Introduction

A burgeoning literature has sought to clarify the association between pornography use and various indicators of relationship quality, both for those in dating and marriage relationships (see systematic narrative reviews in Campbell & Kohut, 2017; Manning, 2006; Newstrom & Harris, 2016; Rasmussen, 2016; and the meta-analyses in Wright & Tokunaga, 2018; Wright, Tokunaga, Kraus, & Klann, 2017). Despite the fairly large number of studies that have examined this issue, there has been a curious lack of consensus on what the data reveal. Some scholars, for example, have concluded that the trends are quite consistent. For instance, in his “historical and empirical review” of research on pornography and romantic and family relationships, Rasmussen (2016, p. 185) argued that the preponderance of evidence suggest pornography tends to be “problematic” for relationships, concluding, “The evidence for pornography’s influence on the stability of romantic and committed relationships is strong. The effects described are grounded in established theory and operate through well-defined processes, and the data produce remarkable agreement.” Similarly, in their meta-analysis of 50 different studies examining pornography’s connection to relational and intrapersonal satisfaction, Wright et al. (2017, p. 336) concluded regarding men specifically, “…the convergence of results across cross-sectional survey, longitudinal survey, and experimental results points to an overall negative effect of pornography on men’s sexual and relational satisfaction.”

And yet, other surveys of the relevant literature have argued that the association between pornography use and relationship quality is far from clear, owing to a number of important limitations in the previous research. In their review of 26 empirically based studies on the topic spanning 15 years, Newstrom and Harris (2016, p. 412) characterized the majority of research as “exploratory and descriptive” and suggested that the findings on pornography’s “effects” on couples’ relationships have been entirely mixed: “Findings indicate that there are both positive and negative effects of pornography use within committed relationships.” They attributed much of the ambiguity in the previous research to the widespread use of convenience sampling, inconsistent units of analysis (focusing on dyads or individuals), as well as under-theorizing. And more recently, in Campbell and Kohut’s (2017) review of research on pornography and romantic relationships, they drew similar conclusions to Newstrom and Harris. Specifically, Campbell and Kohut (2017, p. 6) argued that certain data limitations and assumptions “make it extremely difficult to draw firm conclusions concerning the associations between pornography use and relationship processes and/or outcomes.” Among the limitations they described were using individual rather than dyadic data; differences in measurement of “pornography” making cross-study comparisons difficult (though Wright et al. 2017 recently demonstrated that different pornography measurements do not moderate the association between pornography use and relationship quality); assuming causality in cross-sectional designs; failure to acknowledge gendered assumptions; and “harm-focused” approaches that may bias findings.Footnote 1

The current study aimed to address several (though certainly not all) of these limitations identified by previous researchers in order to establish a dominant trend in the association between pornography use and relationship outcomes. Drawing on data from 30 nationally representative surveys together including 31 measures of relationship quality, this study conducted 29 independent tests of the association between pornography use and various relationship outcomes for unmarried Americans and 57 tests for married Americans. Associations were tested at the bivariate level and with sociodemographic controls, thus accounting for potential spuriousness between pornography use and a variety of relationship outcomes (see Hald, Kuyper, Adam, & de Wit, 2013). Interactions were also tested for gender. While these national surveys cannot overcome the limitation of focusing on individuals (rather than dyads) or being cross-sectional, they did allow for numerous individual tests of the association between pornography use and relationship outcomes using (1) nationally representative samples rather than convenience samples; (2) measures of pornography use that include both continuity and discontinuity across data sets, thus allowing for comparison; (3) numerous non-biased measures of relationship quality that allow pornography use to be associated with positive or negative outcomes; (4) explicit tests for differences in potential associations by gender, thus allowing for the examination of different experiences of pornography use and relationship quality for men and women; and (5) separate analyses for married and unmarried participants thus allowing for a comparison of associations.

As virtually all reviews of previous research acknowledge, the majority of studies focusing on the link between pornography use and relationship outcomes have found that persons who view pornography more often tend to report poorer relationship quality (Campbell & Kohut, 2017; Manning, 2006; Newstrom & Harris, 2016; Rasmussen, 2016; Wright & Tokunaga, 2018; Wright et al., 2017). Importantly, however, the majority of such studies have been cross-sectional and thus cannot demonstrate whether pornography use has an “effect” on relationships or, conversely, whether persons in poorer relationships seek out pornography as a means of release or coping. While some longitudinal and experimental research has suggested that pornography use does indeed seem to have an influence on couples’ relationship quality and stability (Lambert, Negash, Stillman, Olmstead, & Fincham, 2012; Perry, 2017a, 2018; Perry & Davis, 2017; Perry & Schleifer, 2018; Wright & Tokunaga, 2018; Wright et al., 2017), other studies have suggested that pornography use could be a consequence of poorer relationship quality (Muusses, Kerkhof, & Finkenauer, 2015; Peter & Valkenburg, 2009); and still others have argued that the link between pornography use and relationship quality is bi-directional (Muusses et al., 2015).

