Measurement invariance and latent mean differences of the morbid curiosity scale (MCS) across the United States and China

Abstract

In recent years, there has been a growing body of research focusing on morbid curiosity. However, the development of measurement tools has been slow, with only two scales available. Compared to the unidimensional scale of Curiosity About Morbid Events (CAME) proposed by Zuckerman and Little (1986), the recently developed four-factor Morbid Curiosity Scale (MCS) by Scrivner (2021) demonstrates a stable factor structure and good reliability and validity. As the time since the development of this scale is relatively short, its measurement properties have not been widely evaluated. Therefore, this study used exploratory and confirmatory factor analyses to validate the factor structure of the MCS in the Chinese cultural context, and the results supported the four-factor structure of the MCS. Additionally, we established partial scalar invariance of the MCS between Chinese (N = 663) and American (N = 330) cultures, and further analyzed cultural differences in morbid curiosity using latent mean comparison. The results revealed that Chinese individuals had a lower motivation for understanding the minds of dangerous people. This study validated the four-factor Morbid Curiosity Scale across different cultures for the first time, promoting the generalizability of the four-factor MCS and suggesting its potential for use in a wide range of cultural backgrounds. These findings contribute to enriching cross-cultural research on morbid curiosity and its associated psychological factors.

1. Introduction

Morbid curiosity is the extent to which an individual is willing to learn about dangerous situations in life, especially death [1,2]. The level of knowledge about a dangerous situation will affect the psychological response of an individual to the actual danger. In other words, by understanding the factors related to death, people may effectively avoid the negative consequences of these factors [2]. Relatively speaking, individuals who possess a strong inclination towards morbid curiosity demonstrate increased effectiveness in handling diverse hazardous circumstances. Furthermore, studies have shown that individuals with morbid curiosity have greater psychological resilience [3].

The prevalence of research on morbid curiosity has been on the rise in recent years [e.g., Refs. [[13], [14]], [[3], [4], [5]]], indicating an increasing interest in the field. Additionally, research on topics related to morbid curiosity, such as horror amusement [e.g., Refs. [[6], [7], [8], [9]]], aversion [10], violence [e.g., Refs. [11,12]], and dark tourism [e.g., Refs. [13,14]], has become more abundant. However, the development of effective morbid curiosity assessment tools has not kept pace with this growth. To date, only two morbid curiosity scales are available. One is the Curiosity About Morbid Events (CAME) scale developed by Zuckerman and Little in 1986 [15], which has a unidimensional structure and its reliability and validity have not been widely evaluated. Additionally, almost all of the items on the CAME scale only focus on witnessing violence and fail to address other important aspects of morbid curiosity, making it somewhat narrow for studying morbid curiosity. Consequently, the CAME scale has had limited practical use. Recently, Scrivner (2021) developed a new measure of morbid curiosity called the Morbid Curiosity Scale (MCS), which consists of 24 items and measures a four-factor structure including interpersonal violence, body violations, minds of dangerous people, and paranormal danger. The Interpersonal Violence factor measures an interest in watching but not understanding the motivation behind violent behaviors, such as duels. The Body Violation factor measures the extent to which an individual is motivated to understand what happens when the body is damaged, such as in transplant surgery. The Minds of Dangerous People factor reflects an interest in understanding the motivations of dangerous individuals, such as mass murderers. Lastly, the Paranormal Danger factor reflects an interest in phenomena, such as ghosts, that seem to defy scientific understanding or go against the laws of nature. In summary, these factors revolve around individuals' motivation to understand dangerous or threatening phenomena.

Research has shown that the four-factor Morbid Curiosity Scale (MCS) has several advantages over existing scales measuring morbid curiosity, including excellent internal consistency, strong retest reliability, good convergent and discriminant validity, stable factor structure, and the ability to measure different aspects of morbid curiosity. However, due to the limited testing of the scale, further validation is needed to assess its effectiveness in measuring morbid curiosity, particularly in diverse cultural contexts. Given that cultural differences may influence interpretation and response to the MCS across different cultural groups, as well as the consistency of its measurement properties across cultures, it is important to consider such differences [[16], [17], [18]]. For example, collectivistic cultures may prioritize group solidarity and social harmony, resulting in less interest in violent or paranormal phenomena compared to individualistic cultures. Therefore, validating the MCS in different cultural contexts and examining cross-cultural differences in morbid curiosity could enhance our understanding of how individuals from different backgrounds respond to dangerous situations. To address these issues, conducting cross-cultural validation studies of the MCS in diverse cultural groups is crucial in determining the scale's validity and reliability for measuring morbid curiosity across cultures.

