Multicultural Considerations for the Psychometrics of the Brief Pornography Screen

Abstract

Problematic pornography use (PPU) is a burgeoning area of clinical interest. The Brief Pornography Screen (BPS) is a new PPU measure. The BPS has not been psychometrically studied within specific race/cultural groups. We sought to broaden the PPU literature by examining the confirmatory factor analysis (CFA) fit, measurement invariance, and structural invariance of the BPS across eight identity groups. Participants came from a survey administered at three U.S. universities. In total, n = 2,475 participants were analyzed, with the following identity group breakdowns: Asian American/Pacific Islander (AAPI) male = 268, AAPI female = 303, Black male = 101, Black female = 189, Latin male = 208, Latin female = 372, White male = 432, and White female = 602. BPS CFA fit was good across all groups. Measurement invariance analyses suggested metric, but not scalar invariance across all groups. We then split participants by sex assignment, full residual invariance was evident across groups for male participants and partial residual invariance was evident for female participants. Structural invariance analyses indicated anxiety as a weak positive BPS correlate in AAPI, Latin, and White male participants (β’s = 0.25–0.27), but not meaningfully related in the other groups. Pornography viewing frequency was positively correlated with BPS scores across most groups with a wide range (β’s = 0.29–0.52), except for Black male participants (β = 0.15). Our results suggest that the BPS is an appropriate PPU measure across the tested identity groups. While between-group measurement is relatively accurate within sex assignment groups, correlates differed in strength, meaning different variables likely predict PPU for different groups of people.

Introduction

Compulsive Sexual Behavior Disorder (CSBD) is a new psychiatric classification in the International Classification of Diseases-11 (ICD-11^1,2). CSBD involves a persistent pattern of failure to control intense, repetitive sexual impulses or urges resulting in repetitive sexual behavior that causes marked distress or impairment in personal, family, social, educational, occupational, or other important areas of functioning. Problematic pornography use (PPU) is arguably the most common presentation of CSBD, and the most scientifically studied.³ While no comprehensive model of CSBD exists, and classification remains controversial,^4–7 it is commonly agreed that modern technology has proliferated the availability, affordability, and anonymity of pornographic content.^3,8 Importantly, pornography use is not synonymous with PPU,⁹ but is a necessary prerequisite.

Understanding the nature of PPU has become an important empirical endeavor. Central to this aim is having accurate PPU measurement. Many measures exist. These include the PPU scale,¹⁰ problematic pornography consumption scale (PPCS)¹¹ and its various short forms,^12,13 the cyber-pornography use inventory¹⁴ and its short form,¹⁵ among many others. The recently developed Brief Pornography Screen (BPS)¹⁶ represents a more recent scale that can be efficiently utilized in interdisciplinary settings. Importantly, it was originally validated on a clinical sample¹⁶ and has since been validated in many populations of interest, including military veterans¹⁷ and sexual minorities.¹⁸ The BPS was also included in the recent international sex survey involving 45 countries¹⁹ and was recently validated in Bangla.²⁰

The BPS represents an important advance in the PPU psychometric literature. However, areas of continued research exist. Notably, considerations for how PPU might differ across racial/ethnic groups largely remain unknown. While some research exists that shows Black males tend to view more pornography relative to White males and Black and White females,²¹ little current research exists regarding group differences on PPU.

Important to this consideration is the concept of measurement invariance.^22–24 Considerable evidence indicates that latent variable assessments often function non-equivilantely across samples from different cultural backgrounds.^25,26 Cultural differences in terminology processing and perception may be associated with systematic changes in the way participants from different cultural groups respond to items. In turn, between-group differences (or similarities) in measures may not reflect underlying phenomenology due to measurement bias.

Measurement invariance is a statistical technique to assess the degree to which group identity influences measurement function. Measurement invariance can be tested using modern confirmatory factor analysis (CFA) techniques.²⁷ A comprehensive review of measurement invariance testing is beyond the scope of the current article.^22,27,28 However, a brief synopsis is available in Supplementary Data S1. Thus far, gender is perhaps the most common grouping for PPU measurement invariance analyses, with most studies demonstrating that PPU instruments typically achieve metric invariance across genders, but fail scalar.^18,29,30 Although depending on the fit criteria used, other PPU measures might achieve higher levels of invariance for gender.^11–13 The BPS has demonstrated full measurement invariance for veterans and nonveterans,¹⁷ and generally metric invariance has also been shown across gender/sexual identity groups.¹⁸

Considerations for race

Across the research, there has generally been a lack of consideration for how race might influence outcomes. Considerations for race have been identified as a larger problem within the CSBD literature.^31,32 Most authors of PPU scales come from White/European backgrounds. As such, meanings attributable to item wordings (even if they use the same language) might influence psychometric coefficients, and thus outcomes. Accordingly, additional validation information, particularly measurement invariance tests, is needed for PPU scales across race/ethnic groups.

Consistent with this issue, there is a need to understand if correlations between PPU and known outcomes of interest might differ across races. Structural invariance, or the degree to which correlations are equivalent across measures across groups, is a burgeoning psychometric technique. Structural invariance is predicated upon achieving at least metric measurement invariance across groups (or correlation coefficient differences may be attributable to measurement biases rather than latent construct differences).³³ However, little is known about PPU structural invariance in general, with most extant tests being relegated to the same studies in which metric invariance was investigated^17,29; none of which examined how coefficients might change according to race.