While it was expected that the association between pornography use and relationship quality to be negative following the majority of empirical studies and reviews (see Rasmussen, 2016; Wright et al., 2017), the current study made no assumptions about directionality and thus avoided claims suggesting that either pornography use or relationship troubles were the causal agent at work. It also included controls in each analysis in order to mitigate the likelihood that any observed bivariate association was due to some third sociodemographic factor related to both pornography use and relationship outcomes such as gender, race, income, education, or religion.

Complicating our understanding about the link between pornography use and relationship quality has been the consistent assumption and finding that patterns and contexts of pornography use tend to be different for women and men (Campbell & Kohut, 2017; Wright et al., 2017). While men’s romantic and sexual relationship outcomes are often thought to be closely tied to their pornography use, corresponding to higher use frequencies and the assumption that they are using pornography more often to masturbate (Bridges & Morokoff, 2011; Maddox, Rhoades, & Markman, 2011), studies have also found that women’s pornography use can sometimes correspond to higher relationship or sexual satisfaction, possibly a function of their using pornography within the context of a romantic relationship, not as an alternative for sexual intimacy (Bridges & Morokoff, 2011; Poulsen, Busby, & Galovan, 2013). Reversing the causal arrow, it may be that women who are sexually secure and enjoying intimacy with their partner are simply more open to viewing pornography. The current study tested for whether women and men experience the association between pornography use and relationship outcomes differently by testing for interactions between pornography use and gender for each of the 31 measures of relationship quality across the 30 individual surveys. Drawing on the previous literature, the expected direction of the associations was that pornography use would be a stronger indicator of poorer relationship quality for men than for women (see Wright et al., 2017; see also Doran & Price, 2014; Minarcik, Wetterneck, & Short, 2016; Morgan, 2011; Muusses et al., 2015; Perry, 2017a; Perry & Davis, 2017; Stack, Wasserman, & Kern, 2004).

Another dynamic that this study provided for comparison is the potential difference that marital status makes in the connection between pornography use and relationship quality. Because of greater cultural expectations surrounding fidelity in marriage as compared to dating relationships, pornography use might be perceived, by both the consumer and their spouse, as more “off limits” in a way that makes it a stronger indicator of relational dissatisfaction for the viewer or occasion for offense by a spouse who discovers it (see review and meta-analysis in Wright & Tokunaga, 2018; see also Bridges, Bergner, & Hesson-McInnis, 2003; Olmstead, Negash, Pasley, & Fincham, 2013; Schneider, 2000). Thus, it was expected that any associations that might exist between pornography use and relationship quality for married Americans would be weaker or even potentially nonexistent for unmarried Americans.

Method

Subjects

Analyses for the current study were all based on data from representative samples of married and unmarried American men and women, which are described below.

1973–2018 General Social Surveys

The first series of analyses came from 27 independent cross-sectional waves of the General Social Survey (GSS). The GSS is a nationally representative, face-to-face survey of the non-institutionalized, English–Spanish-speaking American adult population in the U.S. The GSS is funded by the National Science Foundation and has been conducted since 1972 (the 1972 wave did not ask about pornography use). While earlier waves of the GSS were administered roughly every year, since 1994 the GSS surveyed roughly 3000 Americans in even numbered years. Unfortunately, not all questions for the GSS are asked of the entire survey sample in a given year, and thus, some years for the GSS include only small samples of men and women who were asked questions about pornography use. Despite several limitations in the actual pornography use measure (a yes/no question asking about viewing an X-rated movie in the previous year), because the GSS has been repeated so often with consistent measures of pornography use and marital happiness, it provides a useful data set to examine trends in associations between pornography use and relationship quality. Additionally, in Wright et al.’s (2017) meta-analysis comparing associations between various measures of pornography use and relationship quality, they found that the type of pornography use measures (e.g., dichotomous, continuous single-item, or continuous multi-item) did not moderate the association between pornography use and relationship outcomes. All analyses used the “wtssall” survey weight for the GSS.

