The concept of worry was first introduced and defined as a “chain of thoughts and images, negatively affect-laden and relatively uncontrollable; it represents an attempt to engage in mental problem-solving on an issue whose outcome is uncertain but contains the possibility of one or more negative outcomes” (Borkovec, Robinson, Pruzinsky, & DePree, Reference Borkovec, Robinson, Pruzinsky and DePree1983, p. 10).
As a process, worry was characterized as ruminative, consisting mainly of thoughts rather than of images (Borkovec & Inz, Reference Borkovec and Inz1990). In terms of content, worry is related with several themes, including health, social and intimate relationships, finances and work/academic performance, and, for some individuals, worry about worry itself (Papageorgiou, Reference Papageorgiou, Davey and Wells2006). Although worry is a clinical feature typical of Generalized Anxiety Disorder (GAD), it is also associated with many other disorders (e.g. insomnia, obsessive-compulsive disorder, health anxiety, etc.) (Purdon & Harrington, Reference Purdon, Harrington, Davey and Wells2006).
Most people are concerned about future events and may find it helpful to deal or prevent the potential negative consequences of these events and to be prepared for the worst (Borkovec, Ray, & Stöber, Reference Borkovec, Ray and Stöber1998). Therefore, worry is a common phenomenon. Ruscio, Borkovec, and Ruscio (Reference Ruscio, Borkovec and Ruscio2001) analyzed the latent structure of worry and found that the differences between GAD worry and normal worry were quantitative rather than qualitative. These authors suggested that normal and pathological worry represent the opposite ends of a continuum. Individuals with GAD, as opposed to non-clinical individuals, perceive worry as subjectively less controllable and less successfully reduced by corrective attempts (Craske, Rapee, Jackel, & Barlow, Reference Craske, Rapee, Jackel and Barlow1989).
To assess worry, some self-report measures were developed. The Penn State Worry Questionnaire (PSWQ) is intended to assess the traits of pathological worry independent of its content (Meyer, Miller, Metzger, & Borkovec, Reference Meyer, Miller, Metzger and Borkovec1990). Since its publication, the PSWQ has been the most widely used measure of worry, and it has been employed within both clinical and non-clinical populations (Startup & Erickson, Reference Startup, Erickson, Davey and Wells2006).
The PSWQ consists of 16 items worded in a 1-5-point scale. Eleven items are positively worded and the other five in a reversed order to reduce the effects of acquiescence (Meyer et al., Reference Meyer, Miller, Metzger and Borkovec1990). This scoring method seems to have led to different outcomes in Exploratory Factor Analysis (EFA). In the seminal study of PSWQ, Meyer et al. (Reference Meyer, Miller, Metzger and Borkovec1990) obtained two factors but only one was retained. When Brown, Antony, and Barlow (Reference Brown, Antony and Barlow1992) investigated the internal structure of PSWQ in a sample of anxiety disorders patients, Principal Components Analysis (PCA) with varimax rotation yielded a two-factor solution with eigenvalues greater than 1, but again only one factor was retained. However, other studies of the PSWQ performing PCA have found a two-factor solution with the positive items loading on the first factor and the negative items on the second factor (e.g., Castillo, Macrini, Cheniaux, & Landeira-Fernandez, Reference Castillo, Macrini, Cheniaux and Landeira-Fernandez2010; Fresco, Heimberg, Mennin, & Turk, Reference Fresco, Heimberg, Mennin and Turk2002)
Two methodological shortcomings seem to arise from the use of EFA methods. First, as argued by some authors, the second factor may be due to a method artefact associated with the presence of negatively keyed items (e.g., Brown, Reference Brown2003). Second, most of the authors performed EFA with PCA (with few exceptions, such as Rodríguez-Biglieri & Vetere, Reference Rodríguez-Biglieri and Vetere2011) and used scree test examination and eigenvalues greater than 1 to determine the number of factors. The use of these methods may overestimate the number of factors to extract (Gorsuch, Reference Gorsuch1997).
