The Journal of Sex Research

ISSN: (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/hjsr20

Fantasy Made Flesh - A Network Analysis of the

Reciprocal Relationship between Sexual Fantasies,
Pornography Usage, and Sexual Behavior

Maximilian T. P. von Andrian-Werburg, Eric Klopp & Frank Schwab

To cite this article: Maximilian T. P. von Andrian-Werburg, Eric Klopp & Frank Schwab (2023):
Fantasy Made Flesh - A Network Analysis of the Reciprocal Relationship between Sexual
Fantasies, Pornography Usage, and Sexual Behavior, The Journal of Sex Research, DOI:
10.1080/00224499.2023.2170964

To link to this article: https://doi.org/10.1080/00224499.2023.2170964

© 2023 The Author(s). Published with View supplementary material

license by Taylor & Francis Group, LLC.

Published online: 21 Feb 2023. Submit your article to this journal

Article views: 745 View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=hjsr20
THE JOURNAL OF SEX RESEARCH
https://doi.org/10.1080/00224499.2023.2170964

Fantasy Made Flesh - A Network Analysis of the Reciprocal Relationship between

Sexual Fantasies, Pornography Usage, and Sexual Behavior
a b a
Maximilian T. P. von Andrian-Werburg , Eric Klopp , and Frank Schwab
a
Institute Human-Computer-Media, Faculty of Human Sciences, University of Würzburg; bDepartment of Education, Saarland University

ABSTRACT
Based on different theories in media research (3AM, catalyst model of violent crime, reinforcing spirals
model), we further explore the relationship between pornography use, sexual fantasy, and behavior. We
suggest that pornography use appears so persistent across time and culture because it is related to a
human universal, the ability to fantasize. Consequently, pornography use seems to be an opportunity to
acquire media-mediated sexual fantasies, and we believe that pornography use interacts with sexual
fantasies and, to a much weaker extent, with sexual behavior. To assess our assumptions, we conducted a
network analysis with a large and diverse sample of N = 1338 hetero- and bisexual participants from
Germany. Analyses were done separately for men and women. Our network analysis clustered parts of the
psychological processes around the interaction of sexual fantasies, pornography use, and behavior into
communities of especially strong interacting items. We detected meaningful communities (orgasm-
centered intercourse, BDSM) consisting of sexual fantasies and behavior, with some containing porno­
graphy. However, pornography use was not part of communities we perceive to account for mainstream/
everyday sexuality. Instead, our results show that non-mainstream behavior (e.g., BDSM) is affected by
pornography use. Our study highlights the interaction between sexual fantasies, sexual behavior, and
(parts of) pornography use. It advocates for a more interactionist view of human sexuality and media use.

Introduction
Research about the (potential) effects of pornography use is thriv­ reaches a global audience (e.g., Hald & Mulya, 2013), its use
ing in the field of sexual science but is mostly neglected in main­ (and production) appears to be motivated by something uni­
stream media psychology and communication science (Grubbs & versal to human nature. For instance, even the combination of
Kraus, 2021; Kohut et al., 2020). The term pornography refers to very strict anti-pornography legislation and widespread
“any type of sexually explicit material that has the intent of Islamic faith did not make its use disappear in Indonesia
producing arousal in those who consume it” (Lehmiller, 2017, p. (Hald & Mulya, 2013). Instead, erotic cultural output, which
402). A possible reason for this lack of scientific attention is a fear can reasonably be perceived as being created with the intent
of stigmatization, which porn researchers are often confronted of causing sexual arousal in its users and therefore classifies as
with (Kohut et al., 2020). Furthermore, the field is underfunded pornography, frequently (re-)appears independent of time or
(Grubbs & Kraus, 2021). culture (Lehmiller, 2017; Schmidt & Voss, 2000). The Turin
However, given the reach and widespread use of porno­ erotic papyrus drawn in old Egypt displays sex scenes so
graphy (e.g., Hald & Mulya, 2013; Martyniuk & Dekker, 2018; vividly that it gained attention from urology (Shokeir &
Price et al., 2016), it is a genuine mass media content whose Hussein, 2004). This is only one example of a potentially
clicks surpass even those of the most prominent news outlets. endless amount of historical sexually explicit findings
BBC.co.uk is the world’s most clicked news outlet, and in its (Schmidt & Voss, 2000). Recognizing these findings, humans
most successful month ever (March 2020), it had about 1.5 seem to have always tended to externalize their sexual fanta­
billion page views (BBC, 2020). In comparison, the porno­ sies through communication and, subsequently, mediated
graphic content provider Pornhub, which is not even the communication (Ohler & Nieding, 2005; Shokeir & Hussein,
largest provider on the Internet, attracts about 130 million 2004). The phenomenon of “modern” pornography might
visits every day, accumulating to an average of 3.9 billion only be molded but not caused by today’s culture because it
views per month (Pornhub Insights, 2021). This shift in is driven by the basic human need for sexuality (e.g., Lippa,
perspective – highlighting pornography as mass media con­ 2009) and shaped by a universal human capability, which is to
tent – allows us to apply different theories about media use fantasize. If this is the case, we may be able to observe reci­
and effects originating from communication science and procal interactions between sexual behavior, sexual fantasies,
media psychology research on mass media (Ferguson et al., and pornography use. Relations between these dimensions
2008; Slater, 2015; Slater et al., 2020). Given that pornography have been described both in theories about the development

CONTACT Maximilian T. P. von Andrian-Werburg [email protected] Chair of Media Psychology, Institute Human-Computer-Media, Faculty
of Human Sciences, University of Wuerzburg, Oswald-Kuelpe-Weg 82, Ger-97074, Wuerzburg, Germany
Supplemental data for this article can be accessed online at https://doi.org/10.1080/00224499.2023.2170964.
© 2023 The Author(s). Published with license by Taylor & Francis Group, LLC.
This is an Open Access article distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives License (http://creativecommons.org/licenses/by-nc-nd/4.0/),
which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited, and is not altered, transformed, or built upon in any way.
2 M. T. P. VON ANDRIAN-WERBURG ET AL.

