An Updated Validation of Persian Versions of Six Personality Questionnaires: TEMPS-A,AFECTS,TCI,PANAS,NEO-FFI,PDQ-4

Abstract

Background:

The revision and updating of questionnaires are part of the list of comprehensive guidelines for using psychological questionnaires. The present study aimed to test the construct and convergent validity of the Persian versions of six personality questionnaires including the Temperament Evaluation of Memphis, Pisa, Paris, and San Diego-Autoquestionnaire (TEMPS-A: 35 items), the Affective and Emotional Composite Temperament Scale (AFECTS: 48 items), the Temperament and Character Inventory (TCI: 125 items), the Positive Affect and Negative Affect Schedule (PANAS: 20 items), the NEO-Five-Factor Inventory (NEO-FFI: 60 items), and Personality Diagnostic Questionnaire-4 (PDQ-4: 99 items).

Methods:

The data from four independent samples of Iranian adults (N1 = 1137, N2 = 558, N3 = 496, N4 = 478; total N = 2669, 70% female, 32.8 ± 10.7 old years) were analyzed by exploratory structural equation modeling (ESEM). Pearson correlations between constructs of the Personality Inventory for DSM-5-Short Form (PID-5-BF: 25 items) or the ICD-11 compatible Personality Inventory for DSM-5–Brief Form Plus, Modified (PID5BF+M: 36 items) and the personality measures were used to evaluate convergent validity.

Results:

Although the number of factors of TEMPS-A and PDQ-4 were slightly different from the original structures, the results supported the factor structure and reliability (all α between 0.68 and 0.92) of all measures. Most factors of all measures were significantly related to either the PID-5-BF or the PID5BF+M constructs (all P <.05).

Conclusion:

The findings supported both construct validity and convergent validity of all personality measures in Iranian samples. Mental health professionals can use the Persian version of all valid personality measures to draw personality profiles of adult populations.

Key Messages:

• The Persian versions of all six personality scales are valid and reliable measures.

• The TEMPS-A has a six-factor rather than five-factor structure.

• The PDQ-4 is appropriate to evaluate a general factor of personality pathology.

• The results supported the convergent validity of all personality measures.

• The updated psychometrics facilitated applicability in clinical settings.

Self-report questionnaires are the most common method for the measurement of adaptive and maladaptive features of personality.^1,2 Personality questionnaires are very broad and originate from different conceptual models such as the Five-Factor Model (neuroticism, extraversion, openness, agreeableness, and conscientiousness) measured by the NEO-Five-Factor Inventory (NEO-FFI),^3,4 the two-factor model of positive and negative affect/temperament measured by the Positive Affect and Negative Affect Schedule (PANAS), ⁵ the model of affective temperaments (depressive, cyclothymic, hyperthymic, irritable, and anxious temperaments) measured by the Temperament Evaluation of Memphis, Pisa, Paris, and San Diego-Autoquestionnaire (TEMPS-A),^6,7 the model of affective (12 components) and emotional (volition, anger, coping, inhibition, sensitivity, and control) composites of temperament assessed by the Affective and Emotional Composite Temperament Scale (AFECTS), ⁸ the model of temperament and character (novelty seeking [NS], harm avoidance [HA], reward-dependence [RD], persistence [PS], self-directedness [SD], cooperativeness [C], and self-transcendence [ST]) measured by the Temperament and Character Inventory (TCI),^9,10 and the Diagnostic and Statistical Manual of Mental Disorders (DSM) model of personality disorders assessed by the Personality Diagnostic Questionnaire-4 (PDQ-4),^11,12 The Persian version of all questionnaires has been translated and validated during the last three decades.^13-18

