Translation, validity, and reliability of the European Portuguese version of the Touch Experiences and Attitudes Questionnaire

Participants

A total of 384 participants completed the study, 299 were female (77.9%) and 85 (22.1%) were male. Age ranged from 18 to 75 years (M = 24.59; SD = 9.56). Demographic variables are described in Table 1. Inclusion criteria included being 18 years of age or older, being native European Portuguese speakers, and completing all questionnaires (TEAQ, STQ, and NEO-FFI). From an initial sample of 475, 73 participants were excluded for not completing all questionnaires, and 18 for not being native European Portuguese speakers. All the participants voluntarily agreed to answer the questionnaire and provided written informed consent. None of the participants received any monetary compensation, and 67 university students received credits for their participation. The study protocol obtained approval from the ethics committee for Research in Human and Social Sciences of the University of Minho (CEICSH 023/2020) and was in accordance with the Declaration of Helsinki.

Measures

Sociodemographic information

Participants completed a sociodemographic questionnaire assessing personal information, namely age, biological sex, nationality, educational level, civil status, number of people they live with, number of children, and professional situation.

Touch Experiences and Attitudes Questionnaire (TEAQ)

The TEAQ (Trotter et al., 2018a) is a 57-items self-report questionnaire that reflects touch experiences at present and during childhood, as well as attitudes towards touch. The original version (UK validated) includes six subscales: Friends and Family Touch (FFT; with 11 items), relating to interpersonal touch experiences with friends and family; Current Intimate Touch (CIT; with 14 items), determines current levels of intimate touch experiences between people who are emotionally close or in a romantic relationship; Childhood Touch (ChT; with nine items), provides a measure of positive childhood touch experiences; Attitude to Self-Care (ASC; with five items), related to skin care and grooming behaviours associated with positive self-care; Attitude to Intimate Touch (AIT; with 13 items), referring to attitudes towards touch experiences between people who are emotionally close or in a romantic relationship; and Attitude to Unfamiliar Touch (AUT; with five items), that relates to attitudes to touch experiences with someone that the individual is less close to, or even a stranger. Items are answered using a 5-point Likert scale ranging from 1—‘Disagree strongly’, 2—‘Disagree a little’, 3—‘Neither agree nor disagree’, 4—‘Agree a little’ to 5—‘Agree strongly’, with eight reverse-scored items. Therefore, higher scores reflect a more positive attitude toward touch or more frequent experiences. The total score for each subscale is obtained by calculating the mean score of the items for each subscale.

Social Touch Questionnaire (STQ)

The STQ (Wilhelm et al., 2001; Portuguese version, Vieira et al., 2016) is a 20-item self-report questionnaire that measures individual perceptions, behaviours, and attitudes toward the social touch. The items represent a vast variety of social interactions, such as touching versus being touched, touching someone you meet versus touching a stranger, touching someone in a public place versus in a private place, and touching without a sexual connotation versus touching with a sexual connotation. Each item is rated on a 5-point Likert scale to state how far the statements are true, ranging from 0 (‘Not at all’) to 4 (‘Extremely’). The score ranges from 0 to 80 and results from the sum of the items, in which a low score represents the lowest avoidance of social touch, and a high score represents the most avoidance of social touch. The Portuguese version demonstrated good internal consistency (Cronbach’s α = .73). In the present study, this questionnaire had Cronbach’s α = .82.

NEO Five-Factor Inventory (NEO-FFI)

The NEO-FFI (McCrae & Costa, 1987; Portuguese version, Magalhães et al., 2014) is a 60-item self-report inventory that assesses individual personality traits. The NEO-FFI is a short version of NEO PI-R and provides a concise measure of the five basic personality factors (Neuroticism, Extraversion, Openness to Experience, Agreeableness, and Conscientiousness) with 12 items for each factor. Each item is measured on a 5-point Likert scale of agreement (0—‘Strongly disagree’, 1—‘Disagree’, 2—‘Neither agree nor disagree’, 3—‘Agree’, 4—‘Strongly agree’). The score for each personality factor results from the sum of the respective 12 items. The Portuguese version demonstrated good internal consistency (with Cronbach’s alpha values ranging from .71 to .81). In our study it had Cronbach’s alpha values between .66 and .85.

