Descriptive statistics on the participants sociodemographic variables

The 1823 participants varied in age from 19 to 65 years (M = 31.66, SD = 6.74). Males were more heavily represented than females (male = 1155 [63.4%], female = 636 [34.9%], non-binary = 32 [1.8%]). About half were single and half were married or in a relationship (single = 900 [49.4%], in relationship but not married = 567 [31.1%], married = 317 [17.4%], divorced = 37 [2.0%], widowed = 2 [0.1%]). Socio-economic status varies as follows: low = 483 (26.5%), intermediate = 1265 (69.4%), high = 75 (4.1%). Most of the participants (79.0%) were whites, with other ethnic groups distributed as follows: Asian = 9.4%, mixed-race = 4.8%, other = 2.2%. Educational level (years of schooling) varied from a minimum of 4 years to a maximum of 27 (M = 15.94, SD = 3.08).

Participants resided in the United Kingdom (1,482; 77.2%), the United States (342;17.8%), Ireland (31; 1.6%), Australia (29; 1.5%), Sweden (24; 1.3%), and New Zealand (12; 0.6%). They were of diverse nationalities: 27 European countries (1,466; 76.1%), 2 North-American countries (307; 16%), 14 Asian countries (51; 2.7%), 8 African countries (31; 2%), 2 Oceania countries (34; 1.8%), 8 Latin-American countries (13; 0.8%), and 5 Middle Eastern countries (11; 0.8%).

Descriptive statistics of responses to scales’ questions, male vs. female comparisons, and correlation between ACSID-F and ACSID-I dimensions

Table 1 shows the descriptive statistics regarding the dimensions and symptoms from all study scales. As shown, only mood modification (a PPCS symptom) and dyadic and solitary sexual desire (the two SDI dimensions) had a score above the middle of the scale range.

The extent of cyberporn use by male participants (M = 2.15, SD = 0.95; scale 1–6) was significantly higher than by female participants (M = 1.75, SD = 0.79), F(2, 1788) = 43.19, p < 0.001, η2p = 0.045. The total PPCS score for male participants (M = 3.08, SD = 1.36; scale 1–7) was also significantly higher than for female participants (M = 2.31, SD = 1.18), F(2, 1788) = 75.35, p < 0.001, η2p = 0.057. The ACSID total score for male participants (M = 1.66, SD = 0.62; scale 1–4) was significantly higher than for female participants (M = 1.36, SD = 0.48), F(2, 1788) = 54.91, p < 0.001, η2p = 0.076. Similarly, the CSBD total score for male participants (M = 1.75, SD = 0.63; scale 1–4) was significantly higher than for female participants (M = 1.56, SD = 0.59), F(2, 1788) = 17.88, p < 0.001, η2p = 0.019.

The PPCS (scale 1–7 points) percentile distribution was as follows: 25th percentile = 1.66; 50th percentile = 2.50; 75th percentile = 3.83. Notably, 19.4% of participants had a score ≥ 4 points, 7.1% had a score ≥ 5 points, and only 1.5% had a score ≥ 6 points. The ACSID (scale 1–4 points) percentile distribution was as follows: 25th percentile = 1.10; 50th percentile = 1.36; 75th percentile = 1.80. Notably, 18.6% of participants had a score ≥ 2 points and only 3.2% had a score ≥ 3 points. The CSBD (scale 1–4 points) percentile distribution was as follows: 25th percentile = 1.15; 50th percentile = 1.47; 75th percentile = 2.05. Notably, 26.1% of participants had a score ≥ 2 points and only 3.7% had a score ≥ 3 points. The chi-square of independence test conducted on the percentiles’ distribution of these three assessment tools indicated significant statistical difference (X2 = 43.71, p < 0.05, η2 = 0.33).

Table 2 shows the correlations between ACSID-F and ACSID-I dimensions. As it can be seen in Table 2, the four ACSID-F factors are highly correlated to the same ACSID-I factors.

Table 2 Pearson correlation between symptoms of the assessment of criteria for specific internet-use disorders — frequency and intensity formsComparison of PPCS vs. ACSID psychometric properties (response to RQ1)Assumption for factorial analyses

For the PPCS, the assumptions of adequacy (Kaiser–Meyer–Olkin [KMO] test = 0.864) and sphericity (Bartlett’s test [df = 15] = 5176, p < 0.001) for factorial analyses were met [88].

