Schema therapy (also known as schema-focused therapy) was developed by Young, Klosko and Weishaar (Reference Young, Klosko and Weishaar2003). It focuses on the identification and modification of core schemas (also referred to as early maladaptive schemas, EMSs). Young (Reference Young1994) defined EMSs as “extremely stable and enduring themes that develop during childhood and are elaborated upon throughout an individual’s lifetime” (p. 9). The EMSs combine an inherited, vulnerable temperament with a dysfunctional experience with parents, siblings and peers. Young postulated that a schema is located at the deepest level of cognition and that a person is not always conscious of his/her schemas’ impacts on emotions and behaviors. According to Young’s theory, EMSs are comprised of memories, emotions, cognitions and bodily sensations regarding oneself and one’s relationships with others. They are underlying cognitive structures that organize one’s experience of the world, and reflect childhood experiences related to attachment and approval/disapproval experiences. Young hypothesized that a child must negotiate five primary tasks in order to develop in a healthy manner: Connectedness, autonomy, worthiness, reasonable expectations, and realistic limits (Rafaeli, Bernstein, & Young, Reference Rafaeli, Bernstein and Young2011). He proposed that EMSs arise when parents or caregivers make it difficult to achieve one or more of these basic needs. EMSs are described as general vulnerability factors for a broad range of psychopathologic conditions, and constitute predisposing factors for the development and maintenance of clinical symptom states (Young, Reference Young1994; Young et al., Reference Young, Klosko and Weishaar2003). In other words, EMSs are not specific for a given pathology, e.g. borderline personality disorder or obsessive compulsive disorder (OCD), and can be considered to be “transdiagnostic”. It has also been postulated that EMSs are present in non-clinical populations - albeit to a lesser extent than in clinical populations. Examples of EMSs are dependence (the belief that one is unable to have one’s everyday responsibilities without help from others) or failure (the belief that one has failed or is inadequate, relative to others).

Young (Reference Young2005) hypothesized that there are 18 EMSs, which can be allocated to five domains reflecting childhood tasks: (A) Disconnection & rejection, (B) impaired autonomy & performance, (C) impaired limits, (D), other-directedness, and (E) overvigilance & inhibition. These domains are linked to their hypothetical origins in childhood. For example, the disconnection & rejection domain is associated with safe attachment and protection; the impaired autonomy & performance domain is linked to autonomy and competency; the impaired limits domain is linked to self-control; other-directedness is associated with free expression of needs, and the overvigilance and inhibition domain is associated with spontaneity and play (Kriston, Schäfer, Jacob, Härter, & Hölzel, Reference Kriston, Schäfer, Jacob, Härter and Hölzel2013). Five schemas are linked to domain A: Abandonment/instability, mistrust/abuse, emotional deprivation, defectiveness/shame, and social isolation/alienation. Four schemas make up Domain B: Dependence/incompetence, vulnerability to harm or illness, enmeshment/undeveloped self, and failure. Two schemas are linked to Domain C: Entitlement/grandiosity, and insufficient self-control/self-discipline. Domain D includes three schemas: Subjugation, self-sacrifice and approval-seeking/recognition-seeking. Domain E is composed of four schemas: Negativity/pessimism, emotional inhibition, unrelenting standards/hypercriticalness, and punitiveness (Rafaeli et al., Reference Rafaeli, Bernstein and Young2011). The latest short version of the YSQ (YSQ–S3; Young, Reference Young2005) includes 90 items (five for each of the 18 schemas).