The PSWQ showed very acceptable values of internal consistency, ranging from .88 to .95 in the previous studies (Startup & Erickson, Reference Startup, Erickson, Davey and Wells2006). Test-retest reliability values obtained for this inventory across the different studies are also very satisfactory, ranging from .74 to .92 (Startup & Erickson, Reference Startup, Erickson, Davey and Wells2006) and this measure was demonstrated to be useful for screening GAD patients (Behar, Alcaine, Zuelling, & Borkovec, Reference Behar, Alcaine, Zuellig and Borkovec2003).
Psychometric evidence concerning PSWQ suggests a unidimensional structure with satisfactory levels of reliability. However, some methodological shortcomings were pointed out. To overcome these methodological problems concerning EFA, in the present study, we followed Gorsuch’s (Reference Gorsuch1997) recommendations. We performed exploratory factor analysis with a community sample. We also performed Confirmatory Factor Analysis with a different, non-clinical sample to test the most common models obtained in previous studies: A one-factor model (e.g., Brown et al., Reference Brown, Antony and Barlow1992; Meyer et al., Reference Meyer, Miller, Metzger and Borkovec1990), a two-factor model (worry/absence of worry) (e.g., Fresco et al., Reference Fresco, Heimberg, Mennin and Turk2002), a bifactor model with a method effect (e.g., Brown, Reference Brown2003; Hazlett-Stevens, Ullman, & Craske, Reference Hazlett-Stevens, Ullman and Craske2004), and one-factor model of the positively worded items (e.g., Sandín, Chorot, Valiente, & Lostao, Reference Sandín, Chorot, Valiente and Lostao2009).
Thus, the purpose of the present study was twofold: To explore the pathological levels of worry in a Portuguese community sample and to extend existing outcomes in the literature examining the latent structure of PSWQ. Despite the widespread use of this questionnaire, its psychometric proprieties in a Portuguese (Portugal) sample has not yet been examined. So, we aim to investigate the exploratory factor structure of the PSWQ by improving EFA’s methodological options, to test the most common models obtained in previous studies using CFA, and to seek evidence of convergent validity for the Portuguese version of this questionnaire.
Method
Participants
For the sample recruitment and selection, a mixed procedure was used. A sample of psychology undergraduate students collectively answered to the assessment protocol in the presence of the first author of this work. Some of these students were recruited and trained to disseminate the evaluation protocol into the community using a snowball approach. Protocols were composed of an informed consent, a set of self-report instruments, and an envelope identified only with a number. Participants received instructions to insert the corresponding instruments into the envelope and then to close it. Participants who reported, at the time of the investigation, did not have any diagnosed mental disease and who read aloud and adequately understood instructions were included in this work. Trained undergraduates verified the comprehension of the instructions.
A sample of 558 Portuguese community participants took part in this study. Participants ranged between 18 and 68 years of age (M = 31.48, SD = 12.80). Most of the participants were female (66.5%, n = 371) and single (59.5%, n = 330). Undergraduate students accounted for less than half (45.3%, n = 235) of the sample.
Measures
The Penn State Worry Questionnaire (PSWQ) (Meyer, et al., Reference Meyer, Miller, Metzger and Borkovec1990) is a 16-item self-report developed to measure the traits of worry. Items are rated on a five-point scale from 1 (not at all typical of me) to 5 (very typical of me). The total score (ranged between 16 and 80) is obtained by summing the item scores after reversing the negatively worded items. Five items (Numbers: 1, 3, 8, 10 and 11) are worded negatively. The original version of PSWQ showed high internal consistency levels (higher than α = .90) and an excellent one-month test-retest reliability (r = .93). The appendix presents the translated version into Portuguese.