For the ACSID-F and ACSID-I, the assumptions of adequacy (KMO test = 0.906) and sphericity (Bartlett’s test [df = 231] = 39284.45, p < 0.001) for factorial analyses were also met [88].

EFA results and CFA indices

Regarding the PPCS, EFA based on eigenvalue yielded one factor structure explaining 61% of the variance. EFA based on parallel analysis yielded two different results: the oblimin rotation outputted a structure with two factors explaining 66.3% of the variance (Factor 1 [salience and mood modification; 41.5%] and Factor 2 [tolerance, relapse, conflict, withdrawal; 24.08%]); and the varimax rotation outputted a structure with three factors explaining 72% of the variance (Factor 1 [salience and mood modification; 33.8%], Factor 2 [tolerance, relapse, conflict; 21.0% ]), and Factor 3 [withdrawal; 17.4%]).

Table 3 The current study confirmatory factorial analysis results: Fit indices by model

Table 3 represents the results of all CFA conducted. As shown, regarding the PPCS scale, the model with one factor had poor adjustment to the data (Model-A: x2/df = 41.11; CFI = 0.93; TLI = 0.88; SRMR = 0.051; RMSEA = 0.148), whereas the model with two factors (Model-B) and the model with three factors (Model-C) showed relatively good fit. Overall, Model-C seemed to be the best model (x2/df = 4.87; CFI = 0.99; TLI = 0.99; SRMR = 0.046; RMSEA = 0.032). Interestingly, Model-B differs from the two-factor model found by Fournier et al. [6] in an Italian sample, which had: Factor 1 (salience and tolerance) and Factor 2 (relapse, conflict, mood modification, withdrawal), with the following CFA fit indices: x2 = 98.729, df = 8, p < 0.001; CFI = 0.986; TLI = 0.974; SRMR = 0.073; RMSEA = 0.033. Therefore, using the Fournier et al. [6] factorial structure above, we conducted CFA on the present study data (Model-D, see Table 3) and found fit indices that indicated relatively poor adjustment to the data: x2/df = 45.75; CFI = 0.93; TLI = 0.87; SRMR = 0.050; RMSEA = 0.157.

The CFA conducted on the ACSID-F data (Model-E) suggested good adjustment: x2/df = 9.75; CFI = 0.97; TLI = 0.96; SRMR = 0.036; RMSEA = 0.069. The CFA conducted on the ACSID-I data (Model-F) indicated good fit: x2/df = 9.47; CFI = 0.97; TLI = 0.96; SRMR = 0.034; RMSEA = 0.068.

Internal reliability

The α of the PPCS one-factor structure (all six items/symptoms) was: 0.87. The α of the PPCS two-factor structure was: Factor 1 (salience and mood modification) = 0.76 and Factor 2 (tolerance, relapse, conflict, withdrawal) = 0.87.

The α of the ACSID-F was 0.90, and the α of the ACSID-I was 0.92. The α of the ACSID-F factorial structure was: Impaired control (IC) = 0.78; Increased priority (IP) = 0.85, continuation/escalation (CE) = 0.78, and functional impairment (FI) = 0.77. The α of the ACSID-I factorial structure was: IC = 0.80; IP = 0.86, CE = 0.79, and FI = 0.78.

Convergent validity

Table 4 shows the correlations between the six PPCS symptom scores, the PPCS total score, participants’ pornography use extent, sexual desire (dyadic and solitary), and the five factors of the CSBD. All r values are positive and show medium to high effect sizes.

Table 4 Pearson bivariate correlation between the symptom from Problematic Pornography Consumption Scale and from Assessment of Criteria for Specific Internet-use Disorders, sexual desire, and the participants socio-demographic characteristics

Table 4 also shows the correlations between, on the one hand, ACSID-F and ACSID-I factorial scores and ACSID-F and ACSID-I total scores, and, on the other hand, participants’ online pornography use extent, sexual desire (dyadic and solitary), and the five factors of the CSBD scale. Most, but not all, r values were positive and of medium to high effect sizes.