Current study

The present study sought to address these gaps. Two aims guided our analyses. First, we sought to test the measurement invariance of the BPS across four ethnic/racial groups: White, Black, Latin, and Asian American/Pacific Islander (AAPI) participants. Because prior literature has shown that scalar invariance generally fails between men and women,¹⁸ we opted to split all outcomes by sex assigned at birth (e.g., White male, Black female). Second, we tested structural invariance between the BPS, pornography viewing frequency, and anxiety. Anxiety was chosen as historical models have suggested that some expressions of CSBD (and thus PPU) may be related to maladaptive attempts to reduce anxiety.³⁴ Moreover, prior research has shown that anxiety problems modestly correlate with PPU.^29,35 Given that PPU is a new mental health classification, understanding how it converges with general psychological problems, such as anxiety, across groups might help contextualize findings.

Moreover, none of the prior studies have accounted for differences by race. Similarly, while pornography viewing frequency is generally not considered synonymous with PPU,⁹ it would be empirically important to know if the correlation between pornography viewing frequency and PPU differed based on identity group. Because our aims were largely exploratory, we did not specify how race would influence outcomes, but we did hypothesize that anxiety and pornography viewing frequency would be modestly (e.g., standardized correlations 0.2–0.4) positively correlated with PPU across groups. Without strong theoretical indications for differences, we largely assumed that measurement invariance would be achieved between racial groups (within sex classification) and that any structural invariance differences would be small.

Methods

Procedure/participants

Data came from two larger multisite surveys on sexual health. Portions of these data have been published in previous studies^36,37; however, none of the previous studies had aims consistent with the present study. Moreover, no researchers have published data from all participating sites. Participants were college students living in the United States from institutions in the Intermountain West, Texas, and the Deep South regions, gathered between 2020 and 2022. Participants were students offered extra credit for their participation. Ethics review board approval was received for all sites, and all participants provided informed consent before participation. Initially, 4,683 participated. However, we removed those who did not indicate sex assigned at birth, failed to provide a targeted racial identity (e.g., White, Black, Latin, AAPI), view pornography in the past 12 months, and/or complete at least 66 percent of study measures.

In total, we had 2,475 participants with data fit for analyses (see Supplementary Data S2 for flowchart of excluded participants). The primary reason participants were excluded from the original sample was due to no reported pornography use within the past 12 months (n = 1,495). The final sample included n = 432 White males, n = 602 White females, n = 101 Black males, n = 189 Black females, n = 208 Latin males, n = 372 Latin females, n = 268 AAPI males, and n = 303 AAPI females. See Table 1 for demographic breakdown.

Table 1.

Demographics

Identity	AAPI male, n = 268	AAPI female, n = 303	Black male, n = 101	Black female, n = 189	Latin male, n = 208	Latin female, n = 372	White male, n = 432	White female, n = 602
Gender identity, n (percent)
Cisgender men	266 (99.3)	0 (0.0)	99 (98.0)	0 (0.0)	204 (98.1)	0 (0.0)	422 (97.7)	0 (0.0)
Cisgender women	0 (0.0)	299 (98.7)	0 (0.0)	186 (98.4)	0 (0.0)	360 (96.8)	0 (0.0)	574 (95.3)
GNC MAB	2 (0.7)	0 (0.0)	2 (2.0)	0 (0.0)	4 (1.9)	0 (0.0)	10 (2.3)	0 (0.0)
GNC FAB	0 (0.0)	4 (1.3)	0 (0.0)	3 (1.6)	0 (0.0)	12 (3.2)	0 (0.0)	28 (4.7)
Sexual attraction, n (percent)
Only attracted to females	210 (78.4)	6 (2.0)	83 (82.2)	9 (4.8)	157 (75.5)	5 (1.3)	341 (78.9)	15 (2.5)
Mostly attracted to females	24 (9.0)	10 (3.3)	6 (5.9)	12 (6.3)	19 (9.1)	15 (4.0)	37 (8.6)	22 (3.7)
Equally attracted to females and males	6 (2.2)	29 (9.6)	2 (2.0)	24 (12.7)	7 (3.4)	55 (14.8)	7 (1.6)	63 (10.5)
Mostly attracted to males	5 (1.9)	109 (36)	4 (4.0)	52 (27.5)	11 (5.3)	98 (26.3)	7 (1.6)	147 (24.4)
Only attracted to males	15 (5.6)	133 (43.9)	6 (5.9)	82 (43.4)	12 (5.8)	178 (47.8)	23 (5.3)	305 (50.7)
Asexual	4 (1.5)	4 (1.0)	0 (0.0)	0 (0.0)	2 (1.0)	0 (0.0)	2 (0.5)	6 (1.0)
Pansexual	1 (0.4)	7 (2.3)	0 (0.0)	7 (3.7)	0 (0.0)	14 (3.8)	6 (1.4)	32 (5.3)
Not sure	3 (1.1)	6 (2.0)	0 (0.0)	3 (1.6)	0 (0.0)	7 (1.9)	9 (2.1)	12 (2.0)
Religion, n (percent)
Christian	158 (59.6)	155 (51.5)	62 (61.4)	103 (54.5)	116 (55.8)	214 (57.8)	205 (47.7)	291 (48.7)
Jewish	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)	5 (1.2)	7 (1.2)
Muslim	3 (1.1)	3 (1.0)	2 (2.0)	2 (1.1)	0 (0.0)	0 (0.0)	1 (0.2)	1 (0.2)
Buddhist	17 (6.4)	24 (8.0)	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)	1 (0.2)	2 (0.3)
Hindu	3 (1.1)	5 (1.7)	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)
Nonreligious	36 (13.6)	47 (15.6)	10 (9.9)	16 (8.5)	37 (17.8)	60 (16.2)	82 (19.1)	100 (16.7)
Agnostic	26 (9.8)	32 (10.6)	4 (4.0)	6 (3.2)	22 (10.6)	25 (6.8)	65 (15.1)	56 (9.4)
Spiritual but not religious	17 (6.4)	27 (9.0)	19 (18.8)	59 (31.2)	30 (14.4)	66 (17.8)	56 (13.0)	123 (20.6)
Something else	5 (1.9)	8 (2.7)	4 (4.0)	3 (1.6)	3 (1.4)	5 (1.4)	15 (3.5)	18 (3.0)
Age, M (SD)	19.46 (2.08)	19.42 (2.61)	21.01 (5.36)	20.18 (4.39)	20.30 (3.84)	19.80 (3.14)	20.56 (4.34)	20.02 (4.07)
Anxiety (standardized), M (SD)	−0.21 (0.77)	0.22 (0.88)	−0.27 (0.83)	−0.01 (0.88)	−0.22 (0.80)	0.16 (0.89)	−0.35 (0.69)	0.20 (0.91)
Pornography frequency, M (SD)	5.81 (2.82)	3.03 (2.09)	5.23 (2.90)	3.31 (2.20)	5.34 (2.73)	2.72 (1.82)	5.30 (2.69)	2.67 (1.79)
BPS, M (SD)	3.62 (3.02)	2.86 (2.32)	3.28 (2.93)	1.98 (2.56)	3.62 (3.09)	1.74 (2.36)	3.19 (2.94)	1.42 (2.21)
BPS clinical cut (above), n (percent)	141 (52.6)	61 (20.1)	56 (55.4)	46 (24.3)	102 (49)	73 (19.6)	182 (42.1)	97 (16.1)

