A content analysis of pornographic images on the Internet
Michael D. Mehta, Ph.D. and Dwaine E. Plaza, Ph.D.
Queen's University and Oxford University
Original study presented October 1994
Reference:  Mehta, Michael D. and Dwaine E. Plaza (1997)
"Content analysis of pornographic images available on
the Internet."  The Information Society, 13(2):  153-162
* Please note - Tables are not included in this electronic version.
Abstract
We examined the nature and content of 150 randomly selected
pornographic images available through newsgroups located on
the Internet computer network.  Using content analysis, we
identified themes that appear most frequently and explored
differences in the type of material posted by commercial and
non-commercial users.  Results suggest that commercial
vendors are more likely to post explicit pornographic
material in public access newsgroups to attract new
customers to their private, pay-per-use bulletin board
services.
Key words: computer, pornography, Internet, content analysis.
 ========================
Introduction:
  Although there have been content analyses of sexually
explicit magazines (Winick, 1985), pornographic comics
(Scott & Cuvelier, 1993;  Palmer, 1979), soft core magazines
(Malamuth & Spinner, 1980), and erotic videos (Garcia &
Milano, 1990;  Palys, 1986), a paucity of research
addressing the content of computer pornography exists.  This
article provides a study of pornographic images located in
public access UseNet newsgroups prior to the wide-scale
commercialization of the Internet.  Additionally, a
comparison of the type of pornographic images posted by
anonymous, non-commercial users with material posted by
commercial vendors is used to illustrate how the rapid
growth of the Internet is outpacing regulation.
  Social scientists and feminist social commentators have
conducted substantial research on the effects of pornography
on attitudes of men toward women (Zillman & Bryant, 1984),
sexual aggression against women (Baron & Straus, 1989;
Donnerstein, Linz, & Penrod, 1987;  Padgett, Brislin-Stutz,
& Neal, 1989), and the role that pornography plays in
perpetuating sexual discrimination (Dworkin & MacKinnon,
1988;  Dworkin, 1985).   Taken as a body, this work is
inconclusive and often contradictory.
  The potential of the Internet to disseminate vast
amounts of pornographic material within an unregulated
marketplace with a large, international audience needs to be
considered.  Social scientists, however, should be cautious
when assessing the impact of computer pornography on
society, not because this pornography is new and different
from erotic magazines and videos, but rather because the
distribution of pornography via computer networks represents
a new use for the technology of the "global village."
  The Internet has evolved over the past 25 years from a
U.S. Defense Department tool for assisting with scientific
research to a commercially-oriented communications network.
As a research tool, the Internet helps users to share
information with minimal costs.  One of the biggest benefits
of using this network is that long distance charges do not
apply.  Until recently, these costs and others were absorbed
by the U.S. National Science Foundation in its administration
of the central "backbone" of the Internet.   However, this is now
changing.
   The quasi-public "information superhighway" is now
facing new challenges as the first stage of privitization
begins.  Because an ever increasing proportion of the
Internet's rapid growth consists of commercial traffic, a
conflict with the National Science Foundation's "acceptable
use" policy exists.  This policy restricts use of the
Internet to research only.  Consequently, commercial
interests are rapidly transforming the face of the network
by using it as a tool for profit and as a market to exploit.
  Branwyn (1993) estimated that the number of computer
networks linked through the Internet is approximately  1.3
million worldwide, with an estimated 14 to 15 million
individual users growing at a rate of 25% every three
months.  This rapid growth and increased interconnectivity
is providing users with access to several sources of
information.  No longer is it necessary for computer users
to procure software from retail outlets like computer and
software stores.  The Internet allows users the opportunity
to download software without ever leaving the comfort and
privacy of their homes and offices, thus ensuring the
anonymity of users retrieving software, such as pornographic
material.
   Computer pornography has moved from simple images
composed of alphanumeric characters to more sophisticated
digitized, moving images.  A variety of computer pornography
including soft core "erotic" images of male and female
models, animated serials, sexually explicit moving images,
interactive sex games, and virtual reality-based types of
cybersex is available (Robinson & Tamosaitis, 1993).
Essentially, pornographic images that can be viewed on a
computer screen are no different in type or nature from
images readily available in erotic magazines or videos.   In
fact, many images come directly from such magazines and
videos.  However, because users select which images are
posted, it is unlikely that pornographic images on the
Internet will have the same frequency of themes as images in
sexually-oriented magazines and videos.
   Pornographic images may be posted on newsgroups in two
ways.  In one way, material may be posted anonymously by
individual users from any location.  That is, individuals
with access to a scanner or a digital camera can scan images
from pornographic magazines and videos or take photographs
that can then be posted in one or more newsgroups.
Alternatively, images may be posted by businesses such as
bulletin board services, software companies, "sex shops," or
graphic design firms.  Perhaps commercial vendors believe
that public domain space like newsgroups provides a good
opportunity freely to advertise their products or services.
Assuming that commercial distributors have greater
awareness of the legal implications of distributing explicit
pornographic material, and that responsibility for posting
such material can be directly traced to them, we
hypothesized that there is a greater likelihood of finding
explicit material posted by anonymous, non-commercial users.