of sexual behavior (e.g., Wright, 2011, 2014) as well as in We assume an interaction between pornography use, sexual
theories of broader communication science and media psy­ fantasies, and behavior. Sexual fantasies might not only be a
chology (e.g., Ferguson et al., 2008; Slater, 2015; Slater et al., rumination of pornography use, as described by Wright (2011,
2020). 2014). Pornography use might as well be a rumination of sexual
fantasies. This exchange could lead to a potentially long-lasting
interaction where fantasies and previous behavioral experi­
Different Perspectives on the Complex Reciprocal
ences trigger the demand for novel pornography use, and the
Interaction of Pornography Use, Sexual Fantasies, and
use of novel pornographic content triggers more fantasies,
Sexual Behavior
leading to a renewed demand for pornography. To formalize
Different scholars have emphasized the relations between sex­ this process, the reinforcing spirals model (RSM; Slater, 2007,
ual fantasies, sexual behavior, and pornography use (Slater, 2015; Slater et al., 2020) comes to mind, which theoretically
2015; Slater et al., 2020; Wright, 2011, 2014). The script acqui­ supports such an interaction between pornography use, sexual
sition, activation, application model (3AM; Wright, 2011, 2014) fantasies, and (possibly) subsequent behavior. It proposes that
mentions the impact of sexual fantasies on the relations people tend to select communication sources and content that
between pornography use and sexual behavior. It argues that best matches their own beliefs and social identity, and in turn,
pornography provides its users with socially constructed beha­ beliefs and behaviors are reinforced by such communication
vior scripts (acquisition), can activate prior acquired scripts selectivity. In the long run, this can lead to a strong attachment
(activation), and encourages them to utilize scripts by depict­ to the respective communication sources and content.
ing certain sexual behaviors to be appropriate and rewarding Applying the RSM to pornography use in the terminology of
(application; Wright et al., 2021). According to the 3AM, sexual the 3AM, users would be initially prone to select pornographic
fantasies can support the acquisition of behavior scripts by content that matches their existing fantasies and behavioral
pornography use because the media sparks fantasies that scripts. Their pornography use may then act as a source of
allow rumination and rehearse the pornographic content inspiration, possibly reinforcing a particular style of sexuality-
used (Wright, 2011). Furthermore, these fantasies may be inclusive coherent behavior and fantasies (e.g., Bondage-
accompanied by masturbation, a form of enactive rehearsal Dominance-Sadism-Masochism [BDSM] enthusiast, fetishists)
that will increase the accessibility of related scripts because in the long run.
the fantasies yield rewarding outcomes in the form of an In summary of our assumptions, we perceive pornography
orgasm. However, pornography users do not automatically use at its core as a mass media-mediated way to use externa­
acquire or apply all behavior scripts they encounter. All three lized sexual fantasies. Pornography use can both spark sexual
processes of script acquisition, activation, and application fantasies and, in rare cases, “inspire” behavior or will be more
depend on different audience, content, accessibility, and situa­ often a reaction to fantasies “outsourcing” parts of the cogni­
tional variables (Wright, 2011). tive processes around sexual fantasies into media content.
This assumption of mediators being present for a media These processes will create a reciprocal interaction between
effect to occur is reflected in another model that formalizes an individual’s fantasies and pornography use and possibly (if
media use and effects. The catalyst model of violent crime only slightly) affect the audience’s behavior (Ferguson et al.,
(CMVC; Ferguson et al., 2008) assumes that adverse media 2008; Wright, 2011, 2014).
effects are small and have only minor, almost negligible effects
on an audience’s behavior. According to the CMVC, media use
is never the root of specific behavior but acts as an indirect
The Network Approach
influence, altering the visual display of how behavior is con­
ducted. For instance, a person with a disposition for acting We use a network approach to examine the amount of
violently (exemplarily caused by a troubled upbringing or a lot interaction between pornography use, sexual fantasies, and
of proximate life strain) might be interested in violent porno­ behavior (e.g., Newman, 2018). The approach provides
graphy and, due to the mediation of this disposition, acquire excellent potential for a more nuanced view of pornography
and apply adverse scripts of sexual behavior previously usage and effects than previously possible. Different ele­
encountered in pornography. In the original CMVC, fantasies ments of interest are conceptualized within a system of
would be hypothesized to predispose this only stylistic influ­ pairwise interactions. In contrast to directed relations, pair­
ence of media use on behavior, which Ferguson et al. (2008) wise interactions in networks are considered reciprocal and
called a stylistic catalyst. do not bear a causal meaning. Thus, networks represent a
Contrary to this, we do not assume a predisposition of complex system of elements influencing each other. From
fantasy to pornography use and behavior but rather a recipro­ the network perspective, sexual behavior is neither caused
cal interaction between sexual fantasies and pornography use by sexual fantasies nor pornography consumption. Instead,
and behavior. However, fantasies are very volatile and diverse the examined variables influence each other. In this way, a
and might depend on less impactful, if any, mediator variables network allows modeling our assumed reciprocal relations
(Wright, 2011, 2014) to interact with pornography use com­ between pornography use, sexual fantasies, and sexual
pared to actual behavior (Joyal et al., 2015; Lehmiller, 2018; K. behavior. If our data should yield evidence for complex
M. Williams et al., 2009). Therefore, fantasies should occur interactions between pornography use, sexual fantasies,
more often in reciprocal relations with pornography use than and behavior, then a network analysis should provide an
actual sexual behavior. accurate and stable network.
 THE JOURNAL OF SEX RESEARCH 3

In network terminology, the elements of a network that pornography use pointing toward a meaningful interpretable
represent the variables of interest are called nodes, and the style of sexuality?
reciprocal interactions between the nodes are called edges.
An edge connects two nodes in a network, and the set of all
edges forms it. The edges represent the pairwise interaction
Method
between the nodes. From a statistical perspective, edges are
represented by partial correlation (cf., D. R. Williams et al., We advertised an online questionnaire exclusively for adult
2019). The edges represent the degree of association participants in social media groups and online forums in
between two variables (nodes) with the effects of the Germany. Previous results of Hald and Štulhofer (2016)
other variables in the network controlled for. Thus, the showed different pornographic content preferences depending
edges describe the interactions between a pair of nodes on sex and sexual orientation. Therefore, we decided to address
indeed. Edges are typically judged according to their our research question separately by sex and only to choose
strength and if the edge represents a positive or negative participants who were at least somewhat sexually attracted to
relation. the opposite sex (self-labeled hetero- and bisexuals). The ana­
Typically, in a network, there are subsets of nodes that lysis required a large sample size, and the prevalence of parti­
are more closely related to each other than other nodes. cipants with an exclusively homosexual orientation for a
In other words, the distribution of edges is inhomogeneous, separate analysis appeared too low in a convenience sample
i.e., there are groups of nodes so that the frequency of (e.g., Greaves et al., 2017).
edges within a group of nodes is high, whereas the fre­ The questionnaire was initially clicked 28,457 times, which
quency of edges between the group’s nodes and nodes resulted in 1496 complete questionnaires. After a first data
belonging to other groups is low. Such groups with a high screening, we excluded 38 participants that reported to be
number of internal connections are called communities. All younger than 18 years, as this is the legal age in Germany to
communities form the community structure of a network. consent for participating in scientific studies. This yielded a
Communities contain nodes that probably share common sample size of 1458 participants. Of these, we excluded the
properties and play similar roles within the network fastest 2.5%, who answered the questionnaire in less than
(Fortunato, 2010). Regarding the reciprocal interaction 11 minutes as it took more than 20 minutes on average to
between pornography use, sexual fantasies, and behavior, complete the questionnaire. Of the remaining 1448 partici­
one can ask if pornography use, fantasies, and behaviors pants, we excluded 11 participants who did not identify as
belong to the same style of sexuality, e.g., BDSM, and male or female. Furthermore, we excluded 81 participants
should form a community. In sum, the network perspective who did not disclose their sexual orientation and 25 who
allows conceptualizing broad patterns of psychological phe­ exclusively identified as homosexual. 1342 participants
nomena as properties that emerge from the interactions remained. After these steps, further screening was conducted,
among certain behaviors and cognitions (Costantini et al., separated by sex. The male sample contained 546 participants,
2019). and the female sample contained Nfemale = 796 participants. In
Network analyses have recently been used in media research the male sample, three participants had more than 18 conse­
to show that active and passive social media use is, on average, cutive missing values on the scales relevant for this paper, and
very weakly related to depressive symptoms (e.g., Rodriguez et one participant had five missing values spread over all relevant
al., 2022) though, associations differed substantially among scales. We excluded all cases with five or more missing values
individuals, highlighting differences in the susceptibility for a on all items in our analysis, resulting in a new sample with
negative media influence to occur. This importance of indivi­ Nmale = 542 participants. For the remaining participants, miss­
dual differences is comparable to our assumptions about weak ing individual values were replaced with the median. No parti­
media effects derived from the 3AM and the CMVC. A com­ cipant in the female group had more than five missing values,
parable general but crucial individual difference variable would and all participants remained part of the sample. All assessed
be sex. Men appear at a higher risk for adverse effects of demographic characteristics of the final sample are displayed in
pornography use as they show an, on average, higher disposi­ Table 1.
tion for violence (Ferguson et al., 2008). Furthermore, one of
the most replicated findings in pornography research is that
Measures
men use on average more pornography, show different usage
patterns and prefer different content types than women (Hald A professional translator who was not informed about the
& Štulhofer, 2016; Petersen & Hyde, 2010; von Andrian- research objective translated the English scales into German.
Werburg et al., 2022). Adding this to our reasoning about The same 5-point intensity scale was used for the scales assessing
pornography use, sexual fantasies, and behavior, we asked pornography preferences, sexual fantasies, and behavior
separately for each sex: described in this section. Its labels started with the minimum
1 = never, 2 = rarely, 3 = occasionally, 4 = often, to the maximum
RQ1: Can sexual fantasies, sexual behavior, and pornography 5 = always. We treated all items as ordinal because many items
use be modeled in a network? in the sex fantasy questionnaire (Wilson, 1988, 2010) and of the
27 different types of pornography (Hald & Štulhofer, 2016)
RQ2: Can we detect different communities (in the networks) of assess comparable uncommon pornography use and sexual
reciprocally interacting sexual fantasies, sexual behavior, and acts. For instance, the Sex Fantasy Questionnaire (SFQ) asks if
4 M. T. P. VON ANDRIAN-WERBURG ET AL.