2006 Portraits of American Life Study

Fielded in 2006, the Portraits of American Life Study (PALS) was intended to be the first wave of a nationally representative panel survey with questions focusing on a variety of topics including social networks, moral and political attitudes, and religious life. The PALS sampling frame included the civilian, non-institutionalized household population in the continental U.S. who were aged 18 years or older at the time the survey was conducted. Surveys were administered in English or Spanish. From April to October 2006, face-to-face interviews were conducted with 2610 respondents in their homes. Interviewers used audio computer-assisted self-interviewing (ACASI) for more sensitive questions (e.g., how often they view pornography). PALS data included a sampling weight (pawt2 in the data set) that, once applied, brings the PALS sample in line with the U.S. Census Bureau’s American Community Survey (ACS) 3-year average estimates for 2005–2007. For more in-depth information about the sampling and data collection process for PALS, see Emerson, Sikkink, and James (2010).

2012 New Family Structures Study

The New Family Structures Study (NFSS) was completed by 2988 American participants aged 18–39 between July 2011 and February 2012. Data collection was conducted by Knowledge Networks, an independent research firm that is now part of the GfK group. Knowledge Networks recruited the first online research panel, called the “KnowledgePanel,” that is representative of the U.S. population. Members of the KnowledgePanel are randomly recruited by telephone and mail surveys, and households are provided with access to the Internet and computer hardware if needed. Unlike other Internet research panels sampling only individuals with Internet access who volunteer for research, this panel was based on a sampling frame which included both listed and unlisted numbers, those without a landline telephone and was not limited to current Internet users or computer owners, and did not accept self-selected volunteers. An evaluation of the Knowledge Networks’ Internet probability sample survey methodology compared favorably to online non-probability samples as well as random-digit-dial telephone surveys (Regnerus, Gordon, & Price, 2016). The main survey completion rate was 61.6%. Cases in the NFSS were assigned a weight (WEIGHT4 in the data set) to adjust for sampling deviations and ensure that the survey was representative of Americans aged 18–39. For more information about the sampling and data collection of the NFSS, see www.familystructurestudies.com (n.d.).

2014 Relationships in America Survey

The Relationships in America (RIA) survey was distributed to a national probability sample of 15,738 adults between the ages of 18 and 60 years in January and February 2014. Data collection was sponsored by the Austin Institute for the Study of Family and Culture and also conducted by Knowledge Networks/GfK using the Knowledge Panel strategy described above for the 2012 NFSS. The main survey completion rate for the RIA survey instrument was 62%. Cases in the RIA sample were assigned a weight based on the sampling design and their probability of being selected, ensuring a sample that was representative of American adults aged 18–60. These sample weights were used in all analyses. For a more comprehensive discussion of sampling and data collection procedures, as well as key outcomes in the RIA, see Litchi et al. (2014).

Measures

Relationship Quality Outcomes

Each of the surveys included measures that can be used to better understand the experiences of men and women in their committed romantic relationships. Questions were asked of both married and unmarried participants regarding their current relationship. Because these questions were from surveys of the general population and were not designed or worded with the explicit goal of connecting them to pornography use, this study can avoid potential bias that could attend studies that were designed with the goal of detecting pornography’s supposed harmful effects on relationships (see this critique in Campbell & Kohut, 2017). Because there are 31 total (1 in the GSS + 13 in PALS + 12 in NFSS + 5 in RIA = 31 measures), space will not be taken to describe each one here. Table 1 lists all of these relationship outcome measures along with all other variables used in the analyses from the GSS, PALS, NFSS, and RIA.

Table 1 Variables used in analyses
Full size table

To be sure, because these different measures in the PALS, NFSS, and RIA are each capturing some aspect of relationship quality, they are consequently measuring similar constructs, and thus, many of the continuous-level measures could hang together as an index, particularly for the PALS and the NFSS. While I will report outcomes when such measures are included together as an index in ancillary analyses (see Footnotes 3 and 4), because the goal of this study was to present the dominant trend in associations between pornography use and relationship quality, the decision was made in the main analyses presented here to keep these relationship measures separate in order to provide the maximum number of independent tests. This also made more sense given that the measures for the GSS and RIA could not be combined into an index. Finally, this also allowed for more nuance into such associations that have been presented either with multi-item scales of relationship quality using PALS (Perry, 2016, 2017a) or GSS data presented only in aggregate, not by year (Doran & Price, 2014; Wright et al., 2017).

