To address the ambiguity surrounding self-control, Jack Block and Jeanne H. Block developed the concept of ego-control (Block, 1950; Block, 1951). As a personality construct, ego-control affects a considerable array of consequential life outcomes, such as cognition (e.g., Block, 2005), marijuana use (e.g., Shedler and Block, 1990), adaptive functioning (e.g., Cicchetti and Rogosch, 2007), and mental health problems (e.g., Bohane et al., 2017). Ego-control refers to “the individual’s modal or characteristic response to behavioral or attentive impulses” (Block and Block, 2006, p. 317), and it ranges from undercontrol to overcontrol. According to Block and Block (2006), an individual who is relatively expressive or attentive to internal pushes and pulls (e.g., immediate and direct expression of behaviors) can be described as undercontrolled; in contrast, individuals who are relatively restrictive in their behavioral or attentional impulses (e.g., excessive delayed gratification) are referred to as overcontrollers. A growing body of literature indicates that both undercontrol and overcontrol in individuals are related to an increased incidence of health problems (for reviews, see Alessandri and Vecchione, 2017; Bohane et al., 2017). Specifically, undercontrolled individuals tend to show externalizing problems (e.g., aggressive behavior), whereas overcontrolled individuals tend to show internalizing problems (e.g., depression).

Jack Block developed the ego-undercontrol scale (EUC; Letzring et al., 2005) to evaluate ego-control in individuals. This self-report measure comprises 37 items, each of which is rated using a four-point scale ranging from 1 (disagree very strongly) to 4 (agree very strongly). Letzring et al. (2005) reported that Cronbach’s alpha of the original scale was only 0.63 and confirmed that the measure had a unifactorial structure. To the best of our knowledge, only two studies have further developed the EUC, revising it to achieve adequate validity and reliability: one study translated the scale into Swedish (Isaksson et al., 2021), while the other translated into Spanish for use in an Argentinian context (Hess and Mesurado, 2021). The Swedish version of the EUC (EUC-13; Isaksson et al., 2021) includes 13 items and comprises 3 dimensions: Uninhibited behavior (e.g., “When I get bored, I like to stir up some excitement”), planful conscientious behavior (e.g., “I like to stop and think things over before I do them”), and socially restrained behavior (“I am easily downed in an argument”). Unfortunately, Isaksson et al. (2021) did not clarify the meaning of each dimension, and it should be noted that the internal reliability for the socially restrained behavior subscale was only 0.51. Furthermore, the factor loading of the EUC-13’s item 11 (“I am easily downed in an argument”) was 0.26 in Isaksson et al.’s (2021) study, which is below 0.4. The results suggest that the EUC-13 may also fail to achieve satisfactory psychometric properties.

The Spanish version of the EUC (Hess and Mesurado, 2021) contains 14 items and measures 2 dimensions: Behavioral impulsiveness (e.g., “I tend to buy things on impulse”) and cognitive impulsiveness (e.g., “I become impatient when I have to wait for something”). Behavioral impulsiveness refers to a lack of control over impulses related to physical behaviors. Cognitive impulsiveness reflects one’s ability to stop and think about a situation before taking action (Hess and Mesurado, 2021). The factor loading of item 18 (“I like to flirt”) was 0.23 in Hess and Mesurado’s (2021) study, which is again below 0.4; however, the results indicate that the psychometric properties of the Spanish version of the EUC also fail to meet acceptable standards. Furthermore, while the EUC-13 and the Spanish version of the EUC both measure multiple dimensions, the studies that introduced these scales lack theoretical support for the identified dimensions. To return to ego-control theory (Block and Block, 1980), if ego-control refers to one’s ability to express or inhibit their impulses, with one end of the trait spectrum representing undercontrol and the other representing overcontrol, it is logical to conclude that the EUC and its derivatives measure only two dimensions: impulsive behavior and inhibitory behavior.