Table 1. Sample characteristics. indicate how often you act out the statements below in
Variables male female reality.”
Sample size 542 796
Age Groups (M/SD) 30.99/11.28 27.44/7.80
18–19 years 31 57 Network Analysis
20–29 years 289 518
30–39 years 121 151 To analyze the possible reciprocal relations between the items
40–49 years 55 45
50–59 years 28 23
from the TOP and SFQ data, we used a network analysis (e.g.,
>60 years 18 2 Borsboom & Cramer, 2013; Epskamp & Fried, 2018; Epskamp
Education et al., 2017). All TOP and SFQ items were used as single items.
Still attending school 3 9
Secondary school 104 124
Because networks describe pairwise interactions between vari­
High school diploma 237 362 ables in partial correlations, they correspond to a Gaussian
University/college degree 184 279 graphical model described by the precision matrix, i.e., the
Different kind of degree 13 21
Left school without graduation 1 1
inverse of the variables’ correlation matrix (cf, Epskamp
et al., 2017; D. R. Williams et al., 2019). The precision matrix
is a standardized matrix in which the entries with reversed
participants had ever had or fantasized about a sexual encounter signs correspond to a partial correlation matrix. These partial
with an animal (Wilson, 1988, 2010). Consequently, many items correlations describe the pairwise interactions between the
are strongly skewed by nature, and this skewness leads to a variables, i.e., the edges in the network.
severe variance reduction for them.
Network Estimation, Accuracy, and Stability
Concerning RQ 1, we considered the estimation of the network
Sexual Orientation as well as its accuracy and stability. If pornography use and the
We assessed sexual orientation with a continuous scale ranging corresponding sexual fantasy and behaviors form reciprocal
from 0 = exclusively homosexual to 100 = exclusively hetero­ interactions, it should be possible to find networks that are
sexual to allow participants to express nuances regarding their firstly accurate, i.e., not affected by sampling variability, and
inclination toward a sexual orientation. secondly stable, i.e., their interpretation remains stable. To
estimate the network structure, we used a nonregularized esti­
Types of Pornography mation (D. R. Williams et al., 2019) that draws on significance
We assessed 26 of the 27 types of pornography (TOP; Hald tests of the partial correlations representing the edges to deter­
& Štulhofer, 2016). These types were empirically derived mine which edges are included in the network.
and initially clustered by Hald and Štulhofer (2016) to We used Spearman correlations to deal with the ordinal
obtain a more differentiated view on pornography use. characteristic of the rating scale belonging to the TOP and
They denominate pornographic content categories or gen­ SFQ items. For this correlation matrix, the precision matrix is
res and range from “mainstream content” (e.g., “amateur,” estimated that represents the partial correlation between the
“oral sex”) to what Hald and Štulhofer (2016) called “non- variables. For each partial correlation, a significance test is
mainstream/ paraphilic” content (e.g., “Bondage and dom­ conducted to determine if the edges are kept in the network
inance,” “sadomasochism”). We excluded the type labeled or are removed. Because of the exploratory nature of this study,
“other” as its ambiguity did not suit our analysis. Before we used a nominal α = .05 level. Those edges where α is larger
carrying out the TOP assessment, we asked: “How often do than the significance level are discarded from the network.
you watch movies from the following porn categories?” We examined the accuracy of the resulting edge weights.
Accuracy indicates how prone the estimated edge weights are
to sampling variation. The bootstrap method is a way to assess
Sexual Fantasy and Behavior Questionnaire this accuracy (Epskamp et al., 2018). To be accurate, the boot­
To assess sexual fantasies and behavior, we used the 40 strap mean of the edge weights should be close to the estimated
labels of the SFQ (Wilson, 1988, 2010). Daytime fantasies edge weight value. The number of bootstrap samples we drew
and actual sexual behavior are reported on the same 40 was 5000.
items originally distinguished into four factors subsuming We also evaluated the stability of the network. Stability
ten items each: Exploratory sexual fantasies & behavior (e. indicates how similar the interpretation remains when there
g., “Sex with two other people”), intimate sexual fantasies & are fewer observations (Epskamp et al., 2018). A method to
behavior (e.g., “Making love outdoors in a romantic setting examine the stability is the case-dropping subset bootstrap.
like a field of flowers or a beach at night”), impersonal In each bootstrap, a certain proportion of the cases is
sexual fantasies & behavior (e. g., “Intercourse with an dropped from the analysis, and the correlation between
anonymous stranger”), and sadomasochistic sexual fantasies the original estimates and those obtained from the boot­
& behavior (e. g., “Being forced to do something”). Equal to strap sample is calculated (Epskamp et al., 2018). The CS(τ)
the TOP items, all items were treated to be single items for coefficient represents the maximum proportion of cases
the analysis. To introduce the assessment of sexual fanta­ that can be dropped so that the correlation between the
sies, we asked: “Please indicate how often you fantasize original estimate and the bootstrap estimate from the case-
about the statements below on an average day.” To intro­ dropping subset exceeds a given threshold τ. Epskamp et al.
duce the assessment of sexual behavior, we asked: “Please (2018) suggested that CS(τ) should at least be greater than
 THE JOURNAL OF SEX RESEARCH 5

0.25 and preferably greater than 0.50. The parameter τ Excluded items are crossed out in respective tables. In this way,
indicates how strong the correlation between the original from the total of 106 items originally used, 82 items fulfilled this
estimates and those obtained from the bootstrap sample criterion for the male sample, and 62 items fulfilled the criterion
should be. We computed the CS(τ) coefficient for our for the female sample.
analysis of the edge weight. We chose a τ = .50 because,
according to Cohen (1988), a correlation of at least .50
indicates a strong effect. However, Epskamp et al. (2018) Network for Males
chose a τ = .70. Therefore, we considered the CS(τ) coeffi­
cients in the range from τ = .50 to τ = .70 to assess the The network for the male sample is shown in Figure 1. There
stability of our networks. The number of case-dropping are 10 Spinglass communities in the network. The commu­
subset bootstrap samples we drew was 5000. nities are ordered in a counterclockwise direction on a fictive
circle. The colored dots represent the nodes. The letters F, R, or
P denominate if the respective node is either fantasy, behavior
Community Detection and Graphical Display
(in reality: R), or pornographic content, followed by a number
To answer RQ 2, we examined if there were communities
to denominate a respective item label displayed in online
in the network. To detect communities in the networks, we
supplementary Table 3.
used the Spinglass algorithm (e.g., Yang et al., 2016). The
Additionally, the communities are marked by equally
basic principle of the Spinglass algorithm is that nodes
colored nodes and are also grouped in the form of fictive
belonging to the same community should be connected,
geometric shapes. For instance, Community 1 at the noon
whereas nodes belonging to different communities should
position is represented by the orange nodes in a diamond
not be connected. A node can only be in one community.
shape, Community 2, approximately at the 1 o’clock position,
Thus, the Spinglass algorithm provides communities in
is represented by the brown nodes in a hexagon shape, and
which the nodes belonging to the same communities are
Community 3, approximately at the 2 o’clock position, is
more interconnected with each other than with nodes
represented by the tan nodes ordered in a circle. In general,
belonging to other communities. Because the Spinglass
Figure 1 shows connections between the nodes belonging to a
algorithm is not stable, i.e., it provides different results in
community. However, the figure also shows many connections
different runs, we first calculated the number of resulting
between nodes belonging to different communities.
communities for 100 different runs, each having a different
We provide the interpretation of the communities below,
and unique seed (Briganti et al., 2018). For the 100 runs, we
but first, we look at the stability analysis results. Bootstrap
determined the number of communities detected in each
results are shown in Figure 2. The plot shows estimates of the
run, computed the median for the number of communities,
edges on the horizontal axis and the absolute value of the
and selected this number as the final number of
difference between the estimate and its bootstrap mean on
communities.
the vertical axis. The differences are relatively small for almost
Finally, the networks were plotted with grouped and
all estimates of edge weights. In particular, for the positive edge
color-coded nodes belonging to a community. Nodes are
weights that are greater than .15, the difference is below 0.01.
represented by circles, and edges are represented by lines
For the negative edge weights that are smaller than about −.15,
where blue lines indicate a positive pairwise interaction and
the difference is below 0.01. Another obvious result is the tail
red lines indicate a negative pairwise interaction.
on the right side of the graph, indicating a number of edges
For all computations, we used the R software (R Core
with large positive edge weights that are not mirrored on the
Team, 2020). We used the igraph package (Csardi &
side of the edges with negative edge weights.
Nepusz, 2006) for community detection and the package
Given the results in general, the estimates are accurate. Please
qgraph (Epskamp et al., 2012) to plot the networks. The
note that the “vast area” between the negative and positive edge
networks were estimated and bootstrapped with the bootnet
weight estimates results from the exclusion of these edges due to
package (Epskamp et al., 2018).
them not being statistically significant. In this way, the plot also
shows that the minimal partial correlation in the network is a
Final Data Screening little bit below ± .10 and that there is generally a wider range of
Despite our ordinal approach, we recognized that some of the positive partial correlation than negative partial correlation.
items did not show any reasonable amount of variation. Concerning the stability, the stability coefficients are CS
Therefore, we implemented a final screening procedure before (.50) = .59, CS(.60) = .52, and CS(.70) = .44. Moreover, the
we included the items in the network analysis to ensure a plot indicates that the absolute difference between the estimate
minimum variability in the data. Thus, we set the exclusion and its bootstrap mean is generally greater for smaller partial
criterion that all items must have an interquartile range (IQR) correlation coefficients than for the larger ones. However, the
greater than zero. absolute difference is rather small, marginally exceeding 0.03,
and in sum, the network estimates are stable.
Regarding RQ1 for the male network, we can sum up that
Results
for men, the relation between pornography use, sexual fanta­
Descriptive statistics of all items can be found in Online supple­ sies, and sexual behavior can be modeled in the form of a stable
mentary Table 1 for male and Table 2 for female participants. and accurate network.
6 M. T. P. VON ANDRIAN-WERBURG ET AL.