The first study of the YSQ–S3 (Soygut, Karaosmanoglu, & Çakir, Reference Soygut, Karaosmanoglu and Çakir2009) focused on Turkish university students. The researchers performed two principal component analyses: One on the 90 items and the other on first-order factors (in order to explore higher-order factors). They selected 14 first-order factors (emotional deprivation, failure, pessimism, social isolation/mistrust, emotional inhibition, approval seeking/recognition seeking, enmeshment/dependence, entitlement/insufficient self-control, self-sacrifice, abandonment, punitiveness, defectiveness, vulnerability to harm, unrelenting standards) and five higher-order factors. For three EMSs (emotional deprivation, emotional inhibition and self-sacrifice), the items were grouped together in the same way as in Young’s model. The other three first-order factors corresponded to Young’s schemas (failure, approval seeking, and abandonment). The other schemas corresponded to clusters of Young’s theoretical EMSs (i.e. enmeshment, dependence & subjugation; social isolation & mistrust abuse; entitlement & insufficient self-control). Of the five higher-order factors, four had been described by Young (impaired autonomy & performance, disconnection & rejection, impaired limits, and other-directedness). The items (schemas) which compose these higher-order factors generally correspond to Young’s theoretical model, with the exception of the other-directedness domain. The fifth higher-order factor (unrelenting standards) covers the unrelenting standards and approval-seeking schemas. The “overvigilance & inhibition” higher-order factor was not found in the Turkish version. Next, Soygut et al. (Reference Soygut, Karaosmanoglu and Çakir2009) explored the YSQ–S3’s internal reliability, test-retest reliability, convergent validity and criterion-related validity. The researchers studied the convergent validity with the Symptom Checklist–90–Revised (SCL–90–R). A clinical population was compared with a population of students: Nine of the 14 schemas (emotional deprivation, failure, pessimism, social isolation, emotional inhibition, enmeshment, abandonment, defectiveness, and vulnerability to harm) and two of the five domains (impaired autonomy & performance, and disconnection & rejection) differentiated between the two groups. The second study (Saritas & Gençöz, Reference Saritas and Gençöz2011) of the YSQ–S3 also focused on the Turkish version but applied it to an adolescent sample (mean age: 16). The researchers used principal component analysis to examine the higher-order factor structure. Only three higher-order factors were found: Impaired limits/exaggerated standards, disconnection & rejection, and impaired autonomy/other-directedness. The first factor (the domain) included two schemas from Young’s impaired limits domain and three schemas from the overvigilance & inhibition domain. The second factor included three schemas from Young’s disconnection & rejection domain, together with schemas from other domains. The third factor included three schemas from the impaired autonomy domain and some schemas from other domains. Next, the convergent validity of the three higher-order factors was studied with regard to the State-Trait Anxiety Inventory and State-Trait Anger Inventory. The third study of the YSQ–S3 looked at the French Canadian version (Hawke & Provencher, Reference Hawke and Provencher2012). Eighteen confirmatory factor analyses (CFAs) were performed on the five items of each EMS, in a mixed population (96 axis I patients and 973 non-clinical participants). The single-factor model was confirmed for 14 EMSs but not for the punitiveness, enmeshment, entitlement and self-sacrifice schemas. Next, five CFAs were conducted on the higher-order domains. Only three of the five domains demonstrated a good fit to the data (disconnection and rejection, impaired autonomy and performance, and other-directedness). Lastly, the researchers tested two full models in a CFA: the first model brought together the five correlated domains with their respective EMSs, and the second model added the total score as a third-order factor. These two models failed to produce acceptable goodness-of-fit indexes. The researchers concluded that the internal validity of the YSQ–S3’s second-order factors (domains) had not been confirmed. Next, the two groups of participants (clinical and non-clinical) were compared: All but two of the EMSs (entitlement and unrelenting standards) differentiated between the groups. In a fourth study, Kriston, Schäfer, Jacob, Härter, and Hölzel (Reference Kriston, Schäfer, Jacob, Härter and Hölzel2013) studied the factorial validity of the German version of the YSQ–S3 in a community sample (n = 1,150). They performed a CFA of the 18 correlated factors corresponding to Young’s model. This version showed “acceptable” factorial validity. The reliability of 17 of the 18 schema scales was sufficient. The convergent validity (studied with the SCL–K–9) and the criterion-related validity were satisfactory. The clinical sample was composed of inpatients and outpatients. All the schemas differentiated between clinical participants and those reporting no mental health problems. In the fifth study, the Spanish version of the YSQ–S3 were studied in a population of 971 students (Calvete, Orue, & Gonzalez-Diez, Reference Calvete, Orue and Gonzalez-Diez2013). Several CFAs were performed, in order to probe the 18 EMSs and the five domains proposed by Young. The correlated 18-factor first-order structure yielded better goodness-of-fit indexes than the uncorrelated 18-factor first-order structure. Next, the researchers studied the second-order structure of the YSQ–S3 in an exploratory factor analysis (EFA). Three factors were found. The first second-order factor (an unlabeled factor) loaded on the scales of self-sacrifice, approval seeking, unrelenting standards, negativity, punitiveness and entitlement. The second factor (impaired autonomy & performance) loaded on the scales of failure, dependence, vulnerability to harm, enmeshment, abandonment, subjugation and insufficient self-control. The third factor (disconnection & rejection) loaded on the scales of emotional deprivation, mistrust, defectiveness, social isolation and emotional inhibition. Next, the researchers performed several CFAs to compare this solution with previous second-order solutions. The researchers found that all the hierarchical models showed acceptable goodness-of-fit indexes but that their own solution was better (for three domains). They concluded that only two domains (disconnection & rejection, and impaired autonomy) may have been validated, with slight changes. The internal reliability was not adequate for six EMSs (dependence, vulnerability to harm, enmeshment, subjugation, unrelenting standards, and entitlement). The test-retest reliability (6 months) was relatively adequate. Convergent validity was studied with the SCL–90–R and the Social Anxiety Questionnaire for Adults. The results of this study should be interpreted with caution because all the analyses (i.e. the CFAs and EFAs) were performed on the same sample. Lastly, Bach, Simonsen, Christoffersen, and Kriston (Reference Bach, Simonsen, Christoffersen and Kriston2017) examined the factorial validity of the Danish version of the YSQ–S3 in a mixed clinical and non-clinical sample by using the same methodology as Kriston et al. (Reference Kriston, Schäfer, Jacob, Härter and Hölzel2013). The researchers concluded that the evidence for the postulated first-order 18-factor structure was weak but acceptable. The criterion-related validity was satisfactory for the 18 schema scales. The internal consistency of the 18 EMSs was good.

