Formatting note: Significant effort has been made to make these on-line documents appear similar to the printed version in the Georgetown Law Journal, while making this document fully readable by a large majority of Web browsers. As compatibility issues, or in possible rare instances, content differences, become apparant or resolved, changes will be made to this document to improve its similarity to the printed copy.

Context note: This article appeared on pages 1849 through 1915, and the appendicies on pages 1916 through 1934. Page numbers will be added as HTML anchors at a later date.

Finally, Reader Advisory: Due to the nature of the studied material, Minors and Those Who Are Going To Be Easily Offended are hereby warned of strong language and sexual content. Leave now if dealing with such things is going to be a problem for you.

Last updated 7/11/95 8:35pm
[*] Researcher and Principal Investigator, College of Engineering, Carnegie Mellon University. This interdisciplinary project was made possible by four grants from Carnegie Mellon University. The author [hereinafter "principal investigator" wishes to thank members of the research team for their encouragement, patience, and support. Principal faculty advisor: Dr. Marvin Sirbu, Department of Engineering and Public Policy. Faculty advisors: Dr. David Banks, Department of Statistics; Dr. Timothy McGuire, Dean, Charles H. Lundquist School of Business, University of Oregon; Dr. Nancy Melone, Associate Professor of Management, Charles H. Lundquist School of Business, University of Oregon; Carolyn Speranza, Artist/Lecturer, Department of Art; Dr. Edward Zuckerman, Department of Psychology. Senior Programmer: Hal Wine. Programmers: Adam Epstein, Ted Irani. Research Assistants: Patrick Abouyon, Paul Bordallo, G. Alexander Flett, Christopher Reeve, Melissa Rosenstock. Administrative Assistant: Timothy J. Burritt. Administrative Support: Dr. Chris Hendrickson, Associate Dean, Carnegie Institute of Technology; Robert P. Kail, Associate Dean, Carnegie Institute of Technology; Barbara Lazarus, Ph.D., Associate Provost for Academic Projects; Jessie Ramey, Director, SURG. Contributors: Lisa Sigel, C.J. Taylor, Erikas Napjas, John Gardner Myers. Special thanks to Ron Rohrer, Wilkoff University Professor, Department of Electrical and Computer Engineering; and Daniel Weitzner, Deputy Director, Center for Democracy and Technology, for review of the legal notes.

In an effort to present an informative and balanced report, members of the Carnegie Mellon research team (the principal investigator, his faculty advisors, and research assistants) have consulted with organizations and experts who hold a wide variety of viewpoints about pornography, although the overwhelming majority of contacts have been with the pornography industry itself. While this article discusses a number of different viewpoints on significant legal and policy issues related to the regulation of pornographic material, the research team does not advocate or endorse any particular viewpoint or course of action concerning pornography on the Information Superhighway.

[1] "Pornography" stems from the Greek words, porno, meaning prostitutes, and graphos, meaning writing. Over the course of history, it has assumed many definitions and meanings. See generally Lynn Hunt, The Invention of Pornography (1993). Many historians have commented on the difficulty of defining pornography. See, e.g., Walter Kendrick, The Secret Museum: Pornography in Modern Culture (1987). The Carnegie Mellon study adopts the "definition" utilized in current everyday practice by computer pornographers. Accordingly, "pornography" is defined here to include the depiction of actual sexual contact [hereinafter "hard-core"] and depiction of mere nudity or lascivious exhibition [hereinafter "soft-core"]. The courts and numerous statutes concur with the distinction presented here between "hard-core" and "soft-core." See, e.g., Miller v. California, 413 U.S. 15, 24 (1973); Ballew v. Georgia, 435 U.S. 223, 228 (1978); Ark. Code Ann. +s 5-68-302(2) (Michie 1987); La. Rev. Stat. Ann. +s 106 (West 1994). By this definition, not all pornography meets the legal test for obscenity, nor should all depictions of sexual activity be construed as pornographic. Accordingly, data was collected for this article only from bulletin board systems (BBS) which clearly marketed their image portfolios as "adult" rather than "artistic." Any BBS or World Wide Web site which made even a modest attempt to promote itself as "artistic" or "informational" was excluded.

"Pornographer" is defined to include BBS operators who do any of the following: commission photographers to provide new pornographic images; scan pornographic images from magazines; pirate pornographic images from other boards; or purchase adult CD-ROMs for distribution via modem to their customers. "Adult" is the term used by most BBS system operators who market pornography.

[2] The question of whether a sexually explicit image enjoys First Amendment protection is the subject of much controversy and reflects a fundamental tension in contemporary constitutional jurisprudence. While this article discusses only the content and consumption patterns of sexual imagery currently available on the Internet and "adult" BBS, the law enforcement and constitutional implications are obvious. Thus, it is necessary to briefly discuss the constitutional status of sexually explicit images.

Obscene material does not enjoy First Amendment protection. See Roth v. United States, 354 U.S. 476 (1957); Miller v. California, 413 U.S. 15 (1973). In Miller, the Supreme Court established the current tripartite definition for obscenity. In order to be obscene, and therefore outside the protection of the First Amendment, an image must (1) appeal to a prurient (i.e., unhealthy or shameful) interest in sexual activity, (2) depict real or simulated sexual conduct in a manner that, according to an average community member, offends contemporary community standards, and (3) according to a reasonable person, lack serious literary, artistic, political, or scientific value. Id. at 25-27; see also Pope v. Illinois, 481 U.S. 497, 500-01 (1987) (rejecting "ordinary member of given community" test, in favor of "reasonable person" standard for purposes of determining whether work at issue lacks literary, artistic, political, or scientific value); Pinkus v. United States, 436 U.S. 293, 298-301 (1978) (excluding children from "community" for purpose of determining obscenity, but allowing inclusion of "sensitive persons" in the "community"); Ginzburg v. United States, 383 U.S. 463, 471-74 (1966) (allowing courts to examine circumstances of dissemination to determine existence of literary, artistic, political, or scientific value); see also United States v. Orito, 413 U.S. 139, 143 (1973) (holding that constitutionally protected zone of privacy for obscenity does not extend beyond the home).

To complicate matters, all adult pornographic material is initially presumed to be nonobscene. Cf. Fort Wayne Books, Inc. v. Indiana, 489 U.S. 46, 62 (1989) (requiring judicial determination of obscenity before taking publication out of circulation); Marcus v. Search Warrant, 367 U.S. 717, 730-31 (1961) (requiring procedures for seizure of obscenity which give police adequate guidance regarding the definition of obscenity to ensure no infringement on dissemination of constitutionally protected speech). Accordingly, law enforcers and prosecutors attempting to pursue an obscenity investigation or prosecution face constitutionally mandated procedural obstacles not present in other criminal matters. See New York v. P.J. Videos, Inc., 475 U.S. 868 (1986). For instance, the so-called "plain view" exception to the Fourth Amendment warrant requirement, whereby contraband plainly visible to a law enforcement officer may be seized, does not apply to allegedly obscene material because, prior to a judicial determination, nothing is obscene and therefore, a fortiori, nothing be can be considered contraband. See Lo-Ji Sales, Inc. v. New York, 442 U.S. 319, 325 (1979) (requiring that search warrants contain specific description of allegedly obscene items to be seized).

In addition to obscenity, one other type of sexually explicit material does not enjoy constitutional protection. In New York v. Ferber, 458 U.S. 747 (1982), the Supreme Court explicitly removed pornography depicting minors from the protective aegis of the First Amendment. That is, obscene or not, visual depictions of children engaged in sexual conduct are not constitutionally protected. Because the government interest identified by the Supreme Court as justifying removing child pornography from the protection of the First Amendment is more urgent than the government interest which justifies denying protection to obscenity, and because the child pornography standard is far less vague than the obscenity standard, law enforcers and prosecutors are not bound by any unique procedural burdens here. See United States v. Weigand, 812 F.2d 1239 (9th Cir.), cert. denied, 484 U.S. 856 (1987).

In sum, the constitutional regime that the Supreme Court has established for pornography creates two distinct categories of sexually explicit imagery that are not protected by the First Amendment. While ascertaining whether a particular digital image contains a minor is not a Herculean labor, ascertaining whether a particular digital image is obscene in the abstract is well-neigh impossible. Accordingly, the research team will not attempt to pass on the question of obscenity as it applies to the digital images that are the subject of this article.