Figure 1. Network for the male sample with color-coded communities. Blue edges represent positive relations and red edges negative relations. Note. The letters F, R or
P denominate if the respective node (colored dots) is either a fantasy, a behavior (Reality) or a pornographic content followed by a number to denominate a respective
item label displayed in Table 2. Community 1 (C1) is female-centered; C2 is about stimulating objects and places; C3 is about an orgiastic setting, C4 about non-
mainstream coprophilic sexual acts, C5 about a fetish (material or clothing); C6 about a promiscuous and diverse sexuality; C7 about sexual failure; C8 about BDSM; C9
about “vanilla/mainstream” sexual acts; C10 about looking at obscene pictures or films.

Community Interpretation of Men’s Network use is caused by TOPs, not including fitting pornography types
For the male sample, there was indeed a community structure. (e.g., outdoor).
An interpretation table containing the communities, abbrevia­ Community 3 (approx. 2 o’clock position, tan nodes in a
tions, and labels of the items can be found in Table 2. The table circle shape) is about an orgiastic setting with mate swapping,
indicates the community number in the first column. The group sex, watching others have sex, homo- and bisexual
following columns contain the item number depicted in the activities, bukkake, and cumshots. These labels compose
nodes and the meaning of the individual item scenes of intercourse during an orgy. Of the matching beha­
Community 1 (noon position, orange nodes in a diamond vior, four of the five items were excluded in the final data
shape) of the male network is maturity-centered with fantasies screening. Therefore, most men seem not to act out on such
and pornography use relating to female sex partners’ post- behavior. It should be mentioned that R12 (watching others
juvenile age and appearance. The pornography nodes physiog­ have sex) is attached to community 6 (promiscuous and
nomically form an overweight mature female body with big diverse sexuality).
breasts. Matching behavior was excluded in the final data Community 4 (approx. 4 o’clock position, green nodes
screening. in a circle shape) is about non-mainstream coprophilic sex­
Community 2 (approx. 1 o’clock position, brown nodes in a ual acts, including pornography use about fisting, anal sex,
hexagon shape) is about stimulating objects and places like and golden showers/enemas. Men whose questionnaire
outside a bed or using objects to possibly further increase reports fit this community tended to expose themselves
sexual pleasure. Subsuming the items, one would imagine a provocatively in fantasy and behavior. Behavior that
vivid sex scene at a beach or in a house outside the bedroom. matched the fantasy of being seduced as an “innocent”
For this community, we suspect the absence of pornography was screened out prior to analysis.
 THE JOURNAL OF SEX RESEARCH 7

Table 2. Interpretation table of the male network including communities, node abbreviations and item-labels.
Community Node Item-Label
1 F31 Having sex with someone much older than yourself
1 P14 Big breasts
1 P16 Fat girls (including “BBW”)
1 P20 Mature/MILF
2 F1 Making love out of doors in a romantic setting e.g., field of flowers, beach at night
2 F18 Making love elsewhere than bedroom. (e.g., kitchen, bathroom)
2 F37 Using objects for stimulation (e.g., vibrators, candles)
2 R1 Making love out of doors in a romantic setting e.g., field of flowers, beach at night
2 R18 Making love elsewhere than bedroom. (e.g., kitchen, bathroom)
2 R37 Using objects for stimulation (e.g., vibrators, candles)
3 F12 Watching others have sex
3 F22 Mate-swapping
3 F5 Sex with two other people
3 F6 Participating in an orgy
3 F9 Homosexual activity
3 P1 Amateur
3 P12 Gang bang (one woman and three or more men)
3 P15 Large penises
3 P21 Oral sex
3 P22 Orgy (≥ 2 persons of each gender)
3 P26 Vaginal sex
3 P4 Bisexual
3 P7 Bukkake
3 P8 Cumshot
3 P9 Threesomes
4 F27 Exposing yourself provocatively
4 F33 Being seduced as an “innocent”
4 P10 Fist fucking
4 P17 Lesbian
4 P19 Masturbation (including sex toys)
4 P2 Anal sex
4 P25 Softcore (nonexplicit sex)
4 P3 Golden showers and enemas
4 R27 Exposing yourself provocatively
5 F19 Being excited by material or clothing (e.g., rubber, leather, underwear)
5 R19 Being excited by material or clothing (e.g., rubber, leather, underwear)
6 F29 Being promiscuous
6 F3 Intercourse with someone you know but have not had sex with
6 F30 Having sex with someone much younger than yourself
6 F34 Seducing an “innocent”
6 F36 Having sex with someone of different race
6 F4 Intercourse with an anonymous stranger
6 P18 Lolita/teen
6 R12 Watching others have sex
6 R29 Being promiscuous
6 R3 Intercourse with someone you know but have not had sex with
6 R30 Having sex with someone much younger than yourself
6 R36 Having sex with someone of different race
6 R4 Intercourse with an anonymous stranger
7 F35 Being embarrassed by failure of sexual performance
7 R35 Being embarrassed by failure of sexual performance
8 F14 Whipping or spanking someone
8 F20 Hurting a partner
8 F24 Being tied up
8 F25 Tying someone up
8 F7 Being forced to do something
8 F8 Forcing someone to do something
8 P11 Fetish (including latex)
8 P13 Violent sex (simulated rape, aggression, and coercion)
8 P23 Sadomasochism
8 P5 Bizarre/extreme
8 P6 Bondage and dominance
8 R24 Being tied up
8 R25 Tying someone up
9 F10 Receiving oral sex
9 F11 Giving oral sex
9 F16 Taking someone’s clothes off
9 F17 Having your clothes taken off
9 F2 Having intercourse with a loved partner
9 F32 Being much sought after by the opposite sex
9 F38 Being masturbated to orgasm by a partner
9 F40 Kissing passionately
9 R10 Receiving oral sex
9 R11 Giving oral sex
(Continued)
8 M. T. P. VON ANDRIAN-WERBURG ET AL.

Table 2. (Continued).
Community Node Item-Label
9 R16 Taking someone’s clothes off
9 R17 Having your clothes taken off
9 R2 Having intercourse with a loved partner
9 R32 Being much sought after by the opposite sex
9 R38 Being masturbated to orgasm by a partner
9 R40 Kissing passionately
10 F39 Looking at obscene pictures or films
10 R39 Looking at obscene pictures or films

Figure 2. Bootstrap results for accuracy estimation of the male network. The x-axis displays the size of the estimated partial correlation. The y-axis displays the estimates
deviation from the bootstrap mean (m = 5000). Note. The differences between the estimates of the edges (x-axis) and the absolute value of the difference between the
estimate and its bootstrap mean (y-axis) is rather small. Thus, the estimates are accurate.