Therefore, the aim of this study is to investigate the applicability of the four-factor MCS (Scrivner) in an oriental culture without English-speaking samples and to examine the scale's reliability in this context. Specifically, this study seeks to validate the structure of the MCS in the Chinese cultural context and compare differences in morbid curiosity between Americans and Chinese individuals. To achieve these aims, we plan to (i) validate the MCS for use in the Chinese culture, (ii) test the measurement invariance of the MCS in two cultural samples (including configural, metric, scalar, and strict invariance), and (iii) compare latent mean differences in morbid curiosity between Americans and Chinese individuals.

2. Methods

2.1. Participants

Chinese participants. The survey included a sample of 906 Chinese participants who were invited to complete the questionnaire online at their convenience. To ensure data quality, attention check items were included in the scale (three times) to identify participants who did not answer carefully. After conducting statistical analysis, we excluded data from 243 participants, resulting in a final sample size of 663 participants. Among these participants, 39.52% were female and 60.48% were male. The age of all participants ranged from 18 to 60 years, with a mean age of 26.51 years. In terms of education level, 14.18% had high school education or below, while 85.82% had college education or higher.

American participants. Data from American participants used in this study were obtained from https://tinyurl.com/mtuhtzun, which were originally collected for the paper "The psychology of morbid curiosity: Development and initial validation of the morbid curiosity scale" [1]. The survey included a total of 340 participants from the United States, who were recruited online through Prime Panels. After statistical analysis, data from 10 participants were removed due to failure to pass the attention check task. As a result, the final sample size used for analysis consisted of 330 participants. Among these participants, 47.58% were female and 52.12% were male. All participants in the study had an age range of 18–96 years, with an average age of 42 years.

2.2. Ethical consideration

Ethical approval and informed consent were obtained from all individual respondents included in the research, in accordance with the ethical standards of the Research Center of Psychological Health Education, Jiangxi Normal University (No. JXSD-20211201), as well as the 1964 Helsinki Declaration and its subsequent revisions, or similar ethical standards.

2.3. Measure

All participants completed the MCS (Scrivner), which includes four subscales: interpersonal violence, body violations, dangerous people's minds, and paranormal danger. Items were scored on a six-point Likert-type scale ranging from 1 (strongly disagree) to 6 (strongly agree). The internal reliability of the overall scale was 0.94, with the internal reliability of each subscale being 0.89, 0.87, 0.92, and 0.90, respectively. The subscales demonstrated retest reliabilities of 0.86, 0.84, 0.81, and 0.80, respectively. Convergent validity and divergent validity were reported as 0.56 and 0.42, respectively [1].

The MCS was only available in English, so it needed to be translated into Chinese before it could be used in China. With the assistance of several professional English teachers and graduate students majoring in psychology, the MCS was translated from English to Chinese and then from Chinese to English. After discussing the existing differences, the final Chinese version of the MCS was formed. After completing the translation, a pilot test was administered to several college students to confirm their accurate understanding of the item meanings expressed by the MCS.

2.4. Statistical analysis

Preliminary estimates were made for all items. It is important to note that the estimation method used is influenced by whether the data conforms to a normal distribution. Typically, maximum likelihood with robust estimators is appropriate if any of the items exhibit high skewness or kurtosis [19]. The correlation between the items and the total score was calculated to assess the scale items' quality. To gain a more comprehensive understanding of the MCS's reliability, Cronbach's alpha and McDonald's ω internal consistency indices were calculated using both the Chinese sample (N = 663) and the U.S. sample (N = 330). The Cronbach's alpha coefficient is a commonly used measure of internal consistency, reflecting the extent to which the items on a scale are interconnected. However, it assumes that all items are equally essential for measuring the construct of interest. When items are not equally essential, McDonald's omega coefficient provides a more realistic estimate of the true reliability of the scale [20]. By calculating the alpha and omega coefficients, the MCS's reliability can be more accurately and thoroughly evaluated.