For several reasons, the present study aimed to update the Persian versions of some validated personality questionnaires (i.e., TEMPS-A, AFECTS, TCI, PANAS, NEO-FFI, and PDQ-4) in Iran. First, validation studies of Persian versions of questionnaires reported the results between 2001 and 2018 (with a median of 2012). Although there are still no clear criteria for when to revise and update psychological tests, ¹⁹ there are considered time-dependent changes of variables that are measured using earlier validated questionnaires. Second, the validation studies included a small sample and did not collect data from the population of western Iran. In addition, five out of six samples were college students rather than the community population.^13,14,16-18 Third, none of the studies used exploratory structural equation modeling (ESEM), which provides a confirmatory assessment of the initial factor structures while also enabling the estimation of all cross-loadings. ²⁰ Instead, two studies reported EFA results, and only one study used both exploratory factor analysis (EFA) and confirmatory factor analysis (CFA).^14,16,18 Other studies did not report model fit indices at all.^13,15,17 Fourth, only one of the studies that applied EFA extracted factors using an oblique rotation. ¹⁸ Usually, there is a moderate to strong correlation between the items of each personality questionnaire. Therefore, oblique rotation methods such as Oblimin are preferable to orthogonal rotations. Fifth, all validation studies were low-quality reports according to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE). ²¹ Sixth, none of the validation studies reported correlations (i.e., convergent validity) between the extracted factors and current transdiagnostic personality constructs (e.g., the DSM-5 and the International Classification Diseases-11 trait models). Finally, the studies did not comprehensively include several adaptive and maladaptive personality questionnaires.

In sum, revision and updating of questionnaires are part of the comprehensive guidelines for using psychological tests. ²² According to the limitations highlighted in the research literature, the present study aimed to evaluate the psychometric properties of the Persian versions of six personality questionnaires, including TEMPS-A, AFECTS, TCI, PANAS, NEO-FFI, and PDQ-4. These six scales were chosen because they are commonly used in research and clinical settings in Iran. Although PANAS is mainly used to measure affect states, it was included in the list of personality scales because it can measure temperament when a person is asked to answer questions regarding his/her features during adulthood. In detail, the first objective of the present study was to test the factor structure of six personality questionnaires. We originally focused on measuring construct validity because half of the initial validation studies neglected it,^13,15,17 and the other half needed to use well-defined methods.^14,16,18 The second objective was to evaluate the reliability of the factors identified by Cronbach alpha and McDonald’s Omega methods. The last objective was to examine convergent validity by the correlations between the constructs of all personality scales and other personality questionnaires proposed by the DSM-5 and ICD-11. This purpose was important since none of the initial validation studies evaluated it.

Methods

Design and Samples

The current cross-sectional study included data from several large samples that were used by some previous studies for other research purposes.^23-26 The data of all studies were collected mainly from the independent samples from the west of Iran. Almost all samples were from community or university students, and none of the data used included clinical samples. In sum, there was access to the data of 2,669 adults 18 years and older (N1 = 1137, N2 = 558, N3 = 496, N4 = 478). The number of participants who responded to each of the personality measures were: TEMPS-A N = 1633,^23,25 AFECTS, TCI, and PANAS N = 1137, ²⁵ NEO-FFI N = 1054,^23,24 and PDQ-4 N = 558.^24,25 The sample sizes are higher than 15–30 cases for each item that the literature highlights for factor analysis methods. The mean and standard deviation of the age of the full sample were 32.8 ± 10.7 years, ranging from 18 to 80 years. Most subjects were female (N = 1854, 70%), single (N = 1393, 52%), and had a university degree (N = 1681, 63%). All methodological details of the present study sample, including the samples and sampling and data collection methods, can be accessed elsewhere.^23-26 The data on the TEMPS-A, AFECTS, TCI, PANAS, NEO-FFI, and PDQ-4 were used to assess construct validity. The Personality Inventory for DSM-5-Brief Form (PID-5-BF: 25 items) and the ICD-11 compatible Personality Inventory for DSM-5 – Brief Form Plus, Modified (PID5BF+M: 36 items) were used to test the convergent validity of all measures. These two questionnaires were used for two reasons. First, these are standard personality scales over the last decade that have evolved according to two internationally accepted manuals of psychopathology (i.e., the DSM-5 and the ICD-11). Second, both questionnaires are short and include few items. This issue is important to avoid fatigue of the participants since each of the four independent samples answered a range of 200–400 items. All participants gave written informed consent to participate in the study. The present study was approved by the Ethics Committee of Mind GPS Institute of Kermanshah (Code: MGPSI.EA.IR.1402.6) and follows the Declaration of Helsinki.