Community 5 (approx. 5 o’clock position, only two light- interpret this community as related to fears and experiences of
green nodes in a bar shape) represents the fetish of “being failing during intercourse.
excited by material or clothing” with the matching fantasy Community 8 (approx. 8 o’clock position, blue-gray nodes
and behavior nodes. We assessed potentially fitting pornogra­ in a circle) is about BDSM. The fantasies (whipping or spank­
phy use with P11 (Fetish [including latex]). However, this node ing someone, hurting a partner) and pornography use (violent
is attached to community 8 (BDSM). sex [simulated rape, aggression, and coercion]) appear to be
Community 6 (6 o’clock position, turquoise-green nodes in more extreme compared to actual behavior reported, such as
a circle) is about promiscuous and diverse sexuality. It seems to being tied up or tying someone up. The remaining behaviors
reflect the desire and enactment of promiscuous sex with that match those fantasies were screened out prior to analysis.
younger partners who are pictured as inexperienced and part­ Community 9 (approx. 10 o’clock position, violet nodes in a
ners of varying physical appearance. Given the appearance of circle) appears to be men’s “vanilla” community with nodes
R12 in this community, it appears that watching others have about sexual fantasies like receiving oral sex, taking someone’s
sex (R12) creates a stronger attachment through its imperson­ clothes off, or having intercourse with a loved partner. The
ality compared to its “natural” cooccurrence when participat­ fantasy nodes are perfectly reflected by their eight behavioral
ing in an orgy (community 3). counterparts. It is worth mentioning that assessed pornogra­
Community 7 (approx. 7 o’clock position, only two blue phy use was not associated with these “every day” sexual
nodes in a bar shape) is about sexual failure containing two behaviors and fantasies. Potentially matching pornography
matching fantasy and behavior nodes labeled “being embar­ use (e.g., P1: Amateur, P21: Oral Sex, P26: Vaginal Sex) was
rassed by failure of sexual performance.” In our sample, we included in the network but attached to different communities.
 THE JOURNAL OF SEX RESEARCH 9

Community 10 (approx. 11 o’clock position, only two rose a community but also edges between the nodes belonging to
nodes in a bar shape) is about looking at obscene pictures or different communities. The results of the bootstrap are
films. Therefore, it assesses the frequency of pornography use. It shown in Figure 4. The absolute values of the difference
should be mentioned that the community does not include any between the estimate and the bootstrap means are smaller
of the TOP nodes. for the female network than the male network, indicating an
Regarding RQ2 for the male network, we can state that it even higher accuracy. Like in the male network, the differ­
was possible to detect meaningful communities of reciprocally ences for the edge weights larger than 0.20 have a minimal
interacting sexual fantasies, sexual behavior, and pornography difference not exceeding 0.005. Additionally, the same pat­
use of the same style of sexuality. However, one should bear in tern of a set of positive edges with a distinct small difference
mind that due to the exclusion of some items, not all corre­ also emerged in the female network, i.e., the tail on the right
sponding sexual fantasies, behaviors, and types of pornography side. Therefore, the estimated edge weights for this network
have corresponding elements in the communities. are accurate, too. Concerning the stability, the stability coef­
ficients are CS(.50) = .67, CS(.60) = .59, and CS(.70) = .44. As
for the male network, the estimates for the female network
Network for Females
seem stable.
The network for the female sample is shown in Figure 3. Comparable to the male sample, females’ sexual fantasies,
There are 8 Spinglass communities in the network. The pornography use, and sexual behavior result in an accurate and
structure of the figure is comparable to Figure 1. As in the stable network so that we can positively answer RQ1 for the
male network, there are not only edges between the nodes of female sample.

Figure 3. Network for the female sample with color-coded communities. Blue edges represent positive relations and red edges negative relations. Note. The letters F, R
or P denominate if the respective node (colored dots) is either a fantasy, a behavior (Reality) or a pornographic content followed by a number to denominate a
respective item label displayed in Table 3. Community 1 (C1) is about the fetish of being excited by material or clothing; C2 is about orgasm focused “vanilla/
mainstream” intercourse; C3 is about BDSM, C4 is about penetration centered group sex (pornography nodes only), C5 about passionate outdoor sex; C6 is about
masturbating with objects while watching porn; C7 about being desired in post juvenile age; C8 is about an orgiastic and impersonal sexuality.
10 M. T. P. VON ANDRIAN-WERBURG ET AL.

Figure 4. Bootstrap results for accuracy estimation of the female network. The x-axis displays the size of the estimated partial correlation. The y-axis displays the
estimates deviation from the bootstrap mean (m = 5000). Note. The differences between the estimates of the edges (x-axis) and the absolute value of the difference
between the estimate and its bootstrap mean (y-axis) is rather small. Thus, the estimates are accurate.

Community Interpretation of the Women’s Network Community 4 (approx. 5 o’clock position, light green nodes on
We now turn to the communities of the female network and a circle) is a community exclusively consisting of pornography
their interpretation. An interpretation table of their eight com­ nodes that center on the interaction around penetration-centered
munities can be found in Table 3. group sex with TOP labels like orgy, large penises, gang bang, and
Community 1 (noon position, only two orange nodes in a bukkake. Potentially matching fantasies (e.g., giving oral sex) and
bar shape) of the female sample consists of two nodes repre­ behavior were assessed but did not attach to this community.
senting the Fetish of being excited by material or clothing in Community 5 (6 o’clock position, turquoise nodes in a
fantasy and behavior. Comparable with community 5 of the heptagon shape) reflects the desire to have passionate outdoor
male sample (Fetish), none of the pornography use assessed sex consisting of sexual fantasy nodes only about making love
was related to the community, even though suitable types outside the bedroom or outdoors and having intercourse with a
remained in the analysis (e.g., P11 Fetish (including latex)). loved partner. Potentially matching pornography use was, as
Community 2 (approx. 1 o’clock position, brown nodes in a for the men, not part of the TOP items.
circle) centers around orgasm-focused “vanilla” intercourse. It is Community 6 (approx. 6 o’clock position, blue nodes in a
worth mentioning that women reported having fewer fantasies diamond shape) is masturbation centered revolving around
regarding this community compared to the male sample (com­ masturbation with objects while looking at obscene pictures
munity 9). Women reported being embarrassed by the failure or films. The community assesses (parts of) the frequency of
of sexual performance, which in our interpretation, reflects the pornography consumption and females’ masturbation beha­
pressure to orgasm during intercourse. Again, potentially vior. Interestingly, no specific TOP items are attached to the
proper pornography use was part of the analysis but did not community, comparable to the male sample.
attach to the community. It subsumes in community 4 (pene­ We interpret community 7 (9 o’clock position, violet nodes
tration-centered group sex). in a diamond shape) to be centered around our participants’
Community 3 (3 o’clock position, green nodes in a circle) is wish of being desired in the post-juvenile age. In the interaction
centered around BDSM. A rough interplay of pornography use of all nodes, the community represents a mature woman with
(e.g., violent sex, sadomasochism, bondage, and dominance) big breasts that is attractive and much sought after by men.
accompanies the fantasies of “being tied up” and “tying some­ Community 8 (approx. 11 o’clock position, rose nodes in a
one up.” Interestingly, only the active action of “tying someone circle) is about orgiastic and impersonal sexuality and centers
up” is part of this community as the only behavior. The passive around fleeting encounters and orgies with various sex part­
action of “being tied up” was excluded in the final data ners. What is noteworthy is that the fantasies seem much more
screening. diverse than the actual behavior.
 THE JOURNAL OF SEX RESEARCH 11