Note: The anxiety indicator is a composite of two standardized scales. The anxiety measures standardized as different scalings had been used across sites.

AAPI, Asian American/Pacific Islander; BPS, Brief Pornography Screen; FAB, female assigned at birth; GNC, gender nonconforming; MAB, male assigned at birth; SD, standard deviation.

Measures

Pornography viewing frequency

Pornography viewing frequency was assessed with a single item: “Within the past 12 months, how often do you watch pornography?” Participants selected from 0 (“Never”), 1 (“Several times a year or less”), 2 (“Once a month or less”), 3 (“Several days a month”), 4 (“One day a week”), 5 (“Two days a week”), 6 (“Three days a week”), 7 (“Four days a week”), 8 (“Five days a week”), 9 (“Six days a week”), or 10 [“Seven days a week (every day)”].

Problematic pornography use

The BPS¹⁶ was used to assess PPU. Participants were provided the following prompt, “In the last six months, have any of these situations happened to you in regard to your use of pornography?” and rated the five items with 0 (“Never”), 1 (“Occasionally”), or 2 (“Very often”). Example item: “You find yourself using pornography more than you want to.” A score of ≥4 is considered the clinical cutoff for probable PPU. Internal consistencies are available in Table 2.

Table 2.

Fit Indices from Individual Group Confirmatory Factor Analyses

Group	n	χ²	df	CFI	TLI	RMSEA	90 percent CI RMSEA	SRMR	Cronbach's α
BPS
AAPI male	268	14.37^*	5	0.979	0.985	0.084	0.035 to 0.136	0.024	0.873
AAPI female	303	14.84^*	5	0.972	0.943	0.081	0.035 to 0.129	0.035	0.830
Black male	101	12.62^*	5	0.950	0.901	0.123	0.038 to 0.209	0.036	0.879
Black female	189	11.89^*	5	0.971	0.943	0.085	0.019 to 0.149	0.030	0.861
Latin male	208	9.49	5	0.987	0.974	0.066	0.000 to 0.129	0.024	0.882
Latin female	372	1.06	5	1.00	1.00	0.000	0.000 to 0.000	0.008	0.851
White male	432	1.00	5	0.993	0.985	0.048	0.000 to 0.092	0.018	0.872
White Female	602	21.05^***	5	0.973	0.947	0.073	0.043 to 0.107	0.027	0.858

^*p < 0.05; ^***p < 0.001.

χ², scaled chi-square; CFI, comparative fit index; CI, confidence interval; df, degrees of freedom; RMSEA, root mean square error of approximation; SRMR, standardized root mean square residual; TLI, Tucker–Lewis index.

Anxiety

Anxiety was assessed using the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) cross-cutting measure. Participants were provided the following prompt, “During the past two weeks, how much (or how often) have you been bothered by the following problems?” and rated items from 0 (“Not at all”) to 4 (“Nearly every day”). Anxiety was assessed with three items (e.g., “Feeling nervous, anxious, frightened, worried, or on edge?”). Because some sites (involving n = 1,189 participants) scaled the anxiety items using the generalized anxiety disorder-7 (GAD-7)³⁸ scaling (i.e., 0 [“Not at all”] to 3 [“Nearly every day”]), all anxiety items were standardized (z-scored) within a scaling group. The standardized indicators were then used in general analyses. This was done to establish a consistent anchor allowing for a single set of anxiety indicators.