The White Bear Suppression Inventory (WBSI) (Wegner & Zanakos, Reference Wegner and Zanakos1994; Portuguese version by Jiménez-Ros, Orgambídez-Ramos, & Pascual, Reference Jiménez Ros, Orgambidez Ramos and Pascual2015) is a 15-item self-report instrument to measure people’s general tendency to suppress unwanted thoughts. Items are rated on a 5-point scale from 1 (strongly disagree) to 5 (strongly agree). A total score is obtained by summing the item scores. The Portuguese version showed adequate levels of internal consistency (α = .88), and reliability, ICC = .61, 95% CI [.31, .78].
The Thought Control Questionnaire (TCQ) (Wells & Davies, Reference Wells and Davies1994; Portuguese version by Jiménez-Ros & Pascual, Reference Jiménez-Ros and Pascual2016) was developed to assess the strategies used by individuals to control unwanted thoughts. It is composed of 30 items rated on a 4-point scale from 1 (never) to 4 (almost always).The Portuguese version maintained the original structure of five factors (distraction, social control, worry, punishment and reappraisal) with adequate levels of internal consistency ranging from a Cronbach’s alpha of .70 for worry to a Cronbach’s alpha of .81 for distraction and test-retest reliability ranging from ICCreappraisal = .55, 95% CI [.26, .73] to ICCpunishment = .71, 95% CI [.52, .83].
Procedure
Following Hambleton’s (Reference Hambleton1996) recommendations, the Portuguese version of PSWQ was obtained by a backward-forward translation process, which ensures conceptual equivalency. Forward translation into Portuguese was carried out by two independent, native-speaking translators who were proficient in English. The investigation team compared and corrected discrepancies of the translations obtaining one Portuguese version. The Portuguese version was backward translated into English by two independent English native-speaking translators who were fluent in Portuguese. No significant changes were suggested. The final version was applied to a small undergraduate sample. After minor modifications, the Portuguese final version was obtained.
Psychology students in the last year of their degree programs were selected and trained to apply the protocol twice to a sample of 50 volunteers from the community. Participants completed the protocol for the second time, from 3 to 6 weeks after the first time.
Data Analysis
The total sample was randomly split into two groups, each with 279 participants, to perform factor analysis. Exploratory Factor Analysis was carried out with Principal Axis Factoring in one half of the sample, using SPSS (v25.0). According to Gorsuch (Reference Gorsuch1997), Principal Axis Factoring should be preferred to Principal Components in order to avoid the inflation of factor loadings. To determine the optimal number of factors to retain, we considered the following criteria: (a) Optimal parallel analysis obtained by the FACTOR program (Lorenzo-Seva & Ferrando, Reference Lorenzo-Seva and Ferrando2006) (b) eigenvalues higher than 1, and (c) the scree plot. We chose optimal parallel analysis as it is more accurate than Cattell’s scree and Kaiser-Guttman’s methods because it compares the eigenvalues of randomly generated data with those for the actual data, while additionally taking into account sampling error possibilities (Wilson & Cooper, Reference Wilson and Cooper2008).
Confirmatory Factor Analysis was performed in the other half of the sample through STATA 14. We tested the following models by confirmatory factor analysis to determine which of them obtained the best fit to our data: (a) Model 1 replicated the initial unidimensional structure obtained by Meyer et al., (Reference Meyer, Miller, Metzger and Borkovec1990); (b) Model 2 tested the latent structure of two factors (with the 11 positively worded items loading in the first factor and the 5 negatively worded items loading in the second) found by Fresco et al. (Reference Fresco, Heimberg, Mennin and Turk2002); (c) Model 3 assumed the latent structure of PSWQ as an unique factor with the negatively worded items grouping in a method factor (Brown, Reference Brown2003); and finally, Model 4 tested the unidimensional structure with only the 11 positively worded items proposed by Sandín et al. (Reference Sandín, Chorot, Valiente and Lostao2009). Goodness-of-fit indices were calculated, including the Satorra–Bentler chi-square/degrees of freedom (SBχ 2/df), Comparative Fit Index (CFI), Root Mean Square Error of Approximation (RMSEA), Non-Normed Fit Index (NNFI) and Standardized Root Mean-Square Residual (SRMR). A χ2/df value < 5 was adequate, ≤ 2 was good, and values = 1 were considered very good (West, Taylor, & Wu, Reference West, Taylor, Wu and Hoyle2012). Fit indices were considered adequate if CFI ≥ .90 (Byrne, Reference Byrne2006), NNFI ≥ .90, SMRS < .10, and RMSEA < .08 (West et al., Reference West, Taylor, Wu and Hoyle2012). To compare the relative fit of models, we performed the Satorra-Bentler chi-square difference test (Satorra & Bentler, Reference Satorra and Bentler2001).