Table 3. Interpretation table of the female network including communities, node abbreviations and item-labels.
Community Node Item-Label
1 F19 Being excited by material or clothing (e.g., rubber, leather, underwear)
1 R19 Being excited by material or clothing (e.g., rubber, leather, underwear)
2 F10 Receiving oral sex
2 F38 Being masturbated to orgasm by a partner
2 R10 Receiving oral sex
2 R11 Giving oral sex
2 R16 Taking someone’s clothes off
2 R17 Having your clothes taken off
2 R2 Having intercourse with a loved partner
2 R32 Being much sought after by the opposite sex
2 R35 Being embarrassed by failure of sexual performance
2 R38 Being masturbated to orgasm by a partner
2 R40 Kissing passionately
3 F24 Being tied up
3 F25 Tying someone up
3 P11 Fetish (including latex)
3 P13 Violent sex (simulated rape, aggression, and coercion)
3 P17 Lesbian
3 P19 Masturbation (including sex toys)
3 P23 Sadomasochism
3 P25 Softcore (nonexplicit sex)
3 P4 Bisexual
3 P5 Bizarre/extreme
3 P6 Bondage and dominance
3 R25 Tying someone up
4 P1 Amateur
4 P12 Gang bang (one woman and three or more men)
4 P15 Large penises
4 P18 Lolita/teen
4 P2 Anal sex
4 P21 Oral sex
4 P22 Orgy (≥ 2 persons of each gender)
4 P7 Bukkake
4 P8 Cumshot
5 F1 Making love out of doors in a romantic setting e.g., field of flowers, beach at night
5 F11 Giving oral sex
5 F16 Taking someone’s clothes off
5 F17 Having your clothes taken off
5 F18 Making love elsewhere than bedroom (e.g., kitchen, bathroom)
5 F2 Having intercourse with a loved partner
5 F40 Kissing passionately
6 F37 Using objects for stimulation (e.g., vibrators, candles)
6 F39 Looking at obscene pictures or films
6 R37 Using objects for stimulation (e.g., vibrators, candles)
6 R39 Looking at obscene pictures or films
7 F31 Having sex with someone much older than yourself
7 F32 Being much sought after by the opposite sex
7 P14 Big breasts
7 P20 MILF/mature
8 F12 Watching others have sex
8 F22 Mate-swapping
8 F29 Being promiscuous
8 F3 Intercourse with someone you know but have not had sex with
8 F30 Having sex with someone much younger than yourself
8 F34 Seducing an “innocent”
8 F36 Having sex with someone of different race
8 F4 Intercourse with an anonymous stranger
8 F5 Sex with two other people
8 F6 Participating in an orgy
8 R1 Making love out of doors in a romantic setting e.g., field of flowers, beach at night
8 R18 Making love elsewhere than bedroom (e.g., kitchen, bathroom)
8 R3 Intercourse with someone you know but have not had sex with

Regarding RQ2 for the female network, we can state the sexuality. Again, one should bear in mind that due to the
same as for the male network: It was possible to detect mean­ exclusion of some items, not all corresponding sexual fantasies,
ingful communities of reciprocally interacting sexual fantasies, behaviors, and types of pornography have corresponding ele­
sexual behavior, and pornography use of the same style of ments in the communities.
12 M. T. P. VON ANDRIAN-WERBURG ET AL.

Discussion (Ohler & Nieding, 2005). Their reciprocal interactions lead to

a structured, complex web of drafts of sexual scripts that is
We asked if pornography use, sexual behavior, and fantasies
neither completely random nor predetermined in fixed action
could be modeled in a network (RQ 1). Additionally, we asked
patterns.
if communities of reciprocally interacting sexual fantasies,
For men and women, the communities we labeled “vanilla”
sexual behavior, and pornography use exist in these networks
for men (community 9 in Table 2) or orgasm-centered
(RQ 2).
“vanilla” intercourse for women (community 2 in Table 3)
Regarding RQ1, networks exist that subsume the assessed
did not include any pornography items at all. Additionally,
sexuality in our sample. In line with previous studies (e.g., Hald
the more extreme forms of sexual behavior (e.g., forms of
& Štulhofer, 2016; Martyniuk & Dekker, 2018; Price et al.,
violent sex) remained exclusively in the domain of fantasy
2016), men and women are different in terms of the frequency
and pornography use without any large-scale occurrence of a
of pornography use but also differ in the content of their
(reported) behavior (for instance, community 8 in Table 2). We
networks. The network’s heterogeneity indicates sex/gender-
think these results align with our assumption that pornography
specific differences in how sexuality is lived. These differences
use affects sexual fantasies but not behavior to the same extent.
are not only caused by different behavioral or usage frequen­
The fact that sexual fantasies tend to be more extreme and
cies (e.g., Petersen & Hyde, 2010) but account for qualitative
volatile has been shown in previous research (Joyal et al., 2015;
differences in the sexuality of men and women. In this study,
Lehmiller, 2018; K. M. Williams et al., 2009).
we did not align toward a cultural (e.g., Eagly & Wood, 2012)
Given our theoretical considerations, the cross-sectional
or evolutionary (e.g., Buss & Schmitt, 1993) explanation
nature of our data only allows for a brief snapshot of our
approach for these differences and made no predictions
assumptions regarding the 3AM, RSM, and CMVC (Ferguson
regarding them. However, a key message from the network’s
et al., 2008; Slater, 2015; Slater et al., 2020; Wright, 2011, 2014).
heterogeneity might be that univariate differences in sexuality
The networks show clear interactions. Still, much stronger
might often be small (Petersen & Hyde, 2010), but these single
longitudinal data would be necessary to test the assumption
differences affect a larger behavioral structure which causes, in
that pornography use plays a role in a reinforcing spiral process
its sum, completely different outcomes. Therefore, a more
and that interaction occurs over time between sexual fantasies,
multivariate perspective regarding sex/gender should be con­
pornography use, and sexual behavior. However, the reported
sidered in future research about sexuality (e.g., Del Giudice,
pornography use very often matched up with the fantasies and
2022).
behavior in our detected communities well, which is in line
Related to RQ 2, men had ten communities of matching
with the 3AM and the RSM. From a mere 3AM perspective, one
sexual fantasies, behavior, and pornography use, and women
could interpret a rumination of pornography use and fantasies.
had eight communities. These communities are established
Regarding the RSM, media use complements a user’s behavior,
along all three, two, or even one aspect of sexuality.
beliefs, and attitudes. Including the CMVC, we see “escapist”
Pornography use can be an integral part of these communities,
handling of fictional media messages or fantasies. These find­
together with behaviors and fantasies, but it does not have to be
ings contradict a mandatory behavior-causing effect, which is
like in women’s community 4 in Table 3. It centers only around
in line with the basic assumptions of the CMCV (Ferguson
pornography use unrelated to fantasies or behavior. We
et al., 2008). Furthermore, the variables in the network affect
advanced Hald and Štulhofer’s (2016) results. The authors
each other as we expected.
ran an EFA on pornography use items to see which types of
Having mapped out crucial parts of the sexuality of a large
pornography did cluster. Our approach also included fantasy
and diverse German sample, the result that appears proble­
and behavior items and used a superior method compared to
matic to us from a health policy point of view is that many
an exploratory factor analysis, which appears unsuitable for the
women seem to put themselves under pressure to orgasm
often non-normal distributions of pornography use items.
during intercourse. If they do not, this pressure may accom­
The communities are strongly interrelated. This high degree
pany a related feeling of shame (see also: Lavie-Ajayi & Joffe,
of community interaction is reflected in many connections
2009). This finding made us pause and think about problematic
between the nodes belonging to one community. However,
sexual scripts (Gagnon & Simon, 2017) and how strongly some
our analysis shows many connections between nodes belong­
participants seem to have internalized these. However, not a
ing to different communities. These interactions between
single assessed variable about pornography use was related to
nodes of different communities demonstrate that many
this phenomenon.
detected patterns show interactions with other communities
of sexual fantasies, pornography usage reports, and behaviors.
This interconnectedness may be caused by basic mental struc­
Limitations
tures (motives, emotions, instincts) that partly influence these
relations between communities or by cultivated narrative We tried to obtain the best grasp on our arguments with the
structures staged in pornographic content (e.g., reciprocal data available. This study was exploratory by nature, but we are
interaction between nodes of fetish and BDSM). Since we the first to use a network approach to map the relations
assume that pornographic content is the externalization of between human sexual fantasies, behavior, and pornography
sexual fantasies in humans, it is most plausible that commu­ use. Still, it was limited in evaluating the actual theoretical
nities’ interconnections result from human mental nature and reasoning. For instance, the Sexual Behavior Sequence
(co-evolved) cultivated pornographic narrative structures (Byrne, 1976) would describe pornography use as some vent
 THE JOURNAL OF SEX RESEARCH 13