Hence, the six studies of the YSQ–S3 have yielded contradictory results. An exploratory method (principal component analysis) has been applied once to the full set of 90 items (Soygut et al., Reference Soygut, Karaosmanoglu and Çakir2009) and three times to the 18 schemas (Calvete et al., Reference Calvete, Orue and Gonzalez-Diez2013; Saritas & Gençöz, Reference Saritas and Gençöz2011; Soygut et al., Reference Soygut, Karaosmanoglu and Çakir2009). Fourteen schemas were identified in a Turkish non-clinical population but only six of these were similar to those postulated by Young (Soygut et al., Reference Soygut, Karaosmanoglu and Çakir2009). In this new structure, the number of items per factor varied. The 14 schemas were clustered into five higher-order factors: four had been described by Young, and one was new (with unrelenting standards replacing Domain E, overvigilance and inhibition). The other two studies to have explored the schemas’ domains (Calvete et al., Reference Calvete, Orue and Gonzalez-Diez2013; Saritas & Gençöz, Reference Saritas and Gençöz2011) found three-factor solutions. Two domains (impaired autonomy, and disconnection & rejection) were selected in both studies. The third factor was referred to as “impaired limits/exaggerated standards” in the Turkish version (Saritas & Gençöz, Reference Saritas and Gençöz2011) and as an “unlabeled factor” in the Spanish version (Calvete et al., Reference Calvete, Orue and Gonzalez-Diez2013). These two studies yielded different schema clustering patterns. In both studies, the impaired autonomy Domain (B) encompassed three EMSs from Young’s theoretical model (dependence, enmeshment and vulnerability), and two schemas from other domains (subjugation in the other-directness domain, and abandonment in the disconnection & rejection domain). The disconnection & rejection domain (A) covered three EMSs from Young’s postulated domain (emotional deprivation, social isolation and defectiveness) and one EMS from another domain (emotional deprivation). In other words, the results of the various exploratory analyses differed from the Young’s theoretical model of schemas and domains, and also varied from one study to another. Three CFAs have been performed on the 18 schemas (Bach et al., Reference Bach, Simonsen, Christoffersen and Kriston2017; Calvete et al., Reference Calvete, Orue and Gonzalez-Diez2013; Kriston et al., Reference Kriston, Schäfer, Jacob, Härter and Hölzel2013). The structure with 18 correlated first-order factors postulated by Young showed acceptable factorial validity in two studies (Calvete et al., Reference Calvete, Orue and Gonzalez-Diez2013; Kriston et al., Reference Kriston, Schäfer, Jacob, Härter and Hölzel2013) but was weakly supported in one study (Bach et al., Reference Bach, Simonsen, Christoffersen and Kriston2017). Two studies included a CFA of the 18 SMSs grouped into five domains (Calvete et al., Reference Calvete, Orue and Gonzalez-Diez2013; Hawke & Provencher, Reference Hawke and Provencher2012). The correlated five-factor structure failed to produce acceptable goodness-of-fit indexes in the two studies. In the study of the French Canadian version (Hawke & Provencher, Reference Hawke and Provencher2012), the single-factor model was confirmed for 14 EMSs but the structural validity of the second-order factors was not confirmed. Only one study (Calvete et al., Reference Calvete, Orue and Gonzalez-Diez2013) used a confirmatory method to test the hierarchical structure of the YSQ–S3 by comparing several models (including that Young’s model). The various solutions gave acceptable goodness-of-fit indexes but the researchers concluded that their own solution was better. Only two domains (disconnection & rejection, and impaired autonomy & performance) may have been validated.