To the best of our knowledge, no scale has been developed as of yet to evaluate ego-control in a Chinese context. Therefore, to further research ego-control in China, as well as to extend Block and Block’s theories about ego function, this study aimed to translate and validate the EUC into Chinese. As exploratory structural equation modeling (ESEM; Asparouhov and Muthén, 2009; Marsh et al., 2014) incorporates exploratory factor analysis into the framework of structural equation modeling, ESEM was utilized to assess the factorial validity of the newly developed measure.

We first translated the English-language EUC into Chinese to create the Chinese version of the EUC (EUC-C), after which we tested its psychometric properties using a non-clinical Chinese sample. We then assessed its factorial validity using ESEM and confirmatory factor analysis (CFA). We hypothesized that the EUC-C would have two dimensions, measuring impulsive behavior and inhibitory behavior. Finally, we assessed the construct validity of the EUC-C by investigating its correlations with other constructs. We hypothesized that ego-undercontrol would be positively related with loneliness and depression and negatively associated with self-esteem and life satisfaction.

A sample of 1,823 freshmen, sophomores, and juniors was recruited (M_age = 19.70; SD = 1.42; 44.2% males) from three general universities in Shanghai. There were 944 freshmen (42.4% male), 686 sophomores (47.8% male individuals), and 193 juniors (40.4% male individuals).

The questionnaire was administered to the students with the help of their counselors in a classroom setting. All participants who took part in this study provided their consent to participate. Each participant completed a paper-and-pencil questionnaire in class and submitted it upon completion. The study was reviewed and approved by the ethics committee of the corresponding author’s institution.

We employed two methods for item analysis. First, we assessed the item-total correlation (ITC), which measures the relationship between the total score and individual item scores for the total dataset. An ITC value is considered appropriate when it is greater than 0.2, due to the fact that it is likely to be small in large samples (Kline, 2015; Yu et al., 2021). Second, we used the critical ratio (CR), which is a more sensitive and stable item discrimination index (Crocker and Algina, 1986; Yu et al., 2021). The total score for all items was ranked from the lowest to highest. The top 27% were classified as the high group, and the bottom 27% as the low group. An independent samples t-test was conducted between the high and low subgroups of the 37 items to find if there was a significant difference in the scores for each item.

The total sample was randomly divided into two subsamples using SPSS (version 22.0). We used Mplus (version 8.3) to conduct ESEM and CFA on the data. The first group (N = 911) was used for ESEM. Factor analyses were conducted using the Kaiser–Meyer–Olkin (KMO) test and Bartlett’s test to investigate sample adequacy. We used the oblique rotation to conduct the factor analyses. At the same time, a parallel analysis (O’Connor, 2000) was conducted using SPSS syntax to determine the upper limits of potential factor solutions, based on the comparison of actual eigenvalues and those generated through simulation using 1,000 samples of identical characteristics. The second group (N = 912) was used to confirm the factor structure.

Given that the EUC is a 4-point scale and not strictly a continuous variable, we used weighted least squares means and variance adjusted (WLSMV) estimation in Mplus software. This approach provides more accurate data results (Beauducel and Herzberg, 2006; Rhemtulla et al., 2012). To evaluate model fit, different fitting metrics were considered: the χ²/df goodness-of-fit index, the Comparative Fit Index (CFI), the Tucker–Lewis Index (TLI), Root Mean Square Error of Approximation (RMSEA), and standardized root mean square residual (SRMR). CFI and TLI values greater than 0.90 indicate acceptable fit (Hu and Bentler, 1999). RMSEA and SRMR values less than 0.08 indicate acceptable fit (Hu and Bentler, 1999). Model fit was assessed by comparing the goodness-of-fit indices between constrained and unconstrained measurement models, rather than comparing the respective chi-squared fit statistics. Both criteria have been widely used in previous studies. However, since Δχ² is sensitive to the sample size, Cheung and Rensvold (2002) proposed a more practical criterion for comparing nested models, which is to consider changes in the goodness-of-fit indices. An increase of at least 0.010 in the CFI and the TLI or a decrease of at least 0.015 in the RMSEA indicates that one model outperformed the other model. In addition, given that missing data were minimal (<0.06%), we used full information maximum likelihood (FIML; Graham, 2009) to deal with missing values.