The Chinese sample was randomly divided into two sub-samples for conducting exploratory factor analysis (EFA) and confirmatory factor analysis (CFA) on the MCS. Sample 1 ($N_1$ = 331) was subjected to EFA, and Bartlett's Test and the KMO Test were used to assess the suitability of the data for factor analysis. Then, the factorial structure of the MCS was validated using CFA separately in both cultural samples, namely Sample 2 ($N_2$ = 332) and the U.S. sample (N = 330). The Comparative Fit Index (CFI), Tucker-Lewis Index (TLI), Root Mean Square Error of Approximation (RMSEA), and Standardized Root Mean Square Residuals (SRMR) were used as model fit indexes in this study. In general, the model fit was considered reasonably acceptable when CFI >0.90, TLI >0.90, RMSEA <0.1, and SRMR <0.08 [[21], [22], [23], [24]]. Considering that χ2 is sensitive to sample size and not precise enough as a model fit indicator, it was not adopted [[25], [26], [27]].

Measurement invariance is a prerequisite for conducting latent mean analysis [28]. The measurement invariance of the MCS was tested using the full Chinese sample (N = 663) and the U.S. sample (N = 330), including tests for configural invariance, metric invariance, scalar invariance, and strict invariance. Configural invariance was assessed to determine the equivalence of the factor structure between the two groups, while metric invariance tested the equivalence of factor loadings across groups. Scalar invariance evaluated the equivalence of both factor loadings and item intercepts between groups, and strict invariance assessed the equivalence of factor loadings, item intercepts, and item residual variances across groups. The four invariance tests were carried out in a nested manner, with the model from the former step being nested within the model from the latter step. As the baseline model for the tests, configurational invariance is a prerequisite for the other invariances. By constraining more parameters, more stringent invariance tests can be established. Metric invariance, scalar invariance, and strict invariance are commonly referred to as measurement invariance [29]. However, achieving strict invariance in practice is often challenging [30]. Some scholars argue that meaningful comparisons between different groups can be made when partial invariance is met, allowing for a small portion of items to be non-invariant [31]. Although strict invariance is not a necessary condition for testing latent means, we reported strict invariance in our current study. Considering the sensitivity to sample size, the values of ΔCFI and ΔRMSEA were used in this study as criteria for judging measurement invariance. Measurement invariance was not supported if there was a change in CFI (ΔCFI) greater than 0.01 or a change in RMSEA (ΔRMSEA) greater than 0.015 [25]. As complete measurement invariance is difficult to achieve in reality, researchers often resort to measures such as releasing the loadings or intercepts of non-invariant items to achieve partial invariance and continue latent mean analysis based on partial invariance [28]. Some researchers suggest that the highest modification index can be used to sequentially release constraints on non-invariant indicators, which produces smaller Type I errors compared to releasing all non-invariant items at once [32].

If scale invariance (at least partial scalar invariance) is satisfied, differences in latent means between Americans and Chinese will be compared [33,34]. In this step, the latent mean for the U.S. sample is constrained to zero, and the latent mean for the Chinese sample is estimated freely. It has been argued that structural equation models are more robust and yield less bias in the mean estimates than general linear models when comparing differences in means between groups [35,36].

3. Results

3.1. Descriptive analysis and reliability

The results of descriptive statistics for the MCS were presented in Table 1, which showed that the skewness coefficient of all items was < |1.009| and the kurtosis coefficient was < |1.535|, indicating that the data satisfied the normal distribution. Therefore, we used maximum likelihood estimation in the subsequent analysis. Additionally, as can be seen from the correlation analysis results in Table 1, all items had high item-total correlation values of more than 0.58. Generally speaking, if the correlation coefficient between an item and the total score is greater than 0.4 at a significant level, it indicates a high correlation as well as homogeneity. Otherwise, items that do not meet this criterion are considered for deletion. As a result, the quality of the MCS items was excellent, and all items were retained.

Table 1.

$Item’$ Mean (SD), Skewness, Kurtosis and Item-total correlation of MCS.

