Freedom of Speech in Cyberspace from the Listener's Perspective: Private Speech 
Restrictions, Libel, State Action, Harassment, and Sex

Eugene Volokh*


I. Edited Conferences

II. Government Protection of Listeners Against Offensive Messages III. Sexually Explicit Material and Minors Conclusion
       Speakers' desires are fairly simple: generally, they want more listeners.1 But listeners don't just want more speakers talking to them. Listeners want more control over their speech diet--a larger range of available speech coupled with greater ease of selecting the speech that's most useful or interesting to them. 
       The success of the new electronic media in the "marketplace of marketplaces" of ideas--where information providers compete for that scarcest of resources, the attention span of modern man--will turn on how well they can satisfy listeners' desires.2 The new media have one significant advantage: they can give listeners many more choices. But for listeners, that's not enough. For listeners, what the new media omit--time-wasting junk, insults, material that might be harmful to their children--is just as important as what they include. Listeners care about this outside the online world, and they care about it just as much online. 
       In the following pages, I will discuss three categories of online speech issues and look at them partly, though only partly, through the lens of the listeners' interests: 
       1. Edited Electronic Conferences: One of the most significant features of the new media is the interactive electronic conference bulletin board, newsgroup, discussion list, or the like. People who listen in on these conferences (and most participants spend much more of their time listening than speaking) want speech that's relevant to their interests, readable, reliable, and not rude. Sometimes an open, unedited electronic conference can provide this, but often it can't. Often--as conference operators have been learning--editing is critical to making online speech worth listening to.
       At the same time, editing is content control, the sort of thing that, if the government did it, would be called "censorship." It includes limitations on who may speak, removal of people who speak badly (in the editor's opinion), the deletion of inappropriate messages, and automatic screening of messages for profanities. Many have expressed concern about this sort of private speech restriction.
       In Part I, I will defend the propriety of private, nongovernmental, content control on electronic conferences. I'll argue that: 
       - Conference operators should generally have the right to decide what's said on their conferences and who says it. 
       - Libel law generally ought not penalize conference editors by imposing on them extra risk of defamation liability beyond what operators of unedited conferences must bear. 
       - The conference operator's power to edit should remain even if the conference is run on a government-owned computer or if the operator is a government employee.
       2. Avoidance of Offense: Listeners don't want to hear material that offends them. This doesn't mean they only want to hear what they agree with; controversy is usually more fun than agreement. But some speech is offensive enough that its emotional cost to the listeners can exceed the informational benefit they derive from the conversation.
       No one likes to be personally insulted. No one likes to hear one's race, sex, religion, or deeply held moral beliefs rudely attacked. Often, we're discomfited even by watching others argue rudely with one another. Some speech like this can be annoying; some can ruin one's mood for hours. People don't go to parties where they think it likely that the other guests will be rude to them--neither do they want to participate in electronic conferences where this happens.
       In Part I, I'll argue that private editing is an important tool for giving people the opportunity to interact in the polite environment they may prefer. In Part II, I'll discuss what the government may do to protect people from speech that offends them when private editors can't or won't edit it out. I'll suggest that: 
       - In general, the government ought not be able to restrict offensive speech in electronic conferences (unless it's a threat or falls into some other Free Speech Clause exception). Some telephone harassment laws, and possibly some aspects of hostile environment harassment law, seem to already impose restrictions on online speech,3 but to the extent they do, they're unconstitutional. 
       - On the other hand, the government should be more able to restrict one-to-one speech--such as personal e-mail--that's aimed at unwilling listeners; such restrictions protect unwilling listeners while still leaving speakers able to communicate with willing ones. The best way to implement such a restriction would be to let listeners demand that a speaker stop sending them direct e-mail, a power people already enjoy with regard to normal mail.4 Such a rule would be better than direct extensions of telephone harassment laws, which often embody dangerously vague prohibitions on speech that's "annoying" or "harassing."
       3. Giving Parents Control Over Their Children's Access: Finally, online as well as offline, parents are concerned about their children gaining access to sexually explicit materials; and, online as well as offline, the question becomes how the law can restrict children's access without also restricting the access of willing adult listeners. In Part III, I'll suggest that: 
       - Some laws may already prohibit the online posting of nonobscene sexually explicit material that might be "harmful to minors" (a term of art described below).5 
       - These laws have been upheld in the offline world, partly because they've been seen as not imposing much of a burden on adults who want to get the material. Online, though, where it's hard to tell who's a child and who's not, these laws are much more burdensome to adult viewers. 
       - These laws ought to be unconstitutional online because there's a less speech-restrictive alternative which could protect children while maximizing the choices available to adult listeners: a self-rating system that would identify which images or discussions are sexually explicit. This approach will still impose something of a burden on speakers and adult listeners, but this burden should be constitutionally permissible.
       In focusing on listeners, I don't mean to suggest that listeners' rights are generally more important than the rights of speakers. After all, the Free Speech Clause guarantees "the freedom of speech," and much of the Court's doctrine has--in my view correctly--protected speakers, even where most listeners might object to what the speakers are saying.
       But this very emphasis on the rights of speakers can lead people to ignore the rights of listeners, rights the Court has also recognized. And worse still, focusing exclusively on the rights of speakers can make us ignore how critical listener satisfaction can be to the survival of the new media. If we think the new media can be valuable tools for public discourse, it's worth trying to make sure that the law doesn't make them unattractive to the listening public.
A. Electronic Conferences and Their Hazards
       Listeners want to hear more of what interests them and less of what doesn't. This becomes especially important for electronic conferences, some of the biggest speech activities on the infobahn.
       These conferences are electronic "places" where people from all over the world can communicate with one another on a particular topic, from the law of government and religion6 to Jewish issues in Star Trek.7 Electronic conferences can be organized as Internet discussion lists;8 as Internet news groups;9 or as special dial-in services, ranging from the big public ones like Prodigy, America Online, and Compuserve, to the smaller and more specialized ones like Counsel Connect, to single-PC bulletin boards that may have only a few hundred subscribers.10 Regardless of implementation, though, these conferences are means by which each of hundreds or thousands of participants can talk to all the others, and will have to listen to what the others have to say.
       The conferences are like faculty or law firm symposia, where everyone present can speak (though not all at once) and comment on what everyone else has said. They are, however, symposia that go on continuously, and that can include hundreds of people who've never physically met one another. And, as in a symposium, though everyone has an opportunity to speak, the overwhelming majority of participants are "lurkers," people who only listen.
       The advantage of these conferences over the traditional media is their openness and interactivity; but this is also their great risk. An electronic conference is a compilation of the messages (or "posts") written by all its participants, and as with any compilation, its value lies both in the substance of the materials it contains and in their selection. As with other compilation media, such as radio programs, magazines, or live conferences, people look for high ratios of wheat to chaff (or, as computer people say, "signal to noise"): electronic conferences in which they find a large fraction of the speech to be interesting.11
       I've seen many users turn away from electronic conferences, even conferences on topics that interest them, because the signal-to-noise ratio was too low. Conference users have limited time, and messages that are irrelevant to the conference topic, messages from people who don't know what they're talking about, and messages that are repetitive all make the conference less valuable. People want conversations of higher quality than talk radio.12
       Conference users also want an emotionally congenial environment. It may be a pleasure to listen to people discuss an issue civilly, but a strain to listen to them yell at each other. Even if the intellectual content is the same, the tone of the speech can be a serious burden.
       Few physical conferences, for instance, invite speakers who insult one another. Many newspapers refuse to print certain profanities. We don't go to clubs or parties where we know boors are likely to be declaiming; similar online conduct can make an electronic conference much less valuable for us. One bad apple on a discussion list can spoil many people's enjoyment. And as more people get online and begin to use online resources, the risk of information overload and of the occasional rude participant escalates.
       Most electronic conferences try to keep both the signal-to-noise ratio and civility high by moral suasion.13 Each conference has an official topic; people generally know not to stray too far from it, and if they do, others might ask them to "take it off-list", continue the discussion in personal e-mail rather than on the electronic conference. When people start getting rude, others might chime in to quiet everyone down. Most conferences have conference operators who are in charge of the technical details of conference administration; they often also take responsibility for informally keeping everyone in line.
       Increasingly, though, conference operators have begun to edit more coercively, in several ways:
       1. They may limit who can access their conference. Counsel Connect, for instance, is limited to lawyers; the LAWPROF Internet discussion group14 is limited to law professors. The theory is that those are the people who are most likely to contribute something valuable to the discussion.15 List operators have discretion, of course, to waive these rules either in favor of admission or of exclusion in particular cases.
       2. They may kick out troublemakers, people who have proven to be consistently off-topic, or rude, or kooky as always, in the operator's judgment.16
       3. They may automatically filter all messages to exclude particular words, such as profanities.17
       4. They may manually screen each message that's sent to the conference before actually passing it along to all conference participants. This can, of course, be a time-consuming process, though for many conferences it might not be prohibitively so.
       And, as with newspapers and physical conferences, editorial decisions make a big difference in the conference content on three different levels. First, editing decisions directly dictate what can be said. A conference that's open to everyone will be different from one that's open only to law professors. A conference that allows militant and even rude debate will be different from one that requires more gentility. A conference on, say, religious freedom that doesn't allow posts which rely on explicitly religious foundations (on the theory that such posts are likely to distract discussion into unresolvable theological arguments) will be different from one that takes another approach.
       Second, editing one post can cut short a whole discussion. Because conferences are interactive, one post leads to others. Excluding one rude message may avoid dozens of messages responding to it, responding to the responses, commenting on the merits of polite debate over rude debate, and so on.
       Third, including or excluding certain messages will change who participates. Some people will get involved in polite discussions, both speaking and listening but will turn away if the discussion turns nasty. Others will participate only so long as there are comparatively few posts that they see as irrelevant. A law professor might be willing to read a law professors-only conference but not be interested in a general public conference, or vice versa.
B. The Right to Edit
    1. The right to exclude content.
       Where there are editors, there are speakers who resent being edited out. Prodigy, for instance, has been criticized for its editing practices, which have at times included automatically screening out profanities, deleting messages that denied the existence of the Holocaust, deleting messages criticizing Prodigy's pricing policies (and urging a boycott of Prodigy), and kicking off users who persisted in posting these messages.18 Similarly, operators of Internet discussion lists who try to restrict what's said on the list, and by whom it's said, can expect a good deal of resistance.19
       Private conference operators clearly don't violate the Constitution by editing, just like newspaper editors don't abridge free speech by refusing to publish a letter to the editor. Some have argued that private providers' use of government-funded or government-regulated Net backbones makes them state actors and thus bound by the First Amendment; this, though, is certainly not so under existing state-action doctrine.20
       But editing is not only constitutional, it is constitutionally protected. A law that would, for instance, prohibit conference operators from screening messages, or even allow screening for relevance but prohibit screening for viewpoint, would violate the First Amendment. As the Court held in 1974 in Miami Herald Publishing Co. v Tornillo21 and recently reaffirmed in Hurley v Irish-American Gay, Lesbian and Bisexual Group,22 the freedom to speak includes the freedom to create one's own mix of speech.
       A parade, a magazine, the playlist of a music radio station, and an edited electronic conference are all speech products created by an editor. "Rather like a composer, the [editor] selects the expressive units of the [publication] from potential participants" to create a particular work.23 Restricting the editor's right to edit would be restricting his right to create a particular speech product: it would make illegal the production of certain speech mixes and require instead the production of others. "The choice of material to go into a [publication], and the decisions made as to limitations on the [publication's] size and content . . . constitute the exercise of editorial control and judgment,"24 and this editorial judgment is constitutionally protected from government interference. This is the rule for newspapers and parades, and it should be the same for electronic conferences.
       Of course, many conference operators don't have a specific ideological perspective they want to communicate. They may just want to spread information about a certain topic and may exclude material only because they think it's irrelevant, impolite, or inaccurate, not because it clashes with their viewpoint.
       But "a narrow, succinctly articulable message is not a condition of constitutional protection."25 The decision by the St. Patrick's Day parade organizers to exclude one group was protected even though the remaining parade wasn't communicating much by way of a specific ideology. The editorial choices of a nonpartisan newspaper that assiduously tries to avoid all political bias are as protected as other papers' choices. The editing decisions of a conference operator are equally a conscious attempt to create a speech product with a particular content; their claim to being protected exercises of editorial judgment is as strong as that of the decisions of the newspapers or parade organizers.
       Nor should it be relevant that electronic conferences are less selective than newspapers and magazines. It's true that newspapers and magazines publish only a small fraction of what might be submitted to them, while even edited electronic conferences tend to let through almost all the messages that people try to post.26 But this lower selectivity shouldn't keep the editors' editorial judgments from being protected.