Internal consistency was analyzed by the Cronbach’s alpha coefficient. We considered values of alpha higher than .70 as indicating acceptable internal consistency (Nunnally & Bernstein, Reference Nunnally and Bernstein1994). Due to the disadvantages of the Cronbach’s alpha coefficient (Dunn, Baguley, & Brunsden, Reference Dunn, Baguley and Brunsden2014), we also computed omega coefficients to estimate the internal consistency of the PSWQ. Statistical Comparison of Cronbach’s alpha coefficients were performed using Cocron package in R (Diedenhofen & Musch, Reference Diedenhofen and Musch2016).
The temporal stability was analyzed by the Intraclass Correlation Coefficient (ICC, Bartko, Reference Bartko1966), which indicates that the proportion of total variability is due to differences between participants. ICC may take values from 0 to 1. Values ranging from .50 to .75 were considered as indicating moderate reliability, values from .75 to .90 indicate good reliability, and values higher than .90 indicate excellent reliability (Koo & Li, Reference Koo and Li2016).
Results
Distribution of PSWQ scores
Our first step in examining the psychometric properties of the Portuguese version of the PSWQ was to carefully analyze primary data. All items reached the minimum and the maximum in the range of possible scores. Means of all items were around 3, indicating that the participants had a moderate tendency to excessive worry. Standard deviation of all items was higher than 1, showing an adequate dispersion of data. Most of the items (11 in 16) showed non-skewed distributions. The mean of the total score was 44.55 (SD = 9.15).
Internal Structure of PSWQ
Because this is a new version of the PSWQ, we first explored the internal structure of PSWQ, conducting exploratory factor analysis in a random half of the sample (n = 279). The Kaiser-Meyer-Olkin index showed an excellent value of .90 and the Bartlett test was statistically significant, (χ2(120) = 1,940.41, p < .001). Conducting EFA with this data seems appropriate. Two dimensions were retained, and they accounted for 50.94% of the total variance.
Using SPSS (v25), we performed principal axis factoring with varimax rotation to extract the factor matrix. Table 1 presents the factor loadings and eigenvalues for the PSWQ. The eleven positively worded items displayed higher loading on the first factor, while the negatively worded items also showed sizable loadings in the second factor.
Table 1. Factor Loading for the PSWQ Items using a two-factor Model (n = 279)
Note: h 2= communalities.
The latent structure of PSWQ remains unclear. For this reason, we performed CFA with the other half of the random sample (n = 279) with the aim to determine if the results obtained by EFA could be attributed to a systematic bias of the measurement method or to conceptual criteria.
Table 2 presents model-fit indices for all models tested. It is not possible to statistically compare models including different variables such as Models 1, 2, and 3 vs. Model 4, so we only compared the relative fit for the 16-item models. Model 3 produced the best relative fit to the data, since it showed a significantly better fit than Model 1 (SB χ2 difference test = 198.30, df = 5, p < .01) and Model 2 (SB χ2 difference test = 11.58, df = 4, p = .02). These results indicate a lack of support for the unifactorial conception of the 16-item version of the PSWQ. We should then choose between: (a) A bifactor structure with a general factor and a method factor associated to the negatively worded items (Model 3) or (b) a unifactorial structure for the 11-item version of the PSWQ (Model 4). Considering the CFI, NNFI and SRMR indices, Model 4 produced an adequate fit to the data, but not when considering RMSEA.