where unfulfilled sexual desires are lived out only in (externa­ mainstream sexuality. The presented analysis can only be the
lized) thought. Before data collection, we did not assume por­ first step toward a methodologically sound cartography of the
nography could be used this way. However, if pornography use interactions between pornography use and human sexual fan­
would indeed be a vent for unfulfilled desires, we would tasy and behavior. In future research, longitudinal data are
assume that there should be close to no matching behavior urgently needed, ideally without the possibility of a social
items in our communities where pornography use is included, desirability bias (e.g., longitudinal web-tracking data).
which is not the case. Still, it is arguable if assessed pornogra­ Furthermore, the next step should include the assessment of
phy use and behavior items matched indeed enough in parti­ vulnerabilities (Ingram & Luxton, 2005) that might be inter­
cipants’ interpretation to rule out this conclusion. twined with the adverse effects of pornography use.
Because we surveyed the use of pornographic movies only, Despite these limitations, we think a good starting point is
our results cannot be applied to other distribution channels for detailed cartography when the aim is to sail through tricky
pornography. This exclusive focus is a limitation of scope and waters. The explored landscape of communities can be read as
constrains the transferability of our results to video-based empirical sedimentation of one’s sexual identity produced by
pornography only. self-reinforcing spirals (RSM; Slater, 2015; Slater et al., 2020),
Another limitation of this study was the lack of assessment like footsteps in the sand, as a snapshot of rumination pro­
of covariates for pornography use, for instance, sexual desire or cesses as described by Wright (2011, 2014) or as a cue about
sexual sensation seeking (e.g., Esplin et al., 2021), whose poten­ potential (unassessed) mediator variables as suggested by
tial influence on the networks should be put greater emphasis Ferguson et al. (2008). Moreover, hypothesis-driven research
on in future research. on the matter is urgently needed.
Furthermore, the use of a convenience sample is proble­
matic. Though our participants had a comparably diverse
educational background, our data had limitations in general­ Acknowledgments
izability. Moreover, the questionnaire method is prone to a We would like to thank our colleague Dr. Michael Brill for critically
social desirability bias, especially when applied to porn reading the manuscript’s draft and for his well-reasoned comments and
research (Kohut et al., 2020; Rasmussen et al., 2018; suggestions.
Willoughby & Busby, 2016). However, recent findings of von
Andrian-Werburg et al. (2022) suggest that major findings of
pornography research can be replicated with web tracking data. Disclosure Statement
At least in Germany, social desirability appears not to be a No potential conflict of interest was reported by the authors.
considerable issue in questionnaire-based research about sexu­
ality. Our results show that “socially undesirable” topics like
being ashamed due to a (self-perceived) lack of sexual func­ ORCID
tioning because of not having an orgasm during intercourse Maximilian T. P. von Andrian-Werburg http://orcid.org/0000-0001-
appeared in the female network. Furthermore, comparing 8713-8237
men’s and women’s networks showed that the communities Eric Klopp http://orcid.org/0000-0002-6638-8912
matched each other well. Frank Schwab http://orcid.org/0000-0001-5488-4321
Despite our efforts to control for skewed distributions and
the ordinal nature of the TOP and SFQ data (Hald & Štulhofer, References
2016; Wilson, 1988, 2010), methodological biases remain. The
items used to measure fantasy and behavior were assessed with BBC. (2020, May 18). New data shows BBC is the world’s most visited news
the same labels. This considerable similarity may cause some site. BBC Media Centre. https://www.bbc.co.uk/mediacentre/world
news/2020/worlds-most-visited-news-site
artificial relations. Furthermore, we cannot assess the whole Borsboom, D., & Cramer, A. O. (2013). Network analysis: An integrative
variety of human sexuality and therefore cannot rule out that approach to the structure of psychopathology. Annual Review of
some unassessed pornographic content or sexual fantasy and Clinical Psychology, 9(1), 91–121. https://doi.org/10.1146/annurev-
behavior might exist that would have been detrimental to our clinpsy-050212-185608
networks. However, many items in both the SFQ and TOP Briganti, G., Kempenaers, C., Braun, S., Fried, E. I., & Linkowski, P.
(2018). Network analysis of empathy items from the interpersonal
already yielded extremely skewed distributions because their reactivity index in 1973 young adults. Psychiatry Research, 265, 87–
labels applied only to very few participants. 92. https://doi.org/10.1016/j.psychres.2018.03.082
Buss, D. M., & Schmitt, D. P. (1993). Sexual strategies theory: An evolu­
tionary perspective on human mating. Psychological Review, 100(2),
Study Implications 204–232. https://doi.org/10.1037/0033-295X.100.2.204
Byrne, D. (1976). Social psychology and the study of sexual behavior.
In sum, our networks provide a snapshot that allows the Personality and Social Psychology Bulletin, 3(1), 3–30. https://doi.org/
assumption that pornography use is in reciprocal relations 10.1177/014616727600300102
with a user’s fantasies and behavior. Different communities Cohen, J. (1988). Statistical power analysis for the behavioral sciences.
are detectable that seem to be in line with common assump­ Erlbaum.
tions about ways to engage in sexuality (e.g., Vanilla, BDSM, Costantini, G., Richetin, J., Preti, E., Casini, E., Epskamp, S., & Perugini,
M. (2019). Stability and variability of personality networks: A tutorial
Fetishism). Pornography use plays a role in some of these on recent developments in network psychometrics. Personality and
communities. However, the assessed types more strongly affect Individual Differences, 136(1), 68–78. https://doi.org/10.1016/j.paid.
the more “uncommon” communities and are no part of 2017.06.011
14 M. T. P. VON ANDRIAN-WERBURG ET AL.