The study of convergent validity was performed twice with the full SCL–90 R (Calvete et al., Reference Calvete, Orue and Gonzalez-Diez2013; Soygut et al., Reference Soygut, Karaosmanoglu and Çakir2009) and once with a shorter version (SCL–K–9; Kriston et al., Reference Kriston, Schäfer, Jacob, Härter and Hölzel2013). One study (Saritas & Gercoz, 2011) focused on the relationships between the schemas on one hand and anxiety and anger on the other. Another study has looked at the relationships between the schema scales and social anxiety (Calvete et al., Reference Calvete, Orue and Gonzalez-Diez2013). The YSQ–S3’s convergent validity has not been extensively studied. Four studies have examined the criterion-related validity (Bach et al., Reference Bach, Simonsen, Christoffersen and Kriston2017; Hawke & Provencher, Reference Hawke and Provencher2012; Kriston et al., Reference Kriston, Schäfer, Jacob, Härter and Hölzel2013; Soygut et al., Reference Soygut, Karaosmanoglu and Çakir2009).

The YSQ–S3 has been translated into several languages (Turkish, German, Dutch, Finnish, Greek, Spanish and French). However, studies of the factorial structure have given some contradictory results. Hence, a firm basis for the structure of the EMSs needs to be developed. Studies of the YSQ–S3’s convergent validity and criterion-related validity are thus required. The objective of the present study was therefore to investigate the YSQ–S3’s first-order and higher-order factorial structure in a French population. Internal consistency, convergent validity and criterion-related validity were assessed. The links between some personality traits considered to be stable in adulthood (neuroticism, extraversion, and psychoticism) and the (YSQ–S3’s) SMSs have never been studied. We thought that it would be valuable to study the SMSs that are constituted in childhood and their links with the memories of an adult concerning dysfunctional parental attitudes (overprotection or rejection) or functional parental attitudes (warmth).

Results

Study 1

Exploratory factor analyses of the domains

Our results for the disconnection & rejection Domain A are shown in Table 1. Five factors were found, as in Young’s theoretical solution (2005). The first factor corresponded to defectiveness/shame (S4), and comprised five items (5, 23, 41, 59 and 77). The second factor corresponded to abandonment/instability (S1), and comprised only four items (2, 20, 38 and 56). The third factor corresponded to emotional deprivation (S3), and comprised four items (1, 37, 55 and 73). The fourth factor corresponded to mistrust/abuse (S2), with five items (3, 21, 39, 57 and 75). The fifth factor corresponded to social isolation/alienation (S5), and comprised five items (4, 22, 40, 58 and 76). Two items (19 and 74) were deleted due to dual saturation. The confirmatory indices obtained for this five-factor solution were satisfactory (χ² = 4,937.03, df = 300, RMSEA = .05, CFI = .97, TLI = .96 and SRMR = .03).