Except for items 21 (r = 0.19, p < 0.001), 23 (r = −0.15, p < 0.001), 24 (r = −0.04, p = 0.09), 31 (r = 0.09, p < 0.001), 33 (r = 0.01, p = 0.77), 35 (r = 0.05, p = 0.02), 36 (r = 0.02, p = 0.32), and 37 (r = 0.03, p = 0.28), the ITC values of the other items were between 0.27 and 0.59, sp<0.001 (see Supplementary Table 1). In the item discriminant analysis, there were 486 and 493 participants in the low- and high-score groups, respectively. The independent samples t-tests indicated that, except for items 24 [t(967) = 0.86, p = 0.39], 33 [t(970) = 0.42, p = 0.67], 35 [t(971) = 1.96, p = 0.051], 36 [t(972) = 1.07, p = 0.28], and 37 [t(973) = 0.71, p = 0.48], the CR values of the other items were significant, sp<0.001.

In addition, in terms of content, items 21 (“In a group of people, I would not be embarrassed to be called on to start a discussion or give an opinion about something I know well”), 23 (“I am against giving money to beggars”), 24 (“It is unusual for me to express strong approval or disapproval of the actions of others”), 31 (“I keep out of trouble at all costs”), 33 (“I am easily downed in an argument”), 35 (“My conduct is largely controlled by the customs of those about me”), 36 (“It makes me uncomfortable to put on a stunt at a party even when others are doing the same sort of thing”), and 37 (“I find it hard to make small talk when I meet new people”) had low content discrimination validity among Chinese college students. Therefore, according to the above results, the items 21, 23, 24, 31, 33, 35, 36, and 37 were deleted.

In the first group (N = 911), the KMO test (0.91) and Bartlett’s test of sphericity (χ²/df = 21.88, p < 0.001) showed a good data fit for EFA. To decide the maximum number of factors, we performed a parallel analysis. The results showed that the maximum number of factors was four (see Table 1; Supplementary Figure 1). Subsequently, we used ESEM to evaluate one-factor, two-factor, three-factor, and four-factor models (see Supplementary Table 2). The ESEM results showed that, except for item 17, only the two-factor model had factor loadings greater than 0.30 for all items, and there were no cross-loading items (see Supplementary Table 3). More importantly, previous researchers (e.g., Bandalos and Gerstner, 2016; Lim and Jahng, 2019; Reio and Shuck, 2015) have suggested that the structure of a model should be consistent with the theory. Given that Block defined ego-control as the ability to express or inhibit impulses (Letzring et al., 2005), we finally extracted two factors: impulsive behavior and inhibitory behavior. Considering that the factor loadings were greater than 0.6, the structure of the factor was relatively stable and independent of the sample size (e.g., Beavers et al., 2013; Guadagnoli and Velicer, 1988). More importantly, the factor loadings were above 0.60, which indicated the unidimensionality of the measures (e.g., Asnawi et al., 2019; Mahfouz et al., 2020). This means that the items loaded on each dimension were relatively stable, and there was no cross-loading. Thus, in line with previous research (e.g., Awang, 2015; Wang et al., 2022), we removed items with factor loadings <0.60, starting with the lowest loading item and proceeding one item at a time, until there were no items with factor loadings <0.60. Finally, a total of 11 items were retained (see Table 2). Factor 1, named impulsive behavior, contained items 12, 13, 14, and 15. Factor 2, named inhibitory behavior, contained items 25, 26, 27, 28, 29, 30, and 32. The ESEM results showed a good model fit, χ²/df = 4.62, p < 0.001, CFI = 0.968, TLI = 0.948, RMSEA = 0.063, 90% CI = [0.053, 0.073], SRMR = 0.025 (see Supplementary Table 4).