       Technology eliminates the need to edit for paper-saving reasons, but the editors' desires to edit for germaneness, civility, and even viewpoint remain as legitimate as they are in the newspaper context. Compelling operators to give access to the conferences to everyone probably won't cost the operators anything out of pocket, but it will cost them the ability to create the speech product they prefer. As the Court has held, "[e]ven if a newspaper would face no additional cost to comply with a compulsory access law and would not be forced to forgo the publication of [its own chosen materials] by the [compelled access]," the First Amendment would still prohibit the compelled access law's "intrusion into the function of editors."27 And, of course, Hurley made clear that the editorial function remains protected even when, as in a parade, the editors rarely exercise it.28
       The Court has, in two contexts, upheld laws that require private property owners to let others speak on their property, but neither of these narrow exceptions should be applicable to electronic conferences:
       Broadcasting: The Court has tolerated "more intrusive regulation of broadcast speakers than of speakers in other media";29 in particular, Red Lion Broadcasting Co. v FCC30 upheld a rule requiring broadcasters to give time to opposing views. But the Court has refused to extend its relatively deferential scrutiny of broadcasting controls to other media, such as newspapers or cable TV.31 Red Lion has been read as turning entirely on "the unique physical limitations of the broadcast medium"--the physical scarcity of available broadcast channels.32 No such limitations exist for electronic conferences; there are thousands of conferences available to everyone who has Internet access (which includes users of Prodigy, CompuServe, America Online, and similar services). Even if one counts only the three big services, three is still more than the number of cable operators or large local newspapers that serve the typical city. Given that the Court has refused to apply Red Lion to cable and newspapers, I don't see how it could justify applying Red Lion to electronic conferences.
       Content-Neutral Access Mandates: In two cases, the Court at least partly approved content-neutral speaker access mandates. Turner Broadcasting System v FCC33 indicated that the government may in some circumstances require cable-system operators to leave open some channels for local broadcasters. PruneYard Shopping Center v Robins34 held that state law may require shopping center owners to let members of the public speak in the center's public areas. The more recent Hurley decision, though, makes clear that these cases wouldn't justify even content-neutral access mandates to electronic conferences.
       An electronic conference, like the parade involved in Hurley or the newpaper in Miami Herald, is a more or less coherent speech product, one whose content is a function of all its components. A parade organizer, newspaper editor, or conference operator may solicit speech from the public, and may decide to let much of it through unedited; but this is just one possible choice on his part, and he might equally well choose to fashion his speech product out of only a certain set of messages. When the government requires a conference operator to include speech that he would prefer to exclude, it's ordering the operator to change the character of the information the conference conveys. Just as a St. Patrick's Day parade which includes an "Irish American Gay, Lesbian and Bisexual Group of Boston" banner communicates something different from a St. Patrick's Day parade which excludes this banner,35 so an unedited conference communicates something other than an edited one.
       PruneYard, as Hurley pointed out, "did not involve `any concern that [mandated access by other speakers] might affect the shopping center owner's exercise of his own right to speak.'"36 Shopping centers aren't usually in the speech business; "[t]he selection of material for publication is not generally a concern of shopping centers."37 The speech by members of the public didn't interfere with any messages the shopping center was trying to communicate. "The principle of speaker's autonomy was simply not threatened in that case."38
       It's true that in some cases compelled access to a shopping center can indeed interfere with the center owner's speech. If the owner, for instance, decided to have a patriotic Fourth of July festival, letting flagburning protesters on the property might affect the center's speech as much as would letting unwanted signs into a parade. But such coordinated expressive activity--the use of the shopping center to communicate a message or a set of messages--is the exception rather than the rule. PruneYard didn't foreclose the possibility that, in such a context, government-mandated inclusion of other speakers in the festival would be unconstitutional; as Justice Marshall said in his concurrence, the shopping center owners were "permitted to impose reasonable restrictions on expressive activity."39 And the Court has never suggested that the government could compel access to, for instance, bookstores or holiday displays or other places where access might seriously interfere with the owner's own message.40
       Turner Broadcasting did involve a limitation on the property owner's speech--the cable operator could no longer use the channels that were set aside by the law to carry the materials it preferred.41 But, as Hurley pointed out, cable operators have a monopoly; the justification supporting the law in Turner Broadcasting was the survival of broadcast stations that might be threatened when a monopolist excludes them.42 This was the interest that made the law valid, and this interest is absent in the electronic-conference context. The interest served by restrictions on conference editing--the interest in "requir[ing] speakers [here, conference operators] to modify the content of their expression to whatever extent beneficiaries of the law choose to alter it with messages of their own"--is, according to Hurley, "exactly what the general rule of speaker's autonomy forbids."43
       Of course, speakers would prefer to have access to an existing conference, with its established pool of listeners, even when setting up new conferences isn't hard. But, as in Hurley, though "the size and success of [an existing conference] makes it an enviable vehicle for the dissemination of [disparate] views, . . . that fact, without more, would fall far short of supporting a claim that [the conference] enjoy[s] an abiding monopoly of access to spectators."44
    2. The right to exclude speakers.
       Some conference operators may want to limit access based on who a person is--for instance, on the person's occupation, professional standing, political opinions, religion, sex, or race--and not just on what he posts. Many traditional conferences certainly select their speakers this way. Of course, it's often hard to tell these things about a person online, so someone might lie his way into a closed group without much difficulty. But few people generally want to do this, and in any event, even online the truth might come out.
       Sometimes, the person's status may be used as a proxy for his knowledge: Counsel Connect, for instance, is generally restricted to lawyers, largely because lawyers are more likely to talk and think in particular ways, ways useful to other lawyers. If a lawyer asks a question about First Amendment law, other lawyers are more likely to respond by citing cases; laypeople may instead respond with textual arguments ("Congress shall make no law") or moral arguments that lawyers may know are generally not accepted by courts.45
       A person's identity can also be directly relevant to the conference operator's purposes. Some conferences are aimed at dealing with issues facing a particular group. Democrats might want to argue about what the Democratic Party platform should be. Homosexuals might want to debate what the homosexual community's stance ought to be on a particular issue. Southern Baptists might want to discuss what stance Southern Baptist churches should take on homosexuality. Blacks might want to argue how they as blacks should react to Louis Farrakhan; whites might want to debate how whites should deal with problems of police racism; men or women might want to share thoughts on why their own sex is superior. In each situation, people might specifically want to hear the voices of their fellow group members (whatever they have to say) and not of others (no matter how sympathetic to the group they might be). We see these sorts of group-limited symposia often in the offline world, where they are sometimes praised and sometimes condemned.
       Finally, a person's identity might more subtly influence the conference operator's actions. An operator might be more willing to bend the rules for, say, academics or women or whites or agnostics than he would be for others. An operator might more quickly kick off a misbehaving person with a male-sounding name than with a female-sounding name, or vice versa. And regardless of the operator's actual reasons, a person who's denied access to a conference--or who's kicked off after having had access--might believe that his group membership was the reason.
       Can the government bar conference operators from discriminating based on, say, race, sex, religion, age, political opinion, marital status, sexual orientation, profession, or education? The law already bars various forms of discrimination in public accommodations. Private clubs,46 parades,47 and the Boy Scouts48 have all been viewed by at least some government agencies as places of public accommodation; the same logic might be applied to electronic conferences.49 And some jurisdictions prohibit much more than just discrimination based on race, sex, religion, or national origin: The District of Columbia, for instance, also bars discrimination based on "age, marital status, personal appearance, sexual orientation, family responsibilities, disability, matriculation, political affiliation, source of income, or place of residence or business."50
       The Court has never squarely dealt with this. Hurley held only that the parade organizers had the right to exclude speech that endorsed homosexuality--it didn't decide whether parade organizers had the right to exclude homosexuals.51 The Court has several times confronted the question whether barring sex discrimination by private clubs would abridge the club members' rights to freedom of expressive association, but this too is a somewhat different matter. When a club is forced to admit unwanted members, the danger is the possibility that "admission of [the unwanted people] as voting members will change the message communicated by the group's speech."52 When a conference is forced to accept unwanted speakers, the danger is the certainty that admission of the speakers will change the message communicated within the conference.
       Still, Hurley did suggest that the expressive association cases may be relevant to determining whether parades--or, presumably, conferences, whether electronic or not--can discriminate in selecting their participants.53 The rule the Court announced in those cases has two components.
       First, the First Amendment is implicated only when the law will indeed "change the content or impact of the organization's speech."54 To determine whether this is so, a court can't just consider generalizations, even statistically accurate ones, about the beliefs shared by most women or most men (or, presumably, by most members of other classes defined by quasi-suspect or suspect attributes such as race).55 When a law "requires no change in the [association's] creed . . . and . . . imposes no restrictions on the organization's ability to exclude individuals with ideologies or philosophies different from those of its existing members,"56 the organization must "show that it is organized for specific expressive purposes and that it will not be able to advocate its desired viewpoints nearly as effectively if it cannot confine its membership to those who share [a particular attribute]."57
       Second, even if the First Amendment is implicated, the regulation may still be upheld if it's narrowly tailored to a compelling state interest which is unrelated to the suppression of ideas.58 The prevention of sex discrimination, even in organizations which do not actually engage in commerce but only provide leadership skills and business contacts, is such a compelling interest.59 Presumably the same would be true of the prevention of race and national origin discrimination.60 And a prohibition on discrimination in membership is by definition narrowly tailored to the interest.
       Justice O'Connor, recently joined by Justice Kennedy, has taken a different view. Under her approach, the only inquiry must be whether the organization is primarily commercial, which is true "when, and only when, the association's activities are not predominantly of the type protected by the First Amendment"--are not predominantly expressive.61 If the organization's activities are primarily commercial, then it has only minimal expressive association rights,62 and government interference with its membership criteria would be permissible.63 If the organization's activities are primarily expressive, then "both the content of its message and the choice of its members" are protected.64
      Thus, if Justice O'Connor's approach is transplanted from the expressive association context to electronic conferences, most conference operators--all those whose conferences are devoted to something other than commercial transactions--would have the unlimited right to discriminate in membership.
       If the majority view is applied, however, the result is less clear. To begin with, the conference operator will have to show that "the content or impact of the [conference's] speech"65 will be changed if the participant restrictions are lifted. Under one definition of "content" this shouldn't be hard; certainly the content of a conference will be changed whenever new speakers are allowed. On the other hand, if courts insist that the change not just be to the exact words the conference contains, but to some substantive aspects of the discussion, the operator will have to show that, say, nonlawyers or women or blacks as a group will probably have different views on various topics than lawyers or men or whites as a group. And when the group is defined by a suspect or quasi-suspect attribute, such as race or sex, the operator will have to show this using more than just statistical generalizations (even empirically valid ones).66
       Next, the court would ask whether the government has a compelling interest in barring the particular form of discrimination. Such a compelling interest quite likely exists for race, national origin, sex, and probably religion. For other attributes, it's less clear. State courts are, for instance, split on whether preventing discrimination based on marital status, sexual orientation, and age are compelling interests,67 and I know of no decisions dealing with whether there's a compelling interest in barring private discrimination based on, say, political affiliation.
       In my view, Justice O'Connor's framework is the better one, especially when one is dealing with choice of speakers and not just with choice of members in an organization.68 An all-lawyer, all-Republican, all-female, allwhite, or all-Catholic electronic conference presents a unique speech mix for its participants. We might not be entirely happy that some people prefer to talk only to members of their own group, but--especially where no salary or other tangible economic benefits are directly involved69--people's choice of correspondents seems as much a part of the freedom of speech as their choice of what to say or listen to.
       Nor is it proper to allow only those exclusions that are in some way germane to the conference topic--to say that, for instance, women might be excluded from an electronic conference for discussion of men's issues, but not from an electronic conference on, say, bankruptcy law. It shouldn't be the government's job to determine what's germane to an expressive association's purposes and what's not. Excluding women (or men) from a bankruptcy law discussion will definitely change the discussion content; we might not think it will change the content in any remotely interesting way, but presumably the discussion organizers disagree. Hurley tells us it's up to the parade organizer, not the government, to decide whether including a certain message would unacceptably change the parade's message. If Justice O'Connor is right to equate an expressive association's interest in "the content of its message" and "the choice of its members," then the decisions about membership in an expressive association--or an electronic conference--should likewise be in the organizer's hands.
       In at least one area, in fact, antidiscrimination law has been appropriately trumped by the First Amendment; despite Title VII, churches continue to have the right to discriminate based on race and sex in their choice of clergy. "The right to choose ministers without government restriction underlies the well-being of religious community, for perpetuation of a church's existence may depend upon those whom it selects to preach its values, teach its message, and interpret its doctrines both to its own membership and to the world at large."70 Though the interest in stopping race and sex discrimination is normally compelling, in the context of clergy selection it must yield to the church's rights under the Free Exercise Clause.71 Conference participants are to a conference what clergy are to a church: The perpetuation of a conference's distinctive content depends on those whom the operator selects to contribute to it.
    3. Possibly permissible requirements.
       All the above has turned on the conference operator's right to create a coherent speech product. Regulations that don't jeopardize this right are a different story. For instance, a law that required online service providers to offer person-to-person e-mail services to everyone and barred providers from restricting the content of such message would probably be constitutional.
