Our study included 4,116 heterosexual Spanish adults from the general population (1,867 men, 2,249 women) aged between 18 and 83 years (M = 40.58; SD = 12.24). Table 1 offers the participants’ socio-demographic characteristics. Of all the participants, 57.30% had university studies, and women’s level of education was higher than that of men. Most (72.40%) had a partner at the time the evaluation was made, with a slightly higher percentage for men, and 92.80% of the people in a relationship practiced sex, while 67.10% with no partner reported having at least one sexual relationship in the past 6 months. Practically all the participants had masturbated on some occasion (94.30%), but men more frequently than women. The age of their first masturbation experience was younger for men (M = 13.05 years) than women (M = 16.52 years). Finally regarding their religiousness, 69.40% of the participants never prayed.

Background Questionnaire. This instrument was used to collect information about sex, age, education level, nationality, partner relationship, current sexual activity, sexual orientation, age of their first sexual experience, masturbation experience, age of their first masturbation experience, and religiosity.

The Negative Attitudes Toward Masturbation Inventory (NATMI; Mosher, 2011). This inventory is a 30-item, 5-point Likert-type scale anchored by 1 (not at all true for me) and 5 (extremely true for me). Higher scores indicate a more negative attitude towards masturbation. Its psychometric properties are described in the Introduction.

The Spanish version of the Sexual Opinion Survey-6 (SOS-6; Vallejo-Medina et al., 2014). It measures erotophilia with six items on a 7-point Likert-type scale ranging from 1 (I strongly disagree) to 7 (I strongly agree). The higher the score, the higher the degree of erotophilia. Its internal consistency reliability is adequate (α = .74) and it evidences convergent validity by correlating its scores with sexual satisfaction, sexual desire, sexual functioning, sexual assertiveness and positive attitudes towards sexual fantasies (Vallejo-Medina et al., 2014). In the present study, the ordinal alpha coefficient was .82.

The Spanish version of the Hurlbert Index of Sexual Fantasy (HISF; Sierra, Arcos-Romero et al., 2020). It consists of ten items to evaluate the positive attitude towards sexual fantasies on a Likert scale from 0 (never) to 4 (all of the time). Higher scores indicate more positive attitude towards sexual fantasies. Its internal consistency reliability is .94 and it presents suitable evidence of validity by being related to similar measures. In this study, the ordinal alpha coefficient was .91.

The Solitary Sexual Desire Subscale of the Spanish version of the Sexual Desire Inventory (SDI; Moyano et al., 2017). Its four items evaluates interest in solitary sexual activity. It presents different scale kinds with Likert-type responses depending on the item (e.g., from 0 = no desire to 8 = strong desire). High scores indicate higher levels of desire for solitary sexual activities. Its internal consistency reliability is .90 for men and .93 for women. There is evidence of suitable convergent validity. Cronbach’s alpha in the present study was .84.

The Spanish version of the Sexual Inhibition/Sexual Excitation Scales-Short Form (SIS/SES-SF; Moyano & Sierra, 2014). It is composed of 14 items that assess sexual excitation (SES), sexual inhibition due to the threat of performance failure (SIS1) and sexual inhibition due to the threat of consequences (SIS2). Responses are provided on a Likert scale ranking from 1 (strongly disagree) to 4 (strongly agree). Higher scores indicate greater excitation and inhibition proneness. The internal consistency reliability of its subscales is over .60 and its authors provide evidence of the validity of its measures. In this study, the ordinal alpha values were .83 for SES, .72 for SIS1 and .71 for SIS2.

The Spanish version of the Arizona Sexual Experience Scale (ASEX; McGahuey et al., 2000; Sánchez-Fuentes et al., 2019). It is made up of six items that evaluate sexual functioning in the last 7 days in terms of desire, excitation, orgasm, erection (in men), vaginal lubrication (in women), facility of having an orgasm and orgasm satisfaction. It employs a Likert-type scale from 1 (good functioning) to 6 (bad functioning). High scores reflect worse sexual functioning. The scale shows suitable psychometric properties: Cronbach’s alpha of .81 and .79 in men and women, respectively, and suitable evidence of validity. In the present study, the ordinal alpha was .81 for men and .86 for women.