Analytic plan

For aim one, we constructed eight CFAs to ensure adequate fit for each group. Adequate fit was determined by a nonsignificant chi-square goodness-of-fit test, but supplemented by established alternative fit indices (AFIs; as the chi-square test is often significant by virtue of power³³). These included the comparative fit index (CFI; >0.90 acceptable fit), Tucker–Lewis index (TLI; >0.90 acceptable fit), root mean square error of approximation (RMSEA; with 90 percent confidence interval [CI] low values of <0.06 and high values <0.10 indicate acceptable fit), and the standardized root mean square residual (SRMR; <0.08 acceptable fit).³³ We then tested measurement invariance across groups that evidenced acceptable fit. We tested measurement invariance in four steps corresponding to configural, metric, scalar, and residual invariance.

First, we constructed a freely estimated model and evaluated fit (configural). Then constrained factor loadings to be equal across all groups (metric) and tested that model against the configural model for fit degradation. Then we constrained the factor loadings and intercepts to be equal across groups (scalar) and tested it against the metric model for fit degradation.

Finally, we constrained factor loadings, intercepts, and error variances to be equal (residual) and tested it against the scalar model for fit degradation. As we anticipated measurement invariance would not be achieved beyond metric invariance due to sex differences,¹⁸ we allowed for a contingency in which we proceeded with within-sex measurement invariance analyses should scalar invariance failure be evident (as between race-within-sex measurement invariance might still be achieved). For all invariance tests, significant fit degradation was determined by a significant chi-square difference test between models (at p < 0.001). Given power sensitivity of the chi-square test,³³ we supplemented fit degradation determination with considerations of changes in CFI and RMSEA where degradation was considered by a CFI decrease ≥0.01³⁹ or an RMSEA increase ≥0.01.⁴⁰

For aim two, a full measurement model was estimated that involved the BPS items (PPU latent variable), standardized anxiety items (anxiety latent variable), and the pornography viewing frequency indicator. We then specified a multigroup regression with anxiety and pornography use simultaneously predicting PPU across all groups. We then subtracted the unstandardized regression coefficients from each other across groups (White males–White females, White males–Black males, etc.) and calculated the 95 percent bias-corrected bootstrapped CIs of the differences.⁴¹ A 95 percent CI that does not include “0” signifies statistical significance at the p < 0.05 level. Descriptive statistics were calculated in SPSS v.28. Measurement invariance analyses were conducted in Mplus v.8. All CFAs were estimated using a maximum likelihood estimation with robust standard errors. Full information maximum likelihood estimation was used to estimate any remaining missing inputs.

Results

Means and standard deviations across groups are available in Table 1. To help illustrate, figures are available in Supplementary Data S3. Observably, pornography viewing frequency and BPS scores varied in accordance with sex, but not race, with males having higher mean scores than females. Within-group analyses of variance confirmed these observations. No significant variation was evident across racial groups for males on the BPS F(3, 1,005) = 1.61, p = 0.186 or pornography viewing frequency F(3, 1,005) = 2.23, p = 0.084, and only minimal variation was evident for females on the BPS F(3, 1,462) = 4.27, p = 0.005, η² = 0.01 and pornography viewing frequency F(3, 1,461) = 6.81, p < 0.001, η² = 0.01.

In both cases, White females reported the lowest BPS scores and least viewing, with Black females reporting the most. The differences were small though, d's = 0.23 and 0.32, respectively. Participants above the BPS clinical referral score also varied by group, with Black: male participants having the highest base rate (55.4 percent) compared with White female participants with the lowest (16.1 percent).

Primary analyses

Results from the initial within-group CFAs are available in Table 2 and the corresponding standardized factor loadings are in Table 3. Most items loaded above 0.70 for all groups, except for item 4 [“You find yourself using pornography to cope with strong emotions (e.g., sadness, anger, loneliness”)], which was below 0.70 for all groups (particularly female participants). Good fit was generally evident for all groups, notwithstanding that the RMSEA had a wide CI across groups (possibly attributable to the BPS scaling and groups with relatively smaller sample sizes). The full configural invariance model in which all eight groups were estimated simultaneously evidenced good fit (see Table 4 for all measurement invariance fit indices). The metric invariance model also evidenced good fit. Moreover, the chi-square difference test failed to indicate a significant degradation χ²(28) = 47.76, p = 0.011, suggesting metric invariance.

Table 3.

Brief Pornography Screen Item Factor Loadings from Individual Groups

Factor loadings	AAPI male	AAPI female	Black male	Black female	Latin male	Latin female	White male	White female
You find yourself using pornography more than you want to.	0.755	0.732	0.741	0.800	0.791	0.764	0.821	0.774
You have attempted to “cut back” or stop using pornography, but were unsuccessful.	0.866	0.844	0.795	0.784	0.825	0.861	0.831	0.814
You find it difficult to resist strong urges to use pornography.	0.841	0.739	0.825	0.799	0.826	0.785	0.828	0.773
You find yourself using pornography to cope with strong emotions (e.g., sadness, anger, loneliness).	0.605	0.553	0.672	0.624	0.654	0.521	0.573	0.615
You continue to use pornography even though you feel guilty about it.	0.737	0.656	0.811	0.725	0.787	0.753	0.755	0.752

Note. All factor loadings were statistically significant at p < 0.001.

Table 4.