       The same should even be true of a law that required service providers (Prodigy and the like) to give their users the ability to create new electronic conferences. Such laws wouldn't prohibit the creation of any speech products; operators could still edit their own conferences any way they please. (One could, of course, still oppose these laws on the policy grounds that the government generally shouldn't interfere with private businesses, or even argue that such laws may sometimes be takings of private property without just compensation.72)
       It may also be permissible to restrict operators from changing posts (as opposed to deleting them) without the authors' permission. Changing people's posts essentially represents them as having said something they didn't say--it implies a false statement of fact, "X said Y" where in reality X said Z. This sort of knowingly false statement should be constitutionally unprotected.73 
       In fact, such modifications of others' posts may already be prohibited in many situations. Under the Copyright Act,74 both copying someone's work and transforming or abridging it are presumptively infringements. By posting to an electronic conference the author obviously gives the conference operator an implied license to copy the work in order to forward it to the other conference members, but he probably doesn't give an implied license to change it. Putting words into someone's (electronic) mouth may also risk a false-light invasion of privacy lawsuit, and, in a commercial context, a misattribution claim under the Lanham Act.75 If an operator wants to change a conference participant's words, the operator should get the person's agreement, either at the time of the post or beforehand (for instance, when the person signs a Terms of Service agreement that makes clear that certain words will be deleted from all posts).
C. Editing from the Listener's Perspective
       The Value of Editing: I've focused on the interests of the conference editor as speaker, largely because this is what the doctrine has generally done. But, as I mentioned earlier, editing is also critical to the interests of listeners.
       As the Court has recognized, listeners have substantial claims to autonomy in their selection of the speech they hear: "no-one has a right to press even `good' ideas on an unwilling recipient."76 In many contexts, this autonomy can't justify silencing a speaker, because other, willing listeners might be present. But listener choice remains an important value; practically, if a medium can't give listeners what they want, listeners aren't likely to use it.
       Private intermediaries are a vital tool for listener choice. Listeners who want more options are generally better served not by a media outlet which carries everything submitted to it, or even by many such outlets, but by many edited media outlets, each with its own editorial judgment. I'd rather have access to twenty radio stations, each with its own playlist, than to twenty (or even fifty) stations that are all open to all comers. And just as speakers' rights to speak can often be fully realized only by their right to associate to form a more powerful speaker (albeit one that might not always perfectly track the ideas of each of its individual members),77 so listeners' control over what they listen to is often made possible by the editors' right to edit (even though the result might not perfectly track the interests of each of the editor's customers).
       Nor would it be wise to prohibit viewpoint-based editing, while allowing editing based on subject matter. Some of the most useful forms of editing are at least partly based on viewpoint. This is common in the print world: The New Republic and The National Review, for instance, are useful to their readers precisely because they have particular outlooks on the world. Likewise, conventional conferences often invite speakers precisely because of the viewpoints they express.
       The same goes for electronic conferences. A biology discussion group might, for instance, reject messages that take a creationist perspective. A gay-rights discussion group might reject messages that argue that gay rights are a bad idea because homosexuality is evil. A Christian theology discussion group might reject messages that try to prove there is no God. Even the most open-minded of us can't devote our time to debating everything. Once we're confident enough about a certain proposition--that evolution is correct, that homosexuality isn't immoral, that God exists--we may want to spend time discussing its implications rather than rehashing the arguments about whether it's correct. Messages that express contrary viewpoints, messages that respond to them, further responses to the responses, and so on, will be useless to us, and by decreasing the signal-to-noise ratio will make the whole conference less useful.
       Of course, this doesn't mean that most electronic conferences will be completely doctrinaire. Debate is the lifeblood of electronic conferences; few conferences of which I know have much of an ideological litmus test. But while the typical conference may tolerate a wide range of opinions, the editor may decide that certain perspectives are beyond the pale. He may do participants a service--making the conference more valuable to them--by excluding those perspectives.
       The Drawbacks of Editing: Of course, editorial judgment itself limits listener choice, by depriving listeners of access to voices which they might like. Perhaps I might be disappointed by the law professor-only conference, and wish that my friend the layman could participate. Perhaps a biologist might think his colleagues could profit from learning more about the creation scientists' arguments, even if many of his colleagues might think they've heard enough of them.
       But because these limitations are a matter of private decision, not of government rule, they are generally easier to avoid. If enough listeners want to hear a particular view and one conference doesn't carry it, others probably will. And if no conference is interested in carrying the view, then chances are that this is because too few listeners want to hear it. In that case, the only way that those who are interested can be satisfied is by imposing on the greater number who aren't interested.
       We see this happening already. Among the big services, Prodigy advertises its editing, while CompuServe generally imposes no content controls. Counsel Connect provides lawyer-only discussions, while many conferences on the big services and on the Internet are open to everyone. There's an AMEND1-L Internet free speech discussion list78 that's open to all and a CLSPEECH list79 just for law professors. New Internet discussion lists are cheap to start; people who already have accounts with certain Internet providers such as Netcom--accounts that cost about $10 per month--can set up such lists for free. And these discussion lists will be open to all Prodigy, America Online, and CompuServe users, as well as those who have direct Internet access.
       Of course, editing won't always be beneficial to listeners. To take one example, Prodigy's notorious removal of messages critical of Prodigy's pricing policies was in no one's interests but Prodigy's own.80 One could defend this on Miami Herald grounds as part of Prodigy's editorial control rights, but it's hard to justify it on listener autonomy grounds. This, though, was an unusual incident, and one that had little overall negative impact on speakers or listeners. Banning this sort of conduct might not hurt listeners, but it wouldn't have helped them much, either.
       On the other hand, Prodigy's removal of anti-Semitic messages from some bulletin boards and its automatic editing of offensive words may be of significant value to many listeners. Just as I'm entitled to avoid magazines that print anti-Semitic propaganda, I have a legitimate interest in having magazine editors, acting as my agents, exclude the anti-Semitic material for me. Having Prodigy impose this editing policy gives me as listener a choice: be exposed to a restricted set of views on Prodigy, or to an unedited set on CompuServe or on the Internet. Barring Prodigy from editing would deprive me of that choice.81 On government property, we may have no choice but to suffer offensive speech, but there's no reason this has to apply to privately owned fora.
       Naturally, by increasing listener choice, editing also increases listeners' ability to choose unwisely. Listeners who choose conferences that tolerate only their own viewpoint, or those that shut down passionate debate, or even those that exclude racist speech, might be sealing themselves off from important arguments, arguments they might find persuasive (or at least worth knowing) if they saw them. Some might see this danger as a justification for laws that would open up the conferences and make sure that listeners don't shut themselves off from balanced debate.
       But though listeners may make the wrong decision, I believe it's better to leave these decisions to the listeners rather than to the government. It seems morally troubling for the government to force unwanted speech onto listeners; and I'm skeptical that even a well-motivated government can be good at determining what listeners really ought to hear and what they can legitimately seek to avoid.
       Equally importantly, I doubt that any attempts to save listeners from their narrow-mindedness will really work. If listeners want to cocoon themselves from opposing ideas, it's hard to see what can be done about that. Maybe compelled access will give some listeners some information for which they'll ultimately be grateful, even though they didn't at first want it. But I doubt this will often happen; you can make people receive messages, but you can't make them read them. And in some situations, for instance, if the government bars editors from screening out insults or racial attacks or even ignoramuses, many listeners may just stop reading the conference altogether. In trying to make people more informed, the government might cause them to become less informed.
D. Defamation Liability
       There have so far been no direct legal threats to conference operators' right to edit. There has been, however, one recent indirect threat: the assertion by one court--and some commentators--that editing should increase the conference operator's exposure to defamation liability.82
       People who run electronic conferences, edited or not, may be liable for defamatory statements posted on those conferences. As a general rule, those who participate in distributing a libel are liable together with the original author; the same may apply to the conference operator.
       A republisher of a libel--for instance, a newspaper printing an op-ed or even a letter to the editor--is liable under the familiar constitutional framework:
       Public Figure/Public Concern: If the false statement is about a public figure and is on a matter of public concern, the republisher is liable if he knows the statement is false or recklessly disregards the possibility that it's false.83
       Private Figure/Public Concern: If the false statement is about a private figure and is on a matter of public concern, the republisher is liable if he acts negligently in publishing the statement (for instance, doesn't do the factual investigation that a reasonable person would have done).84
       Private Concern: If the false statement is on a matter of private concern, the republisher might theoretically be strictly liable.85 In practice, though, few states impose strict liability even on the person who originally makes the statement,86 and I've seen no recent case that imposed strict liability on the republisher. Whether such strict liability would be constitutionally permissible is an open question.87
       A distributor who isn't a republisher--for instance, a bookstore or a newsstand--can be held liable, too, but such a distributor is given an extra immunity: It isn't liable if it "neither knows nor has reason to know of the defamatory article," and it is generally "under no duty to examine the various publications that [it] offers . . . to ascertain whether they contain any defamatory items," unless a particular publication "notoriously persists in printing scandalous items."88
       Much of the recent controversy about online libel turns on whether conference operators should be seen as republishers or distributors; but in practice this distinction isn't that important here. Whether he's a republisher or a distributor, a conference operator will be liable for public figure/public concern statements only if he's acting with actual malice. And regardless of whether he's a republisher or a distributor, an operator will be liable for private figure/public concern statements under some sort of negligence standard--if an operator "has reason to know" that a post is defamatory, he is vulnerable to a lawsuit even if he is seen only as a distributor. The one area of possible difference would be statements on matters of private concern, but even there it seems unlikely that a conference operator would be held strictly liable even as a republisher.89
       The conference operator's main worry, then, has to be about what the negligence standard--the lowest standard with which he'll have to deal--means in practice. Is it reasonable to let all posts through, or does the duty of care include a duty to prescreen? If the operator does have a chance to screen the messages, must he read them carefully, or is it reasonable to adopt a "let-it-through-unless-it's-clearly-off-topic" policy?
       It's here that a court's attitude towards editing becomes important, because the choice of a duty of care is in large part a policy decision for the court to make. What level of expense and effort is "reasonable" to expect can turn, rightly or wrongly, on the court's view of the social utility of the underlying conduct. If a court believes that editing is a bad thing, it might impose more liability on conference operators who edit than it would on those who don't edit.
       Stratton Oakmont, Inc. v Prodigy Services Co.,90 a libel case which held Prodigy to a higher standard than CompuServe because Prodigy edited and CompuServe didn't, is a case in point. Formally, the court claimed it was simply determining whether the better analogy for Prodigy was the bookstore (a distributor) or the newspaper (a publisher). "Prodigy's conscious choice, to gain the benefits of editorial control," the court concluded, "has opened it up to a greater liability than . . . other computer networks that make no such choice," and than "bookstores, libraries, and network affiliates."91 The "decision to regulate the content of its bulletin boards . . . simply require[s] that . . . [Prodigy] also accept the concomitant legal consequences"92,--with editorial control comes increased liability. Prodigy was more like a publisher than a distributor because it "[had] uniquely arrogated to itself the role of determining what is proper for its members to post and read on its bulletin boards."93
       Now it's true that, under conventional negligence principles, one's ability to avoid harmful conduct (here, defamation) is relevant to whether one has a duty to try to avoid it, so situations in which the operators have the opportunity to exercise editorial control may indeed properly lead to greater liability. If operators actually read all incoming messages before distributing them to the conference, then it becomes more likely that they'll know about the defamatory statement; in that case, the operator may be liable regardless of whether it's viewed as a distributor. And even if the operator doesn't read the messages, the fact that it has an "editorial staff . . . who have the ability to continually monitor incoming transmissions"94 might make it fair to impose on it the duty to read the messages.
       But the Stratton Oakmont court didn't limit its discussion of Prodigy's editing to manual editing: It also referred to the automatic software screening program and to Prodigy's practice of deleting some messages after they've been posted, presumably once they have triggered subscriber complaints.95 These practices don't change the cost to the operator (and, indirectly, to its customers) of greater monitoring for libel; they don't change the benefits to defamation victims that such greater monitoring would bring. There's no inherent reason that these sorts of editing decisions should affect the negligence calculus.
       Thus, the court's decision wasn't an application of settled negligence principles. But neither was it an application of a settled libel law distinction between publishers and distributors. The court's models of distributors--bookstores and network affiliates--do exercise editorial control; they, no less than Prodigy, determine what their customers and viewers will see. They select which books they'll carry or which shows they'll broadcast. Sometimes they refuse to carry an item if it seems to them to contain offensive, ideologically unpalatable, or just unpopular material. They may lack the ability to edit books or TV shows line-by-line, but for practical purposes Prodigy lacks this ability, too. Actually, Prodigy is much less selective about its posts than a bookstore is about the books it carries: Bookstores choose to stock only a fraction of all the books that are available to them, while Prodigy lets through virtually all the posts submitted to it.