Following the International Tests Commission’s guidelines (see Hernández et al., 2020; Muñiz & Fonseca-Pedrero, 2019), the NATMI items were translated from English to Spanish by two researchers specialised in sexuality with a high level of English. A consensus on the translations was reached with a bilingual psychologist, who is also an expert in sexuality. The obtained version was judged by four experts in Psychometry and Sexuality, who scored to what extent, from 1 (not at all) to 4 (considerably), all 30 items matched the four criteria (representativeness, comprehension, clarity and ambiguity), and suggested improvements if they deemed them necessary. An analysis with Aiken’s V with a 95% confidence interval was carried out. The results showed that a suitable consensus had been reached with good construct representativeness as the lower Aiken’s V limit was found in all the items over .50. For comprehension, items 3, 10, 17, 20 and 27 had values below .50 and items 10 and 20 were evaluated as not very clear. Finally, items 1, 3, 5 and 10 were found to be ambiguously worded. After revising the seven different items following the experts’ recommendations, the pilot version of the scale was applied to a group of 50 Spanish adults from the different sex-paired ages to see if they understood each item or not. Only item 20 obtained a consensus below 90% by those surveyed. Thus it was revised to improve its comprehension and the definitive version of the instrument was obtained.

The battery of instruments was applied online. This is the usual method in studies that evaluate sexual behaviours (Arcos-Romero & Sierra, 2019; Calvillo et al., 2020; Sánchez-Mendoza et al., 2020; Tavares et al., 2019) and it use is also recommended to study masturbation (Burri & Carvalheira, 2019; Carvalheira & Leal, 2013). Former research has indicated no differences between the responses obtained by this method and the traditional paper-and-pencil kind (Álvarez-Muelas et al., 2021; Carreno et al., 2020; Sierra et al., 2018). The online battery was distributed using virtual platforms (Facebook®, Twitter®, WhatsApp® groups, and e-mail), using LimeSurvey® software. The IP address of the responses was controlled; to avoid automated responses, participants were asked to confirm their access to the survey by responding to a security question consisting of a simple randomized arithmetic operation. Anonymity was guaranteed to all participants, as well as the confidentiality of their data, and their participation was voluntary. Before responding, participants were asked to read and accept an informed consent form, which described the purpose of the study and provided information on data confidentiality and privacy. The study was approved by the University of Granada Human Research Ethics Committee.

Our sample was divided into two random subsamples: Subsample 1 (n = 2,060) was used to carry out the Exploratory Factor Analysis (EFA), while Subsample 2 (n = 2,059) was employed for the Confirmatory Factor Analysis (CFA). All the other analyses were performed with the whole sample. To process any missing data, an algorithm for non-parametric distributions was applied to create a random forest model for each variable by means of the other variables from the database. The estimated error range for the imputation was 26%. With subsample 1, the factorial reduction of the inventory was explored by considering 23 methods to obtain the number of factors. Next according to this estimation, an EFA was performed on the polychoric matrix for which maximum likelihood was employed as the estimation method. Once the EFA results were known, the CFA was performed in the polychoric matrix in Subsample 2. The Weighted Least Squares Means and Variance Adjusted (WLSMV) estimation method was used, which is suitable for non-parametric samples comprising ordinal or categorical data (Lara Hormigo, 2014). The following fit indices were contemplated for the CFA: (a) root mean square error of approximation (RMSEA); (b) comparative fit index (CFI); (c) Tucker-Lewis Index (TLI). The RMSEA values were lower than .06 (Browne & Cudeck, 1993), and the CFI and TLI values exceeded .95 (Kline, 2011), which are indicators of a good model fit. Then internal consistency was examined by the ordinal alpha. Finally to examine discriminant validity, two age-paired groups were formed: (1) 136 cases (Mage = 38.24; SD = 13.78) with two standard deviations over the mean score (i.e., score that equals or exceeds 15: a more negative attitude towards masturbation); (2) 102 cases (Mage = 38.83; SD = 11.94) with scores on the opposite pole (score of 10: a possibly more positive attitude towards masturbation). The differences between both groups were examined in terms of religious frequency and masturbation, erotophilia, attitude towards sexual fantasies, solitary sexual desire and sexual functioning (sexual desire, excitation, erection/vaginal lubrication and orgasm). To examine the differences, the data fit was compared according to the null hypothesis and the alternative hypothesis by applying Fisher’s ANOVA Bayesian analysis. An rJZS = 0.71 was employed. A positive outcome contributes a higher likelihood to favour the null hypothesis, while a negative outcome favours the alternative hypothesis. An outcome is considered more robust the further away it is from zero by assuming the following intervals (Jeffreys, 1961): 1-3 anecdotal, 3-10 substantial, 10-20 strong, 20-30 strong, 30-100 very strong, 100-150 decisive, >150 decisive.