[3] China, for example, recently executed a man found guilty of producing and selling pornographic books. See China Executes Man for Selling Pornographic Books, Reuters World Service (Beijing), Feb. 25, 1995 available in LEXIS Library, Reutrs File.
[4] See Phillip Robinson & Nancy Tamosaitis, The Joy of Cybersex (1993).
[5] The first three points noted were advertised on the introductory menus of several BBS.
[6] Encryption technology also allows pedophiles to conceal their sexually explicit images of children from law enforcers. Even after child pornographic GIFs or JPEGs (image files) are located and seized, they may be unreadable. Due to increasing availability of sophisticated encryption technology, all computer files, including GIFs, can be "scrambled" so thoroughly that they are virtually impossible for law enforcers to unscramble. See Ivars Peterson, Encrypting Controversy, 143 Sci. News 394 (1993). There is evidence that pedophiles and child pornography consumers are increasingly using "public key" encryption technology to avoid detection individually and when sending images to (or communicating with) one another. A child molester recently thwarted efforts to identify his victims when he stored their names in an encrypted computer file. See Steven Levy, Battle of the Clipper Chip: The Cypherpunks vs. Uncle Sam, N.Y. Times, June 12, 1994, +s 6 (Magazine), at 44.

Public key encryption involves utilizing one of a few commonly available software encryption programs which will scramble/encode the contents of a computer file. Every user has two keys, a public one and a private one, which are created unique to their owner. Users can distribute their public keys to other computer users without compromising the security of their private key. Those users can then use the public key to encrypt a message or image which will be sent to the original owner of the public key. Only the original owner can then decode the message with his private key. Id.

[7] "Information Superhighway" and "Cyberspace" are used to refer to any of the following: Internet, Usenet, World Wide Web, BBS, other multimedia telephone, computer, and cable networks.
[8] "Study" hereinafter refers to all data collected by the research team, whereas "article" refers only to those findings reported here.
[9] As a result of federal legal action against a few well known "adult" BBS operators, including Robert and Carleen Thomas (Amateur Action) and Robert Copella (Pequena Panacha), some systems have removed their paraphilic, pedophilic, and hebephilic imagery from public display. This has created a thriving underground market for "private collections" and anonymous ftp sites on the Internet, which cannot be studied systematically. Thus, it may be difficult for researchers to repeat this study, as much valuable data is no longer publicly available. See infra notes 89-95 and accompanying text.
[10] The original number of downloads tabulated was 6.4 million. A total of 5.5 million downloads are analyzed here; the other 0.9 million concern animations, text, and other miscellaneous files. Information concerning an additional 2.1 million downloads was later obtained from the market leader, Amateur Action BBS. Thus, the total number of downloads tabulated is 8.5 million.
[11] Any description with the same name and approximate file size (to +/-1,024 bytes) was identified as a duplicate. A random sampling of 100 suspected duplicates was downloaded to confirm the validity of this method, with one important limitation; if pornographers change the image names, the process becomes far more complex. Thus, 36% is a minimum estimate.
[12] Content analysis studies have depended upon classification schemes developed by researchers or law enforcers, rather than pornographers. While pornographers lack the formal tools necessary to perform reliability and validity testing on their written descriptions, they make their profits satisfying their customers, and one could argue that their financial success suggests no small modicum of understanding of the images they market.
[13] See, e.g., U.S. Public Health Service, Report of the Surgeon General's Workshop on Pornography and Public Health (1986) (suggesting links between childhood involvement in pornography, continued use of violent pornography, and uncommon sexual practices); Edward Donnerstein et al., The Question of Pornography: Research Findings and Policy Implications (1987) (cataloguing various experiments attempting to study the possible link between exposure to pornography and antisocial behavior); Neil M. Malamuth & Joseph Ceniti, Repeated Exposure to Violent and Nonviolent Pornography: Likelihood of Raping Ratings and Laboratory Aggression Against Women, 12 Aggressive Behav. 129-37 (1986) (finding no significant link between continued exposure to violent and nonviolent pornography and laboratory aggression toward women); Dolf Zillmann & Jennings Bryant, Pornography, Sexual Callousness and the Trivialization of Rape, J. Comm., Autumn 1982, at 10 (noting changes in attitude toward rape, women's "liberation" movement, and callousness towards women, with prolonged exposure to pornography); Dolf Zillmann, Effects of Prolonged Consumption of Pornography, in Pornography: Research Advances and Policy Considerations (Dolf Zillmann & Jennings Bryant eds., 1989) 127-57 (linking prolonged use of pornography with trivialization of rape and sexual child abuse as criminal offenses); J.V.P. Check & T.H. Guloien, Reported Proclivity for Coercive Sex Following Repeated Exposure to Sexually Violent Pornography, Nonviolent Dehumanizing Pornography & Erotica, in Pornography: Research Advances and Policy Considerations, supra at 159-84 (distinguishing between types of pornography which are sexually violent, nonviolent but dehumanizing, and nonviolent but erotic in their effect on user's subsequent behavior).
[14] "Consumption" is defined by this study merely as the purchase or download of pornographic products. The issue of consumption admittedly goes further than mere downloads to include why they are purchased (or pirated), and how they are utilized by the receiver. While the notion of consumption has been explored in the area of the written word by Umberto Eco, Janice Radway, and Wolfgang Iser, among many others, no known studies focus on Cyberspace. No data was available to indicate whether such materials are consumed for erotic, curiosity, or other purposes.
[15] The results of the largest content-based study of pornography known to have been conducted were published in 1988. Park Elliot Dietz & Alan E. Sears, Pornography and Obscenity Sold in "Adult Bookstores": A Survey of 5132 Books, Magazines, and Films in Four American Cities, 21 U. Mich. J.L. Ref. 7 (1987-88) [hereinafter Dietz-Sears]. Comparing the present study and the Dietz-Sears study illustrates the manner in which powerful new technologies have transformed the pornographic landscape in the past seven years:
The current study [hereinafter Carnegie Mellon study] is considerably larger than the Dietz-Sears study. The Carnegie Mellon study examines 917,410 images, image descriptions, short stories, and short films, whereas the Dietz-Sears study examined 5132 book, magazine, and film covers, as evidenced by its title.

The Dietz-Sears images were selected by a stratified random method. Id. at 12. The Carnegie Mellon study examines all of the suppliers meeting minimum scale criteria.

The Dietz-Sears study described what was offered for sale; it did not attempt to indicate which products the customers actually purchased, and in what quantities. Id. at 9.

The Dietz-Sears study did not address any changes in pornographic image repertoires and purchases over time. Id. at 11-12. The Carnegie Mellon study tracks image repertoires over a fourteen-year period.

The Dietz-Sears study concentrated only on four cities. Id. at 11 n.11. The Carnegie Mellon study covers pornography offered in thirty-two states representing all regions of the country, both urban and rural.

The Carnegie Mellon study uses largely objective measures, rather than the subjective criteria of many researchers, to classify the images. It makes no attempt, for instance, to determine what is, or what is not, "degrading" in pornography. This enables independent organizations and policymakers to examine the data "raw"(as provided directly by the pornographers) and draw their own conclusions.

The Dietz-Sears study, in examining only box or magazine covers, gave little indication to what extent the contents deviate from the cover. Id. at 12. The covers could conceal or exaggerate the contents.

A number of criticisms of the Dietz-Sears study, and of content analysis of pornography in general, were suggested in Daniel Linz & Edward Donnerstein, Methodological Issues in the Content Analysis of Pornography, 21 U. Mich. J.L. Ref. 47 (1987-88). The Carnegie Mellon research team carefully considered these criticisms in preparation of this article. There are difficulties inherent in a much smaller dataset, as with Dietz-Sears. See also infra notes 65-76 and accompanying text for discussion of the reliability and validity procedures utilized for the Carnegie Mellon study.
[16] Robert T. Michael et al., Sex in America: A Definitive Survey (1994).
[17] One theory may help explain the differing results. The Carnegie Mellon study presents an exhaustive analysis of the consumption habits of a subset of the general population, whereas the recent Sex in America study bases its results on a random sampling of the reported sexual activities of the entire population. This is because the demographics of computer users do not currently mirror those of the general population. The average age of the consumer population in this study is thirty-one, and the vast majority are male. "Adult" BBS system operators report that of those subscribers that are female, many (and in some cases all) are paid by their BBS to "chat" online with male customers. About half the BBS studied offered women free access.

One of the more intriguing questions raised by this study is whether the general population will demand the same types of imagery currently in high demand among computer users.

[18] For example, Senator Jim Exon (D-Neb.) has succeeded in persuading the U.S. Senate Commerce Committee to address access to pornography on computer networks as part of the pending telecommunications reform legislation. S. 314, 104th Cong., 1st Sess. (1995).
[19] See infra note 20 and accompanying text.
[20] Child pornography must now contain visual depiction of an actual child. See New York v. Ferber, 458 U.S. 747 (1982). Thus, the sexual abuse or exploitation of an actual child has necessarily occurred. It is possible that a visual depiction, generated entirely by computer, of a "child" engaged in sexual conduct (without the use of an actual child) could enjoy First Amendment protection. Cf. John C. Scheller, Note, PC Peep Show: Computers, Privacy and Child Pornography, 27 J. Marshall L. Rev. 989 (1994) (arguing that the state's interest in prohibiting child pornography extends beyond protection of the initial victim, and includes protection of potential future victims, obviating the need for the presence of an actual victim to prosecute those in possession of child pornography).