Results and Fit Indices from the Multigroup Invariance Analyses

Grouping	Model	n	χ²	df	CFI	TLI	RMSEA	90 percent CI RMSEA	SRMR	Δχ²	Δdf	ΔCFI	ΔRMSEA
All	Configural	2,475	96.88^***	40	0.983	0.966	0.068	0.051 to 0.085	0.025	—	—	—	—
	Metric	2,475	147.65^***	68	0.976	0.972	0.062	0.048 to 0.075	0.045	47.76^*	28	0.007	−0.004
	Scalar	2,475	456.60^***	103	0.893	0.917	0.105	0.096 to 0.115	0.165	373.34^***	35	0.083	0.043
Male only	Configural	1,466	46.05^***	20	0.984	0.968	0.072	0.045 to 0.099	0.024	—	—	—	—
	Metric	1,466	62.43^**	32	0.981	0.977	0.061	0.038 to 0.084	0.034	11.60	12	0.003	−0.011
	Scalar	1,466	74.76^**	47	0.983	0.985	0.048	0.026 to 0.068	0.045	9.81	15	−0.002	−0.013
	Residual	1,466	80.20	62	0.989	0.993	0.034	0.000 to 0.054	0.046	7.98	15	−0.006	−0.014
Female only	Configural	1,009	50.43^***	20	0.982	0.964	0.064	0.042 to 0.087	0.026	—	—	—	—
	Metric	1,009	66.30^***	32	0.980	0.975	0.054	0.035 to 0.072	0.041	15.01	12	0.002	−0.010
	Scalar	1,009	96.09^***	47	0.971	0.975	0.053	0.038 to 0.069	0.057	29.14^*	15	0.009	−0.001
	Residual	1,009	130.34^***	62	0.959	0.974	0.055	0.042 to 0.068	0.068	33.49^**	15	0.012	0.002

^*p < 0.05; ^**p < 0.01; ^***p < 0.001.

χ², scaled chi-square; ΔCFI, freely estimated model-constrained model; ΔRMSEA, constrained model-freely estimated model.

We then constrained all factor loadings and intercepts to be equal. However, fit degradation was obvious (Table 4), and confirmed by the chi-square statistic, χ²(35) = 373.34, p < 0.001. As anticipated, we used our contingency procedure by performing within-sex group comparisons. For the male participants, full residual invariance was evident. For the female participants, scalar invariance was evident, but residual invariance failed. A post hoc bootstrap test of the difference between residual variances for all items across groups indicated that AAPI females had significantly greater error variances relative to White females for items 4 and 5. The size of these differences appeared small (b_res = −0.08, 95 percent CI: [−0.16 to −0.03] and b_res = −0.07, 95 percent CI: [−0.17 to −0.01]).

We proceeded with Aim two by constructing a full measurement model with an anxiety latent variable, PPU latent variable, and the pornography use frequency item. We again tested measurement invariance to ensure metric invariance held (as anxiety and pornography viewing frequency were now specified). The full configural model evidenced good fit: χ²(200) = 418.46, p < 0.001, CFI = 0.969, TLI = 0.955, RMSEA = 0.059 (90 percent CI: 0.051 to 0.067), SRMR = 0.043. The metric invariance model also evidenced good fit: χ²(228) = 467.88, p < 0.001, CFI = 0.966, TLI = 0.957, RMSEA = 0.058 (90 percent CI: 0.051 to 0.066), SRMR = 0.049. No meaningful degradation in fit was evident: χ²(28) = 49.67, p = 0.007, ΔCFI = 0.003. As such, we proceeded to test our regression model.

Bivariate relationships between pornography viewing frequency and anxiety with PPU scores are available in Table 5, with standardized adjusted regression coefficients among the latent variables in Table 6. Results suggested pornography viewing frequency to be a positive predictor of PPU scores, but with ranging strength. Black males evidenced a weak nonsignificant relationship (β = 0.15, p = 0.17), whereas AAPI females demonstrated a robust relationship (β = 0.52, p < 0.001). Similarly, anxiety was generally a positive predictor of PPU, but with much less strength. Latin females failed to demonstrate any relationship (β = 0.03, p = 0.17), whereas AAPI males demonstrated a modest (but strongest of the groups) relationship (β = 0.29, p < 0.001).

Table 5.

Problematic Pornography Use Correlations by Identity Group

Bivariate correlations
	AAPI male	AAPI female	Black male	Black female	Latin male	Latin female	White male	White female
	PPU	PPU	PPU	PPU	PPU	PPU	PPU	PPU
Anxiety	0.35^***	0.08	0.17	0.15^*	0.30^***	0.10	0.27^***	0.18^***
Pornography viewing frequency	0.34^***	0.47^***	0.19	0.43^***	0.46^**	0.40^***	0.31^***	0.30^***

^*p < 0.05; ^**p < 0.01;^***p < 0.001.

PPU, problematic pornography use.

Table 6.

Regression Coefficients of Anxiety and Pornography Viewing Frequency Predicting Problematic Pornography Use Across Identity Groups

Adjusted unstandardized, SEs, and standardized regression coefficients
	AAPI male			AAPI female			Black male			Black female			Latin male			Latin female			White male			White female
	PPU			PPU			PPU			PPU			PPU			PPU			PPU			PPU
	b	SE	β	b	SE	β	b	SE	β	b	SE	β	b	SE	β	b	SE	β	b	SE	β	b	SE	β
Anxiety	0.23^a,b	0.06	0.29^*	0.03	0.03	0.05	0.06	0.07	0.11	0.06	0.05	0.12	0.19^b	0.05	0.27^*	0.02	0.03	0.03	0.17^a,b	0.04	0.25^*	0.09	0.02	0.17^*
Pornography viewing frequency	0.06	0.01	0.29^*	0.11^c,d	0.02	0.52^*	0.03	0.02	0.15	0.10^d	0.02	0.46^*	0.10^d	0.02	0.45^*	0.11^c,d	0.02	0.42^*	0.06	0.01	0.30^*	0.07	0.01	0.29^*

Significantly stronger effect relative to AAPI females.