To carry out the analyses, the R® environment was used (version 3.6.3; R Core Team, 2020) with its RStudio® interface (version 1.2.5042; RStudio Team, 2020). To impute any missing values, the missForest package was employed (version 1.4; Stekhoven & Bühlmann, 2011). To explore the factorial structure, the Parameters package was resorted to (version 0.8.0; Lüdecke et al., 2020). The EFA and the ordinal alpha calculation were performed with the Psych package (version 1.9.12.31; Revelle, 2019). For the CFA, the lavaan package was utilised (Rosseel, 2012). Finally, the tidyBF package (version 0.4.0; Patil, 2018) was used for the Bayesian analyses.

First of all, in Subsample 1 the factorial structure underlying the original 30-item version was examined. We observed that most methods (four: acceleration factor, VSS complexity 1, TLI, RMSEA) supported the unifactorial structure. In fact the second proposal supported by the three methods suggested a 29-factor structure, which would indicate very little cohesion with items. This would be consistent with the fact that the variance explained by the unidimensional version was low: 41%. What all this evidences is the need to reduce the number of items in order to seek a more consistent structure.

In order to improve the explained variance in NATMI, all the items whose communality (h2) was above .50 were selected by means of the EFA. With this criterion, ten items were selected with which a short version was proposed (items 10, 30, 23, 12, 4, 18, 6, 20, 7, and 27). The factorial structure of this short version was explored by the above-described parameters where, once again, a single factor was proposed (backed by nine methods: 39.13%). Given the greater consensus reached with the unifactorial 10-items version, the EFA for one factor was applied, which explained 66% of total variance (see Table 2).

Next a CFA was performed in Subsample 2 to test the fit of the 10-item version of the unifactorial model. This single-factor structure showed an acceptable fit: RMSEA = .071; 90%CI RMSEA = .065-.078; CFI = .930; TLI = .910; χ2(35) = 400.06, p <  .01. Figure 1 illustrates the flow chart of the unifactorial model where the standardised loadings fell within the .68 (item 27) and .90 (item 23) range.

Figure 1.

Standardised estimates of the 10-item unifactorial model.

Note. The items maintain the numbering of the original 30-item version.

(0.07MB).

Having confirmed the structure of one factor, its internal consistency was analysed by the ordinal alpha for the total inventory score. The 10-item version gave an excellent internal consistency reliability coefficient (ordinal alpha = .95). As we can see in Table 3, eliminating any of these 10 items did not improve this alpha. It is worth stressing the high kurtosis for most items as most responses were for response 1 option (not at all true for me) of the instrument’s Likert-type scale, which would be expected in a non-clinical Spanish general population.

Finally given the high kurtosis, the NATMI scores were compared between those people with a negative attitude and those with a positive attitude towards masturbation. As Figure 2 shows, both groups proved significantly different for masturbation frequency, religiousness, erotophilia, positive attitude towards sexual fantasies, sexual inhibition due to the threat of performance failure and solitary sexual desire, and effect sizes ranged between moderate and large. Significant differences were also found between both groups for sexual functioning, with generally moderate effect sizes (see Figure 3). All these differences were corroborated with Bayesian probability.

Figure 2.

Comparisons between groups with positive and negative attitudes towards masturbation in frequency of masturbation, religiosity, erotophilia, positive attitude towards sexual fantasies, sexual inhibition due to fear of failure in sexual performance and solitary sexual desire.

Note. The red dot shows the population mean (µ) of that group together with the corresponding confidence interval. Below each figure are the results of the Bayesian analysis.

(0.4MB).

Figure 3.

Comparisons between groups with positive and negative attitudes towards masturbation in the dimensions of sexual functioning: desire, arousal, erection/vaginal lubrication, ease of reaching orgasm and orgasm satisfaction.

Note. The red dot shows the population mean (µ) of that group together with the corresponding confidence interval. Below each figure are the results of the Bayesian analysis.

(0.37MB).