Table 2. Model fit indices for all CFA models tested in the subsample (n = 279)
Notes: Model 1 One factor (16 items), Model 2 Two related factors (worry/not worry), Model 3 One factor (+ method effect), Model 4 One factor (11 items). CFI = Comparative fit index; NNFI = Non-normed fit index; SRMR = Standardized root mean-square residual; RMSEA = Root mean-square error of approximation; BIC = Bayesian Information Criterion; AIC = Akaike Information Criterion; SB χ 2 (df) = Satorra-Bentler chi-square (degrees of freedom); CI = Confidence interval.
* p < .05. ** p < .01.
According to our data, the PSWQ could be conceptually considered a unidimensional measure of worry with Model 3 producing an adequate fit to the data and Model 4 a partially adequate fit. For that reason, we will compare which of the versions of this instrument (11 or 16 items) offer the best psychometric properties.
Reliability
We assumed the unidimensionality of the PSWQ and conducted a reliability analysis for the pools of the 16 and the 11 items of the scale.
The PSWQ 16 items showed an average inter-item correlation of .32, which could be considered adequate since it is in the interval of .20–.40, indicating that the items are related but not redundant. Inter-item correlations ranged from .01 and .73.
The Cronbach’s alpha coefficient for the PSWQ–16 was excellent (α = .88). This coefficient remained equal or even decreased if any of the items were removed, except for four of the negatively worded items whose elimination improved the alpha value (Items 1, 3, 8 and 11). However, we decided not to eliminate these items because the improvement introduced by this elimination would be minimal (less than .003).
The 11 positively worded items of the PSWQ–11 version showed an average inter-item correlation of .49. This high average indicates strong relationships between items, but it also alerts us to the possibility of the existence of a slight redundancy between them. Inter-item correlations ranged from .19 to .73. Its Cronbach’s alpha value was again excellent (α = .91), and it remained equal or even decreased if any of the items were removed.
The statistical difference between both versions’ (Items 16 and 11) Cronbach’s alpha coefficients was significant (χ2(1) = 46.76, p ≤ .001).
Performing McDonald’s omega coefficient, we found similar results: Reliability for the 16-item version was lower (ω = .87) than for the 11-item version (ω = .92).
A subset of participants (n = 50) completed the PSWQ a second time after a period ranging from three to six weeks. Temporal stability was analyzed by the Intra-class Correlation Coefficient. The ICC for both versions of the PSWQ indicated good temporal stability, with ICC = .79, 95% CI [.63, .88] for the PSWQ–16 version and ICC = .81, 95% CI [.64, .89] for the PSWQ–11 version.
Our results showed that the reliability levels of the full and short versions of PSWQ are good and could be indistinctly used in Portuguese-speaking populations. To choose one of them, however, we needed to take into account two issues: The PSWQ–16 version includes double negatives in two items, which may contribute to lowering its internal consistency; and, on the other hand, the item-total correlations obtained for the PSWQ–11 version indicate a high homogeneity of the items and could be pointing to this scale’s reduced construct validity.
Convergent Validity
Convergent validity was analyzed by Pearson’s correlation coefficient of the PSWQ with other measures in the study (TCQ and WBSI).