However, there is language in Ferber and Osborne v. Ohio, 495 U.S. 103 (1990), which would support the notion that child pornography produced entirely by digital imaging technology may not be protected by the First Amendment. In both decisions, the Supreme Court used, as a secondary justification for denying constitutional protection to child pornography, the role that child pornography often plays in the molestation of a child. In Osborne, the Supreme Court noted that child pornography is often used to lower the sexual inhibitions of children. 495 U.S. at 109-10. See also Ferber, 458 U.S. at 759; T. Christopher Donnelly, Protection of Children from Use in Pornography: Toward Constitutionally Enforceable Legislation, 12 U. Mich. J.L. Ref. 295, 300-04 (1979) (noting the use of child pornography in child molestation, both to arouse the molester and the child, and to overcome the child's inhibitions); W.D. Erickson et al., Behavior Patterns of Child Molesters, 17 Archives Sexual Behav. 77-78 (1988).

[21] The question of whether access by minors to pornography can be effectively limited has been the subject of much debate among members of the research team and many network administrators. There are currently many ways for minors (and adults) to access Usenet newsgroups which have been "banned" or are not carried by a user's host site. The issue is particularly difficult in light of First Amendment and privacy considerations.

A popular electronic pamphlet, "How to Receive Banned Newsgroups Frequently Asked Questions (FAQ)," which is posted monthly on the Usenet illustrates many of the difficulties. Users may access "banned" Usenet newsgroups through any of the following means:

Not-for-profit news providers: Users may telnet to various hosts, such as Freenet, Hermes, Nyx, Prairienet, Um-m-net, or UNC BBS. From there, they can easily access Usenet newsgroups directly.

Commercial news providers: If users have difficulty with not-for-profit news providers, they can turn to commercial news providers. These news providers will provide shell access, news, dialup lines, and more. There are many national service providers both in the U.S. and around the world.

Open netnews transfer protocol (nntp) sites: If users do not want to use a not-for-profit or commercial news provider, there are open nntp sites that allow almost anyone with shell access and a news reader to read from and post to Usenet newsgroups.

Gopher: Usenet newsgroups are available through gopher. To find them, users may search for "Usenet news -t7" in veronica, or use sites recommended in the FAQ.

Mail to news gateways: Mail to news gateways are sites that will take any article given to them and forward it to Usenet newsgroups. These are listed in the FAQ.

Internet services list: This is not a direct way to access Usenet newsgroups, but it does list many places, and it changes often. To get the Internet services list, users may look in alt.internet.services, comp.misc, biz.comp.services, alt.bbs.internet, news.answers, comp.answers, alt.answers, or ftp to rtfm.mit.edu or archie for inet.services.txt.

Telnet: With telnet, users can access any of the free services listed above. They can also access one of the many gopher servers that will allow them to read Usenet newsgroups. These are listed in the Internet services list. They can also access the World Wide Web via telnet.

Email: Many Usenet newsgroups are mirrored in a mailing list. In addition, some sites have ftp archives. Users may determine whether the group they are interested in has this service by checking the FAQ; these FAQs are available through the ftp site rftm.mit.edu.

Ftp: Ftp service is available through e-mail. For information on ftp-through-mail, users may look in the Internet services list.

Perl Scripts: It is also possible for one user to mail another user the contents of certain Usenet newsgroups with a simple Perl script, provided they have access to a shell account.

Mailing Lists: Users may also subscribe to mailing lists that mirror Usenet newsgroups. To get a list of these, they can ftp to rtfm.mit.edu and get /pub/Usenet/news.answers/mail/news-gateway/partX, where "X" is the number of separate parts.

Archived Newsgroups: Many Usenet newsgroups are archived. Users can often examine the FAQ for that group, available at rtfm.mit.edu.

These user "work arounds" are relevant principally to Usenet access providers (commonly referred to as "host sites"), which do not wish to act as a republisher or prevent users from exploring alternate avenues for retrieving pornography. Most of the "work arounds" discussed here also assume that the host is providing unlimited general internet (IP) access to off site locations which might contain pornographic Usenet newsgroups or gopher sites. Parents and schools are not obliged to provide unlimited access to arbitrary IP addresses and port numbers. They can implement IP address filters which limit access only to known, acceptable locations. They can also implement filter programs which will look for the string "alt.sex.*" and either drop the connection or log a child's access for the parents to deal with later. Accordingly, while acting in locus parentis, schools may be justified in logging all actions by minors at their computers. This would create an "after the fact" method of determining what students are doing. It is possible that employers may also attempt to monitor their employees" logfiles as a means of insuring that company equipment is utilized for business, as opposed to recreational, purposes.

To complicate matters, however, these strings might block desired informational newsgroups such as alt.sex.safe. Many network observers argue that such a strategy thwarts the most fundamental benefits of the Internet. Moreover, the research team and many network administrators have discovered pornography on alt.test and other supposedly "general interest" newsgroups. Given the approximately 50 new World Wide Web sites and 20 new Usenet groups announced each day, the cost in administering restrictive access may be prohibitive. Administrators would have to check each of the new sites and newsgroups daily and make decisions concerning access. They would then need to propogate the restrictions to every machine or account on the local net. Though technically possible, such filter programs are computationally expensive and do not resolve encryption issues (very common with PGP and other PEM products) or protocols that do not use standard ASCII.

In general, primary and secondary schools, as well as most parents, do not have the capability or resources to implement these safeguards. These difficulties might necessitate the review of thousands of newsgroups on a case by case basis, which has led some network analysts to suggest that a national or international rating system be established. See infra note 137. At least one third party vendor has begun to provide such blocking services, but it is not clear to what extent the services will be effective. See Carla Koehl et al., Policing for Porn, Newsweek, May 22, 1995, at 8. Furthermore, consumers can usually block access to 900 numbers at no cost, and they may be angered at the prospect of having to pay for services (and software updates) which attempt to filter out pornography from their home.

At the present time, it is also difficult to make distinctions between multiple users at a given site. For instance, a university could not easily set up filters so that the same public cluster machines could be used by persons of legal age to access anything, while at the same time restricting access to certain groups by minors.

Whatever procedures are implemented, many users may develop increasingly innovative techniques to bypass such blockades and trade pornography through underground channels. Although typical firewalls would block most (but not all) of the alternative paths discussed here, they raise additional concerns that those who manage the firewalls could act as censors by determining what their users can and cannot access. Routers could be programmed to restrict any traffic from a given site, but this implies a very large administrative cost to keep such blockage up to date, perhaps as many as several full-time staff per Internet node. Moreover, because current generation routers were not designed for restricting traffic, they may not have enough capacity for enough filters to block all restricted sites.

An intriguing solution to the difficulties raised here may involve some promising new technologies currently being developed at several universities and research labs, including the IBM Almaden Research Center. See Will Equitz & Wayne Niblack, Retrieving Images from a Database Using Texture-Algorithms from the QBIC System, Research Report, IBM Research Division (1994). Researchers are currently developing algorithms which would automatically scan through large online image databases to identify specific colors, texture (including contrast, coarseness, and directionality), and patterns. Dubbed "Query by Image Content (QBIC)," the technology might assist network administrators in locating the presence of nudity, genitalia, breasts, ejaculate, feces, and various activities in the images that might suggest sexually explicit content.

Even if the method were only moderately reliable, it might substantially narrow the number of suspected images a human monitor would need to sift through to locate pornography. While it may not be possible in the next decade for such technology to automatically classify images with the same precision as the Carnegie Mellon linguistic parsing software, it does not appear exceedingly difficult to develop algorithms that scan through images and check for textures and colors that might suggest sexual explicitness. However, such technology may not work if the images are encrypted, and significant privacy issues are also implicated.

[22] Paul N. Bloom et al., Avoiding Misuse of New Information Technologies: Legal and Societal Considerations, J. Marketing, Jan. 1994, at 98.
[23] Some members assisted on the condition of anonymity and are not listed in the biographical footnote.
[24] Three leading market research firms, Nielsen Media Research, Yankelovich Partners, and ASI Market Research, have recently formed a joint venture to analyze online services. See Research Firms Team Up to Analyze Cyberspace-Paper, May 17, 1995, posted to clari.tw.new_media.
[25] This article assumes that the reader has a basic understanding of Usenet and BBS. Only the technical aspects of BBS which relate to pornography will be explained in detail. Many fine books deal extensively with the technical aspects of the Information Superhighway. See, e.g, Paul Abrams & Bruce Larson, Unix for the Impatient (1992); Alan D. Bryant, Creating Successful Bulletin Board Systems (1994); Adam C. Engst, Internet Starter Kit (1993). The reader is also referred to Boardwatch magazine, a monthly guide to electronic bulletin boards and the Internet.
[26] Proposals for New or Revised Classes of Interstate and Foreign MTS and Wats, 56 F.C.C.2d 593 (1975).
[27] BBS Direct, Brochure (on file with The Georgetown Law Journal).
[28] See supra text accompanying notes 25-27 for discussion of the distinction between the Usenet and commercial BBS.
[29] FTP stands for File Transfer Protocol. It is one of the primary ways users can send and retrieve information from computers on the Internet.
[30] Each of the datasets has limitations. The university data, while perhaps unique in that it offers an accurate, if not exhaustive, picture of the monthly access habits of an entire community, has two limitations. First, 11% of the computer users in this study block the site statisticians from monitoring their online activities. Second, some users have multiple accounts and avoid detection by using a second account to access the Usenet.