Measures

TEMPS-A

The Persian version of the questionnaire includes three versions of 110, 39, and 35 items. ⁷ The questions of all versions are scored directly, and a score of one is given for each “yes” answer. In the present study, the 35-item version was used for two reasons. First, this version is derived from the 110-item Persian version, ¹⁶ while the 39-item version is an independent scale. ⁷ Second, the psychometric properties of the 39-item version were recently reported,^27,28 while the 35-item version has yet to be validated recently. The reliability and validity of the 35-item Persian version used in the present study were acceptable in the previous study. ¹⁶

AFECTS

The scale consists of two sections for evaluating emotional and affective temperaments. The emotional section uses a 7-point bipolar scale covering six subscales, including adaptive (volition, coping, and control) and maladaptive predispositions (anger, inhibition, sensitivity). The emotional section of AFECTS also measures the desire factor using four questions. The second part comprises 12 descriptive items to assess 12 affective temperaments, including depressive, anxious, apathetic, cyclothymic, dysphoric, volatile, obsessive, euthymic, hyperthymic, irritable, disinhibited, and euphoric temperaments. Each item is scored on a 5-point Likert scale (1=, it does not look like me at all to 5=, it looks exactly like me). ⁸ The questionnaire has previously demonstrated acceptable validity and reliability in Iranian samples. ¹⁴

TCI

The TCI is a tool to measure four temperaments, including NS, HA, RD, and PS, as well as three character traits, including SD, C, and ST. ¹⁰ It is a self-report questionnaire with four versions of 56, 125, 140, and 240 items. However, the longer versions of TCI are strongly influenced by cultural differences and have yet to be officially validated in Iran. The 125-item Persian version was frequently used in Iran,^25,29 with subscales that evaluate using 20 (NS1-NS4), 20 (HA1-HA4), 15 (RD1, RD3, and RD4), 5 (RD2 or PS), 25 (SD1-SD5), 25 (C1-C5), and 15 (ST1-ST3) items, respectively. ¹⁰ Respondents answered using “yes/no,” with 64 items scored in reverse. The TCI has acceptable reliability and validity among Iranian samples. ¹⁵

PANAS

There is a 20-item scale that assesses both personality temperaments (in your lifetime…) and mental states (currently…). Half of the questions evaluate positive affect/temperament (items 1, 3, 5, 9, 10, 12, 14, 16, 17, and 19), while the other half evaluate negative affect/temperament (items 2, 4, 6, 7, 8, 11, 13, 15, 18, and 20). The answers are graded on a five-point Likert scale ranging from “not at all” to “very much” (1–5 points). ⁵ The reliability and validity of this questionnaire are acceptable in Iranian samples. ¹⁸

NEO-FFI

The NEO-FFI consists of 60 items to evaluate five personality domains: neuroticism (anxiety, depression, and self-reproach), extraversion (positive affect, sociability, and activity), openness (aesthetic interests, intellectual interests, and unconventionality), agreeableness (non-antagonistic orientation and prosocial orientation), and conscientiousness (orderliness, goal-striving, and dependability). Each domain is measured using 12 items, with all questions scored directly except for 24 items. Responses are scored on a five-point Likert scale ranging from strongly disagree (score 0) to strongly agree (score 4).^30,31 In Iranian samples, the questionnaire has demonstrated acceptable validity and reliability.^13,32

PDQ-4

Hyler developed a 99-item self-report questionnaire, PDQ-4, to assess and diagnose symptoms of personality disorders. ¹² This dimensional scale evaluates symptoms of 12 personality disorders (schizotypal, paranoid, schizoid, borderline, antisocial, histrionic, narcissistic, obsessive-compulsive, avoidant, dependent, depressive, and negativistic types), with specific item counts for each disorder. All questions with a positive answer are scored 1, resulting in a total score between 0 and 99. ¹² Some studies have reported the reliability of the Persian version of PDQ-4 as acceptable.^17,24,33

PID-5-BF

The PID-5-BF is a self-report questionnaire used to evaluate maladaptive personality domains based on Criterion B of the DSM-5 model. ³⁴ It consists of 25 items and encompasses five subscales: negative affectivity, detachment, antagonism, disinhibition, and psychoticism, with five items for each (five items for each). All items are scored directly using a four-point Likert scale ranging from “often false” to “often true” (0–3 points). The validity and reliability of this tool are acceptable in Iranian samples. ³⁵

