Psychometric Properties of the Young's Early Maladaptive Schema Questionnaire – Short Form (YSQ-S3) in Colombia (English Version)

Title

Psychometric Properties of the Young's Early Maladaptive Schema Questionnaire - Short Form (YSQ-S3) in Colombia

Complete Reference

González-Cifuentes, C. E., Ruiz, F. J., & Quiceno, J. M. (2024). Psychometric properties of the Young's Early Maladaptive Schema Questionnaire - Short Form (YSQ-S3) in Colombia. Behavioral Psychology / Psicología Conductual, 32(1), 65–87. https://doi.org/10.51668/bp.8324103n

Type of Study

Psychometric study with two independent cross-sectional studies

---

Context and Objectives

The research addresses the construct validity and reliability of the Young's Early Maladaptive Schema Questionnaire - Short Form (YSQ-S3), an instrument developed by Young (2005). Early maladaptive schemas (EMSs) are dysfunctional patterns of thinking, memories, and bodily sensations learned in childhood and adolescence, which are activated by events congruent with their contents, generating intense experiences of negative emotions (fear, sadness, anger, shame) and counterproductive or significant interpersonal problems.

EMSs have been theoretically and empirically linked to personality disorders, interpersonal problems, emotional and affective symptomatology, phobias, substance abuse, behavioral addictions, and general psychiatric symptoms. The YSQ-S3 was designed as a theoretically pure instrument for clinical assessment of these constructs, incorporating the 18 first-order factors from the original YSQ-L3 within a more practical brief format.

Although the YSQ-S3 has been studied in various languages and countries (Denmark, Germany, Italy, Spain, Chile, Mexico), no published study to date has evaluated its psychometric properties in the Colombian general adult population. Therefore, the objectives were: (1) to obtain evidence of validity and reliability of the YSQ-S3 in Colombia, (2) to re-explore its structure using a novel technique called exploratory graph analysis (EGA), and (3) in Study 2, to compare the best-fitting model from Study 1 against an alternative model.

---

Method

Participants

Study 1: A non-probabilistic sample of 1,004 participants who met criteria of being Colombian, residing in Colombia, and over 18 years of age was collected. The mean age was 27.33 years (SD = 11.80), with a range of 18 to 84 years. The sample included 564 women, 438 men, and 2 participants who identified as another gender. The most frequent educational level was high school completion (n = 421), followed by technical or technological level (n = 239), university degree (n = 221), postgraduate specialization training, master's or doctorate (n = 114), and finally 9 participants with elementary school education. The most frequent marital status was single (n = 732), followed by free union (n = 124), married (n = 118), and other (n = 30). The most frequent current occupation was university student (n = 425), private sector worker (n = 266), unemployed (n = 139), and other (n = 184).

Study 2: A non-probabilistic sample of 806 Colombian adults was collected under conditions similar to Study 1. The mean age was 29 years (SD = 12.80), with 464 women, 340 men, and 2 participants who identified as another gender. Most participants had completed high school (n = 320), followed by technical or technological level (n = 225), university degree (n = 174), and postgraduate training (n = 70). Only a small proportion had elementary school education (n = 17). The most frequent occupation was university student (n = 282), followed by private sector worker (n = 207), self-employed (n = 126), public sector worker (n = 91), technical or technological worker (n = 25), and other (n = 75).

Instrument(s) Evaluated

Young's Schema Questionnaire - Short Form 3 (YSQ-S3): A self-report instrument with 90 items answered on a six-point Likert-type scale (1 = completely untrue of me, 6 = describes me perfectly). The YSQ-S3 assesses dysfunctional patterns of thinking, memories, and bodily sensations in interaction mainly with the family of origin (Young et al., 2015). It is organized into 18 early maladaptive schemas theoretically grouped into five domains: (a) Disconnection and Rejection, (b) Impaired Autonomy and Performance, (c) Impaired Limits, (d) Other-Directedness, and (e) Overvigilance and Inhibition. Each of the 18 schemas is measured through 5 items (total 90 items).

Other Measurement Instruments

Ad hoc Sociodemographic Data Questionnaire: Information on sex, age, educational level, occupation, and marital status was collected through a closed questionnaire in an online form.