Item	Mean (SD)		Skewness		Kurtosis		Item-total correlation
	China	The U.S.	China	The U.S.	China	The U.S.	China	The U.S.
V1	3.023(1.552)	2.824(1.843)	0.203	0.506	−1.129	−1.247	0.705	0.726
V2	2.968(1.564)	3.458(1.873)	0.283	−0.045	−1.100	−1.492	0.686	0.730
V3	3.193(1.611)	3.373(1.722)	0.130	−0.036	−1.157	−1.326	0.814	0.735
V4	3.463(1.510)	3.700(1.750)	−0.103	−0.270	−1.008	−1.219	0.779	0.753
V5	3.131(1.671)	3.664(1.750)	0.234	−0.215	−1.218	−1.208	0.787	0.694
V6	3.775(1.521)	3.845(1.799)	−0.273	−0.419	−0.976	−1.218	0.756	0.758
B1	3.023(1.702)	3.064(1.930)	0.297	0.295	−1.246	−1.486	0.754	0.756
B2	3.241(1.678)	3.524(1.865)	0.078	−0.142	−1.276	−1.452	0.789	0.761
B3	3.087(1.668)	3.024(1.833)	0.222	0.265	−1.270	−1.402	0.796	0.727
B4	3.048(1.644)	3.082(1.832)	0.293	0.278	−1.143	−1.416	0.779	0.788
B5	3.136(1.642)	3.488(1.799)	0.223	−0.092	−1.181	−1.376	0.796	0.756
B6	3.623(1.627)	3.545(1.899)	−0.235	−0.108	−1.143	−1.506	0.785	0.815
M1	3.170(1.616)	3.800(1.696)	0.178	−0.436	−1.202	−1.069	0.799	0.798
M2	3.579(1.628)	4.333(1.666)	−0.249	−0.845	−1.167	−0.497	0.788	0.718
M3	3.789(1.617)	4.361(1.609)	−0.412	−0.926	−0.982	−0.190	0.746	0.696
M4	3.522(1.628)	4.127(1.721)	−0.164	−0.670	−1.178	−0.843	0.816	0.779
M5	3.502(1.655)	4.185(1.738)	−0.130	−0.692	−1.208	−0.813	0.807	0.710
M6	3.409(1.594)	4.058(1.696)	−0.059	−0.651	−1.165	−0.824	0.765	0.801
P1	4.077(1.503)	4.636(1.465)	−0.579	−1.009	−0.637	0.192	0.616	0.586
P2	3.433(1.610)	3.409(1.868)	−0.089	−0.020	−1.181	−1.489	0.778	0.783
P3	4.142(1.510)	3.355(1.788)	−0.734	0.082	−0.392	−1.357	0.641	0.715
P4	3.449(1.598)	3.821(1.827)	−0.119	−0.422	−1.151	−1.263	0.758	0.742
P5	3.861(1.575)	3.367(1.931)	−0.452	0.022	−0.879	−1.535	0.723	0.693
P6	3.706(1.580)	3.764(1.890)	−0.312	−0.262	−0.997	−1.421	0.750	0.684

Notes: V1–V6= Violence subscale. B1–B6= Body subscale. M1-M6 = Minds subscale. P1–P6= Paranormal subscale.

Table 2 presents the Cronbach's alpha and McDonald's ω values for the MCS and its four subscales. In the Chinese sample, the Cronbach's alpha value for the entire scale was 0.968, and the McDonald's ω value was 0.973. The Cronbach's alpha values for the four subscales ranged from 0.898 to 0.918, and the McDonald's ω values ranged from 0.900 to 0.919. In the U.S. sample, the Cronbach's alpha value for the entire scale was 0.964, and the McDonald's ω value was 0.972. The Cronbach's alpha values for the four subscales ranged from 0.892 to 0.932, and the McDonald's ω values ranged from 0.895 to 0.933. The reliability coefficients for both cultural samples being greater than 0.85 indicated that the measurements were highly consistent and dependable across both cultural contexts, demonstrating that the MCS is a reliable measurement tool with good internal consistency.

Table 2.

Cronbach's α and McDonald's ω coefficients of MCS-24.