Procedure

The translation of the original English version of the TEAQ to European Portuguese was performed following previous recommendations (see Fig. 1) (Hilton & Skrutkowski, 2002). First, the original version of the TEAQ was translated to European Portuguese independently by two translators (psychology researchers), whose native language is European Portuguese. After that, a reconciled version was made by three researchers that compared the two versions of the translation and the original one and built a unique version. This version was back-translated from European Portuguese into English by a bilingual translator (psychology researcher), whose native language is English, and without any knowledge of the original items of the TEAQ. To analyse the equivalence between the original version and the retroversion, we also consulted the original authors. A preliminary version of the European Portuguese TEAQ was developed, and a pilot test was conducted with six individuals, three male and three female, aged between 22 and 25 ( M = 23.83; SD = 1.07), to verify any difficulty or doubt when completing the items, and based on their feedback, a final version was developed and released (no changes were made). Participants were recruited via social media platforms and university credit system, through which they were asked to complete an online questionnaire conducted via Qualtrics software (Qualtrics, Provo, UT). All data were collected between April 2020 and January 2021. The final European Portuguese version of the TEAQ is included in Appendix A. The authors have permission to use this instrument from the copyright holders (Trotter et al., 2018a).

Statistical analysis

The descriptive and inferential statistical analyses were conducted using IBM SPSS 28. The multidimensional Rasch model and the differential item functioning (DIF) were estimated using the ACER Conquest 5.19.5 software. Confirmatory factor analysis was carried out using IBM SPSS Amos™ 28.

To assess normality, we analysed the asymmetry values, skewness and kurtosis, revealing no significant biases in relation to the averages, with values lower than —3— and —10— for skewness and kurtosis, respectively (Kline, 2016). Also, to examine the suitability for conducting a confirmatory factor analysis (CFA), the factorability of the correlation matrix was explored. Bartlett’s sphericity test proved to be statistically significant (p < .001) and the Kaiser-Meyer-Olkin (KMO) value was .90, above the recommended criterion of .60 (Pallant, 2011). An analysis of the correlation matrix also revealed several correlation coefficients greater than .30 (Tabachnick & Fidell, 2013).

The parameters of the multidimensional Rasch model were estimated with a marginal maximum likelihood estimation method for item parameters and a Monte Carlo-based approach with 2000 nodes for person parameters. Through this model, we obtained information about item fit, with the scores of the infit and outfit mean square (MNSQ) between .60 and 1.4 indicating that the item was fit for the model with Likert scales (Wright & Linacre, 1994). Items outside this range were removed from the questionnaire. Additionally, the performance of the 5-point Likert scale was examined using an outfit MNSQ <2.0, demonstrating that the response category was functioning well (Linacre, 1999). Furthermore, we examined the item separation reliability, in which values above .67 were considered acceptable (Utari, Liliawati & Utama, 2021). Finally, we obtained information about person separation reliability for each TEAQ subscale in the form of plausible values (PV) reliability indices, in which values above .70 were considered acceptable (Fauth et al., 2019).

The original six-factor model and the parcelled six-factor model of the TEAQ, as originally proposed by Trotter and colleagues (Trotter et al., 2018a), were tested through a CFA, according to the maximum likelihood method. The parcelled six-factor model was used since the reliability of structural equation modelling is reduced when there are a large number of variables. Using the item parcelling method (Nasser & Wisenbaker, 2003) to produce three measures for each factor, it was determined whether the model fit indexes would be affected, as suggested by Yang, Nay & Hoyle (2010). In this procedure, the same item parcelling suggested by Trotter et al. (2018a) in the original version of this questionnaire was used, excluding only the items removed in the multidimensional Rasch model analysis. Thus, we explored if the parcelling method improved the criteria for a good model. The following indexes were considered to determine the goodness of model fit: Relative Chi-Square (x2/df), Normed Fit Index (NFI), Comparative Fit Index (CFI), Goodness-of-Fit Index (GFI), and Root Mean Square Error of Approximation (RMSEA). We considered values of Relative Chi-Square ≤ 5 acceptable fit and ≤ 2 perfect fit; NFI values ≥ .90 acceptable fit and ≥ .95 good fit; CFI values ≥ .90 acceptable fit, ≥ .95 good fit, and ≥ .97 perfect fit; GFI values between .85 and .89 acceptable fit, ≥ .90 good fit, and ≥ .95 perfect fit; and RMSEA values between .10 and .08 weak fit, ≤ .08 good fit, and ≤ .06 perfect fit. The Akaike Information Criterion (AIC) was also examined to compare models (Arbuckle, 2013; Hooper, Coughlan & Mullen, 2008).