Importantly, as can be seen in Table 4, the values of the ACSID-F and ACSID-I correlations with the extent of online pornography use, SDI, and CSBD dimensions are generally lower (although not in a statistically significant manner), compared with the values of the correlations between PPCS and the online pornography use, SDI, and CSBD variables.

Discriminant validity

Table 5 displays the results of the logistics regression models (Model 1–4). They represent a discriminant analysis testing. It shows:

Table 5 The current logistic regression results: Estimated beta coefficients of the associations between the dimensions/symptoms from the Problematic Pornography Consumption Scale and the Assessment of Criteria for Specific Internet-use Disorders and compulsive sexual behavior “low clinical risk” vs. “high clinical risk” cases 1)

To what extent the participants’ scores on the PPCS and ACSID symptoms/factors discriminated between (a) participants with low CSBD total-scores (fist quartile = “low clinical risk” [coded “0”]) and (b) participants with high CSBD scores (fourth quartile = “high clinical risk” [coded “1”]);

2)

To what extent participants’ total scores on PPCS and ACSID-F/ACSID-I discriminated between (a) participants with low CSBD total-scores and (b) participants with high CSBD total-scores.

The “low clinical risk” group (n = 499) had a CSBD total-scores ≤ 1.16. Among them, 275 were male, 210 were female, and 14 were non-binary. The “high clinical risk” group (n = 446) had a CSBD total-scores ≥ 2.05. Among them, 323 were male, 118 were female, and 5 were non-binary. There was a statistical difference between the number of males and females present in each group (X2[df = 2] = 31.05, p < 0.001).

The descriptive statistics for the CSBD score were as follows: scale = 1–4 points; min score = 1, max score = 3.79, mean = 1.68, median = 1.47. The percentiles were: 25% = 1.16, 50% = 1.47, 75% = 2.05. The number of participants by quartile were: first quartile = 499 (27.4%), second quartile = 425 (23.3%), third quartile = 453 (24.8%), fourth quartile = 446 (24.5%).

The omnibus test model coefficients for Model 1 (PPCS total score and ACSID total score; X2 (df = 2) = 564.87, p < 0.001; Nagelkerke R2 = 60%), Model 2 (each of the PPCS symptoms score; X2 (df = 6) = 551.71, p < 0.001; R2 = 59%), Model 3 (each of the two PPCS factors total score; X2 (df = 2) = 545.99, p < 0.001; R2 = 57%), and Model 4 (each of the four ACSID factors total score; X2 (df = 4) = 441.36, p < 0.001; R2 = 50%) were significant, indicating a good fit to the data [89].

Model 1 results (see Table 5) suggests that the ACSID scale (odds-ratio [OR] = 3.88) performed better than the PPCS scale (OR = 2.92) at discriminating between “low clinical risk” of CSBD and “high clinical risk” of CSBD. Among the PPCS six symptoms (see Model 3), withdrawal (OR = 1.65), tolerance (OR = 1.41), and relapse (OR = 1.30) were the dimensions most able to discriminate between participants with “low clinical risk” and “high clinical risk” of CSBD. Among the ACSID four factors (see Model 2), increased priority (OR = 3.50) was the dimension that was the most able to discriminate between participants with “low clinical risk” and “high clinical risk” of CSBD.

The R2 values indicate the percentage of change in the dependent variable (CSBD) explained by the predictor variables in the model. The OR can be read here as the “effect-size” associated with each predictor in the model.

EFA results (response to RQ2)

For the EFA on the PPCS and ACSID dimensions total scores, assumptions of adequacy (KMO = 0.919) and sphericity (Bartlett’s [df = 120] = 11692.43, p < 0.001) were met [88]. The α were: ACSID-dimensions-total-score = 0.88 and PPCS = 0.87.

The EFA based on eigenvalue (including rotation) yielded a two-factor structure explaining 66.29% of the variance (Factor 1 [all ACSID-F five dimensions; 35.49%], Factor 2 [all PPCS six dimensions; 30.8%]). The EFA based on parallel analysis and varimax rotation yielded a four-factor structure explaining 66.4% of the variance (Factor 1 [ACSID-Increased priority, ACSID-Continuation/escalation, ACSID-Functional impairment, ACSID-Marked distress; 24.57%], Factor 2 [PPCS-Salience and PPCS-Mood modification; 23.88%], Factor 3 [PPCS-Tolerance, PPCS-Relapse, PPCS-Conflict, PPCS-Withdrawal; 12.72%], Factor 4 [ACSID-Impaired control; 5.19%]).

