specific domain of life satisfaction relevant to university students. Also, based on past research on the associations between curiosity and personality traits (e.g., Kashdan et al., [28]; Litman & Silvia, [33]), we hypothesized that curiosity would be positively associated with extraversion, openness to experience, conscientiousness, and openness to change, but negatively associated with neuroticism. As for the relationship between curiosity and human values, we predicted that curiosity would be positively related to openness to change and negatively related to conservation, because highly curious people are characterized as having an open and embracing attitude toward uncertainty and ambiguous stimuli in daily life (Kashdan et al., [28]). Besides, as prior research has supported the incremental validity of trait curiosity over personality traits (e.g., Kashdan et al., [27]), we predicted that curiosity would predict human values and university life satisfaction over and above the Big Five personality traits.

A sample of 294 Hong Kong Chinese first-year undergraduate students (76 male, 218 female) was recruited from six colleges in a local university (business, liberal arts and social sciences, science and engineering, creative media, law, and energy and environment). Their ages ranged from 16 to 21 years (M = 18.04, SD =.68). The same group of students was invited to participate in the study again 4 months after Time 1. Among them, 242 students (59 male, 183 female) completed the questionnaires at Time 2.

All of the first-year undergraduate students from a university in Hong Kong were invited to participate in this study via electronic mail, in which a link to the online survey was provided. All links were unique and personalized such that only invited participants had access to the survey. The content of the survey included instructions and measures of curiosity, satisfaction with university life, personality traits, human values, and demographic characteristics. As the focus of this study, the curiosity measure was presented before the other scales to minimize potential order effects. All of these measures were originally developed in English, and were translated into traditional Chinese for this study by a translator and back-translated by a second translator to ensure accuracy and equivalence, according to the International Test Commission test translation and adaptation guidelines (Hambleton, [20]). The participants were invited to complete the measure of curiosity again 4 months after the first survey. Participation was voluntary, and each participant received a cash coupon on completion of the survey at both time points.

The CEI–II (Kashdan et al., [28]) contains 10 items, with 5 items measuring the stretching dimension (e.g., "I actively seek as much information as I can in new situations.") and 5 items measuring the embracing dimension (e.g., "I am the type of person who really enjoys the uncertainty of everyday life."). All items are rated on a 7-point Likert scale ranging from 1 (strongly disagree) to 7 (strongly agree).

The Satisfaction with Life Scale (SWLS; Diener, Emmons, Larsen, & Griffin, [15]) was revised to measure the participants' satisfaction with university life, with the aim of making the measure more specific and relevant to the sample. For instance, the original item, "In most ways my life is close to my ideal," was revised to read "In most ways my university life is close to my ideal." The scale includes 5 items assessed on a 7-point Likert scale ranging from 1 (strongly disagree) to 7 (strongly agree).

The Big Five Inventory (BFI; John, Naumann, & Soto, [24]) consists of 44 items measuring five dimensions of personality traits, including Extraversion (8 items), Agreeableness (9 items), Conscientiousness (9 items), Neuroticism (8 items), and Openness to Experience (10 items). All items are evaluated on a 5-point Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree). Sample items include "talkative," "tends to find fault with others," and "does a thorough job."

Human values were measured by the European Social Survey (ESS) version of the Portrait Values Questionnaire (Schwartz et al., [43]). This measure contains 21 items tapping 10 basic human values (i.e., self-direction, stimulation, hedonism, achievement, power, security, conformity, tradition, benevolence, and universalism). A sample self-direction item is, "Thinking up new ideas and being creative is important to him. He likes to do things in his own original way." All items are rated on a 6-point Likert scale ranging from 1 (very like me) to 6 (not like me at all). The 10 first-order values were further grouped into 4 higher order values (i.e., openness to change, conservation, self-enhancement, and self-transcendence). As suggested by Schwartz et al. [43], the scores of these four values were centered around the mean score to correct for individual differences in response scale.

ESEM was used to examine the original two-factor structure of the CEI–II. This method has the advantages of both exploratory and confirmatory factor analysis, as it allows flexible model specification (e.g., cross-loadings, error covariance, etc.) and provides model fit evaluations (Asparouhov & Muthén, [2]). In case the results from ESEM did not support the original two-factor structure, parallel analysis (Horn, [21]) would be applied to examine the number of underlying factors. The parallel analysis utilizes O'Connor's ([39]) syntax to estimate the distribution of the eigenvalues based on random permutations of the raw data, which can preserve the distribution of the original data and provide more accurate estimation when the normality of the raw data is not met (Longman & Holden, [37]). In this situation, the eigenvalues of the raw data were estimated and compared against the eigenvalues obtained from random data. A factor can be retained if its eigenvalue is larger than its corresponding mean and 95th percentile eigenvalues of the random data. Based on the findings from the parallel analysis, competing models can be specified to examine the structure of the scale using CFA. The time invariance of the best model will then be tested with the longitudinal data.