       It therefore seems to me that the court's decision reflects not the commands of established libel or negligence principles, but rather a policy judgment about the propriety of editing. Consider the court's assertion that Prodigy "has uniquely arrogated to itself the role of determining what is proper for its members to post and read on its bulletin boards";96 its stress that editing "may have a chilling effect on freedom of communication in Cyberspace, . . . [a] chilling effect [which appears to be] exactly what PRODIGY wants";97 and its two references to Prodigy's conduct as "censorship."98 The court seemed to be exacting greater liability as the price for bad, or at least suspect, behavior.
       For the reasons I mentioned above, this is the wrong policy choice to make. Editing is a valuable service, and conference operators shouldn't be discouraged from performing it. Depending on how you weigh the interests in private reputation and in uninhibited speech, some sort of operator liability may be appropriate.99 The economic feasibility of editing might play a role in the balance, just as the economic feasibility of preventive measures is generally relevant in negligence analyses. But whether the operator actually edits shouldn't affect the place the line is ultimately drawn.100
       The Communications Act of 1996101 can be read as prohibiting courts from penalizing conference operators for editing. The Act says:
       No provider or user of an interactive computer service shall be treated as the publisher or speaker of any information provided by another information content provider. . . . 
       No provider or user shall be held liable on account of . . . any action voluntarily taken in good faith to restrict access to or availability of material that the provider or user considers to be obscene, lewd, lascivious, filthy, excessively violent, harassing, or otherwise objectionable . . . . 
       The term "information content provider" means any person or entity that is responsible, in whole or in part, for the creation of development of information provided through the Internet or any other interactive computer service.102 
I think that the best reading of the statute is that no conference operator shall be held liable as a publisher or republisher for defamatory speech by conference participants; that the operator's editing can't be considered in determining whether he should be exposed to defamation liability; and that the "or otherwise objectionable" clause gives protection to any of the operator's editing choices.
       On the other hand, one can at least argue that: (1) a conference operator can be held liable as transmitter of a defamatory statement on the same terms as the original speaker, so long as state law does not label him a "publisher" or a "speaker"; (2) considering an operator's editing decisions as a factor in determining the operator's standard of care does not constitute imposing liability "on account of" his editing; (3) the "or otherwise objectionable" proviso only protects editing choices that turn on the offensive form of speech and not, say, on viewpoint-based or subject matter-based editing choices; or (4) "information content provider," despite its breadth, doesn't include individual contributors to a conference.
       These latter arguments, I believe, are something of a stretch, which is why I think the Communications Act does bar Stratton Oakmont-type reasoning. But there's enough ambiguity in the statute that the matter is not free from doubt.
E. Edited Conference Groups on Public Computers or Run by Public Employees 
       Many of the computers that make up the Internet are run by public institutions, generally public universities. Many Internet electronic conferences are operated from those computers, and many are operated by public employees, especially academics.
       Does the editing of such conferences stand on a different footing from the editing of private conferences? Does the public-forum doctrine, for instance, make certain forms of editing unconstitutional? Conversely, does the government's ownership of the computers or control over its employees give it the power--even if not the duty--to restrict their editing activities? 
    1. No constitutional barriers to editing.
       A private person operating an electronic conference on a public computer is not bound by the requirements of the Free Speech Clause; he may restrict speech even based on its viewpoint. Private speakers don't become state actors just because they're speaking on or using public property.
       This is most obvious with regard to the traditional public forum--the organizers of a rally or a parade may control the speech that goes on there, even though they're using public property for this speech103--but it should be equally true for other public property. A student group meeting in a public school, for instance, should still be able to control its own speaker selection. Use of a public classroom, or of time and space on a public computer, is a valuable government subsidy, but taking subsidies (even large ones) doesn't turn a private organization into a state actor for Free Speech Clause purposes.104 So long as the group's speech-based decisions aren't dictated by the government, there's no state action.
       The same should be true when the subsidy comes in the form of the government letting an employee edit a conference on government time. Books or journals edited by public university professors, for instance, have never been thought of as involving state action. Certainly the editors routinely make viewpoint-based decisions about what gets published and what doesn't, something state actors generally can't do even in a nonpublic forum.105
       A government employee isn't always a state actor, even when he's acting on government time. So long as he isn't "exercising power `possessed by virtue of state law and made possible only because the wrongdoer is clothed with the authority of state law,'"106 he's acting as a private person. Like a public defender, a professor editing an electronic conference is engaging in "essentially a private function, traditionally filled by [private persons], for which state office and authority are not needed," though in this case state dollars are paying his salary.107
    2. No constitutional right to edit.
       On the other hand, the government may let someone set up an electronic conference on its computers, or allow him to operate it on government time, only on condition that he not edit, or that he edit only in certain ways. This is akin to the government's power to create a designated public forum limited to the discussion of particular subjects. The government may conclude that, if it's going to let public property be used, it should be used only in the way that best serves the public--for instance, for a conference that's open to everyone, or a conference that, even if edited, is edited only in viewpoint-neutral ways.
       Nonetheless, I'd recommend that the government be hesitant to restrict editors' power. As I mentioned above, even viewpoint-based editing may often create a more valuable speech product. For instance, a scholarly biology list that accepts messages based on evolutionary premises but refuses to accept messages based on creation science premises would probably be discriminating based on viewpoint: The theory that man was created directly by God is certainly an alternate viewpoint to the theory that man naturally evolved from other animals.108 But such a restriction may well be quite appropriate. Even an open-minded group of scientists may reasonably conclude that they'll use one theory as their operating assumption. At that point, further arguments based on other theories may just become distractions from the business at hand. The scientific community, it seems to me, is better served by one conference devoted to discussion of biological problems from an evolutionary perspective, another from a creation science perspective, and a third for arguing about which is the better perspective, than by three conferences on which creationists and evolutionists fight it out.
       If the concerns about limited access can be allayed not by restraining the editors but by providing more discussion lists on the same topic (but edited in different ways), then providing more lists should be the preferred alternative. Providing the extra list doesn't cost anything by itself; the list is just another entry in the computer's tables. Sometimes adding the new edited list may lead to more messages coming through the computer, but often it won't: The separate edited lists may end up having fewer bandwidth-consuming flame wars and fewer digressions. Editing is a good thing; as a general rule, government computer owners (and especially academic institutions) should encourage it, not discourage it. 
       Another desire of listeners is a pleasant, polite speech environment, both one in which they aren't personally insulted, and one in which they don't have to hear more general statements that they might find offensive (for instance, because the statements are profane, racist, or sexually explicit). As I argued above, editing decisions by conference owners can be valuable to listeners for precisely this reason. Just as many people prefer a "family newspaper," or a newspaper run by polite editors rather than by bigots or fanatics, so many people would prefer a list where flame wars and other forms of abuse are screened out, or at least quickly suppressed. But what if the conference operators choose not to intervene, if they decide not to edit generally, or if they agree with the abusive messages or at least find them valuable enough not to edit out?
A. Protecting Some Listeners Without Burdening Other Listeners
       In my view, the government may generally restrict speech to protect unwilling listeners only if the restriction doesn't interfere with the flow of speech to willing listeners. Thus, speech on electronic conferences should be protected even if it's offensive, insulting, profane, or bigoted,109 because restricting such speech would "permit[] majoritarian tastes . . . to preclude a protected message from [reaching] a receptive, unoffended minority."110 In this, electronic conferences are like billboards, demonstrations, and newspapers. The Court has made clear that restricting offensive speech in these media would impermissibly impoverish public discourse,111 and there's no reason the rule should be different online. The interests of the speaker and of the willing listeners must, I believe, prevail over those of the offended listener.
       On the other hand, some restrictions on unwanted one-to-one communications, such as physical mail, phone calls, and e-mail, should be constitutionally sound. For one-to-one communications, it's possible to create laws that are "narrowly tailored to protect only unwilling recipients of the communications."112 A law that, for instance, stops mailers from sending material to people who've already expressed a desire not to get it is constitutional;113 the same should be true for e-mail.
       Of course, speakers might still want to communicate even to unwilling listeners, and imposition on a speaker's self-expression ought not be taken lightly.114 Nonetheless, especially when a listener has already told the speaker that he's not interested in hearing more, I don't believe the speaker's desire to keep talking should be treated with much solicitude. In such a context, the speech is likely only to annoy or offend, and not enlighten or persuade anyone.115 I agree with the Court that "no one has a right to press even `good' ideas on an unwilling recipient,"116 so long as the unwilling recipient is the only listener involved.
       Space constraints keep me from defending this theoretical position in detail here, though I've talked about it at some length elsewhere.117 Moreover, a broad theoretical defense is probably premature, since there've been few explicit proposals for regulating offensive online speech (other than sexually-themed speech, which I discuss in Part III). Instead, I'll focus more specifically on the two sorts of fairly broad restrictions on online speech that may already exist, though they generally aren't enforced this way today: telephone harassment statutes and hostile-environment harassment law.
B. Telephone Harassment Laws
       In recent years, some telephone harassment statutes, which are today generally used to stop indecent, threatening, and otherwise annoying phone calls--have been specifically extended to online communications. Others had always been broad enough to include online messages. These statutes vary by jurisdiction, but they tend to prohibit some mix of the following: 
       - Threats, a prohibition that generally raises no First Amendment problems.118 
       - "[R]epeated telephone calls with intent to annoy another person," sometimes limited 
          to phone calls directed to the other person's home or work.119 
       - Use of "indecent or obscene language,"120 some times with "intent to annoy, abuse, 
          or harass."121 
       - Any communication made "with intent to harass, annoy or alarm another person,"122 
          sometimes limited to those made "in a manner likely to cause annoyance or 
       - Anonymous communications made "with intent to annoy, abuse, . . . or harass."124 
       These statutes have generally been upheld against First Amendment challenges, a result which is defensible (though, as I discuss below in Part II.E., still problematic).125 But extending them literally to online communications causes significant problems.
C. "Electronic Harassment" in Electronic Conferences
       Consider, for instance, the Connecticut telephone harassment statute, which has recently been amended to say:
    (a) A person is guilty of harassment in the second degree when: . . . 
    (2) with intent to harass, annoy or alarm another per son, he communicates with a person by telegraph or mail, . . . by computer network . . . or by any other form of written communication, in a manner likely to cause annoyance or alarm . . . .126 
Read literally, the statute would prohibit me from posting any message to an electronic conference "with intent to . . . annoy" one of the participants (say, someone with whom I'm arguing). After all, by posting a message to a conference, I'm communicating with each of the conference participants; for instance, if the conference is a distribution list, I'm causing a message to be e-mailed to everyone on the list--"communicat[ing] with" each of them "by computer network." This includes the person I'm trying to annoy.
       How broadly "annoy" would be read is anybody's guess, but a lot of things said in online conversation are intended at least in part (and sometimes entirely) to annoy one's opponents. Perhaps electronic conferences would be better if everyone intended only to enlighten, and never to annoy, but annoying and offensive speech is nonetheless constitutionally protected. Leaflets, newspaper articles, books, and movies can all be annoying (sometimes intentionally) to parts of their audiences; despite this, it seems clear that a ban on "harassing, annoying, or alarming" speech in them would be unconstitutional.127 I see no reason why electronic messages should be less protected.
       The problem is that a statute which originally applied to one-to-one communications is being applied to one-to-many communications. Keeping me from sending annoying messages to one particular person doesn't severly restrain public discourse; if the message is meant to irritate the recipient, it's unlikely to persuade or enlighten him. Its only likely consequence is annoyance. But in a one-to-many context, a message that's annoying, even intentionally so, to one person may indeed be valuable to others.
       I'm not sure that the extension of telephone harassment laws to online communications was meant to cover electronic conferences. Quite possibly the drafters of the laws were only contemplating direct e-mail, a one-to-one medium not much different from conventional phone calls. In this context, as I mention below, restrictions similar to those imposed on telephone harassment might indeed be permissible. But whether intentionally or not, some of the laws on their face sweep considerably further than they should.
D. Hostile-Environment Harassment
       Hostile-environment harassment law is a very different creature from telephone harassment law, but it too might have unexpected consequences in cyberspace.
       The most familiar form of hostile environment harassment is workplace harassment: speech or conduct that is
    - "severe or pervasive" enough to 
    - create a "hostile or abusive work environment" 
    - based on race, sex, religion, national origin, age, disability, veteran status, or, in 
       some jurisdictions, sexual orientation, citizen/alien status, political affiliation, 
       marital status, or personal appearance 
    - for the plaintiff and for a reasonable person.128 
       An employer is liable for hostile environment harassment perpetrated by its employees--and even its customers129--so long as it knows or has reason to know about the conduct.
       This is a broad definition, and it has in fact been applied to a broad range of speech. A state court, for instance, has held that it was religious harassment for an employer to put religious articles in its employee newsletter and Christian-themed verses on its paychecks.130 The EEOC has likewise concluded that a claim that an employer permitted the daily broadcast of prayers over the public address system was "sufficient to allege the existence of a hostile working environment predicated on religious discrimination."131
       Similarly, a court has characterized an employee's hanging "pictures of the Ayatollah Khome[i]ni and a burning American flag in Iran in her own cubicle" as "national-origin harassment" of an Iranian employee who saw the pictures.132 Courts have used harassment law to enjoin people from making "remarks or slurs contrary to their fellow employees' religious beliefs,"133 displaying materials that are "sexually suggestive [or] sexually demeaning,"134 or uttering "any racial, ethnic, or religious slurs whether in the form of `jokes,' `jests,' or otherwise."135 A federal agency has likewise characterized anti-veteran postings at Ohio State University as harassment based on Vietnam Era veteran status.136