Sample	V-subscale		B-subscale		M-subscale		P-subscale		MCS-24
	α	ω	α	ω	α	ω	α	ω	α	ω
China	0.898	0.900	0.918	0.919	0.909	0.910	0.899	0.901	0.968	0.973
The U.S.	0.892	0.895	0.918	0.919	0.932	0.933	0.900	0.904	0.964	0.972

Notes: V-subscale = Violence subscale. B-subscale = Body subscale. M-subscale = Minds subscale. P-subscale = Paranormal subscale. MCS-24 = 24-item morbid curiosity scale.

3.2. Factor structure of MSC-24

EFA (Sample 1;${\mathit{N}}_{\mathbf{1}}$=331). The value of Bartlett's test was χ2 (276) = 5577.87, p < .001, and the value of KMO was MSA = 0.96, indicating the suitability of the data for factor analysis and the adequacy of the sampling. Considering that parallel analysis is the most accurate method for factor retention in EFA [40], the present study utilized the psych package in R to conduct parallel analysis. The parallel analysis suggested that retaining four factors (CFI = 0.94; TLI = 0.92; RMSEA = 0.070, 90% CI [0.063, 0.078]; SRMR = 0.03) was appropriate, which is consistent with the results of a study conducted in the United States. This indicates that there is a similar understanding of the MCS factor structure in both Chinese and American samples, and that four factors can better explain the relationships among the observed variables.

CFA (Sample 2;${\mathit{N}}_{\mathbf{2}}$=332). The four-factor structure of MCS was validated using Sample 2 and American samples respectively and was compared with the one-factor model proposed by Zuckerman and Litle (1986). Table 3 presents the results of the CFA for the two cultural samples. In the one-factor model, the fit indices for the China sample were χ2 (252) = 1233.2, p < .001; CFI = 0.87; TLI = 0.86; RMSEA = 0.108, 90% CI [0.102, 0.114]; and SRMR = 0.053. The fit indices for the US sample were χ2 (252) = 1643.7, p < .001; CFI = 0.78; TLI = 0.76; RMSEA = 0.129, 90% CI [0.123, 0.135]; and SRMR = 0.070. In the four-factor model, the fit indices for the China sample were χ2 (246) = 844.4, p < .001; CFI = 0.92; TLI = 0.91; RMSEA = 0.086, 90% CI [0.079, 0.092]; and SRMR = 0.045, while those for the US sample were χ2 (246) = 778.2, p < .001; CFI = 0.92; TLI = 0.91; RMSEA = 0.081, 90% CI [0.075, 0.087]; and SRMR = 0.054. Analysis and comparison of all model fit indices revealed that the four-factor model provided a better fit to the data compared to the one-factor model, and met the criteria for good model fit [37]. This finding is consistent with the results of Scrivner (2021), indicating that the four-factor model is viable and more suitable for the concept of morbid curiosity. Additionally, the model fit indices were almost identical in both cultures. For example, the values of CFI and TLI were both greater than 0.9, the values of SRMR were both less than 0.8, and the values of RMSEA were both less than 0.1, suggesting that the four-factor model of morbid curiosity fits well and can be applied to different cultures.

Table 3.

Confirmatory factor analysis results of the MCS.

Models	Sample	N	χ2	df	CFI	TLI	RMSEA (90% CI)	SRMR
One-factor	China	332	1233.2	252	0.87	0.86	0.108 (0.102–0.114)	0.06
	The U.S.	330	1643.7	252	0.78	0.76	0.130(0.123–0.135)	0.07
Four-factor	China	332	844.4	246	0.92	0.91	0.086(0.079–0.092)	0.05
	The U.S.	330	778.2	246	0.92	0.91	0.081 (0.075–0.087)	0.06

Notes: RMSEA = root mean square error of approximation with 90% confidence interval; SRMR = standardized root mean square of residuals; CFI = comparative fit index; TLI = Tucker-Lewis Index.

The factor loading results and path model of the 24 items of MCS in two cultures are presented in Table 4 and Fig. 1, respectively. For each item in the MCS, the factor loading values in Chinese culture ranged from 0.70 to 0.89, and the factor correlations ranged from 0.78 to 0.97. In American culture, the factor loading values ranged from 0.66 to 0.89, and the factor correlations ranged from 0.71 to 0.88. The factor loadings for both cultures exceeded 0.6 [38], indicating that the MCS is a measure characterized by structural stability and robustness.

Table 4.

Standardized factor loadings for 24 items of the MCS.