The CFA conducted on the above mentioned two-factor structure (Table 3, Model-G) showed a relatively poor fit to the data: x2/df = 24.06; CFI = 0.91; TLI = 0.89; SRMR = 0.055; RMSEA = 0.112. The CFA conducted on the above mentioned four-factor structure (Table 3, Model-H) showed acceptable adjustment to the data: x2/df = 15.02; CFI = 0.95; TLI = 0.93; SRMR = 0.036; RMSEA = 0.087.

Network relationships between symptoms (response to RQ3)Full zero-order Pearson correlations between the symptoms

Table 6 displays the correlations between all symptom score values. As shown, all correlations were medium or large. However, the correlations were stronger (e.g., r > or = 0.50) within the same category of symptoms. Among symptoms belonging to the two different assessment tools, the correlations were stronger between: ACISD-IP x PPCS-C = 0.63; ACSID-IC x PPCS-R = 0.59; ACSID-IP x PPCS-W = 0.59; ACSID-IP x PPCS-T = 0.57; ACCSID-IF x PPCS-C = 52; and ACSID-IC x PPCS-C = 0.50.

Table 6 Correlations between symptoms of Problematic Pornography Consumption Scale and the symptoms of Assessment of Criteria for Specific Internet-use Disorders -- Frequency formSymptoms network relationships

The stability metrics of this network were as follows: strength coefficient = 0.72, expected influence coefficient = 0.75, and edge coefficient = 0.75. All these indices indicate that the model had excellent stability [86, 87].

Fig. 1

The problematic pornography use variables structural and interactional network: all-sample. Circles are called “nodes” and they represent each variable (which in psychological network analysis are called “symptoms”) included in the model. The lines are called “edges” and they indicate the association between the symptoms and represent the standardized partial correlations. Blue edges indicate positive correlations, while orange edges indicate negative correlations. Thicker edges indicate larger partial correlations

PPCS = Problematic Pornography Consumption Scale. PPCS-S = Salience, PPCS-MM = mood modification, PPCS-C = conflict, PPCS-T = tolerance, PPCS-R = relapse, PPCS-W = withdrawal

ACSID = Assessment of Criteria for Specific Internet-use Disorders. ACSID-IC = Impaired control, ACSID-IP = Increased priority given to the activity, ACSID-CE = Continuation/escalation of use despite negative consequences, ACSID-FI = Functional impairment in daily life; ACSID-MD = Marked distress

Figure 1 shows the results of the psychological network analysis. The circles are referred to as “nodes” and represent each variable “symptom” included in the model. The lines are referred to as “edges” and represent the standardized partial correlations among symptoms (Spearman correlations, in this case). Blue edges indicate positive correlations, whereas orange edges (if any) indicate negative correlations. Thicker edges indicate larger partial correlations.

Fig. 2

Strength centrality of the modeled symptoms: all-sample. The symptoms are in the vertical axe. The Strength Centrality (SC) values are in horizontal axe

PPCS = Problematic Pornography Consumption Scale. PPCS-S = Salience, PPCS-MM = mood modification, PPCS-C = conflict, PPCS-T = tolerance, PPCS-R = relapse, PPCS-W = withdrawal

ACSID = Assessment of Criteria for Specific Internet-use Disorders. ACSID-IC = Impaired control, ACSID-IP = Increased priority given to the activity, ACSID-CE = Continuation/escalation of use despite negative consequences, ACSID-FI = Functional impairment in daily life; ACSID-MD = Marked distress

Figure 2 shows the centrality strength of ACSID and PPCS symptoms. The vertical axis represents symptoms, and the horizontal axis the corresponding centrality strength values. As seen, the central symptoms (the two most extreme-right line peaks) are ACSID-IP (SC = 1.95) and PPCS-T (SC = 1.44). Furthermore, there is a bridge pathway between the two categories of symptoms (ACSID and PPCS) that goes from ACSID-IC to PPCS-R and vice-versa.