Correlations between PSWQ and thought-control strategies were positive, moderate, and significant for the total of the scale (r PSWQ_16–TCQ_Total = .30, p < .01, r PSWQ_11–TCQ_Total = .29, p < .01) and for two subscales: Worry (r PSWQ_16–TCQ_Worry = .29, p< .01, r PSWQ_11–TCQ_Total = .24, p < .01) and punishment (r PSWQ_16–TCQ_Punishment = .38, p < .01, r PSWQ_11–TCQ_ Punishment = .41, p < .01). Correlations were also positive and significant but low for reappraisal (r PSWQ_16–TCQ_Reappraisal = .24, p < .01, r PSWQ_11–TCQ_ Reappraisal = .23, p < .01) and distraction (r PSWQ_16–TCQ_Distraction = .10, p < .05, r PSWQ_11–TCQ_ Distraction = .10, p < .05). There were not significant correlations between PSWQ and social control strategies (r PSWQ_16–TCQ_Social control = –.04, p > .05, r PSWQ_11–TCQ_ Social control = –.07, p > .05). PSWQ and thought suppression correlations were positive and high (r PSWQ_16–WBSI = .47, p < .01, r PSWQ_11–WBSI = .50, p < .01).
Normative Data
The range of scores for the PSWQ–16 was from 16 to 80. The mean of the sample was 44.52 (SD = 9.27). For the PSWQ–11, the mean score was 32.48 (SD = 9.69), with a range from 11 to 55. These results indicate the participants’ moderate tendency to excessive worry. There were no significant differences between male and female tendencies to worry (For PSWQ–16 items: t(564) = 4.04, p < .01, d = .36; for PSWQ–11 items: t(384.763) = 3.161, p < .01, d =.33) or between age groups (young: 18–30 years, young adults: 30–45 years and adults > 45 years): for PSWQ–16 items, F(2, 563) = 2.23, p > .05, ηp2 = .009 and (F Welch(2, 21.382) = 2.051, p > .05, ηp2 = .01.
Discussion
Worry is a common phenomenon with people who anticipate and want to be prepared for future events. Pathological worry, however, is not limited to anticipating potentially dangerous events, but it is applied to small daily events, too. This kind of worry is perceived as uncontrollable, and it is strongly associated with catastrophism, which leads to increased anxiety and stress levels, and ultimately to worsening the problems.
The PSWQ was elaborated to assess the traits of worry. This measure is one of the most used to assess the frequency, intensity, and uncontrollability of worry (Startup & Erickson, Reference Startup, Erickson, Davey and Wells2006), and it was largely applied in both clinical and non-clinical samples. Although it has good psychometric characteristics, its internal structure is still controversial. Methodological considerations may be responsible for the discrepancies found in different studies. In the present study, we performed EFA and CFA with two independent random subsamples of Portuguese community participants.
The results of the EFA clearly yielded a two-factor structure. All items presented high factor loadings, with the positively worded items grouped in the first factor and the negatively worded items in the second factor. When tested by CFA, the two-factor structure, however, failed to produce a significantly better fit than a bifactor structure with a method effect associated with negatively worded items, so that the item grouping on the second factor seemed to correspond more to the wording form of the items than to an underlying factor. If negatively worded items were removed, the one-factor solution of the 11 positively worded items did produce a generally acceptable fit to the data. Similar results were obtained in previous studies (e.g., Brown, Reference Brown2003; Castillo et al., Reference Castillo, Macrini, Cheniaux and Landeira-Fernandez2010; Sandín et al., Reference Sandín, Chorot, Valiente and Lostao2009).
Psychometricians recommended the inclusion of positively and negatively worded items in self-report measures to reduce response style bias (the tendency to respond to items without paying attention to their content) and acquiescence (the tendency to agree with the sentences) (Nunnally & Bernstein, Reference Nunnally and Bernstein1994). Negatively worded items seemed, however, to require additional cognitive effort to be interpreted and answered than positively worded items. Consequently, they may produce unreliable answers and lead to a method artefact in the measure.