	Items	Factor loadings
		China Sample	The U.S. Sample
Violence	V1	0.74	0.76
	V2	0.74	0.80
	V3	0.87	0.74
	V4	0.81	0.82
	V5	0.82	0.66
	V6	0.79	0.80
Body	B1	0.83	0.82
	B2	0.85	0.83
	B3	0.89	0.78
	B4	0.85	0.86
	B5	0.87	0.78
	B6	0.80	0.80
Minds	M1	0.81	0.81
	M2	0.81	0.83
	M3	0.77	0.82
	M4	0.86	0.88
	M5	0.85	0.78
	M6	0.84	0.89
Paranormal	P1	0.70	0.66
	P2	0.80	0.77
	P3	0.72	0.79
	P4	0.81	0.83
	P5	0.84	0.77
	P6	0.88	0.83

Notes: V1–V6= Violence subscale. B1–B6= Body subscale. M1-M6 = Minds subscale. P1–P6= Paranormal subscale.

Fig. 1.

The Four-factor CFA model for the Morbid Curiosity Scale in China (left) and the United States (right).

3.3. Measurement invariance tests for the MCS

In order to examine the cross-cultural measurement invariance of MCS, a series of increasingly stringent tests were conducted on the entire sample ($N_{Chinese}$ = 663; $N_{American}$ = 330), including configural invariance, metric invariance, scalar invariance, and strict invariance.

Configural invariance. As a starting point for the test of measurement invariance, configural invariance only required that the underlying structural relationships between latent and explicit variables were equivalent, and that parameters in the two cultural samples were not set to be equal. As can be observed in Table 5, the data fit indices for the test of configural invariance were acceptable: TLI = 0.906; CFI = 0.916; RMSEA = 0.081 with a 90% confidence interval of [0.078, 0.085]; SRMR = 0.050, which suggested that the configural invariance of the MCS had been established.

Table 5.

Measurement invariance tests of the MCS across two countries.

Model	χ2	df	TLI	CFI	RMSEA (90% CI)	SRMR	Model Comparison	ΔCFI	Δ RMSEA
M1. Configural invariance	2098.301	492	0.906	0.916	0.081 (0.078–0.085)	0.050	_	_	_
M2. Metric invariance	2173.130	512	0.906	0.913	0.081 (0.077–0.084)	0.055	M2-M1	−0.003	0.000
M3. Scalar invariance	2493.730	532	0.893	0.897	0.086 (0.083–0.090)	0.061	M3-M2	−0.016	0.005
M4. Partial Scalar invariance	2336.218	530	0.901	0.905	0.083 (0.079–0.086)	0.057	M4-M2	−0.008	0.002
M5. strict invariance	2570.910	554	0.895	0.894	0.086 (0.082–0.089)	0.062	M5-M4	−0.011	0.003

Notes: RMSEA = root mean square error of approximation with 90% confidence interval; SRMR = standardized root mean square of residuals; CFI = comparative fit index; TLI = Tucker-Lewis Index.

Metric invariance. Metric invariance was tested by restricting the factor loadings in the two cultural samples to be equal in this step and allowing the other parameters (e.g., item intercepts and error variances) to be freely estimated. As reported in Table 5, the values of ΔCFI (−0.003) and ΔRMSEA (0.000) were both less than 0.01, supporting metric invariance. In other words, full metric invariance of the MCS was established.

Scalar invariance. The item intercepts in the two cultural samples were constrained to be equal on the basis of scale invariance, allowing the error variances to be freely estimated. The fit indices were presented in Table 5: ΔCFI = 0.016 and ΔRMSEA = 0.005, indicating that full scalar invariance was rejected. According to the highest modification index, non-invariant items were progressively freed from constraints, resulting in a significant improvement in the fit indices (CFI increased by 0.015 and RMSEA decreased by 0.006). Compared to the metric invariance model, ΔCFI = −0.008 and ΔRMSEA = 0.002, which indicated that partial scalar invariance was satisfied.

Strict invariance. After configural invariance, metric invariance, and partial scalar invariance were established in turn, the strict invariance test was performed, setting the error variances in the two cultural samples to be equivalent. As shown in Table 5, ΔCFI = −0.011, which was greater than 0.01, indicating that strict invariance was rejected.