Additionally, internal consistency was assessed through Cronbach’s alpha values, with values equal to or greater than .70 considered satisfactory (Bland & Altman, 1997; Streiner, 2003). We also computed Pearson’s correlations between TEAQ subscales, and between TEAQ and the other questionnaires, namely the STQ (for convergent validity), and the NEO-FFI (for discriminant validity) (r ≤ —.12——very weak correlation; —.12—≥ r ≥—.24——weak correlation; —.24—≥r ≥—.41——moderate correlation; —.41—≥ r ≥ —1——high correlation) (Lovakov & Agadullina, 2021). DIF was explored using the Rasch model method to compare differences in the functioning of individual items across the two groups (female and male participants), with the following criteria: DIF contrast <.50 logits—negligible DIF; DIF contrast between .50 and 1 logits—moderate DIF; and DIF contrast >1 logits—substantial; if the DIF was substantial and significant it might mean that each group interprets the scale differently and that confounding factors are influencing that interpretation (Bond & Fox, 2015; Cameron et al., 2014). To examine sex differences in the TEAQ subscales, independent samples t-tests were performed, and effect-sizes were measured through Cohen’s d calculation (Cohen’s d: ≥ .20—small; ≥ .50—medium; ≥ .80—large) (Cohen, 1992).

Multidimensional Rasch model analysis

Table 2 shows the statistics for the TEAQ items, including item difficulty location (endorsability) and item fit statistics. All items fit the multidimensional Rasch rating scale model, with the values of infit and outfit MNSQ within the acceptable range (.60 –1.4). The location of items is in the range of −0.842 (Q1) to 1.020 (Q23) logits with an item separation reliability of .99. Item 1 (“Eu não gosto quando as pessoas são muito afetuosas fisicamente em relação a mim”. [I dislike people being very physically affectionate towards me]) was the easiest to endorse (i.e., the easiest item for obtaining a score of 5) with its location at −0.842 logits, and item 23 (“Eu costumo tomar um duche ou um banho com alguém” [I often take a shower or bath with someone]) was the most difficult to endorse with its location at 1.020 logits. Five items (three items from the CIT subscale - items 23, 53, and 54 from the original version; one item from the ChT subscale - item 42 from the original version; and one item from the AIT subscale - item 20 from the original version) were excluded since the infit and outfit MNSQ values were not within the acceptable range. The item’s threshold (τ) appeared in the analysis from the lowest to the highest (τ1 = −0.921; τ2 = −0.177; τ3 = −0.064; τ4 = 1.161). Additionally, person separation reliability in the form of PV reliability for each subscale were .97 for FFT, .93 for CIT, .93 for ChT, .82 for ASC, .87 for AIT, and .82 for AUT, all with acceptable criteria. The distribution of persons and the 52 items’ threshold estimates on the different TEAQ subscales can be consulted in the wright map (see Fig. 2). Therefore, after the multidimensional Rasch model analysis, the Portuguese TEAQ was left with 52 items: FFT (11 items)—4, 13, 14, 16, 20, 28, 36, 45, 48, 51, and 52; CIT (11 items)—11, 17, 18, 23, 25, 27, 34, 39, 42, 43, and 46; ChT (8 items)—5, 6, 9, 15, 21, 30, 31, and 33; ASC (5 items)—2, 7, 40, 49, and 50; AIT (12 items)—8, 10, 12, 19, 22, 24, 29, 32, 38, 41, 44, and 47; AUT (5 items)—1, 3, 26, 35, and 37; in which 1, 3, 9, 26, 35, and 37 are reverse-scored items.

Confirmatory factor analysis and internal consistency

Results of the 52-items confirmatory factor analyses for the original and the parcelled six-factor model are reported in Table 3. Examination of the goodness of fit indexes of competing models revealed that the parcelled six-factor model had a good fit to data (see Table 3 and Fig. 3), as observed by the authors of the original version of the TEAQ (Trotter et al., 2018a). The six factors of the TEAQ revealed good to excellent values of internal consistency, with the exception of AUT subscale, which revealed acceptable internal consistency (FFT, α = .90; CIT, α = .88; ChT, α = .90; ASC, α = .81; AIT, α = .89; AUT, α = .78).