Pornography Viewing

Along with a variety of different relationship outcomes across the surveys, each survey contained a measure of pornography use. The GSS asked respondents whether they have viewed an X-rated movie in the previous year, to which respondents could answer yes or no. Among married participants, 20% answered “yes” across all years, though this percentage was higher for more recent waves of the GSS and among participants who were younger and male. This is the most limited pornography use measure in that the language of “X-rated movie” is rather dated and the binary yes/no response option precludes researchers knowing how the frequency of pornography use might relate to relationship outcomes. Nevertheless, Wright et al. (2017) have demonstrated that outcomes with this GSS measure did not differ from those data with different measures of pornography use. This, in addition to its considerable strengths as a data set, has made the GSS an often-used source for predicting relationship outcomes and attitudes (e.g., Doran & Price, 2014; Perry & Schleifer, 2018; Wright, Tokunaga, & Bae, 2014).

The other three surveys all included pornography use measures that were each asked in roughly comparable ways with multiple response values, addressing one of the limitations identified by Campbell and Kohut (2017; cf. Willoughby & Busby, 2016), namely that of inconsistent usage measures. PALS asked respondents “In the past 12 months, how often have you viewed pornographic materials?” Responses ranged from 1 = never to 8 = once a day or more. The modal response for both married and unmarried participants was 1, and the mean was approximately 2 = “once or twice.” The NFSS asked a similar question” “During the past year, how often did you view pornographic materials (such as Internet sites, magazines, or movies)?” Participants could answer from 1 = never to 6 = every day or almost every day. The modal response for both married and unmarried participants was 1, and the mean was roughly 2 = “once a month or less.” Asking about pornography use in a slightly different way from studies that used “how often” type measures, the RIA asked about pornography use in terms of participants’ last time intentionally viewing it: “When did you last intentionally look at pornography?” Participants could answer 1 = I’ve never intentionally looked at pornography to 10 = Today. Higher numbers indicate that the participant was more likely to be a frequent consumer of pornography. The modal response for both married and unmarried participants was 1, and the mean response was around 4 = “Over 1 month ago.” While slightly different in wording, responses to this RIA question yielded roughly comparable outcomes to those of the GSS, NFSS (Regnerus et al., 2016), and PALS (Perry 2016, 2017a, 2018; Perry & Davis, 2017) suggesting sufficient continuity.

Controls

Analyses presented in Tables 2, 3, 4, and 5 each included regression models with controls added. While different data sets include other control variables that could be included for theoretical reasons (e.g., sex frequency, masturbation, personality characteristics; see Perry, 2019a), such questions were not available across all data sets and thus the decision was made to include only sociodemographic controls that would be available in all data sets. While differences in the survey samples and measurements result in differences across these controls (see Table 1 for descriptions of each variable), each regression analysis included controls to adjust for potential confounders (age, gender, education, household income, race, parental status, and worship attendance), each commonly identified as correlates of both pornography use and various relationship outcomes. For example, frequent consumers of pornography are more likely to be younger, male, better educated, higher income, black, childless, and irreligious (Perry, 2017b, 2019b; Perry & Schleifer, 2019; Wright, 2013; Wright, Bae, & Funk, 2013). Similarly, people with better relationship quality also tend to be male and higher socioeconomic status. Yet they are also more likely to be older, white, and religious (Doran & Price, 2014; Perry, 2016).

Table 2 Correlation and logistic regression coefficients predicting married participants being “very happy” in their marriage by viewing an X-rated movie (1973–2018 GSS)
Full size table
Table 3 Correlation and regression coefficients predicting relationship outcomes by pornography viewing frequency (2006 PALS)
Full size table
Table 4 Correlation and regression coefficients predicting relationship outcomes by pornography viewing frequency (2012 NFSS)
Full size table
Table 5 Correlation and regression coefficients predicting relationship outcomes by pornography viewing frequency (2014 RIA survey)
Full size table

Statistical Analysis

The analysis proceeded as follows. Tables 2, 3, 4, and 5 present bivariate correlations and regression coefficients for the pornography use measure and relationship measures for each individual survey. This resulted in 29 independent tests of the association between pornography viewing and various relationship outcomes for unmarried Americans (12Footnote 2 in PALS + 12 in NFSS + 5 in RIA = 29 independent tests) and 57 independent tests for married Americans (1 × 27 waves of the GSS + 13 in PALS + 12 in NFSS + 5 in RIA = 57 independent tests). Though this study acknowledges that the data are cross-sectional and the direction of any observed association between pornography use and measures of relationship quality can go in either direction, regressions were modeled with relationship quality measures as the outcome variable and pornography use as the predictor following the vast majority of previous studies.