While there is no evidence to suggest that Usenet and Internet users who block the monitoring of their accounts access pornography more frequently than those who do not, one also cannot assume that a notable difference does not exist. This is especially true in the context of pedophilia and child pornography consumption. Preferential molesters (i.e., pedophiles with a true sexual attraction to children) frequently employ inventive mechanisms to evade discovery, as discovery will likely lead to incarceration. Kenneth Lanning, Child Molesters: A Behavioral Analysis 16-19 (3d ed. 1992). Given this behavior pattern, it is at least possible that some Internet users who block their accounts prefer sexual images of children and wish to avoid detection.

The research team has also mounted a continuous and persistent effort to obtain similar data from other universities. In most instances, the universities contacted indicated that they did not compile such data, while others were uncooperative. Still, there is no reason to believe consumption at the university studied differs from that of other universities from which pornographic Usenet newsgroups can be accessed.

With respect to the worldwide statistics, the "arbitron" script that is used to produce the data for the monthly Usenet readership surveys is supplied by Brian Reid at DEC Network Systems Lab. The system administrators who agree to participate in the survey customize the script for their site, and then run it once a month. Importantly, it automatically produces a readership survey for that site and then mails the results to decwrl.dec.com. These results are run through a statistics program on the first day of each month and include any report that has reached it by that time. Because the worldwide statistics are based on a small sampling of sites (453 or .17% of the total estimated sites), there is no way of knowing whether those who chose to participate in the Usenet survey differ significantly from other sites. Reid notes that:

The most controversial field in the survey report is the "$US per month per reader." It is the estimated number of dollars that are being spent on behalf of each reader, worldwide, on telephone and computer costs to transmit this newsgroup. The rate of $.0025 per kilobyte is the same value used in the UUNET statistics reported biweekly. It is based on discussions among system administrators about the true cost of news transmission . . . [who] treat this value as being accurate to within about 25%.
Brian Reid, USENET Readership Summary Report for Jan. 95, Feb. 1, 1995 (electronic newsgroup posting, on file with author). For a complete description of the methodology used to compile the worldwide statistics, see the periodic postings in news.lists.

The results of the arbitron program are dependent on the rate at which sites expire (delete) newsgroups. If a site is small and expires newsgroups rapidly, the results may indicate fewer active readers than may be the case. Most sites expire news every three days to two weeks, depending on the group. The university arbitron, on the other hand, was modified to count readers who had accessed an article within the last thirty days. However, network administrators at the university also compiled weekly statistics and reported no significant differences between the weekly and monthly statistics.

[31] A content analysis of Usenet imagery is of questionable value. First, relatively few posters are responsible for the majority of posts. Thus, such a study would reflect the tastes and repertoires of only a small number of users. Second, descriptions are usually not provided, which means researchers must often classify the images themselves. Moreover, there is much data which suggests the overwhelming majority of users who read the Usenet bulletin boards do not post to them. For instance, approximately forty new images were posted to alt.binaries.pictures.erotica (a.b.p.e.) each day, yet estimates put the number of readers of a.b.p.e. at 260,000 people per month.
[32] These were the largest available at the research site.
[33] The poster is the person who initially sends the image to the newsgroup site.
[34] Also included were images with identifiable cropped logos or telephone numbers.
[35] The same holds true for "adult" BBS, which Boardwatch magazine estimates at 5% of the total number of BBS in the country. Top 100 B.Boards, Boardwatch, Sept. 1994, at 32-33. Indeed, in a somewhat unscientific survey, at least 23 of the top 100 BBS listed by Boardwatch magazine were found to contain "adult" materials. Id.
[36] According to New Media magazine, "[m]ultimedia is the first phenomenon in the history of computing to be driven by the consumer not corporate markets." Peter Jerram, Who's Using Multimedia, New Media, Oct. 1994, at 48, 49. A study by Dataquest, Demand for Multimedia in Large Business, placed the consumer market for multimedia technology at $16 billion, which dwarfs the $2 billion business market. Id. Given that 83.5% of all current Usenet imagery is pornographic, such commercial and entertainment software may be used largely for pornographic purposes. "I am frequently patted on the back by CEOs of companies privately," said James Erlich, creator of Penthouse Interactive Virtual Photo Shoot, "because their hardware sales have gone through the roof since our titles have been released." David Landis, Regulating Porn: Does It Compute, USA Today, Aug. 9, 1994, at D1.
[37] Heavy pornography traffic over the Internet was reported at a University of Delft computer site. More than 10,000 users from around the world were found to obtain approximately 30,000 images a day using ftp. Other pornography collections obtained or distributed via ftp were reported at the U.S. Department of Energy's Lawrence Livermore National Laboratory, and at France's National Conservatory of Arts and Crafts. See Jared Sandberg, This Is the Tale Of a Mad Scientist And His "Sex Machine", The Reason Why He's Mad: Voyeurs by the Thousands Overloaded His Program, Wall St. J. Eur., Feb. 8, 1995, at 1.
[38] These numbers were calculated from data collected at a sample of 453 sites by Brian Reid and reported in the Usenet newsgroup news.lists for March 1995.
[39] The Usenet protocol is nntp and the World Wide Web protocol is http. Usenet and World Wide Web refer to the "service" which consists of client software, host software, data residing at hosts, and protocols for moving data between clients and hosts; or, in the case of Usenet, between hosts and other hosts.
[40] The research team consulted with several privacy experts and opted not to report detailed demographics of the university population of computer pornography consumers. These demographics included age, sex, nationality, marital status, position (faculty, staff, student), and department. Although the research team obtained such demographics by means available to any authorized user of the campus network, reporting them would raise complex ethical and privacy issues. The data would have to be disguised in a manner that could not be reconstructed to identify individual users.
[41] One must be careful to distinguish between a newsgroup reader who subscribes to a newsgroup and a newsgroup reader who accesses a newsgroup. Those who subscribe may or may not read the board, whereas those who access the board may or may not subscribe.

It should also be noted that Usenet names tend to be euphemistic. The bulletin board alt.binaries.pictures.erotica contained, in addition to soft-core and hard-core imagery as defined by this study, the following paraphilias: pedophilia, hebephilia, fisting, B&D/S&M, coprophilia, urophilia, transvestite, and transsexual. The commercial bulletin board clari.news.sex was clearly not pornographic, although sex-related. The newsgroups identified as pornographic include images, as well as written material and offers for sale of pornographic products.

[42] Some network analysts have interpreted this to indicate that more people are likely to read pornographic bulletin boards if they are given as easy access to them as to non-pornographic bulletin boards.
[43] These images are not actually rated in any official way; the research team has merely chosen "PG" and "R" to describe material that would not generally be considered pornographic but which some viewers might identify as pornographic. See supra text accompanying notes 29-34 for discussion of how this methodology was selected.
[44] The posting of potentially illegal images by BBS users raises complex issues regarding Usenet liability when those images are later downloaded or otherwise transmitted. In such a circumstance, sysops may attempt to allege they lack the requisite scienter to have violated either obscenity or child pornography laws, and that the imposition of criminal liability would thus chill protected expression. See Smith v. California, 361 U.S. 147 (1959) (holding that the First Amendment prohibits imposition of criminal liability without scienter in context of obscenity); United States v. X-Citement Video, 115 S. Ct. 464 (1994) (holding that federal child pornography laws do not raise First Amendment concerns with respect to scienter because they include an express requirement); Cubby, Inc. v. Compuserve Inc., 776 F. Supp. 135 (S.D.N.Y. 1991) (holding that computer service company that provided its subscribers with access to electronic library of third party publications was a mere distribution of information and could not be held liable for disseminating defamatory publications absent showing that it knew or had reason to know of defamation).
[45] See Robinson & Tamosaitis, supra note 4; Billy Wildhack, The "Adult" BBS Guidebook (1993).
[46] Baud is a measure of modem communications and transfer speed.
[47] Regrettably, due to a lack of hard-disk space, time, and available funds, the research team was unable to maintain detailed records on some smaller or newer BBS excluded from this study. This data may have indicated the growth or contraction of the industry, as well as the estimated number of boards in the United States.
[48] These small but aggressive BBS are important to an understanding of the dissemination of digitized pornography. Because their hard disk space is limited to 500-1000 images, these operators must take special care to provide only those pornographic images that they believe will maximize download revenues. Often, they concentrate on a single market niche, such as bondage, anal, or urophilic images.
[49] Whether board operators attempt to identify and then refuse access to minors raises constitutional questions. The Supreme Court has held that government may limit availability of pornographic images, even though they are protected by the Constitution, in order to prevent minors from gaining access to material which society deems inappropriate for the young. See Ginsberg v. New York, 390 U.S. 629 (1968). However, the application of this principle is limited where its effect is to prevent adults, as well as children, from gaining access to constitutionally protected sexual expression. See Butler v. Michigan, 352 U.S. 380 (1957) (finding the prohibition of general distribution of a book because of its potentially deleterious effect on children to be an impermissible restriction on freedom of speech).