As for model estimation, the selection of estimation method was based on the multivariate normality of the data. Maximum likelihood (ML) estimation would be adopted if the data followed multivariate normal distribution, and robust maximum likelihood (RML) estimation would be used if the assumption was violated. Goodness of fit of the model was assessed with several fit indexes, including root mean square error of approximation (RMSEA; Steiger & Lind, [47]), standardized root mean residual (SRMR; Jöreskog & Sörbom, [25]), nonnormed fit index (NNFI; Bentler & Bonett, [5]), and comparative fit index (CFI; Bentler, [4]). It is suggested that SRMR ≤.08 indicates a good model fit (Hu & Bentler, [23]). According to Browne and Cudeck [6], RMSEA ≤.10 demonstrates an acceptable fit and ≤.08 shows a good fit. In addition, NNFI and CFI ≥.95 are required for a well-fitting model (Hu & Bentler, [23]). It has been suggested that model evaluation be based on an overall assessment of various fit statistics instead of a single index (Gerbing & Anderson, [17]). When nested models are compared, the chi-square difference test is applied to evaluate the significance of the change in model fit. In case RML was used, Satorra–Bentler (SB) scaled difference chi-square tests (Bryant & Satorra, [8]) were used for model comparison. The CFA models and time invariance were tested with LISREL 8.8, and the ESEM analyses were conducted in Mplus 7.11.

Table 1 present the means, standard deviations, skewness, and kurtosis of the CEI–II items. The tests of multivariate skewness (z = 16.39, p <.001), multivariate kurtosis (z = 11.66, p <.001), and multivariate skewness and kurtosis (χ2 = 404.63, p <.001) were all significant, indicating that the data did not follow a multivariate normal distribution. Most items (i.e., 9 out of 10) and the total score showed no significant gender difference in the mean level, which was generally consistent with Kashdan et al. [30]. When the mean of the total score was converted (Little, [35]) and compared with the mean derived from the pooled sample in Kashdan et al. [28], it was found that the level of curiosity in this study was significantly higher (t = 8.74, p <.01, Cohen's d =.64).

Table 1. Descriptive statistics and factor loadings of the CEI–II items.

10001 Note. CEI–II = Curiosity and Exploration Inventory–II; ESEM = exploratory structural equation modeling; CFA = confirmatory factor analysis.

In the ESEM, the target loadings of the CEI–II items on the two factors (i.e., stretching and embracing) were specified, with oblique rotation to allow the factors to correlate. Because the data did not meet the assumption of multivariate normality, RML estimation and SB scaled chi-square statistics were used. The model did not generate a clear factor structure as expected, although the model fit was acceptable, χ2sb(39, _I_N_i_ = 294) = 65.13, p <.001, RMSEA =.072, 90% CI [.050,.093], SRMR =.035, NNFI =.94, and CFI =.96. As presented in Table 1, there were several cross-loadings on the two factors (e.g., Items 3 and 9) and larger loadings on nontarget factors (e.g., Items 4, 7, and 10). Given the results, another ESEM procedure was conducted with geomin oblique rotation without any constraint on the target loadings so as to explore the best two-factor model for the data. The results remained largely the same in terms of the factor loadings, factor correlation, and model fit.

Because the two-factor structures did not fit the data, we tested the number of underlying factors using parallel analysis. Results from the parallel analysis showed that the eigenvalues of the raw data were 5.38 and 0.98 for the first and second factors, respectively. The mean and the 95th percentile eigenvalues of the random permutation were 1.30 and 1.38 for the first factor and 1.21 and 1.27 for the second factor, respectively. Because the eigenvalue of the second factor was smaller than the eigenvalue from random data, it was suggested that only one factor should be retained.