Table 1.

Domain A: Exploratory Factor Analysis

Note: The loadings for items selected for the factor are shown in bold type.

Table 2 corresponds to Domain B (impaired autonomy & performance), which covers four factors. The first factor corresponded to enmeshment/undeveloped self (S8), and comprised five items (9, 27, 45, 63 and 81). The second factor corresponded to failure (S9), and comprised five items (6, 24, 42, 60 and 78). The third factor corresponded to vulnerability to harm/illness (S7), and comprised three items (26, 44 and 80). Lastly, the fourth factor corresponded to dependence/incompetence (S6), and comprised three items (43, 61 and 79). Two items (7, 8) were deleted due to dual saturation, and two others (25 and 62) were deleted because of the absence of saturation. The confirmatory indices obtained for this four-factor solution were satisfactory (χ² = 3486.99, df = 190, RMSEA = .05, CFI = .97, TLI = .95 and SRMR = .04).

Table 2.

Domain B: Exploratory Factor Analysis

Note: The loadings for items selected for the factor are shown in bold type.

Domain C (Table 3) included two factors. The first factor (corresponding to entitlement, S10) included three items (32, 68 and 86). The second factor (insufficient self-control, S11) included five items (14, 15, 33, 51 and 87). Two items (50, 69) were deleted due to dual saturation. The confirmatory indices obtained for this solution with two factors are satisfactory (χ² = 794.97, df = 45, RMSEA = .09, CFI = .92, TLI = .86 and SRMR = .06) with the exception of the RMSEA and the SRMR (> .05).

Table 3.

Domain C: Exploratory Factor Analysis

Note: The loadings for items selected for the factor are shown in bold type.

Domain D (other-directedness) encompassed three factors (Table 4). The first factor corresponded to self-sacrifice (S13), and included five items (11, 29, 47, 65 and 83). The second factor corresponded to subjugation (S12), and included six items (10, 28, 46, 52, 64 and 82). Lastly, the third factor (approval seeking, S14) includes four items (16, 34, 70 and 88). The confirmatory indices obtained for this three-factor solution are satisfactory (χ² = 1741.32, df = 105, RMSEA = .04, CFI = .98, TLI = .97 and SRMR = .03).

Table 4.

Domain D: Exploratory Factor Analysis

Note: The loadings for items selected for the factor are shown in bold type.

Domain E (over vigilance and inhibition) was constituted of four factors (Table 5). The first factor (emotional inhibition, S16) was composed of Items 12, 30, 66, and 84. The second factor is negativity/pessimism (S15), which included the Items 17, 35, 53, 71, and 89. The third factor (unrelenting standards, S17) was composed of 3 items 13, 31, 67. Lastly, the fourth factor (punitiveness, S18) included five items (18, 36, 49, 54, and 72). Three items (48, 85 and 90) were deleted due to dual saturation. The confirmatory indices obtained for this four-factor solution were satisfactory (χ² = 2,952.42, df = 190, RMSEA = .05, CFI =.96, TLI = .94 and SRMR = .04)

Table 5.

Domain E: Exploratory Factor Analysis

NB: the loadings for items selected for the factor are shown in bold type.

Confirmatory factor analyses of the domains

We performed three CFAs on a second sub-sample: The solution that we found with the first sample, a one-factor solution, and Young’s theoretical solution. Our solution gave the best indices in all cases (Table 6).

Table 6.

Goodness-of-fit Indexes in the CFAs of the Domains

Note: The selected solution is shown in bold type. RMSEA = root mean square error of approximation; CFI = comparative fit index; TLI = Tucker-Lewis index; WRMR = weighted root mean square residual.