PID5BF+M

This 36-item questionnaire was adapted from the original PID-5 version to measure six maladaptive domains of negative affectivity, detachment, antagonism, disinhibition, anankastia, and psychoticism. Each maladaptive trait is measured using two items, scored from “often false” to “often true” (0–3 points).^34,36 The initial validation study confirmed the reliability and validity of the PID5BF+M in populations from 16 different geographical regions.⁵⁰ The tool’s validity and reliability are acceptable in Iranian samples. ³⁷

Analytic Plan

Since the data were from previous studies, we did not face any missing data. First, the statistical assumptions for the parametric analysis, such as normality (skewness and kurtosis between –1 and +1), were checked (see Table S1). Then, ESEM was utilized to analyze the factor structures of all personality measures. ESEM was used to evaluate the factor structure of all personality measures because it estimates all possible cross-loadings and offers a confirmatory test of extracted latent structures. Well-validated personality inventories often do not fit well in confirmatory factor analysis models. ³⁸ In the personality measures with few items/dimensions, including TEMPS-A, AFECTS, and PANAS, the ESEM was used at the level of items. In the longer measures, including TCI, NEO-FFI, and PDQ-4, the ESEM was used at the level of higher-order dimensions (i.e., the domain scores) than individual item scores to prevent poor-fitting models for parsimonious latent structures with fewer factors. The model fit of the multiple factor structures for all solutions was compared using the chi-square test, Aikake Information Criteria (AIC), Bayesian Information Criteria (BIC), Root-Mean-Square Error of Approximation (RMSEA), Comparative Fit Index (CFI), Tucker-Lewis Index (TLI), and Standardized Root-Mean-Square Residual (SRMR). A CFI ≥ 0.90, ∆CFI ≥ 0.02, RMSEA ≤ 0.08, and SRMR ≤ 0.05 for eigenvalues ≥1.0, and parsimonious models with fewer latent factors were considered to be the acceptable factor solutions. ³⁹ Valid factor structures also had lower poor factor loadings (with less than three items with loading >0.40) and lower poor item loadings (items that were not more strongly loaded on the expected factor).

Because of the moderate to strong intercorrelations between the items in each of the personality measures, maximum likelihood estimations with Oblimin rotations performed by Mplus 7.4 were used to determine the factor loadings on each of the latent factors. Both McDonald’s Omega and Cronbach’s alpha of all personality measures and factor scales were reported. Finally, convergent validity was checked using Pearson correlations between subscales of all personality measures and maladaptive personality constructs measured by the PID-5-BF (five dimensions) or PID5BF+M (six dimensions). Correlation coefficients were classified as weak (r ≤ 0.30), medium (r between 0.31 and 0.70), and strong (r ≥ 0.71). For more knowledge, intercorrelations between all personality scales under validation in the present study also were reported (see Tables S2–S5). All analyses were performed using SPSS-27 software, and a significance level of less than 0.05 was considered.

Results

Validation Metrics of the Factor Structure of all Measures

Table S6 shows ESEM, the model fit indices for the multiple factor structures of all personality measures. The preliminary analysis supported the 5-factor structure of TEMPS-A (CFI = 0.95, ∆CFI ≤ 0.02, RMSEA = 0.03, SRMR = 0.02, eigenvalues ≥1.17). However, there were many items with poor loadings that were not more strongly loaded on the expected factors (18 out of 35 items). Therefore, a 6-factor structure that had both good model fit (CFI = 0.97, ∆CFI ≤ 0.02, RMSEA = 0.02, SRMR = 0.02, eigenvalues ≥1.07) and fewer poor item loadings (12 items) compared to the 5-factor model was selected. Regarding the AFECTS, a 7-factor structure with a good model fit (CFI = 0.93, ∆CFI ≤ 0.02, RMSEA = 0.05, SRMR = 0.02, eigenvalues ≥1.18) and fewer poor factor loadings (all good) compared to other structures were selected. The results showed a good model fit for the 5-factor structure of the temperament dimension of TCI. However, since the eigenvalue of the last factor was less than one (=0.84), we chose a 4-factor structure (CFI = 0.98, ∆CFI = 0.05, RMSEA = 0.04, SRMR = 0.02, eigenvalues ≥1.08). Regarding the character dimension of TCI, the results showed a good model fit for the 4-factor structure. However, since the eigenvalue of the last factor was less than one (=0.80), a 3-factor structure (CFI = 0.98, ∆CFI = 0.05, RMSEA = 0.03, SRMR = 0.02, eigenvalues ≥1.42) was selected. Although the model fit of PANAS for the 2-factor structure is somewhat weak (CFI = 0.86, ∆CFI = 0.41, RMSEA = 0.08, SRMR = 0.05, eigenvalues ≥4.29), this structure was selected because both good item/factor cross-loadings and consistency with the conceptual framework. ⁵ A 5-factor structure to the NEO-FFI with a good model fit (CFI = 0.99, ∆CFI ≤ 0.02, RMSEA = 0.03, SRMR = 0.01, eigenvalues ≥1.00) was selected. Regarding PDQ-4, first, 6- to 13-factor structures for the items were examined, all of which showed weak factor loadings and did not have an acceptable model fit. Thus, ESEM was used to identify the factor structure of 12 subscales, the results of which were acceptable for a 2-factor structure (CFI = 0.96, ∆CFI = 0.05, RMSEA = 0.06, SRMR = 0.03, eigenvalues ≥1.00). Details of all results can be seen in Table S6.