Determining whether an image is obscene partly depends on its appeal and effect on a hypothetical "average person." See Miller v. California 413 U.S. 15, 24 (discussed supra note 2). The average person standard stands in stark contrast to an older approach to obscenity which tested whether a pornographic image impacted the most susceptible, rather than the average person. Regina v. Hicklin, 3 L.R.B.Q. 360 (1868). Cf. Commonwealth v. Friede, 171 N.E. 472 (1930) (focusing standard for determination of obscenity on whether it tends to corrupt youth).

[50] For instance, a particularly aggressive BBS--Amateur Action--displayed a message to each user which asserted that the e-mail messages contained on the BBS were protected under the Privacy Protection Act of 1980, 42 U.S.C. +s2000aa et seq. Section 2000aa protects electronically stored "work product" from search or seizure by law enforcement. The Amateur Action BBS alleged that e-mail constituted protected "work product" and threatened harsh reprisals against law enforcers or others who might surreptitiously read protected e-mail.
[51] While this was the primary format, fourteen other formats were encountered, and had to be converted to this standard form. The record format is determined by the software package used by an operator.
[52] Windy City Freedom Fortress is the only defunct BBS included in this study.
[53] Many sysops acknowledged that at least some of their images were pirated.
[54] Most BBS limit their customers" downloads to a certain number of kilobytes per day. Taking the average number of kilobytes per file, this could be converted to the number of downloaded images. For instance, a $40.00, six month subscription on Amateur Action BBS bought 600 kilobytes per day; with each image an average of 211 kilobytes, the customer could download an (average) maximum of 2.85 images per day, 19.9 images per week, and 85.3 images per month. As will be reported later, based upon information provided by sysops, most customers downloaded substantially less than the maximum number of images permitted by the BBS to which they subscribed.
[55] Conspicuously absent from this "single line" format are the posting dates and download information.
[56] See supra note 15 and accompanying text.
[57] See Dietz-Sears, supra note 15, at 19.
[58] This is further supported by the number of BBS discovered which have "B&D/S&M" portfolios of 50% or more. See infra notes 68-72 and accompanying text.
[59] Consider the description "Young girl with no boobs and no pussy hair gets fucked hard!" Although the judges supported this description as pedophile, the actual image more closely approximated hebephile, or even young "adult." The most well known computer portfolio of hebephile images came from the now defunct Windy City Freedom Fortress BBS, referred to by consumers as the "Junior Miss Series," or simply "JMIS," in "My Private Collection," or simply "MPC." They are now among the most frequently posted genre of pornographic images on the Usenet. The Carnegie Mellon study cannot always differentiate between pedophilic and hebephilic imagery because of the ambiguous manner in which pornographers market their images. Accordingly, the term "pedophilic" is used in this article, whenever the research team, through linguistic analysis and extensive validity testing of actual images, conclusively identified images of pre-pubescent children, as opposed to adolescents. Otherwise, the term pedo/hebephilic is invoked.
[60] See Appendix A for a list of many final categories and their definitions.
[61] 11% of the images studied "overlapped" into two or more categories.
[62] See Dietz-Sears, supra note 15, at 18.
[63] It can thus be assumed that all other classifications depict some form of heterosexual or lesbian activity. This was done to simplify the data for the study and is not meant to suggest any value judgments concerning homosexual imagery.
[64] Every effort was made to correct as many of these as possible.
[65] It should be noted that neither Dietz-Sears, nor any other study known to the research team, has adequately resolved the methodological problems of image classification. The research team views such classifications merely as a convenient way of discussing and describing the availability of, and demand for, computer pornography. Other classification schemes are possible with the current data. See generally David Lyons & John Anderson, The Psychological and Social Consequences of Pornography: Results of a Systematic Review (1993).
[66] This should not be interpreted as a limitation of the Carnegie Mellon methodology. Since reliability procedures were completed in September 1994, the research team has continued to fine tune the computer dictionary with considerable success. The individual classifications may be presented with further statistical analysis in a later paper.
[67] The paraphilia classifications were adapted from an authoritative psychiatric text. American Psychiatric Association, Diagnostic and Statistical Manual of Mental Disorders 522-32 (4th ed. 1994) [hereinafter DSM-IV].
[68] Voyeurism did not account for a significant number of downloads and thus could be discarded without any notable effect. However, voyeurism provides an excellent example of how information collected by both pornographers and researchers will be utilized (although not necessarily by linguistic parsing) in the coming years when studying marketing on the Information Superhighway. The description "He watches a hot blonde get fucked" is illustrative. Strictly following DSM-IV, this description would be classified as paraphilic because it contains the word "watches." However, without the introduction of more sophisticated analytic techniques, it would not be clear which of the following three keywords were the reason a customer downloaded the image: "watches," "blonde," or "fucked." It would also not be clear whether those three words in combination were what especially attracted the consumer.

There are numerous techniques a researcher or pornographer could apply to answer this question. First, one could obtain from the BBS the log file which indicates the download habits for each consumer. Thus, if the word "watches" were the reason for the download, the consumer would indicate a pattern of having downloaded other images which were described as "watches" (or "blondes" or "fucks"). The log file, which is kept privately by most BBS operators, is the best way to develop profiles of individual consumer habits. However, researchers (or sophisticated pornographers) might be more interested in analyzing segments or market niches; this would enable them to draw general conclusions about the industry independent of a few "unusual" customers. Second, if the log file were not available (and it is expected that in most cases it would not be) there would remain at least two other powerful techniques researchers could apply to answer such questions: cluster analysis and contingency table analysis. For a contingency table analysis, one could separately count the percentage of images which contained each of the three words and then predict the expected number of cases of overlap. This expected number could then be compared to the actual number to determine whether an unusually high or low degree of activity were present.