       Hostile-environment law may even cover coworkers' use of job titles such as "foreman" and "draftsman,"137 sexually themed (but not misogynistic) jokes,138 and "legitimate" art.139 The constitutionality of workplace harassment law is being hotly debated,140 but as of this writing the risk of harassment liability is certainly a fact of life.
       What does this have to do with the Internet? Well, the foundation of workplace harassment law is the theory that harassment is itself discrimination: the denial to certain people of a particular kind of employment benefit--a tolerable work environment--based on their race, sex, and so on.141
       This theory is equally applicable to other discrimination statutes, including statutes that bar discrimination in places of public accommodation. Some statutes make this explicit, prohibiting, for instance, "communication of a sexual nature" that creates "an intimidating, hostile, or offensive . . . public accommodations . . . environment."142 Other statutes that speak only of discrimination have also been interpreted as barring harassment: For instance, a recent Wisconsin administrative agency decision has concluded that an overheard (but loud) discussion that used the word "nigger" created an illegal hostile public accommodations environment for black patrons, even though the statements weren't said to or about the patrons.143 Likewise, the Minnesota Supreme Court has held a health club liable for creating a hostile public accommodations environment, based on the club's owners "belittl[ing]" a patron's religious views (expressed in a book the patron had written) and "lectur[ing] her on fundamentalist Christian doctrine."144 And it's fairly well-established that other antidiscrimination statutes, which ban discrimination in education and housing, also apply to hostile environment harassment;145 it stands to reason that the same would be true for public accommodations statutes.
       As Part I.B.2 discusses, it's eminently plausible that commercial online services would be considered places of public accommodation. Given that some judges have seen even noncommercial establishments--such as parades, the Boy Scouts, and private clubs--as places of public accommodation, Prodigy, Counsel Connect, and others would quite likely qualify.146 At least one commentator has in fact suggested this very point.147 Say, then, that someone frequently posts slurs or sexual jokes or sexually explicit messages or sexist or racist or anti-veteran or religiously bigoted statements or even religious proselytizing to an electronic conference. In the eyes of some factfinders, such messages may well create a "hostile or abusive" environment for some of the conference participants. If the conference operator has the power to do something about this--for instance, if the conference is moderated but the moderator lets these messages through, or if the operator can kick off the offender but refuses to do so--the speech could give rise to liability.
       The best real-life online example of this came in the context of hostile educational environment law. In late 1994, in the wake of a controversy about an allegedly sexist ad in the Santa Rosa Junior College newspaper, some students posted sexist remarks about two female student newspaper staffers on a college-run electronic conference.148 Though the female students didn't see the message, they eventually learned about it, and when they did, they filed a complaint with the U.S. Department of Education's Office for Civil Rights.
       The Office concluded that the messages were probably "so severe and pervasive as to create a hostile [educational] environment on the basis of sex" for one of the students.149 A college tolerating speech that creates a sexually hostile educational environment would, in the Office's view, violate Title IX of the Civil Rights Act.150 If this is so, then a service provider tolerating similar speech on its computers would probably be violating public accommodations statutes.
       I believe these sorts of speech restrictions are generally unconstitutional, entirely so in the educational and public accommodations contexts, and partly so in the workplace context. I don't want to go into the details of this here; the arguments have been amply discussed elsewhere.151 Put briefly, I can't deny that "hostile or abusive" speech can greatly diminish the value of an online conference--public or university--for those who are offended; but such speech, even racially, religiously, or sexually bigoted speech, is protected by the Free Speech Clause from government abridgment.152 It's protected on sidewalks, in private homes, in the pages of newspapers. Despite the recent spate of campus speech codes, courts have held that it's protected in universities.153 There's no reason it shouldn't be equally protected in Prodigy, Counsel Connect, and the like.
       Some courts have been willing to uphold hostile-environment harassment law in the workplace, though it bears emphasizing that others have suggested that even there it may face substantial constitutional problems.154 But this has been in large part because of their view that "the workplace is for working," not for debate.155 Electronic conferences are created precisely for debate. Whatever the constitutional status of workplace harassment law, such speech restrictions in places devoted to communication can't be valid.
       But though I'm confident that most restrictions on harassing speech will ultimately be struck down, the fact remains that they are something of a growth field in free speech law today, and that they enjoy a good deal of support. Hostile environment-based restrictions on online speech are likely to arise with some frequency in coming years.
E. One-to-One Online Harassment
       Restrictions on some one-to-one messages--such as personal email--are at least theoretically more defensible because they help insulate unwilling listeners while still protecting the right to communicate to willing ones. Some telephone harassment laws have been upheld precisely on these grounds.
       Still, even in the one-to-one context, these laws pose significant problems. "Annoying," "harassing," and "indecent"--words the laws use to define the speech they bar--are vague terms, and fairly read, the laws can sweep quite broadly. If an acquaintance of mine has botched a task I asked him to do, and I phone him or e-mail him and say, "You idiot, you really fucked this up," I may well have committed harassment; I've said something that's arguably both "annoying" and "indecent,"156 I've said it with the intent to annoy (and perhaps "harass," whatever that means), and my statement is in fact likely to annoy. Under many telephone harassment statutes, I've committed a crime.
       Likewise, I edit an electronic poetry journal,157 and our subscribers sometimes send us messages about our poems. One such message said nothing more than "Your poems suck!"158 Under the Connecticut law, sending that message may well have been a misdemeanor; the message was likely to annoy me and could have been intended to do so (as well as to communicate the sender's views). I can't claim that the messages in these examples are of remarkably great First Amendment value, but it isn't clear that this sort of speech should be criminal.
       Some of the broader telephone harassment statutes have been held unconstitutional for this very reason.159 Other statutes have been upheld, especially if they've required that the speech be intended to annoy (which not all statutes do);160 and some commentators argue that this intent requirement should help save the law from invalidity.161 But I'm not sure that even those statements made with the intent to annoy should be considered criminal--as my example above shows, many of us might say such things in an exasperated moment, with little likely harm to anyone.
       In practice, of course, telephone harassment laws are considerably less menacing than their language suggests. It takes conscious effort to make an annoying call to a stranger; few people complain to the police about the occasional annoying call from an acquaintance; in many such situations, prosecutors may decide not to prosecute; and it's often hard to prove what the caller actually said. In some respects, these checks might limit the law's reach to only the most serious situations.
       But the e-mail environment changes some of these conditions. While annoying phone calls are usually deliberate pranks, thought-through insults, or conscious attempts to menace, annoying e-mail can easily happen on the spur of the moment. A person sees a message he dislikes on an electronic conference, and in a few seconds he can send an angry retort--one intended to and likely to annoy--directly to the author's mailbox. And because e-mail, unlike phone calls, leaves a written record, its content is easy to prove.
       The person who sent me the "Your poems suck!" e-mail probably wouldn't have called the publisher of a print magazine to say the same thing. The ease with which one can reply to an e-mail makes such replies spontaneous and sometimes rude. This sort of conduct seems less like a deliberately harassing phone call, and more like the annoying words said in public--one can imagine someone saying the same thing at a poetry reading in a coffee-house--which are generally not punishable unless they're likely to cause a fight. In the great majority of cases, recipients will still not complain and prosecutors won't prosecute, but in my view these shouldn't be the only barriers between a basically decent computer user and a misdemeanor conviction.
F. The Continued Unwanted Contact Model
       Instead of outlawing a particular category of speech--the annoying or even the intentionally annoying--a better solution might be to leave the decision in the listener's hands.162 If you don't like what I'm e-mailing to you (and to you alone), you should be able to demand that I stop. The demand would put me on notice that my messages have gone beyond the tolerable; the law might even require that the notice specifically alert me that further contact is illegal. Such a law would thus avoid criminalizing the occasional intemperate outburst, while still giving annoyed recipients the power to demand some peace.
       The Court has already, in Rowan v United States Post Office Department,163 upheld such a law for physical mail. The law provided that, if any householder concludes that a sender has been sending him sexually suggestive advertisements, he can notify the Postmaster General that all further mailings from the sender must stop. The Postmaster General would then order the sender to remove the recipient from its mailing list; if the sender kept sending material, it would be committing a crime.
       Though the law focused on sexually suggestive material, the Court didn't ground its (unanimous) conclusion on any supposed "lower value" of such material. In fact, it was clear that the householder could label any material, including political ads and dry goods catalogs, as being sexually suggestive: The householder's judgment was unreviewable, and the very unreviewability of this judgment was, in the Court's view, central to the law's validity because it kept the decision about what's offensive out of the government's hands.164 Though the law kept the sender from "speaking" to the recipient, the Court concluded that "no one has a right to press even `good' ideas on an unwilling recipient."165
       This logic seems eminently applicable to e-mail; I see no reason for treating electronic communications differently from paper ones. True, you can easily delete e-mail, but you can easily throw out paper mail. You can usually even set up your e-mail program to automatically delete messages that come from a particular address, but you can likewise throw out unread letters that come from a particular place. The auto-delete capability might make a Rowan-type law less necessary because the messages would get deleted without the recipient even knowing they arrived; still, the sender could get around this by using remailers or other means that hide his address.
       The reasoning in Rowan rested in large part on the privacy of the home; the unwanted continued mail was seen as particularly offensive because it intruded on this privacy.166 One might therefore argue that any Rowan-inspired law for e-mail should only apply to e-mail that's received at home. But I don't think the concern about unwilling recipients of mail or of e-mail should be any different when they pick up the message at school or at work.
       One of the great features of the new technologies is that one's physical location no longer matters. A person can pick up messages from the office one day and from home the next. The emotional and dignitary burden to listeners of having to see messages to which they've already lodged their objections is the same whether they're in the home or in their office. The burden to speakers of having to stop talking to the listeners is likewise the same regardless of where the offended listener is located. So long as the speakers retain the ability to communicate to unoffended listeners, the offended listener should be entitled to demand that communication to him stop, whether he's at home or not.
       To make sure that speakers do retain the ability to communicate to others, the law should be clearly limited only to those situations where the sender can personally choose whether the message is to go to the recipient. For instance, if the message is e-mailed through a distribution list whose membership is outside the sender's control, then the messages to the recipient can't be suppressed without restricting the messages to other list members.
       The law should also make sure the sender gets adequate notice; a simple "Oh you idiot, stop bothering me" should probably not be seen as a legally enforceable bar to all future communications. But if someone clearly says they don't want any more e-mail from the sender--and e-mail programs can easily have the proper language programmed into them so that a sender can send the right form (which would also be saved somewhere for evidentiary purposes) at the touch of a button--there should be no constitutional difficulty with the law enforcing such a demand.
       All this merely shows that the law is constitutional, not necessarily that it's wise. It's not clear that we need to bring the might of the criminal law down on people just for sending repeated unwanted messages, whether they are ads or personal insults or just pranks. And of course, certain kinds of unwanted messages, such as threats or extortion attempts, would be barred anyway by other laws.
       But there are contexts where repeated unwanted messages can indeed cause more than just mild annoyance. A Rowan-inspired law would apply only when the recipient has specifically asked the sender to stop. When someone keeps contacting you despite your specific requests to the contrary--when he knows you're so uninterested in hearing that you've taken the trouble to tell him to stop talking--this might reasonably cause you some alarm. This can be especially so for repeated romantic overtures; the recent spate of stalking legislation reflects the fact that persistent unwanted romantic interest, expressed in a context where it's clearly not reciprocated, can be quite disconcerting even if no specific threat is present. 
       Besides wanting to control their own speech diet, people may also want to control that of their children, especially with respect to sexually explicit materials. Some are skeptical of the need to do this, and compared to the threats to which children are exposed in the physical world, the danger of exposure to explicit online material indeed seems slight. Still, many people are concerned about it; the legal system has generally decided that they have a right to translate their concerns into law; and outside the online world, laws that restrict minors from accessing sexually explicit material have generally been upheld. For purposes of this discussion, I'll assume that preventing children from seeing such material is indeed an important goal.
A. The Potential Restrictions
      The most prominent restriction on sexually explicit material online is, of course, in the Communications Decency Act of 1996. The Act makes it a crime for people to "use[] any interactive computer service to display in a manner available to a person under 18 years of age, any [material] that, in context, depicts or describes, in terms patently offensive as measured by contemporary community standards, sexual or excretory activities or organs."167 Many electrons have been spilled regarding the Act, and the Court is scheduled to pass on its constitutionality this Term in Reno v. ACLU.168 I don't intend to specifically focus on it, though some of the things I say below may apply to it.
       Rather, I want to focus on a different kind of statute. Many states have laws that bar "public display" of certain kinds of sexually explicit matter, and many such statutes have been upheld against First Amendment challenge.169 For instance, Georgia Code § 16-12-103 provides, among other things, that: 
     (e) It shall be unlawful for any person knowingly to . . . display in public . . . at any . . . public place frequented by minors or where minors are or may be invited as part of the general public: 
    (1) any [visual representation] . . . which depicts sexually explicit nudity, sexual conduct, or sadomasochistic abuse and which is harmful to minors; or 