3.4. Latent mean differences

With partial scalar invariance being established, we compared differences in latent means between Americans and Chinese [39]. Taking the U.S. sample as the reference group, the latent means were set to zero, and the latent means for the Chinese sample were estimated freely. Taking into account the potential impact of partial scalar invariance on latent means analysis, we examined latent mean differences based on both a fully invariant model and a partially invariant model. As shown in Table 6, the results were almost identical between the two models, indicating that non-invariance had little effect on the results. Therefore, we only reported the results based on the partially invariant model. The results indicated that Chinese participants exhibited significantly lower morbid curiosity than American participants on the factor "minds of dangerous people" (standardized fitted mean = −0.50, p < .001), but showed no significant differences on the factors "interpersonal violence" (standardized fitted mean = −0.18, p > .001), "body violation" (standardized fitted mean = −0.07, p > .001), and "paranormal danger" (standardized fitted mean = −0.01, p > .001).

Table 6.

Mean differences for Chinese and Americans on MCS for fully invariant and partially invariant models.

	Fully invariant model		Partially invariant model
	M(SE)	Standard Mean Difference	M(SE)	Standard Mean Difference
interpersonal violence	$-$ 0.20(0.09)	$-$ 0.18	$-$ 0.20(0.09)	$-$ 0.18
body violations	$-$ 0.10(0.10)	$-$ 0.07	$-$ 0.10(0.10)	$-$ 0.07
minds of dangerous people	$-$ 0.64(0.10)	${-0.50}^a$	$-$ 0.64(0.10)	${-0.50}^a$
paranormal danger	0.02(0.08)	0.02	$-$ 0.01(0.08)	$-$ 0.01

Notes: a. Significant mean differences.

4. Discussion

In contrast to the unidimensional CAME scale (1986), which considered only one aspect of morbid curiosity, the Morbid Curiosity Scale (MCS; 2021) assumed that morbid curiosity contained a four-factor structure: interpersonal violence, body violations, minds of dangerous people, and paranormal danger. This four-factor structure had been demonstrated in American cultural studies [1]. The current study investigated the four-factor MCS in Chinese culture to examine whether the MCS could be applied to an oriental culture without English-speaking samples, and whether this scale demonstrated reliability in this context. The results confirmed that the four-factor model introduced by Scrivner was still more suitable for understanding morbid curiosity than the one-factor model proposed by Zuckerman and Little (1986) in Chinese culture. This finding was consistent with previous findings in American culture. Thus, the present study further strengthens previous findings and indicates that, despite cultural differences, the four-factor MCS proposed by Scrivner remains valid and can be applied to different cultural contexts. Additionally, this study explored the factor structure of morbid curiosity in Chinese culture for the first time. This greatly enriches the study of morbid curiosity and hopefully provides valuable insights for future scientific research.

The literature currently lacks any exploration into the measurement invariance of the Morbid Curiosity Scale. In the present study, configural invariance, metric invariance, and partial scalar invariance of the MCS were successively established. First of all, we conducted the first step of measurement invariance: configural invariance. The results supported configural invariance, indicating that the basic structure of the four-factor MCS was the same across the two cultures. Next, the factor loadings of the two groups were set to be the same. Compared with the configural invariance model, there was no significant decrease in the overall model fit. The results supported metric invariance, indicating that the relationship between the observed items and latent traits of the MCS had the same significance in the Chinese and American groups. In other words, subjects in different groups had the same understanding of the same construct. Building on metric invariance, the item intercepts of the two groups were set to be the same. It was found that the overall model fit significantly decreased compared to the metric invariance model, resulting in the rejection of scalar invariance. It has been acknowledged by researchers that achieving full measurement invariance is not an easy task. Therefore, intercepts of items that did not satisfy invariance were sequentially released based on the highest modification index. After releasing constraints on item 4 and item 12 (related to Paranormal Danger), it was discovered that the overall model fit did not significantly decrease compared to the metric invariance model, supporting partial scalar invariance.