Factor Loadings and Reliability of all Measures

Tables S7–S10 show the factor loadings for the subscales of all personality measures. Table S7 shows that most items of TEMPS-A loaded well on at least one factor (|≥0.30|) including (a) depressive (items 1–5; all loadings between 0.40 and 0.70), (b): cyclothymic (items 6–12; all loadings between 0.29 and 0.49), (c): hyperthymic (items 13–22; all loadings between 0.21 and 0.63), (d): dysphoric (items 23–26; all loadings between 0.48 and 0.60), (e): anxious (items 28–30; all loadings between 0.41 and 0.47), and (f): irritable (items 27 and 31–35; all loadings between 0.32 and 0.40) temperaments. The high similarity of the item content of some factors caused some items not to load well on the expected factors. For example, items 6–10, which were developed to measure depressive temperament, loaded more strongly on the cyclothymic factor. We considered all items with very weak factor loadings on the expected factor for other factors with stronger factor loadings (e.g., items 17, 18, 27, 31). Reliability of all scale items was acceptable (α = 0.83), whereas McDonald’s omega of depressive, cyclothymic, hyperthymic, dysphoric, anxious, and irritable factors were 0.78, 0.70, 0.63, 0.73, 0.56, and 0.74, respectively. Except for anxious temperament, the internal consistency of all factors was higher than the metrics of the initial validation study in Iran (see Table S7).

Table S8 shows that most items of AFECTS loaded well on at least one factor (|≥0.40|) including (a): volition (items 1–8; all loadings between 0.62 and 0.88), (b): anger (items 9–15; all loadings between 0.40 and 0.83), (c): coping (items 31–37; all loadings between 0.37 and 0.82), (d): inhibition (items 16 and 18–23; all loadings between 0.15 and 0.71), (e): control (items 38–45; all loadings between 0.28 and 0.81), (f): sensitivity (items 24–30; all loadings between 0.21 and 0.58), and (g): desire (items 17 and 46–48; all loadings between 0.31 and 0.77). Except item 17, which was developed to measure inhibition, but loaded well on the desire factor, other items loaded exactly on the expected factors. We assigned item 17 to the desire factor because its content in the Farsi translation is completely compatible with this factor. McDonald’s omega reliability of volition, anger, coping, inhibition, control, sensitivity, and desire factors were 0.92, 0.85, 0.89, 0.64, 0.91, 0.79, and 0.78, respectively. Cronbach’s alpha of all scale items was equal to 0.89, which is somewhat higher than the reliability reported by the initial validation study in Iran (see Table S8).