    (2) any [printed or audio material] which contains . . . explicit and detailed verbal descriptions or narrative accounts of sexual excitement, sexual conduct,or sadomasochistic abuse and which, taken as whole, is harmful to minors.170 
"Harmful to minors" is defined as: 
     that quality of description or representation, in whatever form, of nudity, sexual conduct, sexual excitement, or sadomasochistic abuse, when it: 
    (A) Taken as a whole, predominantly appeals to the prurient, shameful, or morbid interest of minors; 
    (B) Is patently offensive to prevailing standards in the adult community as a whole with respect to what is suitable material for minors; and 
    (C) Is, when taken as a whole, lacking in serious liter ary, artistic, political, or scientific value for minors.171 
Under this definition, even material that's not obscene as to adults is barred from public places.
       If--as seems quite possible--electronic places that are accessible to the public (such as conferences or Web pages) are found to be "public places," then many items that are posted on the Net are already in violation of Georgia law, and of similar laws in other states.172 And the law isn't just limited to obscenity or even "pornography" (whatever that may be). It could equally apply to "legitimate" art in an online gallery, or to a person's favorite Mapplethorpe photo on his Web page, or to online sex-talk groups, or conceivably even to groups that talk clinically about sex.
       The problem is that, online, most places--electronic conferences, Web sites, and so on--are "frequented" by at least some minors. And while offline one could put the explicit material in a separate room and check people's IDs before one let them in, there's no similar mechanism online. Internet newsgroups are accessible to everyone. So, generally speaking, are discussion lists. While Web pages can theoretically ask for a credit-card number before they let a person into particular places, that won't work if someone wants to display material for free. The Georgia statute and others like it might thus do something quite similar to what the Communications Decency Act is criticized for doing--they might bar a great deal of sexually themed material from most freely available places on the Net.
B. The Least Restrictive Alternative Requirement
       In my view, the state public display laws--and the similar Communications Act provisions--are unconstitutional when applied online, because there are means of protecting children in cyberspace without unduly limiting adults.
       In Sable Communications v FCC,173 the Supreme Court made clear that the government may not, in trying to protect children, bar sexually-explicit speech generally when it could implement a less restrictive alternative that restricts only children. If such an alternative exists, and if it serves the compelling interest as well as would a total ban, then the ban is unconstitutional.174 Butler v Michigan,175 which struck down a law that tried to "shield juvenile innocence" by "reduc[ing] the adult population . . . to reading only what is fit for children," suggests the same.176
       Formally, the rule that the government may prohibit distribution to minors of materials that are harmful to them doesn't include such a proviso; when distributed to minors, material that's harmful to minors is considered essentially obscene and lacking in First Amendment value.177 But if a law that bars distribution to minors also interferes with distribution to adults, the government should have to show--as in Sable--that the law is the means that's least restrictive of adult access.178 As to adults, of course, nonobscene material is constitutionally protected.
C. Ratings
       Fortunately, technology may provide part of the solution here. Unlike in the offline world, a child's online eyes--the software that retrieves and displays the material--can be set up to screen out unwanted material. The difficulty comes in recognizing it: To a computer, a Mapplethorpe and Mickey Mouse are both just 1s and 0s. There has to be (as in the offline world) some human agent that identifies what's suitable for minors and what's not.
       1. The clean list/dirty list models.
       One possible solution is for someone to provide a list of online locations--World Wide Web sites or Internet electronic conferences--that have been checked and certified "clean," together with a Net access program that allows access only to those locations. The program can, when it's first set up, ask the buyer (presumably the parent) to configure a password; then, whenever it's run, it will ask for the password, and if the right password isn't entered, it can run in clean-only mode. Alternatively, if someone comes up with a list of places that are "dirty," the software can allow access to all places except the dirty ones.
       There are already several commercially available products (often called filters) that, among other things, maintain lists of dirty sites and prohibit access to them. The best-known one is called SurfWatch; it costs about $50 179 and for $5.95 per month180 a customer will get the list of dirty sites constantly updated. SurfWatch in fact employs people to look through the Net, always on the lookout for new bad locations. If the government wanted to, it could buy out SurfWatch or one of its competitors--for a fraction of the cost of legally enforcing any sort of speech restriction--and then distribute the software for free to all users.
       This need for updates, of course, shows a weakness of the dirty list approach. The Net is a constantly changing environment; new Web pages and discussion groups are constantly being added, and existing resources are constantly being changed. It may take the filter distributors a while to detect the changes.
       The clean list model does promise to shield children almost perfectly from harmful-to-minors material, because children would be able to access only those pages that the filter distributors have screened. Leakage of dirty material could happen only when a clean page is modified after the filter people check it.
       But precisely because a child can see a page only if it's been certified clean, any clean list program will give children access to only a fraction of the clean material on the Web. Screeners almost certainly couldn't check more than a small fraction of existing Web resources, and any new resources--including new pages at existing sites--might go unchecked for quite a long time.181 Many parents, even those who want to shield their children from harmful matter, might be understandably reluctant to restrict their children's access that greatly.182
       Courts might, despite this, still conclude that a clean list is a less restrictive but equally effective alternative to a ban: After all, the clean list will effectively serve the primary interest, shielding children from material that's harmful to minors. The fact that it does this only by insulating children from a lot of other material, the argument would go, should be ignored for purposes of this inquiry. The cases neither clearly endorse nor clearly foreclose such an argument, largely because there've been so few cases that have seriously elaborated on the meaning of "less restrictive alternative."
       It seems to me, though, that such an argument is ultimately unpersuasive. A proposed alternative may aptly be called "not equally effective" if it serves the compelling interest only through an unacceptable sacrifice of other important interests.183 For example, say that "clean list" filters were unavailable, and opponents of online speech restrictions suggested that the less restrictive alternative would simply be parents keeping their kids off the Internet. This alterative would indeed shield children quite well, but at an unacceptable cost. The clean list model is of course quite different in degree from keeping kids off the Net altogether, but it seems similar in kind.
       Finally, the clean list model appears quite inadequate to the task of filtering newsgroup posts and discussion list messages. The distributors of clean list software can identify a newsgroup that's mostly clean--that generally doesn't carry any harmful-to-minors material--but there can be no assurance of this; anyone can post something dirty at any time.
       Ultimately, even all these considerations might not be much of a problem. The various filters may catch the great majority of offending sites, and, after all, restrictions on the distribution of pornography to minors outside the electronic world are also notoriously imperfect.184
       Nonetheless, the fact that the filters will miss some sites makes it possible to argue that these programs are not the sort of less restrictive alternative that would invalidate a total ban--that they are not as effective as a combination of the ban and whatever technological assists (such as filters) that parents can use. Under the strict scrutiny test the Court applied in Sable, a law is unconstitutional if there are other means that are less restrictive but as effective as the ones being challenged.185 If the proposed means "fall short of serving [the] compelling interests," then the challenged law may be constitutional.186
       2. The ratings model.
       There is, however, a possible supplement to the clean list and dirty list approaches that would probably be at least as effective as a total ban: A rating system by which people can self-identify explicit material that they make available.
       On the computer, pictures are stored in files that are organized in a conventional way. If one establishes a convention that a small part of every file contains a marker indicating whether the file contains sexually explicit material, then the programs that read these files can also look at this marker. If the program is running in child-only mode (something the parents can set), then the program would refuse to display the file.187 Even text posted to electronic conferences might be specially labeled so that programs which read these conferences can filter it out.
       The rating system might even rate the pictures in finer-grained ways, for instance for degree of sexual explicitness, or violence, or what have you; then the parent could set whatever threshold rating he wants. Of course, if the system gets too detailed, the risk of mislabeling might increase.
       3. Enforcing the ratings system.
       The ratings system can work only if the ratings are accurate. The law might make the system "voluntary" in the sense that no one has to attach a rating to his pictures; the child-only programs would then just view any unrated picture as presumptively dirty.188 But there'd still be a problem if a picture that should be X-rated is self-rated as G.
       One can question the propriety of using the criminal law in this context--I agree with Justice Stevens that distribution of obscenity should be at most civilly punishable, and I'd say the same about material that's harmful to minors. But I think that, under the existing doctrine, it would be constitutional to criminalize the display of any harmful-to-minors material which does not carry a correct rating.
       The courts have already agreed that it's permissible to burden such speech with restrictions on its distribution to minors, and with restrictions on its display in places where minors may be present. Requiring that it be labeled seems to me no different than requiring that it be put on a separate shelf which minors can't read. Of course, if a defendant makes an error in labeling, he may be found criminally liable. But the same is true under the laws upheld in Ginsberg: If a defendant erroneously decides that material isn't harmful to minors, he may end up in jail.189
       4. Why this is at least an equally effective alternative.
       The rating system would, of course, not be perfect at screening out material that's harmful to minors. But it would generally be no worse at this than would an outright criminalization of Internet displays of such material.
       Under a rating system, minors would still be able to get access to illegally misrated materials, but under a general prohibition, minors would be able to get access to illegally posted materials. If posters aren't deterred by the risk of criminal punishment for misrating, it seems unlikely that they'd be deterred by the risk of criminal punishment in a general prohibition.
       Likewise, the ratings laws won't protect children from material posted by foreign or anonymous sources; but, of course, neither would a total ban. Foreign and anonymous posts are the weak point of any system for controlling cyberspace materials. Any domestic control regime--of pornography, gambling, or what have you--will at most cut down on the number of places that contain the prohibited material; it can't even come close to eliminating them.
       Rating systems won't shield minors who get access to a computer that isn't running the screening software (for instance, a computer at a friend's house). In that respect, a ban might be more effective; by deterring the posting of certain materials, the ban would decrease the amount of harmful-to-minors matter available even to those children using an unshielded computer.
       But balanced against that is the likelihood that more people will comply with a rating system than a general prohibition. People who post explicit material, even for free, presumably get some value out of doing it. Realistically, criminal liability won't be much of a deterrent--criminal prosecutions will probably be rare, just as prosecutions for distributing material that's harmful to minors are rare now. For foreign and anonymous posts, prosecution may be next to impossible.
       Given the low cost of violating the law, the important issue for many posters will be the cost of following the law. Many people might not want to stop posting explicit material and might therefore be willing to run the low risk of prosecution instead of complying with a total prohibition. But some of these people may gladly attach a "dirty" rating to the material, because it will basically let them do what they want--make the material available. Even people who face no risk of criminal liability may voluntarily comply with a ratings system; many of them might be quite happy to prevent children from accessing their posts.
       If I'm right about all this, then a total ban on display of harmful-to-minors matter in places which might be accessible to minors would be unconstitutional. The alternative--a ratings system--would be both less restrictive and at least as effective.190
    5. Community standards.
       In all this, I've omitted discussion of varying community standards. Under existing law, whether material is obscene (or is harmful to minors) is determined with reference to the standards of the geographical community to which it's being distributed. If a product is being sold by mail order or telephone, the burden is on the seller to determine where it's going and what the relevant standards are.191 In theory, then, any mechanism for dealing with material that's harmful to minors will need to rate each item separately for each possible jurisdiction. This will be much more complicated than the normal clean/dirty determination (which itself might be complicated enough along the edges).
       I don't talk much about this because the whole concept of community standards will in any event have to be rethought for cyberspace, both for conventional obscenity and for material that's harmful to minors. Online, of course, one can't tell a person's place of residence any more than one can tell his age. Making distributors of possibly obscene material--either sellers or people who post it for free--liable based on local community standards will reduce online sexually explicit speech to the level of the most restrictive community.
       One way or another, then, a national online community standard will be implemented. The only question is whether such a national standard will be consciously chosen, or whether the most restrictive local community standard ends up becoming the national one by default. In any event, others have discussed this point at length, and I would rather refer the reader to their treatment of it.192 
       The new information technologies can be a boon for the marketplace of ideas; speakers, listeners, and society at large can all profit from them. Free speech in cyberspace should be protected as zealously as it is in the offline world.
       But the more the new media are viewed as primarily a soapbox, the more likely they are to become as irrelevant as a soapbox. Cyberspace can only thrive if it provides listeners material that's valuable to them: material that's relevant, that doesn't make them feel annoyed or harassed, and that doesn't make them afraid to give access to their children, who may have a great deal to gain from the online world.
       The difficulty, as always, is accommodating the interests of speakers, of listeners who want to hear what a particular speaker says, and of listeners who don't want to hear it. And a general preference for the interests of speakers, at least where government regulation of the speaker's message goes, is justifiable. Despite my endorsement of some speech restrictions, I remain generally skeptical about government speech regulations.
       Still, while the Constitution might require that speakers prevail in many cases, there are ways in which listeners can also be properly protected, even at some speakers' expense. Modest restrictions on persistent unwanted e-mails and modest restrictions on sexually explicit material that might be accessible to children are two such mechanisms; but of all the possible protections for listeners, a thriving market of private editors is by far the most important. The new age will indeed be, in Virginia Postrel's words, "[t]he Age of the Editor."193 The more we interfere with this, the less valuable the new technologies will become.