[69] For an explanation of the distinction in definitions utilized in this study between "hard-core" and "soft-core," see supra note 1.
[70] The presentation of kappa values, which indicate the agreement or disagreement among the judges, was considered unnecessary because of the high level of reliability.
[71] Additional interjudge comments: "incest" is entirely a marketing technique, and its true accuracy cannot be determined by any interjudge reliability test. If the pornographer decides to market a particular image as father-daughter, brother-sister, or sister-sister, it is to assist the consumer in achieving a certain level of fantasy. "amazing" was found to be another marketing ploy. When the pornographer does not have in mind a good, quick description, he will often try to sell his customer with words such as "amazing" or phrases such as "you won't believe this one!" What makes an image intrinsically "amazing" is entirely subjective. famous models: any time a name was used, rather than the pronouns "she" or "her," an image was classified as famous models. The judges, however, were not avid consumers of pornography and thus did not recognize names of particular pornographic "stars." It should be noted here that famous models was defined broadly, to include a great number of "local" stars from BBS across the United States. hebephile included images of young, "flat-chested" females, their genitalia shaved, clad in pigtails and bobby socks, and sometimes surrounded by stuffed animals. Thus, they may be, or appear to be, adolescents. The emphasis of this study, as noted earlier, is to determine how many images the pornographers market as pedo/hebephile, an important area which has not been adequately addressed by the legal system. b&d/s&m: When a rope or harness or whip is visible in the image, it was classified under b&d/s&m. However, the "pain and torture" category caused the most difficulty. Pornographers often market anal sex and fisting images as "She screams in pain as he fucks her ass," or "She is tortured as her girlfriend fists her pussy." It was necessary to permit overlap in these cases. Thus, the former image would be classified both as anal and b&d/s&m, and the latter would be classified as fisting and b&d/s&m.
[72] Courts and statutes do not distinguish between pre- and post-pubescence. Child pornography is defined as depicting persons under age 18. See 18 U.S.C.A. +s 2256 (West 1995).
[73] See New York v. Ferber, 458 U.S. 747 (1982); Osborne v. Ohio, 495 U.S. 103 (1990).
[74] See DSM-IV, supra note 67.
[75] See supra note 1.
[76] Images obtained from the Usenet that had a BBS logo were compared to their corresponding descriptions in the "allfiles" listings of the commercial BBS where they originated. These procedures are defined as informal because the researchers, having seldom encountered images which did not correspond to their descriptions, did not keep exact records in this respect. The main task of the Usenet portion of the study was merely to determine what percentage of Usenet pornography originates from "adult" BBS.
[77] The term "availability" seems more appropriate than "supply," because electronic images can be easily reproduced, and could be in unlimited supply. However, they may be available only through a limited number of outlets.
[78] This article assumes that the market for digital pornography is, for the most part, free and open. Significantly, the high level of demand and small number of suppliers could lead to a less competitive market if a relatively small number of suppliers were to tacitly or overtly agree to fix prices or otherwise hinder upstart competitors. However, the limited barriers to entry should inhibit horizontal concentration in any digital image transmission market.
[79] The remaining downloads involved text and animation files which were not analyzed for this study.
[80] Histograms which examined the number of downloads per file for each BBS for each of the four major classifications strongly support this conclusion.
[81] See supra text following Table 4 for a narrow definition of vaginal sex imagery.
[82] See id. for a broad definition of vaginal sex imagery.
[83] Recall that this data covers only those BBS for which download data was available. The research team estimates there are approximately another half-dozen equally large "adult" BBS not analyzed in this article because no or only partial download information was available.
[84] Paraphilic pornography not depicting children is analyzed under the Miller obscenity standard, which requires proof that an allegedly obscene image offends community standards. 413 U.S. 15, 24-27 (1973).
[85] 458 U.S. 747 (1982).
[86] Four explanations offer possible answers for this disparity. First, it may be that in the case of pedophilic imagery, such images are not widely available, and thus one would expect demand to exceed availability in most instances. However, based on detailed conversations the principal investigator has had with many prominent "adult" BBS sysops, three other explanations appear more viable. First, because the research team has found pedo/hebephilic and paraphilic imagery to be widely available on CD-ROMs and computer networks, many sysops may not provide such imagery out of fear of prosecution. Second, sysops such as Robert Thomas have indicated that they want to give their subscribers a "variety" of choices, regardless of demand for particular classifications. These sysops often view their BBS as extensions of their own interests in pornography. Third, many sysops have indicated either a general ignorance of the power of their data or a lack of adequate time and resources to perform such analysis.
[87] To determine whether market forces common to all "adult" BBS exist, the research team generated scatterplots that examine the number of files offered by each BBS versus number of downloads, for each of the four classifications. The patterns on these scatterplots were muddled, with most "adult" BBS clustering near the origin. The data was then transformed, with the natural logarithm of the number of downloads plotted against the natural logarithm of the number of files. In this scale the muddle disappeared, suggesting that the transformation reflects implicit market forces and is the natural perspective from which to seek structure in the data. For pedophilia, the null hypothesis of no linear relationship between the transformed variables was rejected with a significance probability of 0.0002; for paraphilia, 0.0077; for hard-core, 0.049; and for soft-core, 0.264.

Before transformation, the Amateur Action BBS pedo/hebephile portfolio appears to differ substantially from other "adult" BBS, but on the logarithmic scale it is not an exceptional data point.

[88] Because of the sensitive nature of the consumer data, the research team consulted privacy experts before submitting this article for review and publication. It was decided that once the demographics were tabulated and independently confirmed by two reviewers, the names of all consumers would be permanently deleted from the Carnegie Mellon database. This was done to protect the privacy of consumers in the United States and abroad.
[89] United States v. Robert Thomas et al., No. CR-94-20019-G (W.D. Tenn. filed Jan. 25, 1994). See James Crawley, Memphis Porn Decision is Far-reaching: Ruling Raises Concerns About Rights of Online Computer Users, San Diego Union Trib., Aug. 16, 1994, at 9.
[90] See Joshua Quittner, Computers in the 1990s: Life in Cyberspace, N.Y. Newsday, Aug. 16, 1994, at B27. At the intermediate appellate level, the case has already attracted the attention of several amici interested in the development of Cyberspace. See, e.g., Electronic Frontier Foundation AABBS Amicus Brief in Support of the Thomases (Appellants), April 19, 1995, posted to comp.org.eff.news.
[91] 413 U.S. 15, 25-27 (1973).
[92] 413 U.S. at 30-34.
[93] There are at least two visions of pornography regulation which compete with the traditional test the Supreme Court established in Miller. The most fundamentally different perspective has been proposed by feminist legal scholars. See, e.g., Catharine A. MacKinnon, Not a Moral Issue, 2 Yale L. & Pol'y Rev. 321 (1984). According to this view, pornography is seen as causing and perpetuating patriarchal domination and antiwoman violence and is a practice of sex discrimination. Adopting this approach, two cities attempted to define pornography strictly in terms of the manner in which women are purportedly harmed. See Att"y Gen. Comm"n on Pornography, Final Report 392 (1986). These efforts were provisionally rejected by the courts. See American Booksellers Assoc., Inc. v. Hudnut, 771 F.2d 323 (7th Cir. 1985), summarily aff'd, 475 U.S. 1001 (1986). Another competing vision consists of a revised version of the Miller standard. Instead of using community standards, the proponents of the revised Miller standard advocate the creation of a per se list of sexual activities which are automatically and irrevocably deemed obscene. See Bruce A. Taylor, A Proposal for a Per Se Standard, 21 U. Mich. J.L. Ref. 255 (1987-88).
[94] This question goes to the heart of the Miller standard. In Miller, the Supreme Court approved the notion that communities desiring to exclude marginally valuable sexual expression should be able to do so without regard to the status of similar expression in other communities. Miller, 413 U.S. at 30-34. Implicit in the concept of community, as it relates to obscenity, is the idea that a community necessarily involves a geographic limitation. Whether this geographic limitation should be discarded in the context of cyberspace poses an intriguing set of questions. If, for instance, Cyberspace were held to be a separate and distinct community and that even the most graphic sexual images do not violate Cyberspace community standards, will the traditional, geographically defined community then be compelled to permit such images? This would clearly thwart the Miller Court's attempt to protect geographically defined communities" ability to regulate marginally valuable sexually explicit expression. Conversely, if Cyberspace community standards prove to be more restrictive than the standards of the geographic community where a pornographic image is downloaded, should the government be confined to Cyberspace community standards?

This dilemma is made more complex by the Supreme Court's decision in Stanley v. Georgia, 394 U.S. 557 (1969). In Stanley, the Court identified a penumbral privacy right to possess obscene material within the confines of one's home. Since computer transmitted pornography most often comes directly into the home, it is at least arguable that Stanley operates to immunize computer-transmitted pornographers from criminal liability. If this argument were accepted, the idea that Cyberspace should be considered an independent community would, oddly, be one of boudoirs and bedrooms connected to one another only by telephones.

Lastly, if one chooses to view computer acquisition of obscenity as merely one method, among many, of obtaining obscene material, another logical view is that Cyberspace does not constitute a community at all. A philosophical discussion of "community" is beyond the scope of this article. In sum, the entire conception of community and community standards may be revisited by federal courts as digital pornography becomes more prevalent.