Although the rapid growth of the Internet has created many
regulatory gaps, we believe that commercial vendors will not
risk arousing public concern, precisely because the new
regulations that are bound to accompany increased
commercialization of the network may eventually prove to be
overly restrictive.  Additionally, because these online
services offer more than just access to computer
pornography, the risk of jeopardizing their entire
commercial operation may be too large a gamble.
Sample
   A sample of 150 randomly selected pornographic images
were downloaded from newsgroups on the Internet.  Newsgroups
were selected based on our interpretation of whether the
name assigned to such groups suggested that sexually-
oriented material was present (e.g., alt.binaries.pictures.
erotica, alt.sex.bondage).  Of the 4937 newsgroups available on
our university's system  on 18 April 2024, 17 alternate newsgroups
that provided sexually-oriented image files were identified.
  The sample was obtained using proportionate random
sampling.  The size of the newsgroup in proportion to the
total number of graphic files available in all 17 newsgroups
determined the relative frequency of files selected from
each newsgroup.  A random number table was then used to
select specific graphic files from the available pool of images.
Measures
Frequency and percent of features and themes:
  Because each image coded may contain multiple
combinations of themes, the number of images containing
specific themes plus the percentage of themes present within
the sample were calculated.  We also compared the frequency
of variables from our findings with the frequencies from
other content analyses on pornographic magazines and videos.
Comparison between commercial and non-commercial images:
  We used the correlation coefficient phi to compare each
variable along the mode of distribution dimension.  Phi is a
symmetric measure of association for 2 X 2 crosstabulations
used when comparing non-parametrically distributed
variables.  A one-tailed test of significance using .05 as a
cut-off value was employed.
Intercoder reliability:
 Because images were dichotomously coded and nominally
scaled, it was necessary to use kappa as a measure of
agreement.  Kappa was calculated using the crosstabulation
function in SPSS PC for Windows (Version 6.0), taking into
account the amount of agreement expected by chance.
According to Landis and Koch (1977), kappa values greater
than .75 indicate excellent agreement beyond chance, values
between .40 and .75 indicate fair to good agreement beyond
chance, and values less than .40 indicate poor agreement
beyond chance.  Mean kappas for each variable by each pair
of coders was then calculated.  Variables with mean kappas
below .40 were dropped from subsequent analysis.
Procedure
  We developed a coding scheme by adapting several
content analysis categories used in previous research to
analyze erotic videos, magazines, and cartoons (e.g., Garcia
& Milano, 1990;  Palys, 1986;  Winick, 1985;  Palmer, 1979).
In addition, new categories related to how computer
pornography is displayed were also developed (e.g., color
versus black-and-white images, relative quality of images,
digitized or animated images).  Each category is briefly
described in Table 1.
   A two-stage coding procedure was used in this study.
In the first stage, we individually rated the sample of 150
images by assigning one or more content categories, based on
central themes, to each image in the sample.  Content
categories were coded for either their presence or absence.
Concern about intercoder reliability led to a second round
of coding of a random subsample of 35 images.  We and two
female graduate students in the social sciences worked
independently to code this subsample.  The two female coders
were not told the hypothesis but were made aware of coding
procedures by both oral and written instructions.  Both
coders had little prior exposure to pornography in any form.
Results
  Kappa reliability coefficients were calculated for each
variable among the four coders on a sample of 35 images.
The overall mean kappa for all raters across the 22 themes
is .67 indicating a good to fair degree of intercoder
reliability (Landis & Koch, 1977).  Table 2 shows some
discrepancy between male and female coders on the variables
for quality, close-ups, homosexual sex, use of a foreign
object, images of children/adolescents, anal penetration,
bondage and discipline, and ejaculation.  We hypothesized
that these discrepancies arose from limited coder training
and distress in viewing explicit images that one female
coder particularly mentioned.  Nevertheless, the mean
intercoder reliability fell below .40 only for the quality
and frontal nudity variables, which we dropped from the
analysis, as they fall into Landis and Koch's poor agreement
category.
   Table 3 shows the percentage of features and themes
found in a sample of pornographic images on the Internet.
Of the 150 pornographic images analyzed, 65% are distributed
non-commercially by anonymous network users, 81% are color,
and 92% are digitized.  The most prevalent themes are close-
ups (43%), erect penises (35%), fetishes (33%), and
masturbation (21%).
   We compared our results with other content analyses to
determine whether computer pornography differed in theme
from magazine and video pornography.  We found that fellatio
was present in 15% of our sample, compared with Garcia and
Milano's (1990) finding of 8.1% for fellatio.  For
homosexual sex, we found that 18% of our sample contained
this theme, whereas Winick (1985) and Garcia and Milano
(1990) obtained much lower levels, in the 2-4% range.
Finally, group sex appeared in 11% of computer pornography
but in only  1-3% of other pornography (Winick, 1985;
Garcia & Milano, 1990).  These results suggest that the
criteria for selecting what material to post on the Internet
are different from those used by magazine and video
producers.
   Our hypothesis -- that explicit pornographic material
is more likely to be posted by anonymous, non-commercial
users -- is not supported by analysis.  Table 4 shows
bivariate cross-tabulations using mode of distribution as
the independent variable.
  This analysis shows six significant differences between the
type of material posted by commercial and non-commercial
users.  Contrary to our hypothesis, commercial vendors
generally post more images that contain explicit themes than
do non-commercial distributors.  For example, themes like
close-ups (phi=.16, p

Return to Critical the Criminology Homepage