 * Acting Professor, UCLA Law School ( Many thanks to the members of Trotter Hardy's discussion list, and especially to Solveig Bernstein, Evan Caminker, Michael Froomkin, Robert Goldstein, Kenneth Karst, and Daniel Lowenstein.

 1.     Sometimes they want more influential listeners, or listeners who are most likely to buy certain products, but as a rule more is better.

 2.     I follow the convention of free speech jurisprudence by talking about "speech," "speakers," and "listeners," but of course in the online world this generally refers to writings, writers, and readers.

 3.     See notes 113-20 and accompanying text (Part II.B).

 4.     See 39 USC § 3008 (1988) (allowing an addressee to notify the Postal Service that he does not want to receive pandering materials and to demand that the Postal Service tell the sender to refrain from sending such materials).

 5.     See notes 135-39 and accompanying text (Part III.A).

 6.     See

 7.     See

 8.     To use an Internet discussion list, you have to "subscribe" to it by sending a subscription request to a particular e-mail address (the so-called "list server address"). Once you're subscribed, any message you send to another e-mail address (the "list address") will get forwarded to all the other subscribers. The message will arrive in their mailboxes just as if you had sent each of them a personal e-mail. You can talk to the whole list, and you'll get messages from anyone else who talks to the whole list.

 9.     Internet news groups are basically like discussion lists, but (1) you read and write to them not using your e-mail facility but using a special program called a "news reader"; (2) you don't have to be specially subscribed to a news group to access it; and (3) the people who manage your computer system can choose which news groups they'll make available and which they won't.

 10.     These services can set up their own "discussion groups" that are similar to Internet newsgroups, but accessible only to those who use the particular service.

 11.     Conference operators and operators of electronic magazines are rather proud of their high signal to noise ratios: "THINK OF THE NUGGETS THIS WAY: There's no flaming, no endless discussions, no spam and no blather. Stamper's News Nuggets has the best signal-to-noise ratio around. Just a hot cross-section of nifty stuff. Why not tell a friend to subscribe?" Stamper's News Nuggets #29 (Nov 13, 1995) (short electronic magazine). "I want to second Lynda Frost's observation that CRIMPROF is marked by both a strong substantive content and a marked civility. That these traits are not universally found on listservs makes their presence on CRIMPROF a tribute to Steve Sowle (the founder) and to all the subscribers." Post on from Herbie deFonzo (Oct 18, 1995).

 12.     Sometimes conference operators place entire topics off limits as a prophylactic measure. In one list operator's words, "Discussion of the right to keep and bear arms is actually prohibited on CJUST-L [a criminal justice discussion list] because we found it to be too contentious an issue and it was creating way too high a noise-to-signal ratio (and CJUST-L is supposed to be an academic discussion list, too)." E-mail from Alex Rudd to author (Dec 7, 1995); e-mail providing introductory information on CJUST-L listserver (Dec 5, 1995) ("There is only one content-based rule: discussion of the Right to Keep and Bear Arms is not welcome. Refusal to adhere to this guideline may result in your being removed from the list."). See also e-mail announcing new Republicans96 electronic conference (Dec 1, 1995):

 13.     For instance, "I'm in charge here :-), and I have had a number of pleas from subscribers to call for an end to the religious discussions. Naturally, we do not censor exchanges on IFREEDOM, but it may be time to move on to other matters concerned with censorship and intellectual freedom. I invite subscribers to tell us about any new cases in their locations." Post on electronic conference from conference operator (Oct 11, 1995). The ":-)" in the message, called a "smiley," because if you look at it at a ninety degree angle it looks like a stylized smiling face is a conventional symbol for humor or irony; the conference operator seems to be slightly embarrassed by his assertion of control.


 15.     See also e-mail announcing LWSCHOLAR electronic conference (Nov 1, 1995) ("To enhance the value of discussion on the list, membership will be open only to individuals who are faculty members at ABA-accredited or AALS-member schools.").

 16.     See, for example, e-mail announcing BloodQuill electronic conference (Nov 13, 1995) ("[the list] is open and I moderate by knocking rude people off for a week, but we are very broad-minded").

 17.     Daniel Pearl, Internet Cybersmut Crackdown Expected, San Diego Union-Tribune, ComputerLink Section 12 (Feb 14, 1995).

 18.     Jack Rickard, Prodigy Can't Win as a Censor, Computerworld 25 (Nov 11, 1991); IBM's Prodigy Does What?, PC-Computing 56 (March 1991); Evelyn Richards, Dissident Prodigy Users Cut off from Network, Wash Post C1 (Nov 3, 1990).

 19.     Consider, for example, a message posted to on December 19, 1994 by a person whose comments I, as list operator, tried to restrict: The message was titled "Censorship Attempt Here Rejected," and read, in part, "No, I will not acquiesce to Professor Volokh['s] thinly veiled threat. This post is an affirmation of that stubborn support of free[dom] of inquiry. Professor Volokh will have to find a way to bar me from this list in order to gag me. I would hope that many on this list would publicly repudiate this real example of narrow sectarianism."
              Likewise, after I declined to allow a nonacademic to subscribe to a law professor list which I operate, I got the following message: "I bow to your desire for a closed loop of exclusivity. One would not wish to have one's inferiors peeking through one's window and learning one's priestly secrets! :-) Most humbly yours, [name]." E-mail to author, October 24, 1995. The ":-)", which many Net people use as a signal of humor or irony, suggests that the sender was taking the exclusion in better spirits than the person I quoted in the previous paragraph.