The one-factor structure was examined using CFA. Most fit indexes showed that the one-factor model provided acceptable fit to the data, except for the RMSEA, χ2sb(35, _I_N_i_ = 294) = 141.90, p <.001, RMSEA =.102, 90% CI [.085,.120], SRMR =.056, NNFI =.96, and CFI =.97. The modification indexes were inspected to locate the source of misfit. It was found that there was significant error covariance between Items 2 and 8 (r =.17, p <.001), which was due to some similarity in the Chinese wordings of these two items. With the error covariance being included, the revised model achieved a satisfactory fit to the data, χ2sb(34, _I_N_i_ = 294) = 121.52, p <.001, RMSEA =.094, 90% CI [.076,.112], SRMR =.053, NNFI =.96, and CFI =.97. All items loaded highly on the factor, with loadings ranging from.52 to.82 (see Table 1).

The time invariance of the one-factor model of the CEI–II was tested. Because the assumption of multivariate normality was not met, RML estimation and SB scaled difference chi-square tests (Bryant & Satorra, [8]) were used. The two-wave longitudinal data were examined within the same model, in which the items of Time 1 loaded on one factor and the items of Time 2 loaded on the other. To estimate the stability of curiosity over time, the two latent factors were allowed to correlate. Results indicated that the initial longitudinal factor model did not fit the data well, χ2sb(169, _I_N_i_ = 242) = 669.05, p <.001, RMSEA =.111, 90% CI [.102,.120], SRMR =.072, NNFI =.94, and CFI =.94. After including the autocovariances between the measurement errors across the two time points, the model fit the data satisfactorily, χ2;sb(159, _I_N_i_ = 242) = 314.97, p <.001, RMSEA =.064, 90% CI [.053,.074], SRMR =.059, NNFI =.98, and CFI =.98, and achieved a significant improvement over the initial model, Δχ2scaled(10) = 219.21, p <.001. Therefore, this revised model was adopted as the baseline configural model for the subsequent tests of invariance of factor loading and measurement error.

It was found that imposing equality constraints on the factor loadings of the CEI–II items across Times 1 and 2 did not result in significant deterioration in the model fit, Δχ2scaled(10) = 15.32, p =.121, indicating that the factor loadings were invariant over time. Moreover, constraining the measurement errors to be equal across Time 1 and Time 2 did not significantly worsen the model fit, Δχ2scaled(10) = 11.23, p =.340, suggesting the invariance of measurement errors over time. The goodness of fit of the final model was satisfactory, χ2sb(179, _I_N_i_ = 242) = 343.11, p <.001, RMSEA =.062, 90% CI [.052,.071], SRMR =.068, NNFI =.98, and CFI =.98, with factor loadings varying from.52 to.82. The latent variable of curiosity at Time 1 was highly correlated with that at Time 2 (r =.73, p <.001). This finding was very close to Kashdan et al. [30], which obtained.78 and.74 for the two subscales of the original CEI with an interval of 1 month. These results provided clear support for the time invariance of the factor structure of the CEI–II. The final model is presented in Figure 1.

Graph: Figure 1. The final longitudinal factor model for the Curiosity and Exploration Inventory–II (CEI–II). Note. Nonsignificant parameter is underscored.

Based on this model, the latent mean change was examined. Following the procedure suggested in Raykov and Amemiya [40], the corresponding intercepts were fixed to be equal across Time 1 and Time 2. Furthermore, the latent mean of the Time 1 factor was set to zero, whereas the latent mean of the Time 2 factor was freely estimated. Although the overall model fit was satisfactory, χ2sb(188, _I_N_i_ = 242) = 375.11, p <.001, RMSEA =.065, 90% CI [.055,.074], SRMR =.068, NNFI =.98, and CFI =.98, a significant change in the scaled chi-square value, Δχ2scaled(9) = 43.42, p <.01, suggested that the item intercepts were not invariant across Time 1 and Time 2. According to the modification indexes, partial invariance (Byrne, Shavelson, & Muthén, [9]) was achieved, Δχ2scaled(6) = 10.47, p =.11, after relaxing the constraints on three items (i.e., Items 4, 5, and 10). Based on this model with partial invariance, the latent mean at Time 2 was nonsignificant (M = −.11, t = –1.87, p >.05), which suggested no significant change in trait curiosity from Time 1 to Time 2.

Table 2 presents the descriptive statistics, correlations, and reliability of all measures at Time 1, including satisfaction with university life, personality traits, and human values. As expected, curiosity was moderately positively correlated with satisfaction with university life (r =.45, p <.001). As the CEI–II was developed based on a model of positive emotional experience, it was reasonable that students who were motivated to stretch and embrace new and uncertain university experiences tended to be more satisfied with their university lives.