In addition, in terms of content, items 12 (“I do not always tell the truth”), 13 (“My way of doing things can be misunderstood or bother others”), 14 (“Sometimes I rather enjoy going against the rules and doing things I am not supposed to”), and 15 (“At times, I am tempted to do or say something that others would think inappropriate”) represented impulsive behaviors for Chinese college students. In terms of content, items 25 (“I like to stop and think things over before I do them”), 26 (“I do not like to start a project until I know exactly how to proceed”), 27 (“I finish one activity or project before starting another”), 28 (“I am steady and planful rather than unpredictable and impulsive”), 29 (“On the whole, I am a cautious person”), 30 (“I do not let too many things get in the way of my work”), and 32 (“I consider a matter from every viewpoint before I make a decision”) represented inhibitory behaviors for Chinese college students. Therefore, these items are consistent with the assumption of this study. Notably, compared to the original 37-item scale, the shortened 11-item version is more concise (i.e., it determines which items are impulsive behaviors and which items are inhibitory behaviors), easier to administer, and more cost-effective. This facilitates the advancement of research related to ego-control.

In the second subsample (N = 912), a confirmatory factor analysis was performed to test the previously identified factor structure. The two-factor model fit was acceptable: χ²/df = 4.88, p < 0.001, CFI = 0.956, TLI = 0.944, RMSEA = 0.065, 90% CI = [0.057, 0.074], SRMR = 0.039. All standardized factor loadings for the CFA were significant (p < 0.001), ranging from 0.61 to 0.79 (see Supplementary Figure 2). Moreover, we examined the second-order two-factor and bi-factor models and found that the models did not converge with the data.

In the total sample (N = 1,823), the results of the measurement invariance tests across sex are shown in Supplementary Table 4. Measurement invariance was confirmed at all constructs, metrics, and scalar levels, with changes of less than 0.010 in CFI and TLI and less than 0.015 in RMSEA, compared to each model. The results suggested that ego-undercontrol scores had the same basic construct and measurement properties for both male and female participants (see Supplementary Table 5).

Reliability tests were carried out on the total sample. Two months later, we randomly selected a class to conduct the reliability tests. The class consisted of freshmen (N = 55), including 24 male students. There was no evidence to suggest that this class was not representative. The Cronbach’s alpha coefficients for all items, the impulsive behavior factor, and the inhibitory behavior factor were 0.78, 0.77, and 0.88, respectively. Two-month test–retest reliabilities for all items, the impulsive behavior factor, and the inhibitory behavior factor in the sample (N = 55) were 0.74, 0.74, and 0.89, respectively.

As shown in Table 3, ego-undercontrol and impulsive behavior were significantly positively associated with loneliness, depression, and negative affect. However, they were significantly negatively associated with self-esteem, life satisfaction, and positive affect. In contrast, inhibitory behavior was significantly positively associated with self-esteem, life satisfaction, and positive affect, whereas it was significantly negatively associated with loneliness, depression, and negative affect. Finally, ego-undercontrol was significantly negatively associated with inhibitory behavior. There was no significant association between impulsive behavior and inhibitory behavior.

There are some limitations to the current study that need to be addressed. First, the present study focused on the Chinese cultural context, which may limit the generalizability of the findings. Future research should explore whether these dimensions can be found to the same extent in other cultural contexts. Second, data from this study were collected from a sample of college students, raising questions about generalizability to other age groups. It is therefore important to validate the EUC-C in other samples (e.g., adolescents). Finally, further research is needed to explain the underlying mechanisms of the association between ego-control and psychological adjustment indicators, as well as the potential moderating factors of these associations.

There is currently a lack of measures assessing ego-control in the Chinese cultural context. The current study’s results support the EUC-C as a promising tool for measuring ego-control, undercontrol, and overcontrol in a Chinese context. As Block (2002) stated, the potential roles and applicability of the EUC are quite broad, as ego-control is conceptualized as the organizational structure of the personality system. Researchers studying self-control, self-regulation, self-depletion, delayed gratification, and other related areas may find that the EUC can be particularly useful (Letzring et al., 2005). For example, research has found that too much self-control (i.e., overcontrol) potentiates risk (Gilbert et al., 2020; Gilbert et al., 2022). Therefore, future research should investigate the mechanisms through which ego-control influences mental health, as well as the characteristics of the relationship between ego-control and self-control.