Significantly stronger effect relative to Latin females.

Significantly stronger effect relative to AAPI males.

Significantly stronger effect relative to Black males.

p < 0.001.

SE, standard error.

Results from the structural invariance analyses broadly indicated the following notes (Table 6). White males had a significantly stronger relationship between anxiety and PPU relative to Latin and AAPI females; Black males had a significantly weaker relationship between pornography viewing frequency and PPU relative to Black females, Latin males/females, and AAPI females; Latin males had a significantly stronger relationship between anxiety and PPU relative to Latin females; Latin females had a significantly stronger relationship between pornography viewing frequency and PPU relative to AAPI males and a significantly weaker relationship between anxiety and PPU relative to AAPI males; and AAPI males had a significantly weaker relationship between pornography viewing frequency and PPU relative to AAPI females and a significantly stronger relationship between anxiety and PPU relative to AAPI females.

Discussion

Our results reveal important considerations for the psychometrics of the BPS, and for how race (and sex) moderates the relationships between pornography viewing frequency and anxiety as PPU correlates. First, the BPS functions relatively equally across racial groups once sex is accounted. Consistent with past research, the BPS failed scalar invariance.¹⁸ However, once groups were split by sex, full residual invariance was evident across races for males and partial residual invariance for female participants. The lack of full residual invariance for female participants is a relatively minor issue with minimal practical implications since full scalar invariance was evident.³³ The bootstrap analyses indicated that for items 4 and 5, AAPI female participants had significantly greater error variances relative to White female participants. Overall, correlational comparisons are appropriate using the BPS across all of the examined groups, and mean comparisons are appropriate for all groups if examined within sex.

Perhaps more notably, the factor loadings of item 4 appeared weaker than the rest (although statistically significant) for all groups. The factor loading is the relationship between the hypothesized underlying PPU latent variable (shared variance across BPS items) and the actual item. Meaning item 4 is tapping into a slightly different construct (pornography use for coping), opposed to the remaining items, which are more closely related to PPU as a compulsive disorder (perhaps a construct more consistent with CSBD). Interestingly, there was a tendency for some of the model fit indices to improve as constraints were added. This is likely the result of unequal subsamplings. Given that all groups had relatively adequate fit, we do not consider this a concern presently. However, we do recommend further BPS psychometric research with larger subgroups.

Partially consistent with our hypotheses, pornography viewing frequency and anxiety were generally positive correlates of PPU. However, notable divergences in correlate strength were evident. For instance, anxiety was not a significant correlate of PPU for nonWhite female participants or for Black participants generally. Moreover, even though anxiety was significantly related to PPU for White women, the correlation was extremely weak (β = 0.17) and only significant by virtue of our large sample. Indeed, the only groups for which anxiety was a notable PPU correlate were White, Latin, and AAPI men (with no differences in path strength between those groups). Conceptually, these findings suggest that anxiety concerns are probably not an adequate explanation for PPU or likely CSBD. Even when considering the significant paths, they were all modest (below 0.4). In other words, anxiety is probably only a correlate of PPU and not a root source for most cases. That said, our sample was broadly nonpathological. The connection between anxiety and PPU might be stronger in populations with severe psychopathology.

The connection between pornography viewing frequency and PPU was more robust, but also varied across groups. For instance, the connection between pornography viewing frequency and PPU was nonsignificant for Black male participants, notwithstanding the finding that Black males reported viewing approximately the same amount of pornography as the other male participants and had similar BPS scores. Meaning, the predictors of PPU for Black males differ from those of other groups. This is the first research to show this discrepancy. Concurrently, pornography use was a robust predictor of PPU in AAPI, Black, and Latin female groups and Latin males, but was only a weak predictor for White and AAPI males, as well as White females.

Limitations

Our results provide nuanced information regarding BPS psychometrics. However, they are contextualized by several important limitations. First, we only examined homogenous groupings of participants. For instance, cultural differences within specific identity groups (e.g., Latin participants with Mexican vs. Argentinian heritage) may likely be as diverse as those between groups. Future BPS psychometric research is needed with more specific cultural groupings. Second, 639 participants were excluded for identifying more than one race. As this resulted in dozens of multiracial combinations, we opted to not include multiracial as a category for analyses. However, in line with our previous recommendation, researchers should examine the BPS psychometrics in specific racial groups, including those from diverse multiracial identity backgrounds.

Third, we had to classify participants by sex. We did this to aid in our power estimations. This was beneficial in that we retained non-cisgender participants (who are often removed from such analyses) but did not have sufficient power to formally breakdown analyses by gender identity groups. Future researchers are encouraged to examine group differences accounting for non-cisgender participants. Fourth, the cross-sectional nature of our design precludes any firm causal interpretations. Longitudinal designs are recommended for future study, particularly those examining the relationship between anxiety and PPU. Finally, our findings are limited by the convenience sampling procedure, which is inherently associated with selection and response biases.

Conclusion

Pornography viewing frequency and PPU rates differ significantly according to sex, but not race, with females reporting significantly less on measures of both. The BPS demonstrated sound psychometric properties across White, Black, Latin, and AAPI racial/ethnic groups; however, between-group differences were still evident by sex. Because metric invariance held by sex across groups, correlational comparisons using the BPS are appropriate across all groups. That said, because scalar invariance failed, researchers should not conduct mean comparisons between sex groupings without adjusting for measurement biases associated with participant sex differences. All mean comparisons should occur within-sex groupings. Structural invariance analyses indicated anxiety as a relatively weak PPU correlate for most groups, and only meaningful for White, AAPI, and Latin males.