Table 2. Descriptive statistics and correlations among variables.

• 20001 Note. Values on the diagonal are Cronbach's alpha coefficients obtained in this study. CEI–II = Curiosity and Exploration Inventory–II.
• 20002 p <.05. **p <.01. ***p <.001.

Consistent with the findings of Kashdan et al. [28], curiosity was positively correlated with Openness to Experience (r =.52, p <.001), Extraversion (r =.47, p <.001), and Conscientiousness (r =.32, p <.001), but was negatively correlated with Neuroticism (r = −.34, p <.001). Following Cohen's ([12]) approach, these correlations were transformed into Fisher's z and compared against those in Kashdan et al. [28]. It was found that the differences were small and nonsignificant (q = −.08 to.13, p >.05). Unexpectedly, curiosity was positively correlated with Agreeableness (r =.25, p <.001). This result differed from Kashdan et al. [28], which showed no association between curiosity and Agreeableness. Comparison of the correlations showed a significant difference with medium effect size (q =.30, p <.01).

With respect to the relationship between curiosity and human values, curiosity was positively correlated with openness to change (r =.43, p <.001) and negatively correlated with conservation (r = −.43, p <.001). These results supported our hypotheses. No significant correlations were found between curiosity and either self-enhancement or self-transcendence.

Hierarchical multiple regression analyses were performed to assess the incremental validity of the CEI–II for predicting human values and university life satisfaction over and above the Big Five personality traits (see Table 3). In each regression equation, gender and age were entered in Step 1, the Big Five personality traits were entered in Step 2, and the CEI–II score was entered in Step 3. We found that greater curiosity was significantly predictive of greater openness to change (ΔR2 =.03, β =.22, p <.001), weaker conservation (ΔR2 =.04, β = −.29, p <.001), greater self-transcendence (ΔR2 =.02, β =.19, p =.009), and greater satisfaction with university life (ΔR2 =.06, β =.33, p <.001) after controlling for demographic and personality factors. These results demonstrate that the CEI–II score has incremental validity above and beyond the Big Five personality traits in predicting human values and satisfaction with university life.

Table 3. Hierarchical multiple regression analyses predicting human values and university life satisfaction.

• 30001 Note. CEI–II = Curiosity and Exploration Inventory–II. Gender is dummy coded (0 = male, 1 = female).
• 30002 p <.05. **p <.01. *** p <.001.

The CEI–II was originally developed in a Western context with samples of predominantly female undergraduate students (Kashdan et al., [28]). This study assessed the reliability and validity of the CEI–II in a Chinese context using a similar sample in terms of gender and age distributions. The high Cronbach's alpha value (α =.90), interpretable factor solution, and meaningful correlations with relevant constructs all suggested that the CEI–II was a reliable and valid measure of curiosity among Hong Kong Chinese university students. However, the original two-factor structure of the CEI–II was not supported. Different from the findings of previous studies among Western samples (Kashdan et al., [28]), this study showed that a one-factor structure was more appropriate among Hong Kong Chinese university students. This finding could not be attributable to the methodological decision about the number of factors, because when a two-factor ESEM model with either target or geomin rotation was estimated, several items had either cross-loadings on more than one factor or high loadings on nontarget factors. Subsequent analyses suggested a single-factor model of the CEI–II, which was found invariant across the investigated time period. These results provided strong empirical evidence for the unidimensional factor structure of the CEI–II among Chinese university students and called into question the applicability of the two-factor structure of the CEI–II among the sample. Furthermore, the mean difference test revealed that the curiosity level in this study was higher than that in Kashdan et al. [28]. Being educated in a culture that puts much emphasis on innovation and creativity, Hong Kong students' motivation to explore new knowledge and their willingness to manage the tension of novelty and uncertainty might tend to be integrated and synergized in the process of socialization. Future research could collect more data in other Chinese samples or contexts to test the generalizability of this finding.

Consistent with prior research on the associations between trait curiosity and other related constructs, this study found that the CEI–II score was correlated with the Big Five personality traits, human values, and university life satisfaction. The CEI–II score was also predictive of human values and university life satisfaction over and above the Big Five personality traits. Students with higher trait curiosity levels were more satisfied with their university life than those who were less curious. Students with greater autonomy in seeking new knowledge and experiences might find the new teaching and learning environment at university more comfortable and easier to adapt to.