Table S9 shows that most temperament facets of TCI loaded well on at least one factor (|≥0.25|) including (a): NS (NS1–NS4; all loadings between 0.33 and 0.53), (b): HA (HA1, HA3, and HA4; all loadings between 0.43 and 0.55), (c): PS (RD2 with factor loading of 0.27), and (d): RD (RD1, RD3, and RD4; all loadings between 0.17 and 0.57). However, HA2 (fear of uncertainty) and RD4 (dependence) more strongly were loaded on factors (c) and (b), respectively. Table S9 also shows that most character facets of TCI loaded well on at least one factor (|≥ 0.40|), including (a): SD (SD1–SD5; all loadings between 0.44 and 0.73), (b): C (C1–C5; all loadings between 0.40 and 0.65), and (c): ST (ST1–ST4; all loadings between 0.41 and 0.59). Cronbach’s alpha of all scale items for temperament and character dimensions were 0.68 and 0.82, respectively. McDonald’s omega reliability of NS, HA, PS, RD, SD, C, and ST were 0.65, 0.79, 0.55, 0.45, 0.82, 0.88, and 0.95, respectively. Except for the RD, Cronbach’s alpha of all factors was almost the same as in the validation study in Iran (see Table S9).

Table S9 also shows that most items of PANAS loaded well on at least one factor (|≥0.50|) including (a): negative affect (items 2, 4, 6-8, 11, 13, 15, 18, and 20; all loadings between 0.43 and 0.76) and (b): positive affect (items 1, 3, 5, 9, 10, 12, 14, 16, 17, and 19; all loadings between 0.50 and 0.68). McDonald’s omega reliability of negative and positive affect were 0.87 and 0.86, respectively. Cronbach’s alpha of all scale items was equal to 0.81, which is somewhat higher than the reliability reported by the initial validation study in Iran (see Table S9).

Table S10 shows that most facets of NEO-FFI loaded well on at least one factor (|≥0.30|) including (a): neuroticism (anxiety, depression, and self-reproach; all loadings between 0.38 and 0.79), (b): conscientiousness (orderliness, goal-striving, and dependability; all loadings between 0.62 and 0.81), (c): extraversion (positive affect, sociability, and activity; all loadings between 0.16 and 0.51), (d): agreeableness (non-antagonistic orientation and prosocial orientation; all loadings between 0.36 and 0.51), and (e) openness (aesthetic interests, intellectual interests, and unconventionality; all loadings between 0.22 and 0.66). However, sociability and prosocial orientation more strongly were loaded on factors iv and ii, respectively. Cronbach’s alpha of all scale items was equal to 0.68, whereas McDonald’s omega of neuroticism, conscientiousness, extraversion, agreeableness, and openness factors were 0.72, 0.84, 0.68, 0.52, and 0.40, respectively. Except for neuroticism and agreeableness, the internal consistency of all factors was somewhat higher than the metrics of the initial validation study in Iran (see Table S10).

Table S10 also shows the 1–3-factor structures of the PDQ-4. In the one-factor structure, all subscales of PDQ-4 loaded well on general factor of personality disorder (|≥.50|). In the 2-factor structure, all subscales loaded well on at least one factor (|≥.30|) including (a): four symptoms pathology (borderline, avoidant, dependent, and depressive; all loadings between 0.41 and 0.58) and (b): eight symptoms pathology (paranoid, schizoid, schizotypal, antisocial, narcissistic, histrionic, obsessive-compulsive, and negativistic; all loadings between 0.30 and 0.74). We accounted antisocial symptoms for the second factor because other cluster B personality symptoms (e.g., narcissistic and histrionic personality disorders) loaded on it. In the 3-factor structure, all subscales were loaded more strongly on the first and second factors, and the third factor contained unacceptable factor loadings. McDonald’s omega reliability of all scale items for all three structures was between 0.85 and 0.92. The internal consistency of total scale items is somewhat higher than the test-retest reliability reported by the initial validation study in Iran (Table S10).