Correlations between subscales

All TEAQ subscale scores correlated significantly with each other, except the AUT with the CIT and ASC subscales (see Table 4). Significant high correlations were found between FFT and CIT (r = .42; p < .001), FFT and ChT (r = .50; p < .001), FFT and AIT (r = .46; p < .001), and CIT and AIT (r = .58; p < .001) subscales. Significant moderate correlations were found between FFT and ASC (r = .30; p < .001), FFT and AUT (r = .40; p < .001), CIT and ChT (r = .32; p < .001), CIT and ASC (r = .34; p < .001), ChT and AIT (r = .26; p < .001), and ASC and AIT (r = .31; p < .001) subscales. Significant weak correlations were found between ChT and ASC (r = .21; p < .001), ChT and AUT (r = .15; p < .01), and AIT and AUT (r = .19; p < .001) subscales. The strongest correlation was observed between CIT and AIT (r = .58), and the weakest significant correlation between ChT and AUT (r = .15).

Correlations with other scales: convergent and discriminant validity

Regarding convergent validity, the STQ and TEAQ subscales correlations were significant (p < .001). These results revealed high correlations with FFT (r =  − .59), AIT (r =  − .43), and AUT (r =  − .70) subscales, moderate correlations with CIT (r =  − .25), and ChT (r =  − .31) subscales, and a weak correlation with ASC (r =  − .19) subscale (see Table 5). Concerning discriminant validity, we made correlations between the five personality factors of NEO-FFI and the TEAQ subscales (see Table 5). For the Neuroticism (N) factor, a moderate correlation was found with the AUT (r =  − .29; p < .001) subscale, and weak correlations with the FFT (r =  − .16; p < .01), CIT (r =  − .12; p < .05), and ChT (r =  − .15; p < .01) subscales, but no significant correlations with ASC and AIT. For the Extraversion (E) factor, a high correlation emerged with the FFT (r = .45; p < .001) subscale, moderate correlations with the CIT (r = .35; p < .001), ChT (r = .30; p < .001), AIT (r = .35; p < .001), and AUT (r = .34; p < .001) subscales, and a weak correlation with the ASC (r = .15; p < .01) subscale. For the Openness to Experience (O) factor, we observed a moderate correlation with the AIT (r = .25; p < .001) subscale, and a very weak correlation with the FFT (r = .11; p < .05) subscale, but no significant correlations with the CIT, ChT, ASC, and AUT. For the Agreeableness (A) factor, moderate correlations were observed with FFT (r = .34; p < .001), ChT (r = .27; p < .001), and AIT (r = .32; p < .001) subscales, and weak correlations with the CIT (r = .21; p < .001), ASC (r = .14; p < .01), and AUT (r = .17; p < .001) subscales. Lastly, for the Conscientiousness (C) factor, a moderate correlation was found with the CIT (r = .25; p < .001) subscale, and weak correlations with the FFT (r = .19; p < .001), ChT (r = .16; p < .01), ASC (r = .19; p < .001), and AIT (r = .23; p < .001) subscales, but no correlation with AUT.

Sex differences on TEAQ

No substantial DIF was found in the TEAQ items across female and male participants. Moderate DIF was found for item 3—“Eu tenho de conhecer alguém relativamente bem para desfrutar do seu abraço”. [I have to know someone quite well to enjoy a hug from them] (i.e., relatively easier for males than females [0.608 logits]), item 24—“Eu gosto de fazer sexo”. [I enjoy having sex] (i.e., relatively easier for males than females [0.636 logits]), item 26—“Eu não gosto de ter proximidade física com pessoas que não conheço bem”. [I am put off by physical familiarity] (i.e., relatively easier for males than females [0.622 logits]), item 37—“Faz-me sentir desconfortável se alguém que eu não conheço muito bem me toca de uma maneira amigável”. [It makes me feel uncomfortable if someone I don’t know very well touches me in a friendly manner] (i.e., relatively easier for males than females [0.528 logits]), item 40—“Eu gosto de exfoliar a minha pele”. [I like exfoliating my skin] (i.e., relatively easier for females than males [0.532 logits]), and item 50—“Eu gosto de usar máscaras faciais na minha pele”. [I like to use face masks on my skin] (i.e., relatively easier for females than males [0.954 logits]). All other items have DIF <.50, and item 19—“Beijar é uma boa forma de expressar atração física”. [Kissing is a great way of expressing physical attraction] revealed the same level of difficulty for both groups. These findings show that, regardless of other variables, the items performed similarly for people with the same level of ability.

Table 6 shows the sex differences on TEAQ subscales. Females showed higher scores on FFT, CIT, ChT, and ASC subscales than males, with a large effect size on all the subscales. AIT subscale scores for female and male participants were comparable. AUT subscale scores were higher for males than females, with a large effect size.