Pornography viewing frequency was a robust PPU correlate, particularly for AAPI, Black, and Latin females, and Latin males; and to a lesser extent AAPI males and White males and females. Neither anxiety nor pornography use frequency was a relevant PPU correlate for Black males. Further research into the reasonings for these findings is encouraged.

Footnotes

Authors' Contributions

N.C.B. is responsible for the study design, analysis, data gathering, article writing, and project administration. B.M.W. is responsible for article writing, article editing, and table construction. S.W.K. is responsible for data gathering and article revision.

Author Disclosure Statement

All authors declare no conflicts of interest.

Funding Information

Support for B.M.W. and S.W.K. was provided by the Kindbridge Research Institute. The funding agencies did not provide input or comment on the content of the article, and the content of the article reflects the contributions and thoughts of the authors and does not necessarily reflect the views of the funding agencies.

Supplementary Material

References

Reed

, First

, Billieux

, et al. Emerging experience with selected new categories in the ICD-11: Complex PTSD, prolonged grief disorder, gaming disorder, and Compulsive Sexual Behavior Disorder. World Psychiatry, 2022; 21(2):189–213.

World Health Organization. International Statistical Classification of Diseases and Related Health Problems. 11th ed. World Health Organization: Geneva; 2022.

Grubbs

, Hoagland

, Lee

, et al. Sexual addiction 25 years on: A systematic review of empirical literature on compulsive sexual behavior and an agenda for future research. Clin Psychol Rev, 2020; 82:101925; doi: 10.1016/j.cpr.2020.101925

Castro-Calvo

, Flayelle

, Perales

, et al. Compulsive Sexual Behavior Disorder should not be classified by solely relying on component/symptomatic features: Commentary to the debate: “Behavioral addictions in the ICD-11.”. J Behav Addict, 2022; 11:210–215; doi: 10.1556/2006.2022.00029

Bőthe

, Koós

, Demetrovics

. Contradicting classification, nomenclature, and diagnostic criteria of Compulsive Sexual Behavior Disorder (CSBD) and future directions: Commentary to the debate: “Behavioral addictions in the ICD-11.”. J Behav Addict, 2022; 11:204–209; doi: 10.1556/2006.2022.00030

Borgogna

, Aita

. Another failure of the latent disease model? The case of Compulsive Sexual Behavior Disorder • Commentary to the debate: “Behavioral addictions in the ICD-11.”. J Behav Addict, 2022; 11(3):615–619; doi: 10.1556/2006.2022.00069

Sassover

, Weinstein

. Should compulsive sexual behavior (CSB) be considered as a behavioral addiction? A debate paper presenting the opposing view. J Behav Addict, 2022; 11:166–179; doi: 10.1556/2006.2020.00055

Cooper

, Delmonico

, Burg

. Cybersex users, abusers, and compulsives: New findings and implications. Sex Addict Compulsivity, 2000; 7:5–29; doi: 10.1080/10720160008400205

Bőthe

, Tóth-Király

, Potenza

, et al. High-frequency pornography use may not always be problematic. J Sex Med, 2020; 17:793–811; doi: 10.1016/j.jsxm.2020.01.007

10.

Kor

, Zilcha-Mano

, Fogel

, et al. Psychometric development of the Problematic Pornography Use Scale. Addict Behav, 2014; 39:861–868; doi: 10.1016/j.addbeh.2014.01.027

11.

Bőthe

, Tóth-Király

, Zsila Á, et al. The development of the Problematic Pornography Consumption Scale (PPCS). J Sex Res, 2018; 55:395–406; doi: 10.1080/00224499.2017.1291798

12.

Bőthe

, Tóth-Király

, Demetrovics

, et al. The short version of the Problematic Pornography Consumption Scale (PPCS-6): A reliable and valid measure in general and treatment-seeking populations. J Sex Res, 2021; 58:342–352; doi: 10.1080/00224499.2020.1716205

13.

Chen

, Luo

, Bőthe

, et al. Properties of the Problematic Pornography Consumption Scale (PPCS-18) in community and subclinical samples in China and Hungary. Addict Behav, 2021; 112:106591; doi: 10.1016/j.addbeh.2020.106591

14.

Grubbs

, Sessoms

, Wheeler

, et al. The Cyber-Pornography Use Inventory: The development of a new assessment instrument. Sex Addict Compulsivity, 2010; 17:106–126; doi: 10.1080/10720161003776166

15.

Grubbs

, Gola

. Is pornography use related to erectile functioning? Results from cross-sectional and latent growth curve analyses. J Sex Med, 2019; 16:111–125; doi: 10.1016/j.jsxm.2018.11.004

16.

Kraus

, Gola

, Grubbs

, et al. Validation of a Brief Pornography Screen across multiple samples. J Behav Addict, 2020; 9(2):259–271; doi: 10.1556/2006.2020.00038

17.

Borgogna

, Kraus

, Grubbs

. Military veterans' psychological distress associated with problematic pornography viewing. Int J Ment Health Addict, 2022; 20:2194–2207; doi: 10.1007/S11469-021-00508-Z/FIGURES/1

18.