Convergent Validity of all Measures

Table S11 indicates Pearson correlations between all personality measures and transdiagnostic constructs of personality measured by the PID-5-BF and the PID5BF+M. Except for hyperthymic temperament, all TEMPS-A subscales were moderately related to maladaptive personality constructs (r ranging from 0.13 to 0.52, all P <.01). There was also a significant weak correlation between hyperthymic temperament and some of the PID5BF+M constructs (e.g., detachment). Most of the AFECTS subscales were weakly to moderately related to the personality constructs (r ranging from –0.35 to 0.36, all P <0.05). However, the correlations between the AFECTS subscales and anankastia were very weak (r ranging from –0.08 to 0.10). Some of the character dimensions of TCI, including SD and C, showed moderate correlations with personality constructs (r ranging from –0.11 to –0.52, all P <.01), while the temperament dimensions showed milder significant correlations. The correlations were in ranging from 0 to 0.40 for NS, 0.07 to 0.44 for HA, –0.21 to 0.14 for RD, and –0.09 to 0.21 for PS. Negative affect on the PANAS was moderately related to all personality constructs (r ranging from 0.18 to 0.54, all P <.01). At the same time, there was a slight negative correlation between positive affect and some PID5BF+M constructs (r ranging from –0.30 to 0.06, all P <.01). Although openness was only weakly related to the detachment (r = –0.17) and antagonism (r = –0.14) subscales of the PID-5-BF, other NEO-FFI subscales showed moderate correlations with maladaptive personality constructs (r ranging from –0.48 to 0.57, all P <.01). Finally, the results of Table S11 show the moderate correlations between all PDQ-4 and PID-5-BF subscales (r ranging from 0.14 to 0.51, all P <.01). Both factors of the 2-factor structure were moderately related to all maladaptive personality constructs (r ranging from 0.37 to 0.61, all P <.01).

Discussion

The present study aimed to test the construct and convergent validity of the Persian version of several personality questionnaires. Although the number of factors of TEMPS-A and PDQ-4 were slightly different from the original structures, the multiple metrics of model fit and reliability statics fully supported the factor structure of all measures. Because almost none of the initial validation studies in Iran reported the model fit of the factor structure of measures,^13,15-18 it is not possible to compare the current results with the previous ones. Only the AFFECTS validation study reported the model fit derived from a confirmatory factor analysis, ¹⁴ where all indicators indicated a weaker factor structure than the present results. Regarding reliability, 4 out of 5 TEMPS-A factors and 3 out of 5 NEO-FFI factors had higher internal consistency compared to what the initial validation studies reported. Also, the total items of the three measures, including AFECTS, PANAS, and PDQ-4, had higher internal consistency compared to what the initial validation studies found. This evidence supports the effect of the type and size of the population on the validation data of personality measures and highlights the added value of the present study.

Like the multiple model fit indices that helped to identify the number of factors of each measure, the factor cross-loadings and internal consistency of the total scale items supported the validity of all questionnaires. Although some items loaded relatively weakly on all factors, they were assigned to the same factor as expected due to conceptual content (e.g., items 20-22 of hyperthymic temperament in the TEMPS-A). The TEMPS-A was expected to include only five latent factors; however, ESEM supported the six-factor structure, which includes a dysphoric temperament. The dysphoric factor was composed of items of anxious and irritable temperaments that had very similar content. These differences between the Persian version and the original structure of the TEMPS-A probably come from the translation issues and the cultural context of the samples (most of the samples were Kurdish speakers). The lack of access to data from clinical samples could also have affected the results because TEMPS-A was originally developed to identify patients with affective disorders. ⁶ However, all latent factors showed good cross-loadings and acceptable reliability (except for anxious temperament that probably comes from the few items). The present results showed that the internal consistency of most factors was higher than the metrics of the initial validation study in Iran. ¹⁶

Good cross-loadings and acceptable reliability for all factors supported the original structure of AFECTS with seven latent factors. Only one item of the inhibition factor loaded strongly on the desire factor. The inhibition factor also had relatively poor internal consistency. ⁴⁰ However, Cronbach’s alpha of all scale items was somewhat higher than the reliability reported by the initial validation study in Iran. ¹⁴ When the factor structure of TCI was assessed using ESEM, the results supported good cross-loadings on the seven-factor structure and acceptable reliability for most factors (especially character factors). However, HA2 (fear of uncertainty) and RD4 (dependence) more strongly were loaded on unexpected factors. All the character factors and HA showed acceptable reliability, while three factors of temperament, including NS, RD, and PS, showed relatively weaker reliability. However, the internal consistency of all factors (except for RD) was almost the same as in the validation study. ¹⁵ The present results were consistent with the findings of some studies in Iran that showed relatively weak internal consistency for some temperament factors.^25,27,41 Although the model fit of PANAS for the two-factor structure was somewhat weaker than 3- to 5-factor structures, the structure was selected because of both good cross-loadings and consistency with the conceptual framework. ⁵ The structures with more factors are suitable for identifying a set of lower-order factors. In the 2-factor structure, all items of PANAS loaded well on one of the negative affect and positive affect factors. Both factors showed acceptable reliability, and Cronbach’s alpha of all scale items was higher than the reliability reported by the initial validation study in Iran. ¹⁸