Bivariate associations are presented in the left column; regression coefficients from binary logistic or ordinary least squares regression are presented in the middle column; and regression coefficients for interaction terms (pornography use × male) in interaction models are in the far right column. Ns for each analysis are listed in the column next to the correlation or regression coefficient. To conserve space, coefficients for control variables are not presented.

Results

General Social Survey (1973–2016)

Table 2 presents correlations and logistic regression coefficients for married Americans reporting that they viewed an X-rated movie and affirming that their marriage was “very happy” across each of the 27 waves in which the GSS asked these two questions together. At the bivariate level, 63% (17/27) of the associations did not attain statistical significance at the .05 level (56% if the marginal associations at p < .10 are considered due to smaller sample size). Looking at the 37% (10/27) of the bivariate associations that were significant at .05 or better, each one was small in magnitude (r < .15) and negative. Nearly all of the correlations that did not attain statistical significance were negative as well. When GSS years (1973–2018) were aggregated together, the correlation was also significant, small in magnitude, and negative.

When binary logistic regression models were conducted for each year predicting being “very happy” in one’s marriage on viewing an X-rated movie and sociodemographic controls, only 15% (4/27) of coefficients attained statistical significance at .05 (22% if marginal cases are included). The remaining 85% of tests showed a nonsignificant association between viewing an X-rated movie and marital happiness with controls in place. When the same logistic regression model was conducted with all GSS years (adding a control variable for year of GSS), the association between viewing an X-rated movie and being “very happy” in marriage was significant beyond the .001 level and negative.

Were these associations different for men and women? Roughly 11% (3/27) of the interaction terms were statistically significant at .05 (19% if marginal associations are included), and each was in the expected direction with the negative association between watching an X-rated movie and marital happiness being stronger for men. The model from the aggregated GSS showed a highly significant (p < .001) interaction term and confirmed this general trend.

Taken together, the independent-year tests suggested that viewing an X-rated movie in the previous year was more often unassociated with being “very happy” in one’s marriage either at the bivariate level or with controls in place. Most models showed the relationship between viewing an X-rated movie and marital happiness did not differ for men and women, and when it did, the association was stronger for men as expected. All aggregated results showed viewing an X-rated movie was negatively associated with marital happiness and this association was stronger for men.

2006 Portraits of American Life Study

Table 3 presents correlations and regression coefficients for pornography viewing frequency in the previous year and various measures of relationship quality for both unmarried and married Americans in the 2006 PALS. For unmarried Americans, 50% (6/12) of the bivariate associations were nonsignificant, while the other 50% were significant, ranging from small to moderate in magnitude (r between .16 and .31 in absolute value). The direction of each association pointed to poorer relationship outcomes corresponding to higher frequencies of pornography use. When sociodemographic controls were added using regression models, one of the significant associations dropped out, leaving 42% (5/12) of tests showing a significant association with relationship outcomes. Specifically, with controls in place, unmarried persons who viewed pornography more often were more likely to report that their partner harshly criticized them and tended to report lower levels of relationship happiness, satisfaction with the affection they received, their sex life, or decision-making in the relationship. Two significant interactions with gender indicated that the negative association between pornography viewing and relationship happiness and experiencing kindness from one’s partner was stronger for men.

For married participants, another measure of relationship quality (experiencing a marital separation in past 3 years) was added. The results for married persons were not substantively different from those of unmarried persons. Over 69% (9/13) of the bivariate associations were significant, though small in magnitude (r all below .14). The direction of each significant association indicated that more frequent pornography use was associated with poorer relationship outcomes for married Americans. When controls were added in regression models, two previously significant associations became nonsignificant and one became marginal, while other non-associations became significant. This was the case in the two other surveys as well, which suggests suppressor effects, namely when the inclusion of certain controls in a multivariate model removes some of the unexplained variance between the key predictor and the outcome, resulting in a larger, more statistically significant effect for that key predictor (in this case pornography use). With controls in place, married Americans who viewed pornography more frequently reported less affection (marginal) or complements from their spouse; they were less likely to be happy in the relationship or satisfied with the affection, sex life, or decision-making; they were more likely to believe their spouse had cheated; more likely to have cheated themselves romantically (marginal) or sexually; and more likely to have experienced a marital separation in the past 3 years (marginal). None of the interaction terms for pornography use and gender were significant at the .05 level. Two marginally significant associations suggested that the negative association between pornography viewing and receiving compliments or satisfaction with decision-making may be stronger for women than for men.