The reliability of the PSWQ–16 was adequate, and it was in accordance with the reliability obtained in prior research (Startup & Erickson, Reference Startup, Erickson, Davey and Wells2006). Temporal stability was also satisfactory and consistent with values found in previous studies (Meyer et al., Reference Meyer, Miller, Metzger and Borkovec1990). Two of the items (Numbers 1 and 11) included a double negation in the Portuguese translation. Double negatives involve complex grammatical structures and are usually ambiguous. For this reason, inclusion of double negation in psychological measures is usually discouraged. Compared to other studies (e.g., Castillo et al., Reference Castillo, Macrini, Cheniaux and Landeira-Fernandez2010), in our study, these items showed the most reduced item-total correlations. The reliability of the PSWQ–11 version was superior to that obtained in the whole scale. This result was congruent with Sandín et al.’s (Reference Sandín, Chorot, Valiente and Lostao2009).
McDonald’s omega coefficient was proposed as a better estimate for the internal consistency of a self-report instrument (Dunn, et al., Reference Dunn, Baguley and Brunsden2014). Therefore, in addition to Cronbach’s alpha, which allows us to compare our results with those obtained in previous studies, we also computed McDonald’s omega. Compared to Cronbach’s alpha coefficients, the differences between the internal consistency values of PSWQ–16 and PSWQ–11 were intensified when McDonald’s omega was computed, showing an even worse result for the full, 16-item version and a better result for of 11-item version.
Both PSWQ versions showed adequate convergent validity. Associations were significant between this measure and other instruments to assess worry and maladaptive strategies to deal with it, such as punishment or thought suppression. Associations between PSWQ and adaptive strategies such as distraction or reappraisal were, however, low.
Our results are congruent with a recent investigation conducted with the aim to determine the psychometric implications of the use of positive and reversed items in measurement instruments (Suárez-Álvarez, Pedrosa, Lozano, García-Cueto, Cuesta, & Muñiz, Reference Suárez-Álvarez, Pedrosa, Lozano, García-Cueto, Cuesta and Muñiz2018). As in the present study, Suárez-Álvarez, et al., Reference Suárez-Álvarez, Pedrosa, Lozano, García-Cueto, Cuesta and Muñiz2018) found that when reversed items were introduced, the dimensionality of the instrument was jeopardized by another source of variance: The instrument’s measurement precision was lower, and verbal skills seemed to influence participants’ responses. In any case, and given that the two Portuguese versions of PSWQ (16 and 11 items) seem to show adequate levels of reliability and validity, we encourage investigators to report results for both versions of the PSWQ in order to further investigate their characteristics in research and applied settings.
As worry is a common phenomenon, community participants presented, as expected, moderate tendencies to worry. Similar results were found in previous studies (e.g., Sandín et al., Reference Sandín, Chorot, Valiente and Lostao2009). There were no differences between the tendency to worry experienced by men and women. At this point, the results of previous studies are conflicting. Some of them also did not find any differences (e.g., Crittendon & Hopko, Reference Crittendon and Hopko2006), while others found a greater worry tendency for women than for men (e.g., Castillo et al., Reference Castillo, Macrini, Cheniaux and Landeira-Fernandez2010; Sandín et al., Reference Sandín, Chorot, Valiente and Lostao2009). Some studies also found a great tendency to worry in young people (e.g., Gillis, Haaga, & Ford, Reference Gillis, Haaga and Ford1995), but in our study, we did not find differences in worry based on age. Hence, excessive worry in our community Portuguese sample was moderate and did not differ by participants’ age or gender.
The present study was conducted in a non-clinical sample of volunteers. This limits the generalization of the results to the general Portuguese population or to clinical populations. Additionally, the discriminant and criterion validity were not investigated. Future studies should try to replicate these results with anxiety patients and specifically with GAD patients due to the relationship between worry and this disorder.
Despite the disadvantages of the PSWQ–16 version due to the inclusion of negative items, it seems to be a sufficiently robust measure to assess the excessive tendency to worry. The PSWQ–11 items version may be, however, a better alternative in applied settings with the advantage of the reduction in application time.
The study protocol was approved by the Scientific Committee of the Faculty of Social and Human Sciences of the University of the Algarve- Portugal. All participants gave their consent to participate in the online application.
Authors have no potential conflicts of interest to disclosure.