When the factor structure of NEO-FFI was assessed using ESEM, the results supported the good cross-loadings on the five-factor structure and acceptable reliability for neuroticism, conscientiousness, and extraversion factors. Although the internal consistency of all factors was almost similar to the initial validation study in Iran, ¹³ of the findings were consistent with the previous study in Iran, which showed relatively weak Cronbach’s alpha for agreeableness and openness. ³² The ESEM revealed the 2-factor structure with good cross-loadings and acceptable reliability for both factors of PDQ-4. Although the model fit for the one-factor structure was somewhat weak, strong factor loadings also supported the use of the PDQ-4 to assess a general factor of personality disorder. In sum, the well cross-loadings for the 1- and 2-factor structures, as well as relatively poor internal consistency for all original subscales of the PDQ-4, suggested that this measure may be more appropriate for assessing higher-order personality factors and general personality pathology than for specific types of personality disorders. One study in Iran also showed weak internal consistency for most subscales of PDQ-4. ⁴¹

The present research evaluated the convergent validity of the personality measures using Pearson correlations between all factors and transdiagnostic constructs of personality, including negative affectivity, detachment, antagonism, disinhibition, psychoticism, and anaplastic.^11,34-37 The results showed that all personality factors were weakly to moderately -but all significantly- related to the transdiagnostic constructs. As expected, depressive, irritable, and cyclothymic temperaments, negative affect, HA, neuroticism, and borderline and depressive personality disorders were the most important correlates of negative affectivity. Negative affectivity was also negatively related to hyperthymic temperament, positive affect, volition, coping, control, SD, C, conscientiousness, agreeableness, and extraversion. These findings were consistent with previous reports.^25,29,41,42 Positive and negative correlations of all factors with detachment (e.g., avoidant personality disorder, low extraversion, and low positive affect), antagonism (e.g., antisocial personality disorder, low agreeableness, and low C), disinhibition (e.g., borderline personality disorder, low conscientiousness, and low SD), psychoticism (e.g., schizotypal personality disorder, high ST, and low RD), and anankastia (e.g., high PS, high negative affect, and low SD) were expected and consistent with the research literature.^43-45 These findings supported the convergent validity of the Persian version of all personality measures.

Limitations

The present pioneering research focused on the explanation of the validation findings of several personality measures in light of the change in both the population type and the sample size. However, there were some methodological limitations. The present research used data from four studies with completely independent samples.^23-26 The sample size used to validate each questionnaire was different from the others. Therefore, the sample used to validate the PDQ-4 was relatively small, which can be considered in future research. All the data included in the analysis were from community samples, whereas maladaptive measures such as PDQ-4 should be validated in samples with personality disorders as well. Due to the data limitation, the criterion validity of the questionnaires to distinguish clinical from non-clinical samples was not tested.

Additionally, we only assessed construct and convergent validities, which highlights the use of other psychometric methods, such as discriminant and divergent types, in future studies. Time-dependent effects that may lead to some changes in the content of the items of each of the questionnaires were not evaluated. The Initial validation studies^13-18 included populations from the central, northwestern, and south-central regions of Iran. Due to the multiplicity of cultures and ethnicities living in Iran, mental health professionals should be cautious in generalizing data to the Eastern, Northern, and Southern regions of Iran.

Conclusion

The present results supported the factor structure of all personality measures in Iranian populations. The factor loadings and internal consistency of the items of each factor were similar to the initial validation studies. In contrast, the reliability of total scale items of all measures was higher than what was reported by the initial validation studies. However, the cross-loadings for the 1- and 2-factor structures, as well as relatively poor internal consistency for all original subscales of the PDQ-4, suggested that this measure may be more appropriate for assessing higher-order personality factors and general personality pathology than for specific types of personality disorders. The significant correlation patterns between all scales and other personality constructs, such as negative affectivity, supported the convergent validity of all measures. Mental health clinicians can use the Persian version of all valid personality measures with some caution when using the PDQ-4 subscales.

Supplemental Material

Supplemental material for this article is available online.