Practically, this study could provide insights for teachers and education practitioners regarding the functions and benefits of curiosity in the higher education setting. Curiosity, adaptation to the university environment, academic performance, and satisfaction with university life might interact in a dynamic way. On one hand, trait curiosity might enhance academic performance by facilitating adaptation to the new environment and can eventually increase satisfaction with university life. On the other hand, students who are satisfied with their university lives might be more curious and willing to explore the environment that is full of novel experiences. Therefore, these students are likely to adapt better to the new environment and achieve better academic performance. We suggest that future research should examine the causal mechanism behind the association.

As with past research on curiosity and the Big Five personality traits, results from this study showed that curiosity was strongly related to both Openness to Experience and Extraversion, which is reasonable given its conceptual overlap with these two factors. In addition, it was also shown that curiosity was positively related to Conscientiousness and negatively related to Neuroticism. Unexpectedly, the results revealed a positive correlation between curiosity and Agreeableness. Previous studies using the CEI and the CEI–II have yielded inconsistent findings regarding the relationship between curiosity and Agreeableness in Western samples. When the CEI was used, Kashdan et al. [30] found that Agreeableness was significantly positively correlated with one dimension of curiosity. However, when using the CEI–II, Kashdan et al. [28] reported a nonsignificant correlation between curiosity and Agreeableness. In our study, curiosity was found to be positively associated with Agreeableness, which is characterized by interest in others, empathy with others' feelings, and concern for others' well-being (Graziano & Eisenberg, [18]). This is consistent with findings from recent indigenous studies on Chinese personality (e.g., Cheung et al., [10]). For instance, using the Chinese Personality Assessment Inventory, Cheung et al. [10] found that indigenously derived openness scales were closely related to personality factors of social potency and interpersonal relatedness, suggesting that the construct of openness and curiosity in Chinese culture includes not only the approach to ideas and interest as commonly understood in Western cultures, but also the social relationships with other people. Therefore, it is reasonable that curiosity and Agreeableness were correlated among Chinese people.

Furthermore, this study broadens the existing literature of trait curiosity by providing evidence that curiosity and human values are closely related constructs. The results showed that curiosity correlated positively with openness to change and negatively with conservation. In other words, curiosity was positively related to human values that emphasize personal concerns, anxiety-free motivations and growth, and curiosity was negatively related to values that focus on social interests, anxiety avoidance, and self-protection. Further longitudinal research would be helpful to deepen our understanding about the directionality of the relationship between curiosity and human values.

Despite the contributions of this study, there are several limitations that warrant attention and discussion. First, findings of this study are based on a sample of Chinese first-year university students. The generalizability of these findings to other populations such as children, secondary school students, or working adults is not ensured. University students generally possess higher reading capabilities and better knowledge than adolescents and children. It is possible that the structure and function of curiosity among younger Chinese people is different from those found in this study. Furthermore, most work settings are essentially different from the university environment. In different organizations and positions, the level of curiosity and its associations with various outcomes could differ greatly. For example, it is interesting to investigate how trait curiosity is related to job outcomes such as satisfaction and productivity among employees (e.g., Mussel, [38]). Second, the relationships between curiosity and other constructs were investigated using a cross-sectional design. The directionality of these relationships, therefore, remains unclear. Future longitudinal studies would be helpful to disentangle the causal relationships among curiosity, life satisfaction, human values, and other constructs. Third, this study was conducted in Hong Kong, where the culture is a mix of Chinese and Western cultures. It would be interesting to examine the generalizability of the findings to other Chinese cultural contexts such as in Mainland China, where tradition and respect for authority are strongly emphasized. Future research in this area might reveal interesting differences in the factor structure and the associations between curiosity and other constructs. Such knowledge is essential for a full understanding of theory and practice related to curiosity. Fourth, this study did not examine the relationship between the CEI–II score and academic performance. Given curiosity's central role in students' learning, future research should evaluate the effectiveness of the CEI–II in predicting academic success among Chinese students. Fifth, curiosity can have different meanings and be understood differently across cultures. Our findings are based on measures that were developed in Western contexts. To deepen the current understanding, future research could develop an indigenous curiosity scale and relate it to other indigenous measures such as the Chinese Personality Assessment Inventory (Cheung et al., [10]) and the Chinese Values Survey (The Chinese Culture Connection, [11]). Findings from such studies will help us better understand the nature and functions of curiosity among Chinese people.