Depression, Anxiety and Stress Scale (DASS-21): An instrument with 21 items assessing negative emotions answered on a four-point Likert-type scale (0 = did not apply to me at all, 3 = applied to me most of the time). The DASS-21 has three subscales of seven items each: Depression, Anxiety, and Stress. The alpha coefficients for the Spanish version in Colombia were: .88 for Depression, .83 for Anxiety, .83 for Stress, and .95 for the total scale.

Data Analysis

Descriptive psychometric analyses were performed with SPSS v. 26. Confirmatory factor analyses (CFA) were conducted with JASP v. 0.17.1 using the structural equation modeling (SEM) module that runs the R package lavaan. Internal consistency of the YSQ-S3 was estimated using Cronbach's alpha coefficient and McDonald's omega.

Given that items are answered on a six-point Likert-type scale, the alpha coefficient is suitable for evaluating the degree to which the test measures the construct (Muñiz, 1997). For interpretation of the alpha coefficient, interpretation ranges suggested by George & Mallery (2003) were used: excellent > .80, good > .70, acceptable > .60, questionable > .50, poor < .50 unacceptable. The omega coefficient was calculated using the extension macro for SPSS. Chi-square conventions, relative chi-square ratio (with expected values ≤2 good and ≥3 acceptable), significance test, and goodness-of-fit indices were employed: CFI, TLI, NNFI, and SRMR. Values ≥ .95 for CFI, TLI, NNFI, and NFI were considered indicators of good fit, and values ≥ .90 as indicators of acceptable fit. Values ≤ .05 were considered to show good fit according to RMSEA and SRMR indices, while values ≤ .08 indicated acceptable fit (Jordan-Muiños, 2021).

A progressive CFA strategy was used to analyze the construct validity of the YSQ-S3. Four theoretical models were tested: (a) the 18 Early Maladaptive Schemas (EMSs) model generally regarded as best fitting the data (Sakulsipraser et al., 2016); (b) the model of 18 EMSs with a single second-order factor proposed by Kriston et al. (2012); (c) Young's original theoretical model (2005) consisting of 18 first-order schemas and five second-order domains; and (d) the most recent version proposed by Bach et al. (2018), proposing 18 EMSs and four second-order domains.

The R package EGAnet was used to conduct an exploratory graph analysis (EGA). EGA is a novel technique aimed at determining the dimensionality of psychometric measurement instruments by identifying communities of items in a network with powerful visual support (Golino et al., 2020). To our knowledge, this data analysis approach has not been used previously with the YSQ-S3 and might offer insights into factor structures that have been proposed and analyzed (Brown et al., 2023).

Finally, Pearson correlations were calculated between the scales of the YSQ-S3 and DASS-21. These correlations were interpreted following Cohen's (1988) criteria: correlation sizes from .10 to .29 are small, .30 to .49 are medium, and ≥ .50 are large.

---

Results

Validity Based on Internal Structure

Study 1

Table 1 shows that the unidimensional model obtained a good fit for twelve of the eighteen scales (CFI ≥ .95, TLI ≥ .95, NFI ≥ .95, RMSEA ≤ .05) and acceptable fit for five scales (CFI ≥ .90, TLI ≥ .90, NFI ≥ .90, and RMSEA ≤ .08).

For the Mistrust/Abuse scale, fit was acceptable for all but one of the four indicators (CFI = .92, TLI = .84, NFI = .91, RMSEA = .079). An additional confirmatory factor analysis was performed with the ULSMV estimator on the 18 correlated schemas, which showed excellent fit (CFI = .988, TLI = .987, NFI = .981, RMSEA = .042). An additional CFA was performed on the 18 schemas assuming uncorrelated factors; however, in this case the model was completely inadequate according to goodness-of-fit indices (CFI = .080, TLI = .059, NFI = .079, RMSEA = .362).

Given that the general model of 18 first-order correlated schemas showed goodness-of-fit indicators meeting criteria for a good model, factor loadings of items within this structure were analyzed (Table 2). It was observed that almost all factor loadings were ≥ .40, with the exception of two items with factor loadings < .40. The recommended standard in psychometrics (Lloret-Segura et al., 2014) was .30 on the Social Isolation scale, where item 22 ("I am fundamentally different from other people") obtained a factor loading of .32, which is still considered acceptable. On the Unrelenting Standards/Hypercriticalness scale, item 49 ("I must meet all my responsibilities") yielded a factor loading of .15, being an item with low factor loading.