Similar to findings for the GSS, bivariate and adjusted associations between pornography viewing and relationship outcomes tended to be either nonsignificant or significantly associated with poorer relationship outcomes, for both unmarried and married Americans. In only two instances were these associations significantly different for men and women at the .05 level, and each was in the predicted direction.Footnote 3

2012 New Family Structures Study

Table 4 presents correlations and regression coefficients for pornography viewing frequency in the previous year and measures of relationship quality for unmarried and married Americans in the 2012 NFSS. Looking at unmarried Americans, one quarter (3/12) of the associations were statistically significant (50% if marginal associations are included) and all were small in magnitude (r all below .17). Significant associations were in the expected direction with unmarried persons who viewed pornography more often reporting more frequently thinking their relationship was in trouble and discussing ending the relationship with their partner. Frequent viewing was also negatively associated with self-rated relationship happiness. Once controls were included in regression models, nearly 42% (5/12) of associations became statistically significant, and all in the expected direction. Specifically, unmarried Americans who viewed pornography more frequently were more likely to report thinking about ending their relationship, thinking their relationship was in trouble, and discussing ending the relationship. They were also less likely to agree that their relationship was “pretty much perfect” and tended to report lower relationship happiness.

Two significant interactions suggested conflicting trends. The positive association between viewing frequency and thinking one’s relationship was in trouble was stronger for men, but the positive association between pornography viewing and actually talking to one’s partner about separating was stronger for women.

Among married Americans, less than half (5/12) of the bivariate associations between viewing frequency and relationship outcomes were statistically significant (50% including marginal associations). Significant associations were small in magnitude (r all below .13), though all in the expected direction with higher pornography viewing frequency corresponding to poorer marriage outcomes. When controls were added in regression models, a number of the associations either became significant or increased in significance. Over 80% (10/12) associations were statistically significant at the .05 level with one marginal association. All but one association was in the expected direction. Specifically, viewing pornography more often was associated with more frequently thinking one’s marriage was in trouble, discussing ending the marriage, and repeatedly breaking up. It was also negatively associated with characterizing one’s relationship as “good” or “strong,” agreeing that one felt like a team with their spouse, that their relationship made them happy, or that their relationship was nearly perfect. Frequent pornography use was also negatively associated with self-rated marital happiness with controls in place.

The one exception to this trend was that pornography viewing frequency seemed to be negatively associated with a married participant talking with their spouse about separating. This association, though barely attaining statistical significance (p = .049), would contradict the theory that pornography use tends to be associated with more negative relationship outcomes.

Two significant interactions were in the opposite of the expected direction. Specifically, they indicated that the positive association between pornography use and repeatedly breaking up appeared to be stronger for women, while the negative association between pornography use and reporting one’s relationship made them happy was also stronger for women.

Findings from the 2012 NFSS suggest that the association between pornography viewing and relationship outcomes is not as strong or consistent for unmarried persons as for married persons, though this may be due to differences in sample size. Numerous bivariate associations were either nonsignificant or small in magnitude, with more significant associations emerging with controls in place. With one curious exception, the significant associations indicated that more frequent pornography use is associated with poorer relationship quality.Footnote 4

2014 Relationships in America Survey

Table 5 presents correlations and regression coefficients for participants’ most recent pornography viewing and relationship outcomes for unmarried and married Americans in the 2014 RIA. Among unmarried Americans, 80% (4/5) bivariate associations were statistically significant and small in magnitude (r all below .20). Each significant association was in the expected direction. When controls were included in regression models, 100% (5/5) associations were statistically significant and all in the expected direction. Unmarried Americans who viewed pornography more often (as indicated by when they reported last using it) were more likely to think about leaving their partner, talk with their partner about separating, cheat sexually, and experience physical violence. They also tended to report lower relational happiness. The only statistically significant interaction term (at the .05 level) indicated that the positive association between viewing pornography and thinking about leaving one’s partner was stronger for men.

Turning to married Americans, 100% (5/5) of the bivariate correlations were statistically significant, though they were small in magnitude (r all below .16). All regression coefficients were also significant with controls in place. The direction of the effects pointed to the trend that more frequent pornography use was associated with poorer marital outcomes, in this case, being more likely to think about leaving one’s spouse, talk about separating, cheat sexually, or experience physical violence, while also reporting lower marital happiness. The two significant interaction terms were in conflicting directions. The positive association between pornography viewing and thinking about leaving one’s spouse was stronger for men, but the positive association between pornography use and talking about separating was stronger for women.