Borgogna

, Griffin

, Grubbs

, et al. Understanding differences in problematic pornography use: considerations for gender and sexual orientation. J Sex Med, 2022; 19:1290–1302; doi: 10.1016/J.JSXM.2022.05.144

19.

Bőthe

, Nagy

, Koós

, et al. Problematic pornography use across countries, genders, and sexual orientations: Insights from the International Sex Survey and Comparison of Different Assessment Tools. Addiction, 2024 (in press).

20.

Islam

, Tasnim

, Sujan

MSH

, et al. Validation and evaluation of the psychometric properties of the Bangla version of the Brief Pornography Screen in men and women. Int J Ment Health Addict, 2022; 22:1–15; doi: 10.1007/S11469-022-00903-0/TABLES/4

21.

Perry

, Schleifer

. Race and trends in pornography viewership, 1973–2016: Examining the moderating roles of gender and religion. J Sex Res, 2019; 56:62–73; doi: 10.1080/00224499.2017.1404959

22.

Vandenberg

, Lance

. A review and synthesis of the measurement invariance literature: Suggestions, practices, and recommendations for organizational research. Organ Res Methods, 2000; 3:4–70; doi: 10.1177/109442810031002

23.

Milfont

, Fischer

. Testing measurement invariance across groups: Applications in cross-cultural research. Int J Psychol Res, 2010; 3:111–130; doi: 10.21500/20112084.857

24.

, Tracey

TJG

. Use of multi-group confirmatory factor analysis in examining measurement invariance in counseling psychology research. Eur J Couns Psychol, 2017; 6:75–82; doi: 10.5964/ejcop.v6i1.120

25.

Borgogna

, Brenner

, McDermott

. Sexuality and gender invariance of the PHQ-9 and GAD-7: Implications for 16 identity groups. J Affect Disord, 2021; 278:122–130; doi: 10.1016/j.jad.2020.09.069

26.

Kubarych

, Aggen

, Kendler

, et al. Measurement non-invariance of DSM-IV narcissistic personality disorder criteria across age and sex in a population-based sample of Norwegian twins. Int J Methods Psychiatr Res, 2010; 19:156–166; doi: 10.1002/MPR.318

27.

Putnick

, Bornstein

. Measurement invariance conventions and reporting: The state of the art and future directions for psychological research. Dev Rev, 2016; 41:71–90; doi: 10.1016/j.dr.2016.06.004

28.

van de Schoot

, Lugtig

, Hox

. A checklist for testing measurement invariance. Eur J Dev Psychol, 2012; 9:486–492; doi: 10.1080/17405629.2012.686740

29.

Borgogna

, Duncan

, McDermott

. Is scrupulosity behind the relationship between problematic pornography viewing and depression, anxiety, and stress?. Sex Addict Compulsivity, 2018; 25:293–318; doi: 10.1080/10720162.2019.1567410

30.

Borgogna

, McDermott

. The role of gender, experiential avoidance, and scrupulosity in problematic pornography viewing: A moderated-mediation model. Sex Addict Compulsivity, 2018; 25:319–344; doi: 10.1080/10720162.2018.1503123

31.

Borgogna

, Garos

, Meyer

, et al. A review of behavioral interventions for Compulsive Sexual Behavior Disorder. Curr Addict Rep, 2022; 9:1–10; doi: 10.1007/S40429-022-00422-X

32.

Griffin

, Way

, Kraus

. Controversies and clinical recommendations for the treatment of Compulsive Sexual Behavior Disorder. Curr Addict Rep, 2021; 8:546–555; doi: 10.1007/S40429-021-00393-5/TABLES/2

33.

Kline

. Principles and Practice of Structural Equation Modeling. 4th ed. Guilford Press: USA;. 2016.

34.

Coleman

. Is your patient suffering from compulsive sexual behavior?. Psychiatr Ann, 1992; 22:320–325; doi: 10.3928/0048-5713-19920601-09

35.

Levin

, Lillis

, Hayes

. When is online pornography viewing problematic among college males? Examining the moderating role of experiential avoidance. Sex Addict Compulsivity, 2012; 19:168–180; doi: 10.1080/10720162.2012.657150

36.

Habashy

, Benning

, Renn

, et al. Psychometric properties of the eating disorder examination questionnaire: Factor analysis and measurement invariance by race/ethnicity and gender. Eat Behav, 2023; 48:101696; doi: 10.1016/j.eatbeh.2022.101696

37.

Borgogna

, Meyer

, Trussell

, et al. Further understanding the correlations between sexting and mental health: Considerations for sex and sexual identity. Cyberpsychol Behav Soc Netw, 2023; 26:323–331; doi: 10.1089/CYBER.2022.0312

38.

Spitzer

, Kroenke

, Williams

JBW

, et al. A brief measure for assessing Generalized Anxiety Disorder. Arch Intern Med, 2006; 166:1092–1097; doi: 10.1001/archinte.166.10.1092

39.

Cheung

, Rensvold

. Evaluating goodness-of-fit indexes for testing measurement invariance. Struct Equ Model, 2002; 9:233–255; doi: 10.1207/S15328007SEM0902_5

40.

Rutkowski

, Svetina

. Assessing the hypothesis of measurement invariance in the context of large-scale international surveys. Educ Psychol Meas, 2014; 74:31–57; doi: 10.1177/0013164413498257

41.

Cheung

, Lau

. A direct comparison approach for testing measurement invariance. Organ Res Methods, 2012; 15:167–198; doi: 10.1177/1094428111421987

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.01 MB

0.03 MB

0.06 MB