[95] The fact that other Amateur Action digitized images have not been adjudicated obscene does not mean that these images are conclusively non-obscene. Rather, federal prosecutors simply selected a cross-section of representative images available on Amateur Action to bring to trial. Thus, it is probable that a number of Amateur Action images constitute obscenity, but have not as yet been adjudicated as such.
[96] Although it is common for other BBS to offer menus which divide images into basic categories, such as vaginal, anal and lesbian, Thomas purposely avoided any classification of his portfolio. The principal investigator, who has spoken personally with Thomas, will explore the reasons for this in a future paper.
[97] Amateur Action BBS is not a market leader for B&D/S&M imagery. Of the 35 "adult" BBS discussed here, Amateur Action BBS accounted for only 6.1% of available B&D/S&M imagery.
[98] See, e.g., Catharine A. MacKinnon, Feminism Unmodified: Discourses on Life and Law (1987); Andrea Dworkin, Letters from a War Zone: Writings, 1976-1989 (1989); Carlin Meyer, Sex, Sin, and Women's Liberation: Against Porn-Suppression, 72 Tex. L. Rev. 1097 (1994).
[99] Distinctions between different types of sexually explicit imagery and their impact have been made in the past by many observers. The distinction between "erotica" and "pornography" was first made by Gloria Steinem in 1978. See Gloria Steinem, Erotica and Pornography: A Clear and Present Difference, Ms. Mag., Nov. 1978, at 53. Steinem suggested that while "erotica" depicts "mutually pleasurable, sexual expression between people who have enough power to be there by positive choice," "pornography" depicts "domination and violence against women ... unequal power that spells coercion ... sex being used to reinforce some inequality." Id at 54.
[100] The mean popularity index indicates the average number of times an image from a particular classification has been downloaded.
[101] AA-11412; one download.
[102] The reader should keep in mind that due to the precedence scheme used, this low number reports fellatio only as it overlaps with other hard-core classifications, but not with pedo/hebephile and paraphilia. Thus, the actual number of images involving fellatio is slightly higher. The exclusion turns out to be useful because the bestiality image cited infra note 107 does not directly relate to the power imbalance between the sexes. See infra notes 104-10 and accompanying text (describing possible degradation of women in pornography).
[103] More study is needed to determine what effect such marketing techniques have on how consumers perceive or fantasize about the images.
[104] This assumes, of course, that bestiality is a "degrading" act. On several Usenet boards, including alt.soc.feminism, the research team encountered many who insisted that such acts are not degrading, and that at least some women enjoy cohabiting with beasts. It should be noted that the Dietz-Sears study found it necessary to identify three views of degradation: traditional, moderate, and liberal. Dietz-Sears, supra note 15, at 30-33. According to Dietz-Sears, even the most liberal view would find bestiality degrading. Id. at 33. The research team t akes no position on this issue.
[105] A preliminary analysis on urophilia, coprophilia, fisting, B&D/S&M, and enema suggests a similar disproportion.
[106] AA-8630; 672 downloads. Many additional pages could be devoted to discussing this particular image. Specifically, Thomas cleverly interweaves physical attributes, such as hair color; demeaning language, such as "slut;" and value-neutral words indicating magnitude, such as "huge" and "tight."
[107] AA-7806; 597 downloads.
[108] AA-9854; 153 downloads.
[109] With forty telephone lines, each connected to a 28,800 bps modem, Thomas generates approximately 1400 downloads per day. Curiously, there does not appear to be a clear correlation between the number of telephone lines maintained by a BBS and the number of downloads. The Windy City Freedom Fortress BBS generated more than one million downloads with only eleven telephone lines. This may be due to the fact that its operator, Robert Copella, former vice-president of research and development for the Rand-McNally Corporation, devised a means of maximizing customer calls while minimizing the number of available telephone lines. Mr. Coppella is currently in jail in New Jersey, charged with several federal child pornography offenses and awaiting trial related to his alleged operation of a child pornography BBS in Mexico.
[110] AA-16996; 21 downloads.
[111] Thomas does not make it clear whether his promotional descriptions are intended to be taken as truth, fantasy, or both. It should be noted that the First Amendment may not protect his misleading advertisements. States are permitted to regulate commercial speech which is false, deceptive, or misleading. See Virginia State Bd. of Pharmacy v. Virginia Citizens Consumers Council, Inc., 425 U.S. 748, 771 (1976).
[112] AA-16590; 365 downloads.
[113] AA-7480; 219 downloads; paraphilias: voyeurism, foreign objects, incest.
[114] AA-8506; 228 downloads.
[115] AA-8589; 214 downloads.
[116] Because child pornography necessarily entails the direct sexual abuse or calculated sexual exploitation of a child, the U. S. Supreme Court has ruled that child pornography can be prohibited regardless of community standards. New York v. Ferber, 458 U.S. 747 (1982). In order for material to be legally proscribed under child pornography laws, it must be 1) a visual depiction; 2) depicting a minor; 3) engaged in sexually explicit conduct. 18 U.S.C. +s 2256 defines the phrase "sexually explicit conduct." Under +s 2256, sexually explicit conduct includes actual or simulated intercourse, actual or simulated masturbation, and lascivious exhibition of the genitals. The concept of lascivious exhibition has recently been the subject of some debate. See, e.g., United States v. Knox, 32 F.3d 733, 744 (3d Cir. 1994), cert. denied, 115 S. Ct. 897 (1995) (holding the exhibition of a child's covered genitalia may violate federal law because "lascivious exhibition" does not require child nudity).

Robert Thomas is presently under federal indictment in the District of Utah for allegedly selling child pornography over his Amateur Action BBS and is awaiting trial. United States v. Thomas, No. 94-CR-107J (D. Utah filed July 20, 1994).

[117] The research team found that the vast majority of pedo/hebephilic imagery on Amateur Action BBS were of pre-pubescent children. Thus, the term "pedophilic" is used here.
[118] Since his indictment and conviction on obscenity charges in Memphis, Thomas has made changes to his textual descriptions of at least 196 images available on Amateur Action BBS, including many which depict pre-pubescent children. In broad terms, these descriptive changes attempt to deemphasize the focus of these images on the children's genitals or the sexually explicit nature of the images. The actual content of the images themselves have not changed.
[119] This may be due to the fact that pedophiles believe Thomas is practicing "deceitful marketing" whenever he markets hebephilic-hard-core imagery as pedophilic. The significant possibility of law enforcement action is another possible reason why Thomas's pedophile clients choose the soft-core images of children.
[120] The first level comprises the four major classifications. The second layer comprises the specific categories (e.g., orgy, fisting, etc.).
[121] AA-12330; 1495 downloads.
[122] AA-12542; 1024 downloads.
[123] Usenet bulletin boards also serve as a vehicle for computer users to make requests for trades away from public traffic. These trades may involve the use of "drop" or "underground" sites without the owner's knowledge or consent. This may also be done with the full consent of the owner, who wants knowledge of the existence of the site restricted.
[124] AA-042; 1687 downloads.
[125] See supra note 2 for discussion of the difference between obscenity and child pornography and the lack of legal overlap.
[126] A contingency table analysis of pedo/hebephile and bestiality confirmed that Thomas carefully avoids paraphilic descriptions when marketing images to pedophiles.
[127] Because Amateur Action BBS has been in business for more than four years, its cumulative number of subscribers is higher. Other "adult" BBS sysops report the average number of images downloaded by a subscriber to be between 75-150, but no data was available regarding the standard deviation. If one were to assume an average of 112 images downloaded per customer, Thomas would have had a total of 35,644 customers (3,700,000/112). However, some of these may be the same subscribers who renewed their subscriptions. Thus, the estimate of 35,644 cumulative subscribers should be taken only as a very rough approximation of the total number of subscribers to the Amateur Action BBS.

Applying a similar analysis to the other 34 "adult" BBS suggests a very rough estimate of 75,892 subscribers whose consumption patterns were analyzed for this article. This does not include the other 33 "adult" BBS for which no download information was available, including one of the largest "adult" BBS discovered, Event Horizons, which reports more than 30,000 active subscribers. This also does not include those who obtain pornographic images via the Usenet, ftp sites, private trades, or other "adult" BBS.

Usenet statistics estimate the number of active users of alt.pictures.binaries.erotica at 260,000 worldwide. See supra Table 2. Due to overlap of readers within the .binaries hierarchy, there is no way of determining the total number of readers of pornographic Usenet newsgroups, but it appears to be notably higher. See, e.g., supra text accompanying note 126, which suggests that pornography consumers often comprise mutually exclusive subsets according to sexual tastes. Moreover, because this article demonstrates that pornography consumption at at least one university is substantially higher than among the general population, see supra section entitled Comparisons: Univerisity and Worldwide Results, it may be that the worldwide sampling underreports the total number of readers of pornographic Usenet newsgroups because it undersamples university sites. However, because "adult" BBS were found to use the Usenet to advertise their products, the subscribers to commercial "adult" BBS may merely be a subset of those accessing pornography via the Usenet. See supra section entitled Origins of Usenet Pornographic Imagery.

Finally, the "adult" BBS market, as with the BBS market in general, appears to be expanding rapidly. This article reports data from commercial BBS as of June 1, 1994. The most recent editions of Joy of Cybersex, The Adult BBS Guidebook, and the alt.bbs hierarchy of the Usenet contain listings of many "adult" BBS which opened in the past twelve months and were not included in this article. For instance, six months after the first dataset was collected for this study, the research team obtained a list of new BBS in the Chicago area. Twenty-three were advertised as "adult," but none of them had been available for inclusion in the original dataset. In sum, while the reader is cautioned against interpreting the 8.5 million downloads discussed here as alarmingly high, he or she is equally cautioned against attributing the total number of downloads of sexually explicit materials in Cyberspace, both reported and unreported by this article, to a disproportionately small percentage of the computer literate population.