Together, results from the 2014 RIA consistently showed that the association between pornography viewing and relationship outcomes for married and unmarried Americans were statistically significant and all in the expected direction, with pornography viewing frequency being linked with poorer relationship quality.

Analysis of the Analyses

Figure 1 illustrates the general patterns that are observed in Tables 2, 3, 4, and 5 across the regression models (with sociodemographic controls). Because the GSS did not include a relationship outcome for unmarried Americans and PALS also had a question that asked specifically about marital separation, there were only 29 independent tests for unmarried Americans specifically across the 2006 PALS, 2012 NFSS, and 2014 RIA. Roughly 52% of the tested associations were statistically significant and pointed to a negative outcome. In other words, in 52% of the associations for unmarried Americans, those who viewed pornography more frequently reported poorer relationship quality by various measures. The other 48% of associations were nonsignificant, and 0% were situations in which pornography use was associated with a positive relationship outcome.

Fig. 1
figure 1

Breakdown of outcomes from regression models predicting relationship outcomes on pornography use measures

Full size image

Among married Americans, treating the GSS as 27 independent data sets resulted in 57 individual tests of pornography’s association with marriage outcomes. The 2% of associations in which pornography use was significantly associated with a positive outcome was the result of the one association in the 2012 NFSS (Table 4) in which married persons who viewed pornography more often seemed to be less likely to talk to their spouse about separating. By contrast, 44% of the associations were significant and in a direction that suggested pornography use (either at all or in greater frequencies) was associated with poorer marital quality. Ultimately, however, the largest percentage (54%) of associations were not statistically significant. The bar on the far right, however, shows that this preponderance of nonsignificant associations was primarily because of numerous nonsignificant associations in the individual GSS waves, where samples were often quite small. When the GSS was treated as an aggregate (see Doran & Price, 2014; Patterson & Price, 2012), only about one quarter of associations were nonsignificant and 71% were significant signaling a poorer relationship outcome. Thus, in broadest terms, the analyses of nationally representative data with a variety of relationship outcomes have shown that pornography use is generally either unassociated with relationship quality or it is an indicator of poorer relationship quality. By contrast, more frequent pornography use only rarely (one instance in this study) corresponded to better relationship quality in the general population.

Discussion

Despite the numerous studies conducted on the association between pornography use and committed romantic relationships, there remains some disagreement among scholars as to whether there are clear trends. Part of the challenge has been that data were often taken from small, non-representative populations, using measures or designs that could be negatively biased, and findings could often be curiously different for men and women. Using 31 measures of relationship quality across 30 nationally representative surveys, the current study sought to mitigate these issues in order to establish a dominant trend in the association between pornography use and relationship quality for representative samples of unmarried and married men and women. That dominant trend seems to be that pornography use in the general population—either at all or in higher frequencies—is either unassociated with romantic relationship quality or is weakly associated with poorer relationship quality. This was true for married and unmarried Americans alike as well as for men and women. Conversely, more frequent pornography use was almost never associated with better relationship quality, at least on average. Moreover, consistent with Wright et al. (2017), these patterns held across different measures of pornography use, including dichotomous measures (GSS), those asking about general frequency (PALS, NFSS), and those asking about most recent use (RIA).

To be sure, this study has made no claim as to the direction of the association between pornography use and relationship quality nor could it do so with these data. While other studies using the panel component of PALS (e.g., Perry, 2017a, 2018; Perry & Davis, 2017) or the GSS (e.g., Perry & Schleifer, 2018; Wright et al., 2014) have sought to establish a directional “effect” between pornography use and relationship outcomes, the goals of this study were to establish a dominant pattern in associations across a maximum number of relationship outcomes and surveys. Since this study cannot determine directionality, it could very well be that any observed association between pornography viewing and poorer relationship quality can be explained by self-selection (i.e., Americans in struggling relationships seek out pornography as an escape or alternative), just as it could be that frequent pornography use is contributing to the relationship struggles. As suggested by Muusses et al. (2015), it could also be both.