Three hierarchical factor models including the 18 schemas were then compared: the model suggested by Young (2005), proposing five second-order factors; the model indicated by Bach et al. (2018), proposing four second-order factors; and the general maladaptivity model with a single second-order factor (Kriston et al., 2012). Observed variables were the instrument items, first-order factors were the schemas, and second-order factors were the domains. Table 3 shows two models that obtained incremental fit indicators greater than .95 and RMSEA ≤ .05. Unexpectedly, CFA results on second-order domains yielded an inadmissible solution for the four-domain model most recently proposed by Bach et al. (2018) due to a Heywood case. Therefore, the result indicates that the best fit for the data is Young's original model with five second-order domains.

Exploratory Graph Analysis (EGA)

Exploratory graph analysis was conducted in Study 1 and yielded six item communities: Community 1 consisted of 27 items, Community 2 consisted of 25 items, Community 3 consisted of 23 items, Community 4 consisted of 5 items, Community 5 consisted of 5 items, and Community 6 consisted of 5 items. The results of this analysis can be seen in Figure 1.

A theoretical analysis of EGA groupings allowed identification that Community 1 grouped all items from Defectiveness (5), Confused Attachment (5), and Failure (5) schemas, and most of Subjugation (4), Dependence (4), and Social Isolation (3). This item community replicates a mixture of the second-order theoretical dimensions proposed for the YSQ-S3. Community 2 groups all items from Grandiosity (5), Unrelenting Standards (5), most of Approval/Recognition Seeking (4), and Punitiveness (4). This community closely replicates a mixture of the second-order theoretical dimensions of Insufficient Limits and Excessive Standards. Community 3 groups all items from Abandonment (5), and most of Vulnerability to Harm and/or Illness (4), Insufficient Self-Control (4), Mistrust (3), and Negativity/Pessimism (3). This item community has a very mixed theme of insecure attachment, behavioral impulsivity, emotion regulation, and boundaries insecurity with experiences of emotional deprivation. Community 4 captures all Emotional Inhibition items (5). Community 5 contains almost all Emotional Deprivation items (4), and Community 6 captures Self-Sacrifice items.

In summary, exploratory graph analysis suggested three major dimensions proposing a new grouping of scales and domains but left three clusters unclustered. Thus, EGA results suggest an alternative organization of the YSQ-S3 into three broad dimensions that integrate multiple EMSs, though the conceptual analysis suggests these reflect the underlying theoretical dimensions of the YSQ-S3 in a mixture of themes involving insecure attachment, behavioral regulation difficulties, and boundaries insecurity with emotional deprivation experiences.

Validity Based on Relationships with Other Variables

Study 1

Table 4 presents Pearson correlations between YSQ-S3 and DASS-21 scales. Overall, the early maladaptive schemas that yielded the highest correlation sizes with the DASS-21 negative affect total score were Negativity/Pessimism (r = .64), Vulnerability to Harm and/or Illness (r = .64), Social Isolation (r = .61), and Subjugation (r = .61), while the smallest, though still moderate, correlation sizes were with Unrelenting Standards/Hypercriticalness (r = .35), Self-Sacrifice (r = .41), Emotional Inhibition (r = .45), and Approval/Recognition Seeking (r = .49). The other scales showed strong correlations with the DASS-21 total score > .50. For YSQ-S3 domains, all correlations were strong with the DASS-21 total score.

Internal Consistency

Study 1

Alpha coefficients ranged from .67 to .89, with a total coefficient of .97. The omega coefficient ranged from .67 to .86, and a total omega of .97. The lowest discrimination index was .21 for item 22 ("I am fundamentally different from other people"). This was the only item with an index below .20, while the highest discrimination index was .77 for item 38 ("I worry that people I feel close to will leave me or abandon me"). Additionally, alpha coefficients were calculated for items grouped by domains according to Young's (2005) model, obtaining good to excellent coefficients ranging from .81 to .93 (Table 5).