It has been suggested by scholars that when partial scalar invariance is met, comparisons can be made through latent mean analysis [30,39]. In the present study, metric invariance and partial scalar invariance were established. However, analyzing latent means based on partial invariance may result in bias. Therefore, latent means were further tested under the conditions of full scalar invariance and partial scalar invariance, respectively [40]. The results showed that regardless of using the full invariance model or the partial invariance model, significantly lower levels of morbid curiosity were observed in Chinese individuals than in Americans on the "minds of dangerous people" factor, and this difference was consistent. However, there were no significant differences in the factors of interpersonal violence, body violation, and paranormal danger. These findings indicate that it is reasonable to analyze latent means based on partial scalar invariance, which is consistent with the conclusion drawn by Steinmetz [41].

The concept of morbid curiosity primarily encompasses four distinct types of threats that humans face, including the threat of interpersonal violence, the risk of physical harm, the presence of dangerous individuals, and the perceived threat of supernatural or paranormal phenomena. The current study concludes that there are distinct differences between Chinese and American individuals only in terms of the factor of dangerous people. One possible explanation for this finding is related to the socialization process and cultural values in different societies. In Chinese culture, collective interests and common responsibilities are emphasized, whereas in the individualistic culture of the United States, individual achievement and self-expression are highly valued. This cultural disparity may lead Chinese individuals to be more concerned about social stability and personal safety, focusing primarily on avoiding violent and dangerous behavior, rather than attempting to understand the underlying motives behind such behavior. Additionally, this cultural dissimilarity may also be linked to the emphasis placed in Chinese culture on respecting authority and traditional values, as studying dangerous people or phenomena may be perceived as disrespectful or challenging social norms.

Although the four-factor MCS in this study proved to be a robust measurement tool applicable to different cultures, it is important to acknowledge certain limitations. Firstly, there were differences between the American and Chinese samples investigated in this study regarding demographic variables such as age and gender. These differences may restrict the generalizability of the measurement invariance findings of the four-factor Morbid Curiosity Scale. Therefore, future studies should aim to obtain samples that are better matched in terms of demographic variables and evaluate the measurement invariance properties of the MCS using different cultural groups matched on age and gender. Secondly, the current study only conducted measurement invariance analysis on two cultural groups, which cannot represent all cultures and populations. This limitation hinders the generalizability of the findings to populations beyond the research sample. Therefore, future research should implement the measurements in multiple cultural samples to deepen the understanding of the relationship between morbid curiosity and different cultures. Next, individuals with higher levels of morbid curiosity may possess greater resilience, but our study did not explore the structural relationship between the Morbid Curiosity Scale and other scales or concepts such as resilience. Consequently, future research can focus on investigating and establishing a more comprehensive research framework concerning morbid curiosity in a cross-cultural context. In addition, this paper validated the structural stability of the four-factor Morbid Curiosity Scale at the cross-sectional level, but its stability over time remains unknown. Hence, it is necessary to investigate the scale's structural stability over time in future research, starting with longitudinal invariance analysis. Finally, the age sample surveyed by the Morbid Curiosity Scale did not include minors, yet factors such as horror and violence associated with morbid curiosity have always affected the mental and physical health of minors. Therefore, it is worth considering further refining the scale to better capture the experiences of minors in future research.

5. Conclusion

The present study has important implications for the field of cross-cultural psychology. Specifically, our study confirms the robustness of the four-factor Morbid Curiosity Scale (MCS) presented by Scrivner (2021) in the context of oriental culture without English-speaking individuals. The partial scalar invariance of the MCS across different cultures has also been established, indicating that the MCS can reliably measure subjective morbid curiosity in various cultural contexts.

Moreover, significant latent mean differences have been found between Americans and Chinese in their inclination to understand the motivations of dangerous people. These differences may be attributed to cultural factors. Therefore, it is suggested that future research should investigate the cultural factors influencing the perception and understanding of dangerous people, thus enhancing our understanding of the cognitive and behavioral aspects associated with morbid curiosity in different cultures.

Author contribution statement

Xue Wang: Conceived and designed the experiments; Analyzed and interpreted the data; Wrote the paper.

Yan Cai: Performed the experiments; Analyzed and interpreted the data.

Dongbo Tu; Qin Wang: Conceived and designed the experiments; Performed the experiments; Contributed reagents, materials, analysis tools or data.

Funding statement

This research was supported by the National Natural Science Foundation of China (Grant No. 32160203).

Data availability statement

Data associated with this study has been deposited at https://osf.io/m5fnc/

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.