[128] Thus, the total number of people who consume AABBS images is higher. Until "pirating" images in this fashion became commonplace, Thomas used to license each image to other "adult" BBS for $10 each. Last year he indicated to the principal investigator that he no longer had any objections to redistribution without a licensing fee, because it was "good advertising." It is interesting to note that Thomas claims to own copyrights to many images which members of the research team obtained independently from computer sources around the world, many of which were scanned directly from previously copyrighted magazines.
[129] When Thomas was tried in Memphis for transmitting obscenity, prosecutors elected not to charge him with computer transmission of child pornography. Instead, he was charged with receipt of child pornography through the mails after postal inspectors delivered a child pornography magazine to him, allegedly at Thomas's request. This was the sole count for which he was found not guilty. See Quittner, supra note 90, at B27.
[130] The Carnegie Mellon research data suggests that paraphilic pornography has been "democratized" and rendered ubiquitous by computer networks in the late 1980s and early 1990s. According to Lynn Hunt, Annenberg Professor of History at the University of Pennsylvania: "Although desire, sensuality, eroticism and even the explicit depiction of sexual organs can be found in many, if not all, times and places, pornography as a legal and artistic category seems to be an especially Western idea with a specific chronology and geography." Lynn Hunt, The Invention of Pornography 10 (1993). While computer networks may seem at first glimpse as merely an extension of this chronology (e.g., the printing press, photograph, magazine, video, and computer technologies) and geography (globalization), there appears to be little historical precedent for the widespread availability of paraphilic pornography. Most "explicit" depictions discovered among ancient ruins of Greece and Rome would be classified by this study as "hard-core," not paraphilic. For instance, representations of Priapus, known for his gigantic erect phallus, were excavated from the ruins of Pompei, and were often declared "obscene" by the cultures responsible for such excavation. See Walter Kendrick, The Secret Museum: Pornography in Modern Culture 8 (1987). Although statues such as one "representing a satyr in sexual congress with an apparently undaunted goat," were also excavated from Pompei, it does not appear that such works were widely produced or consumed in a fashion analogous to modern pornographic consumption. Id. at 6.
[131] One of the suggestions of the 1986 Attorney General's Commission on Pornography was that pornography had, in comparison to previous decades, become increasingly violent, especially in its depiction of women. Att"y Gen. Comm"n on Pornography, Final Report 386 (1986). While the computer pornography studied here is heavily paraphilic, the research team has not yet determined the extent pornographers rely on violence against women to market their imagery. Such an undertaking would first require agreement on what constitutes a "violent" act. For instance, "She screams in pain as he fucks her pussy with a bat" might convince most judges to classify the image as violent. However, "She smiles as he fists her virgin asshole" poses more complicated questions (e.g., whether the model is really smiling, or really enjoying the act). A dictionary could be developed which identifies and tabulates the frequency of such words as pain, torture, abuse, etc. This data could further be sorted according to gender, race, and ethnic origin. Because there may be some discrepancy between what the judge and the pornographer or model considers a violent act, multiple dictionaries which consider different points of view would likely have to be developed.
[132] The animation was initially distributed by (or pirated from) Davide Toma and Luca De Gregio, Milano, Italy.
[133] An instructor at the Carnegie Mellon College of Fine Arts confirmed that the image was not digitally manipulated or "doctored."
[134] See supra note 21 and accompanying text.
[135] See supra note 9.
[136] These are very tiny images that customers can use to quickly preview larger images while online.
[137] There are many technical problems associated with regulating sexually explicit imagery on the Internet and Usenet. See supra note 21. Two additional possible solutions are discussed here: content flags and authentication procedures. Content flags could identify the content of computer files transferred or posted on the Usenet and could have an internationally agreed upon standard definition, to which all Internet member countries assent. Computer users would be permitted to send whatever files/traffic they liked within their community or country, but transfers between communities or countries could be filtered based on local pornography laws.

Each government and, indeed, each site can regulate what sort of traffic they will receive. For example, a computer network run by a religious organization may elect not to carry any sexually explicit material, while major service providers may choose to carry everything permitted by law. Parents who do not want their children to see pornographic images could establish a filter from their homes. If successfully implemented, content flags might allow the government to regulate obscenity without unduly restricting access to other sexually explicit material which is protected by the First Amendment. The obvious difficulty will likely be enforcing a content flag requirement, both on a national and international basis.

To identify the senders of child pornography, or those sending obscene materials without the correct content information, some form of authentication is required. Simply pointing back to a hostname is not viable, because hostnames might be forged. Similarly, pointing back to an IP address, although more reliable, only identifies the site responsible for that particular block of IP addresses. This may work well for entities such as CompuServe, or Carnegie Mellon University, where complying with the law is generally expected, but many other sites exist on the Internet which may not feel obligated to comply with the law as scrupulously.

Without the help of computer system administrators, law enforcement agencies would have to expend an enormous amount of resources tracing a site, which may be connected to another site which is connected to a third site, etc. For authentication to be practical, sites might have to be connected according to a hierarchy, with responsibility defined at each level.

Furthermore, the implementation of content flags and authentication procedures may raise constitutional concerns. With content flags, the government would essentially be requiring computer users to identify the nature of their speech. One question is whether this constitutes "compelled speech," which the Supreme Court has viewed with some suspicion. See, e.g., Wooley v. Maynard, 430 U.S. 705, 713 (1977) (striking down state statute requiring all license plates to display state motto); West Virginia State Bd. of Educ. v. Barnette, 319 U.S. 624, 642 (1943) (striking down state resolution requiring all public school children to salute the American flag and recite Pledge of Allegiance). But see Red Lion Broadcasting Co. v. FCC, 395 U.S. 367, 375 (1969) (upholding FCC order requiring broadcasters to air replies to political editorials).

Moreover, in the case of authentication procedures, the government may be requiring computer users, arguably engaged in expressive conduct, to identify themselves. Generally, the First Amendment protects the anonymity of those transmitting speech. The Supreme Court has invalidated a requirement that prohibited the distribution of leaflets which did not identify the author of the leaflet. Talley v. California, 362 U.S. 60, 64-65 (1960). While trading digitized images on computer networks is vastly more complex than leafletting, the question of anonymity continues to be important. Thus, the Supreme Court's anonymity cases may apply to Cyberspace. Because the apprehension of child pornography producers, distributors, and consumers is a particularly compelling government interest, see New York v. Ferber, 458 U.S. 747, 761 (1982), carefully drawn authentication procedures may pass constitutional muster.

Even if content flag and authentication requirements are implemented, other forms of electronic communication, such as ftp and World Wide Web sites, pose even more serious challenges to law enforcement. Child pornographers will inevitably develop more effective means of circumventing the law. Currently, they set up anonymous mail bouncers to redirect traffic elsewhere, with a new authenticated identification.

[138] Indeed, several BBS contacted by the research team, when asked if they had "any good beasty pics," offered to sell them privately on diskette.
[139] See, e.g., John Tierney, Porn, The Low-Slung Engine of Progress, N.Y. Times, Jan. 9, 1994, at B1.
[140] Id. In 1978 and 1979, when fewer than 1% of American homes owned VCRs, more than 75% of the videocassettes sold were pornographic. Id.
[141] See Entertainment Adult: Web Access to XXX-Rated Movies and Novelties, Feb. 16, 1995, posted to alt.sex.erotica.marketplace.
[142] Id.
[143] Id.
[144] Anne Wells Branscomb, Who Owns Information? 3-4 (1994).

The collection, retention, and improper use of personal information by businesses is not a phenomenon unique to Cyberspace. In 1972, in California, voters amended the state constitution to include privacy as an inalienable right prohibiting intrusive business practices. See Cal. Const. art. I, +s 1; White v. Davis, 13 Cal. 3d 757, 773-75 nn. 8-9 (1975) (discussing the purposes and applications of the state constitutional amendment). Importantly, one of the reasons California voters felt compelled to amend their constitution is that the federally protected penumbral right to privacy operates only against state action. See Hill v. NCAA, 7 Cal. 4th 1, 28-29 n.8 (1994) (contrasting California right to privacy with federal privacy right).

In hindsight, California appears to have been prescient. With the aid of computer software applications which can quickly analyze and categorize personal information, businesses can become far more intrusive. It would be erroneous to assume that bulletin board operators who offer extremely explicit, possibly illegal digitized pornography for sale will not collect and use personal information. In fact, the potential for blackmail is readily apparent assuming that few, if any, persons would want their personal pornography consumption habits to be made public. As with California, there may be increasing calls for enacting state and local legislation, which reaches beyond state action, to protect against such abuses of personal information.

[145] After seven years of compiling data about its users, the university has since decided to automatically block public access to these logfiles on all new accounts.
[146] This is how the worldwide statistics for this study were compiled. However, the statistics reported here do not implicate any individuals, as network administrators apparently did not abuse their access privileges.
[147] Usenet newsgroup discussion posted to comp.org.eff.news.
[148] Hunt, supra note 1, at 35.
[149] Data collected in May 1995 regarding an additional 2.1 million downloads suggests only minor differences in consumer demand, with the exception of bestiality and fisting imagery.
[150] See Matthew Gray, Growth of the World Wide Web, available online at URL http://www.netgen.com/info/growth.html
[151] See Internet Society Domain Survey, cited in "WebWeek" Weekly Tabloid Introduced, Apr. 4, 1995 posted to listserv <newsbytes@clarinet.com> (NB-MSP) (on file with author).
[152] See The First and Second GVU Center WWW User Surveys, available online at URL http://www.cc.gatech.edu/gvu/user_surveys/User_Survey_Home.html (finding a female user population of only 5-10%).

Copyright (c) 1996, Three Rivers Free-Net,
Carnegie Library of Pittsburgh. All Rights Reserved.