Study 2

Results showed that the two models had good fit, with minimal differences (Table 7). The 18 correlated first-order early maladaptive schemas model was still superior due to its better incremental fit indicators, lower error, and relative chi-square ratio over degrees of freedom below the alternative model. However, the six-factor model performed very similarly being parsimonious, suggesting a direction for future research.

---

Discussion and Conclusions

The results of Study 1 provided support for the validity and reliability of the YSQ-S3 as a measure of early maladaptive schemas in Colombia. The confirmatory factor analyses support the construct validity of the instrument, consistent with test evidence reported by other authors and with the scientific literature available for the YSQ-S3 (Calvete et al., 2013; Kriston et al., 2012; Quiñones et al., 2018). Additionally, the validity of the YSQ-S3 with respect to external criteria was demonstrated, as expected relationships between constructs were found. Finally, the test's internal consistency was adequate for all but one scale. These findings indicate that the YSQ-S3 is a valid and reliable measure of early maladaptive schemas for use in Colombia.

The EGA results suggest an alternative organization of the YSQ-S3 into three broad dimensions integrating multiple early maladaptive schemas. The groupings make sense if analyzed conceptually as mixtures of the theoretical dimensions of the YSQ-S3. The exploratory graph analysis results provide evidence that the hierarchical organization of the instrument into five second-order domains has the best fit indices. Additionally, through second-order analysis, evidence was found supporting the hierarchical organization of the instrument into five second-order domains.

In Study 2, in which the EGA results obtained in the first study were incorporated, the six-dimension model proposed by the analysis was found to fit almost as well as the alternative 18 early maladaptive schemas model. A conceptual analysis suggests that this reflects the underlying theoretical dimensions of the YSQ-S3. Community 1 groups elements of defective selves with a mixture of themes of disconnection, rejection, and impaired autonomy. Community 2 reflects the idea of superiority and a mixture of vital themes of insufficient limits and excessive standards. Community 3 was a very mixed theme of insecure attachment, behavioral impulsivity, emotion regulation, and boundaries insecurity with emotional deprivation experiences. According to schema therapy theory, this latter dimension seems to involve a history of toxic frustration of multiple needs. In any case, it is suggested that a future line of research be deepened in subsequent studies.

Regarding limitations of the present studies: sampling was incidental online, which raises caution regarding the generalizability of population parameters and the lack of a clinical sample. Additionally, it is important to mention that only a brief emotional symptomatology assessment instrument was used as an external criterion. For future studies, it is suggested to use a well-established and adequate personality test to identify personality psychopathology, such as the Exploratory Personality Questionnaire III (EPQ-III) (Caballo et al., 2011), which has also been validated for Colombia (González-Cifuentes & Vera-Maldonado, 2015). It is also suggested to conduct pencil-and-paper sampling of the instrument and to include a sample of patients, especially those with interpersonal problems, persistent emotional or affective personality disorders.

In conclusion, the YSQ-S3 is a valid and reliable instrument for the Colombian adult population. According to CFA results, the presence of 18 correlated early maladaptive schemas was validated with excellent goodness-of-fit indicators in both studies. Additionally, evidence was obtained supporting the organization of first-order factors into a hierarchical structure of five second-order factors or theoretical domains according to schema therapy theory. On the other hand, reliability coefficients were between acceptable and excellent for 17 of the 18 subscales.

---

Significance and Contribution

This study provides the first comprehensive psychometric validation of the YSQ-S3 in Colombia, filling an important gap in the Latin American literature on the measurement of early maladaptive schemas. The findings demonstrate that the instrument possesses solid psychometric properties with excellent internal consistency (total ω = .97) and a confirmed factorial structure. The validation using two independent samples (N = 1,004 in Study 1; N = 806 in Study 2) strengthens the robustness of the study. The innovative use of exploratory graph analysis provides new insights into the dimensional structure of the instrument. This study makes a significant contribution to the field of clinical assessment in Spanish-speaking Latin American populations, providing a reliable and valid tool for the measurement of early maladaptive schemas in clinical research and practice contexts, enabling both identification of dysfunctional cognitive patterns and assessment of therapeutic change.

---

---