Csardi, G., & Nepusz, T. (2006). The igraph software package for complex on poor measurement foundations. The Journal of Sex Research, 57(6),
network research. InterJournal, Complex Systems, 1695(5), 1–9. https:// 722–742. https://doi.org/10.1080/00224499.2019.1695244
igraph.org Lavie-Ajayi, M., & Joffe, H. (2009). Social representations of female
Del Giudice, M. (2022). Measuring sex differences and similarities. In D. orgasm. Journal of Health Psychology, 14(1), 98–107. https://doi.org/
P. VanderLaan & W. I. Wong (Eds.), Gender and sexuality develop­ 10.1177/1359105308097950
ment: Contemporary theory and research (pp. 1–38). Springer. https:// Lehmiller, J. J. (2017). The psychology of human sexuality. Wiley Blackwell.
doi.org/10.1007/978-3-030-84273-4_1 Lehmiller, J. J. (2018). Tell me what you want. Da Capo Press.
Eagly, A. H., & Wood, W. (2012). Social role theory. In P. Van Lange, A. Lippa, R. A. (2009). Sex differences in sex drive, sociosexuality, and height
Kruglanski, & E. Higgins (Eds.), Handbook of theories of social psychol­ across 53 nations: Testing evolutionary and social structural theories.
ogy (pp. 458–476). Sage Publications Ltd. https://doi.org/10.4135/ Archives of Sexual Behavior, 38(5), 631–651. https://doi.org/10.1007/
9781446249222.n49 s10508-007-9242-8
Epskamp, S., Borsboom, D., & Fried, E. (2018). Estimating psychological Martyniuk, U., & Dekker, A. (2018). Adult pornography use in Germany:
networks and their accuracy: A tutorial paper. Behavior Research Results of a pilot study. Zeitschrift für Sexualforschung, 31(3), 237–249.
Methods, 50(1), 195–212. https://doi.org/10.3758/s13428-017-0862-1 https://doi.org/10.1055/a-0664-4441
Epskamp, S., Cramer, A. O. J., Waldorp, L. J., Schmittmann, V. D., & Newman, M. (2018). Networks: An introduction. Oxford University Press.
Borsboom, D. (2012). Qgraph: Network visualizations of relationships Ohler, P., & Nieding, G. (2005). Sexual selection, evolution of play and
in psychometric data. Journal of Statistical Software, 48(4), 1–18. entertainment. Journal of Cultural and Evolutionary Psychology, 3(2),
https://doi.org/10.18637/jss.v048.i04 141–157. https://doi.org/10.1556/jcep.3.2005.2.3
Epskamp, S., & Fried, E. (2018). A tutorial on regularized partial correla­ Petersen, J. L., & Hyde, J. S. (2010). A meta-analytic review of research on
tion networks. Psychological Methods, 23(4), 617–634. https://doi.org/ gender differences in sexuality, 1993–2007. Psychological Bulletin, 136
10.1037/met0000167 (1), 21–38. https://doi.org/10.1037/a0017504
Epskamp, S., Rhemtulla, M., & Borsboom, D. (2017). Generalized network Pornhub Insights. (2021, April 8). The pornhub tech review. Pornhub
psychometrics: Combining network and latent variable models. Insights. https://www.pornhub.com/insights/tech-review
Psychometrika, 82(4), 904–927. https://doi.org/10.1007/s11336-017- Price, J., Patterson, R., Regnerus, M., & Walley, J. (2016). How much more
9557-x XXX is generation X consuming? Evidence of changing attitudes and
Esplin, C. R., Hatch, H. D., Hatch, S. G., Braithwaite, S. R., & Deichman, C. behaviors related to pornography since 1973. The Journal of Sex
L. (2021). How are pornography and sexual sensation seeking related? Research, 53(1), 12–20. https://doi.org/10.1080/00224499.2014.
Sexual and Relationship Therapy, 1–11. https://doi.org/10.1080/ 1003773
14681994.2021.1937977 Rasmussen, K. R., Grubbs, J. B., Pargament, K. I., & Exline, J. J. (2018).
Ferguson, C. J., Rueda, S. M., Cruz, A. M., Ferguson, D. E., Fritz, S., & Social desirability bias in pornography-related self-reports: The role of
Smith, S. M. (2008). Violent video games and aggression: Causal religion. The Journal of Sex Research, 55(3), 381–394. https://doi.org/
relationship or byproduct of family violence and intrinsic violence 10.1080/00224499.2017.1399196
motivation? Criminal Justice and Behavior, 35(3), 311–332. https:// R Core Team. (2020). R: A language and environment for statistical
doi.org/10.1177/0093854807311719 computing. R Foundation for Statistical Computing. https://www.R-
Fisher, W. A., & Barak, A. (2001). Internet pornography: A social psycho­ project.org/
logical perspective on internet sexuality. The Journal of Sex Research, 38 Rodriguez, M., Aalbers, G., & McNally, R. J. (2022). Idiographic network
(4), 312–323. https://doi.org/10.1080/00224490109552102 models of social media use and depression symptoms. Cognitive
Fortunato, S. (2010). Community detection in graphs. Physics Reports, 486 Therapy and Research, 46(1), 124–132. https://doi.org/10.1007/
(3–5), 75–174. https://doi.org/10.1016/j.physrep.2009.11.002 s10608-021-10236-2
Gagnon, J. H., & Simon, W. (2017). Sexual conduct. The social sources of Schmidt, R. A., & Voss, B. L. (2000). Archaeologies of sexuality: An
human sexuality. Routledge. introduction. Routledge.
Greaves, L. M., Barlow, F. K., Lee, C. H., Matika, C. M., Wang, W., Shokeir, A. A., & Hussein, M. I. (2004). Sexual life in pharaonic Egypt:
Lindsay, C.-J., Case, C. J., Sengupta, N. K., Huang, Y., Cowie, L. J., Towards a urological view. International Journal of Impotence Research,
Stronge, S., Storey, M., De Souza, L., Manuela, S., Hammond, M. D., 16(5), 385–388. https://doi.org/10.1038/sj.ijir.3901195
Milojev, P., Townrow, C. S., Muriwai, E., Satherley, N., . . . Sibley, C. G. Slater, M. D. (2007). Reinforcing spirals: The mutual influence of media
(2017). The diversity and prevalence of sexual orientation self-labels in selectivity and media effects and their impact on individual behavior
a New Zealand national sample. Archives of Sexual Behavior, 46(5), and social identity. Communication Theory, 17(3), 281–303. https://doi.
1325–1336. https://doi.org/10.1007/s10508-016-0857-5 org/10.1111/j.1468-2885.2007.00296.x
Grubbs, J. B., & Kraus, S. W. (2021). Pornography use and psycholo­ Slater, M. D. (2015). Reinforcing spirals model. Conceptualizing the
gical science. A call for consideration. Current Directions in relationship between media content exposure and the development
Psychological Science, 30(1), 68–75. https://doi.org/10.1177/ and maintenance of attitudes. Media Psychology, 18(3), 370–395.
0963721420979594 https://doi.org/10.1080/15213269.2014.897236
Hald, G. M., & Mulya, T. W. (2013). Pornography consumption and non- Slater, M. D., Shehata, A., & Strömbäck, J. (2020). Reinforcing spirals
marital sexual behaviour in a sample of young Indonesian university model. In J. van den Bulck, D. Ewoldsen, M.-L. Mares, & E. Scharrer
students. Culture, Health & Sexuality, 15(8), 981–996. https://doi.org/ (Eds.), The international encyclopedia of media psychology (pp. 1–11).
10.1080/13691058.2013.802013 John Wiley & Sons. https://doi.org/10.1002/9781119011071.iemp0134
Hald, G. M., & Štulhofer, A. (2016). What types of pornography do people von Andrian-Werburg, M. T. P., Siegers, P., & Breuer, J. (2022). A
use and do they cluster? Assessing types and categories of pornography reevaluation of online pornography use in Germany using a combination
consumption in a large-scale online sample. The Journal of Sex of web tracking and survey data. PsyArxiv. https://doi.org/10.31234/osf.
Research, 53(7), 849–859. https://doi.org/10.1080/00224499.2015. io/ehqgv
1065953 Williams, K. M., Cooper, B. S., Howell, T. M., Yuille, J. C., & Paulhus, D. L.
Ingram, R. E., & Luxton, D. D. (2005). Vulnerability-stress models. In B. L. (2009). Inferring sexually deviant behavior from corresponding fanta­
Hankin & J. R. Z. Abela (Eds.), Development of psychopathology: A sies: The role of personality and pornography consumption. Criminal
vulnerability-stress perspective (pp. 32–46). Sage. Justice and Behavior, 36(2), 198–222. https://doi.org/10.1177/
Joyal, C. C., Cossette, A., & Lapierre, V. (2015). What exactly is an unusual 0093854808327277
sexual fantasy? Journal of Sexual Medicine, 12(2), 328–340. https://doi. Williams, D. R., Rhemtulla, M., Wysocki, A. C., & Rast, P. (2019). On
org/10.1111/jsm.12734 nonregularized estimation of psychological networks. Multivariate
Kohut, T., Balzarini, R. N., Fisher, W. A., Grubbs, J. B., Campbell, L., & Behavioral Research, 54(5), 719–750. https://doi.org/10.1080/
Prause, N. (2020). Surveying pornography use: A shaky science resting 00273171.2019.1575716
 THE JOURNAL OF SEX RESEARCH 15

Willoughby, B. J., & Busby, D. M. (2016). In the eye of the beholder: Exploring Wright, P. J. (2014). Pornography and the sexual socialization of chil­
variations in the perceptions of pornography. The Journal of Sex Research, 53 dren: Current knowledge and a theoretical future. Journal of Children
(6), 678–688. https://doi.org/10.1080/00224499.2015.1013601 and Media, 8(3), 305–312. https://doi.org/10.1080/17482798.2014.
Wilson, G. D. (1988). Measurement of sex fantasy. Sexual and Marital 923606
Therapy, 3(1), 45–55. https://doi.org/10.1080/02674658808407692 Wright, P. J., Paul, B., & Herbenick, D. (2021). Preliminary insights from
Wilson, G. D. (2010). The sex fantasy questionnaire: An update. Sexual a U.S. probability sample on adolescents’ pornography exposure,
and Relationship Therapy, 25(1), 68–72. https://doi.org/10.1080/ media psychology, and sexual aggression. Journal of Health
14681990903505799 Communication, 26(1), 39–46. https://doi.org/10.1080/10810730.
Wright, P. J. (2011). Mass media effects on youth sexual behavior assessing 2021.1887980
the claim for causality. Annals of the International Communication Yang, Z., Algesheimer, R., & Tessone, C. (2016). A comparative analysis of
Association, 35(1), 343–385. https://doi.org/10.1080/23808985.2011. community detection algorithms on artificial networks. Scientific
11679121 Reports, 6(1), 30750. https://doi.org/10.1038/srep30750