Beyond the fact that all these data were cross-sectional, they are also only of individual Americans rather than dyads. Thus, the study was unable to address one of the primary critiques of the previous research on pornography use and relationship quality (see Campbell & Kohut, 2017; Newstrom & Harris, 2016), in that it cannot examine the relationship quality of someone whose partner is viewing pornography nor is it able to examine relationship outcomes of couples who view pornography together. Some of the confusion about findings linking pornography use with relationship outcomes stems from these two limitations. In their recent narrative review and meta-analysis of literature examining heterosexual men’s pornography use and their female partner’s response, Wright and Tokunaga (2018) demonstrated the general trend that women who perceived their male partner as pornography consumers tended to be less relationally or sexually satisfied, and tended to be more insecure about their own bodies. Moreover, because such Americans who use pornography together with their partner (and thus might experience positive returns to their pornography use) would also be included in these samples, the findings presented here suggest that these are a minority among pornography users. That is, whether or not coupled pornography use might be beneficial for some couples, the stronger pattern among a larger percentage of Americans is that pornography consumption happens more frequently in relationships that are not doing well comparatively.

Interestingly, the tests for interactions also showed that in the vast majority of instances, gender did not significantly moderate the association between pornography use and relationship outcomes. And the relatively few situations where these interactions were significant painted rather inconsistent results. Sometimes, it seemed that the quality of men’s romantic relationships was more closely tied to pornography use, while other times it seemed that the association was stronger for women. At the very least, the consistent lack of a moderating effect for gender would challenge assumptions that women’s pornography use tends to be associated with better relationship quality, while men’s is associated with poorer relationship quality due to different use patterns. Rather, for both men and women, married and unmarried, pornography use tended to be either unassociated with relationship quality or associated with poorer relationship quality.

There also seemed to be little discernable difference between those in marriage relationships verses unmarried romantic relationship in terms of the association between pornography use and relationship outcomes. Despite research suggesting that pornography use might be viewed as more of a violation in marriage relationships perhaps due to more expansive and stringent expectations for sexual “fidelity” (Bridges et al., 2003; Olmstead et al., 2013; Schneider, 2000), there were relatively few instances where associations in the 2006 PALS, 2012 NFSS, or 2014 RIA survey were statistically significant for married Americans and were not significant for unmarried Americans, despite some potentially large differences in sample size.

Despite the broader trend that pornography use tended to be an indicator of poorer relationship quality in the majority of significant associations, the exception (in the 2012 NFSS; Table 4) must be considered as an important qualifier. On the face of it, the finding that married persons who viewed pornography more often were less likely to talk to their spouse about separating would contradict the idea that pornography use is associated with poorer relationship outcomes. Unfortunately, the interpretation of this association is not so clear. It could also be that persons who view pornography more often are simply less likely to talk to their spouse at all, not just about separating. Moreover, given that 9 of the other 12 outcomes for married participants in the NFSS all point to the conclusion that viewing pornography more often is linked with poorer marital quality, this finding is anomalous and perhaps an outlier. However, to the extent that this association is capturing a real relationship, it requires that scholars provide appropriate qualification when drawing conclusions about pornography’s association with relationship outcomes. To the extent that the two are related at all (and in many instances they were not), pornography use tends to be an indicator of poorer relationship quality, though not always.

Conclusion

This study sought to establish a dominant trend in the association between pornography use and relationship quality using representative samples of married and unmarried men and women and 31 measures of relationship quality across 30 nationally representative surveys. Trends demonstrated that pornography use in the general population (whether men or women, married or unmarried) is either unassociated with romantic relationship quality or associated with poorer relationship quality. Pornography use is almost never an indicator of better relationship quality in general. And indeed, these patterns hold true across slightly different measures of pornography use and after adjusting for relevant sociodemographic factors.

There are implications for clinicians and therapists. Despite the inconsistency or ambiguity that some recent scholars have claimed regarding pornography’s association with relationship outcomes (e.g., Campbell & Kohut, 2017; Newstrom & Harris, 2016), the current study in addition to findings of recent reviews and meta-analyses (e.g., Rasmussen, 2016; Wright & Tokunaga, 2018; Wright et al., 2017) suggests that future research may proceed under the general assumption that pornography use, to the extent that it is associated with relational quality at all, is nearly always an indicator of poorer relationship quality on average, regardless of marital status or gender. From there, however, other questions need to be asked. While the findings of this study suggest that, in general, pornography use signals relational problems, other studies have shown that this association is attenuated (or even reversed) among the minority of couples who view pornography as a part of sexual activity (e.g., Maddox et al., 2011) or if persons or their spouses/partners do not feel religiously or morally conflicted about their pornography use (e.g., Perry, 2016, 2019b; Perry & Whitehead, 2019). And still others suggest that the association between pornography use and relational outcomes may be more about the act of solo-masturbation rather than viewing pornography per se (e.g., Perry, 2019a). Thus, despite what this study’s findings tell us about general trends, clinicians and other relationship counselors need to inquire about the specific relational and moral contexts within which pornography is being used.