Confirmatory factor analyses of the higher-order structure

Using the study population as a whole, we performed CFAs on three solutions: Young’s theoretical solution, our five-factor solution and Calvete et al.’s (Reference Calvete, Orue and Gonzalez-Diez2013) three-factor solution (Table 7). The confirmatory indices obtained for Young’s theoretical solution were χ² = 1,024.76, with df = 125, RMSEA =.10, CFI =.78, TLI =.73, SRMR = .06 and AIC = 60,338.82. The confirmatory indices obtained for our five-factor solution were as follows: χ² = 10,41.71, df = 125, RMSEA = .11, CFI =.76, TLI = .71, SRMR = .07 and AIC = 58,566.44. Lastly, the confirmatory indices obtained for Calvete et al.’s three-factor solution were χ² = 967.35, df = 132, RMSEA =.10, CFI =.80, TLI =.77, SRMR =.06 and AIC = 60,338.82. The three models tested gave similar but not fully satisfactory results, although our five-factor solution yielded the smallest AIC. Consequently, we decided to select this solution for our subsequent analyses.

Table 7.

Factorial Solution of Calvete’s Model, Young’s Model and Our Five-Factor Solution

Note ^a: the items added to our factorial solution are indicated between square brackets and in italics; the deleted items are indicated between brackets and in italics.

Mean (SD) and internal consistency of the 18 schemas

The mean (SD) values were calculated for all the participants in the non-clinical group (Table 8). The coefficients for internal consistency were satisfactory for 14 schemas in our factorial solution (S1, S2, S3, S4, S5, S8, S9, S12, S13, S14, S15, S16, S17 and S18), and ranged from .70 to .88. We obtained coefficients below .70 for two schemas (S6 and S7) in the impaired autonomy & performance domain and for two schemas (S10 and S11) in the impaired limits domain.

Table 8.

Internal Consistency, and Means (SD) (whole Sample)

Convergent validity

The product-moment (r) was interpreted according to Cohen’s criteria (1992). The assessment of convergent validity with the EPQR-A revealed some statistically significant correlations between the neuroticism scale and the 18 schemas (Table 9). We noted a strong correlation with S15, nine moderate correlations (the five schemas in Domain A, S6 and S7 in Domain B, and S11 and S12) and eight weak correlations (S8, S9, S10, S13, S14, S16, S17 and S18). We noted four moderate correlations (S4, S5, S12 and S16) and seven weak correlations (S2, S3, S6, S8, S9, S11 and S15) with the extraversion scale. The psychoticism scale showed nine weak correlations with S3, S4, S5, S8, S10, S11 (all from Domain C), S13, S17 and S18. The convergent validity with the s-EMBU revealed statistically significant correlations between the maternal and paternal rejection scales and 12 schemas (S4, S3, S2, S5, S9, S6, S10, S11, S13, S12 S15 and S17). Schemas S7 and S16 were weakly correlated with the paternal rejection scale, and schemas S14 and S18 were weakly correlated with the maternal rejection scale. The maternal and paternal emotional warmth scales were significantly and negatively correlated with 8 schemas (S4, S3, S2, S5, S9, S11, S12 and S16). Four schemas (S1, S6, S13 and S15) were weakly correlated with the paternal emotional warmth scale, and two schemas (S8 and S18) were weakly correlated with the maternal emotional warmth scale. The maternal and paternal overprotection scales were weakly correlated with 5 schemas (S8, S6, S11, S12 and S15). Two schemas (S7 and S14) were weakly correlated with the paternal overprotection scale, and eight schemas (S4, S3, S2, S5, S9, S13, S17 and S18) were weakly correlated with the maternal overprotection scale. All but five of the correlations were weak (Cohen, Reference Cohen1992). Schema 3 was moderately correlated with the maternal and paternal rejection scales and the maternal and paternal emotional warmth scales. Schema 5 was moderately correlated with the paternal rejection scale.

Table 9.

Correlations between the YSQ–S3 and Two Questionnaires (s-EMBU and EPQR-A)

*p < .05.

Study 2

The OCD patients obtained higher scores than healthy participants for the 18 schemas. Fourteen schemas (S3, S2, S5, S9, S8, S7, S10, S11, S12, S13, S14, S16, S17 and S18) did not differentiate between the groups (Table 10). There was a statistically significant intergroup difference for four schemas: Abandonment/instability (S1), defectiveness/shame (S4), dependence/incompetence (S6) and negativity/pessimism (S15).

Table 10.

Mean (Standard Deviation) of the 18 Schemas in the two Groups

*Bonferroni correction: p < .05/18= .003.