 20.     See Rendell-Baker v Kohn, 457 US 830 (1982) (holding that the decisions of a private school which received more than 90 percent of its funding from the government were not state action); Blum v Yaretsky, 457 US 991 (1982) (holding that private decisions are not transformed into state action by the use of government forms or the receipt of government funds).

 21.     418 US 241, 258 (1974).

 22.     115 S Ct 2338 (1995).

 23.     Id at 2348.

 24.     Miami Herald, 418 US at 258.

 25.     Hurley, 115 S Ct at 2345.

 26.     See, for example, Edward V. Di Lello, Functional Equivalency and Its Application to Freedom of Speech on Computer Bulletin Boards, 26 Colum J L & Soc Probs 199, 231 (1993) (arguing that this weakens conference operators' claim of a right to edit).

 27.     Miami Herald, 418 US at 258.

 28.     Hurley, 115 S Ct at 2348-50.

 29.     Turner Broadcasting System, Inc. v FCC, 114 S Ct 2445, 2456 (1994), citing Red Lion Broadcasting Co. v FCC, 395 US 367 (1969), and National Broadcasting Co. v United States, 319 US 190 (1943).

 30.     395 US 367 (1969).

 31.     See Miami Herald, 418 US at 241; Turner Broadcasting, 114 S Ct at 2456-58.

 32.     Id at 2456-57.

 33.     Id at 2445.

 34.     447 US 74 (1980).

 35.     See Hurley, 115 S Ct at 2347.

 36.     Id at 2350, quoting Pacific Gas & Electric Co. v Public Utilities Commission, 475 US 1, 12 (1986) (plurality opinion).

 37.     PruneYard, 447 US at 99 (Powell concurring in part and in the judgment).

 38.     Hurley, 115 S Ct at 2350-51 (describing and distinguishing PruneYard).

 39.     PruneYard, 447 US at 94.

 40.     Some to whom I've talked about this have suggested that electronic conferences are akin to public forums and should therefore be open to all comers. But "public forum" in the Free Speech Clause sense refers only to government property, which the government must open to all; no similar doctrine exists for private property.

 41.     114 S Ct at 2456.

 42.     Hurley, 115 S Ct at 2349.

 43.     Id at 2350. Some might argue that cable-system operators, as opposed to cable-channel programmers, are mere carriers rather than speakers, and thus shouldn't have broad First Amendment protection for their speech selections; but this is not the view that the Turner Court took.

 44.     Id at 2349.

 45.     Of course, Counsel Connect could impose a content requirement instead; it could prescreen all posts to see if they seem useful to lawyers, or it could wait for people to complain about legally ignorant participants. But prescreening takes manpower and may lead to nasty disputes about what deserves to be posted and what doesn't. Kicking off the legally ignorant is likely to be equally unpleasant; and by the time someone overcomes his inhibitions enough to complain about allegedly ignorant posts (something many people will be reluctant to do), the damage to the discussion might have already been done.

 46.     See Louisiana Debating & Literary Association v City of New Orleans, 42 F3d 1483 (5th Cir 1995) (holding city ordinance forbidding discrimination in public accommodations unconstitutional when applied to private social clubs with 325-1000 members that "prohibit the transaction or discussion of any business on their premises").

 47.     Hurley, 115 S Ct at 2338.

 48.     Curran v Mt. Diablo Council of the Boy Scouts of America, 34 Cal App 4th 299, 29 Cal Rptr 2d 580 (1994), review granted and opinion superseded, 874 P2d 901 (1994) (no final published opinion).

 49.     Many state antidiscrimination statutes define "place of public accommodation" quite broadly, as covering any "business [or] entertainment . . . facility of any kind whose . . . services . . . are extended, offered, sold, or otherwise made available to the general public." See, for example, Hawaii Rev Stat § 489-2 (1993); compare, for example, Colo Rev Stat Ann § 24-34-601(1) (West 1990) ("any place offering services . . . to the public"); Ind Code Ann § 22-9-1-3(m) (West 1991 & Supp 1996) ("establishment that caters or offers its services or facilities or goods to the general public"). Many statutes give a list of examples, though stressing that these are only examples and not limitations, which often include "educational institution[s]," "theatre[s]," and "librar[ies]." Colo Rev Stat Ann § 24-34-601(1) (West 1990); see also Hawaii Rev Stat § 489-2 (1993).
              I'm unaware of any cases that have tried to determine whether service providers, or, for that matter, electronic conferences operated on those providers, are places of public accommodation, but such an interpretation seems at least plausible. Moreover, if the claims that cyberspace is the library, theater, and educational institution of the 21st century are taken seriously, then it becomes considerably harder to resist, as a matter of statutory construction, the applicability of the laws to cyberspace accommodations.

 50.     DC Code § 1-2519 (1981 & Supp 1995).

 51.     Hurley, 115 S Ct at 2346-47.

 52.     Roberts v United States Jaycees, 468 US 609, 627 (1984).

 53.     Hurley said that antidiscrimination laws "do not, as a general matter, violate the First or Fourteenth Amendments," 115 S Ct at 2346 (emphasis added), and followed this with a "see, e.g.," citation to the expressive association cases. The "as a general matter" was not a meaningless hedge: As the discussion below shows, it's quite possible that in some situations antidiscrimination laws might indeed violate First Amendment expressive association rights.

 54.     Roberts, 468 US at 628. See also Board of Directors of Rotary Intl v Rotary Club of Duarte, 481 US 537, 548 (1987); New York State Club Association v City of New York, 487 US 1, 13 (1988).

 55.     Roberts, 468 US at 627-28. But see New York State Club Association, 487 US at 13 ("It is conceivable, of course, that an association might be able to show that it is organized for specific expressive purposes and that it will not be able to advocate its desired viewpoints nearly as effectively if it cannot confine its membership to those who share the same sex, for example, or the same religion.").

 56.     Roberts, 468 US at 627.

 57.     New York State Club Association, 487 US at 13.

 58.     Roberts, 468 US at 623; Rotary Intl, 481 US at 549.

 59.     Rotary Intl, 481 US at 549.

 60.     See Bob Jones University v United States, 461 US 574, 602-604 (1983) (holding that prevention of race discrimination is a compelling state interest for Free Exercise Clause purposes).

 61.     Roberts, 468 US at 635 (O'Connor concurring in part and concurring in the judgment). See also New York State Club Association, 487 US at 20 (O'Connor, joined by Kennedy, concurring).

 62.     Roberts, 468 US at 634-35.

 63.     Id.

 64.     Id at 633 (O'Connor concurring in part and concurring in the judgment); id at 638 ("[t]he First Amendment is offended by direct state control of the membership of a private organization engaged exclusively in protected expressive activity").

 65.     Roberts, 468 US at 628.

 66.     Id at 627-28.

 67.     As to marital status, compare Swanner v Anchorage Equal Rights Commission, 874 P2d 274 (Alaska 1994) (holding interest is compelling) with Attorney General v Desilets, 418 Mass 316, 636 NE2d 233 (1994) (remanding for decision on whether interest is compelling), and State ex rel. Cooper v French, 460 NW2d 2 (Minn 1990) (holding interest is not compelling). See also Swanner v Anchorage Equal Rights Commission, 115 S Ct 460 (1994) (Thomas dissenting from denial of certiorari) (expressing skepticism about argument that interest is compelling).
              As to sexual orientation, compare Gay Rights Coalition v Georgetown University, 536 A2d 1 (DC App 1987) (holding interest is compelling) with Curran v Mount Diablo Council of the Boy Scouts of America, 34 Cal App 4th 299, 29 Cal Rptr 2d 580 (1994) (holding interest is not compelling in a noncommercial context), review granted and opinion superseded, 874 P2d 901 (1994) (no final published opinion). As to age, compare Gallo v Salesian Society, Inc., 290 NJ Super 616, 643-44, 676 A2d 580 (1996) (holding interest is compelling in context of employment of teacher at religious school) with Powell v Stafford, 859 F Supp 1343, 1347 (D Colo 1994) (holding interest is not compelling in context of employment of theology teacher at religious school).

 68.     See Invisible Empire of the Knights of the Ku Klux Klan v Mayor of Thurmont, 700 F Supp 281 (D Md 1988) (holding that the KKK's decision to exclude nonwhites and non-Christians from its parade is protected by the freedom of expressive association, despite Roberts).

 69.     Of course, participation in an electronic conference might indirectly provide economic benefits; the connections one makes online can help get one a client or a job or what have you. See Eugene Volokh, Technology and the Future of Law, 47 Stan L Rev 1375, 1401-03 (1995). In my view, though, this possibility isn't enough to justify the infringement on people's choice of whom to speak with. Being invited to my home for a dinner party may help you make business connections. Joining a church can do the same. But despite this, we generally assume the government can't interfere with my choice of guests or a church's choice of congregants that the rights of intimate association and free exercise trump even the government's interests in furthering equality. I agree with Justice O'Connor that the right of expressive association, which goes to the heart of editors' ability to create the speech product they want, should be equally protected.

 70.     Rayburn v General Conference of Seventh-Day Adventists, 772 F2d 1164, 1167-68 (4th Cir 1985) (internal citation omitted) (rejecting sex and race discrimination claim on Free Exercise Clause grounds). Even if one sees conference participants as being more akin to congregants than to the clergy, the analogy still cuts in favor of First Amendment protection; it seems to me that the government may not require religious groups to let in would-be members whom they would prefer to exclude.

 71.     It might also violate the Establishment Clause because it would require the government to entangle itself in decisions about church doctrine, but it seems to me that there's an independent violation of the Free Exercise Clause. A church's right to choose its own clergy seems to me to be quintessentially a matter of religious freedom. Imagine for a moment that the Establishment Clause was never incorporated against the states (a nonlaughable argument, see Akhil Reed Amar, The Bill of Rights as a Constitution, 100 Yale L J 1131, 1157-58 (1991); Note, Rethinking the Incorporation of the Establishment Clause: A Federalist View, 105 Harv L Rev 1700 (1992)); presumably even in its absence a state couldn't order the Catholic Church to ordain women priests.

 72.     See PruneYard, 447 US at 96-97 (Powell concurring); Turner Broadcasting, 114 S Ct at 2480 (O'Connor concurring in part and dissenting in part).

 73.     See Milkovich v Lorain Journal Co., 497 US 1, 20 n 7 (1990) (implied falsehoods, not just express falsehoods, are constitutionally unprotected).

 74.     17 USC § 106 (1988) (giving the copyright holder the exclusive right to prepare or authorize the preparation of copies and derivative works).

 75.     Section 43 of the Lanham Act, 15 USC § 1125(a)(1)(A) (1988 & Supp 1993) (prohibiting misattributions "on or in connection with any goods or services"); Gilliam v American Broadcasting Co., 538 F2d 14 (2d Cir 1976).

 76.     Rowan v United States Post Office Department, 397 US 728, 738 (1970).

 77.     NAACP v Alabama, 357 US 449, 460 (1958).



 80.     See Richards, Wash Post at C2 (cited in note 18).

 81.     For a more ambiguous example, see Daphne Patai, There Ought to be a Law, 22 Wm Mitchell L Rev 491, 515-16 (1996) (discussing the ejection of various posters from the FEMISA discussion list). Quite a few editing decisions may indeed be unnecessarily closed-minded and thus ultimately against the listeners' (or even the list owners') interest. Wise list owners should be careful when they edit, especially when they edit on political grounds. I argue only that editing is often valuable, not that it is always valuable.

 82.     Stratton Oakmont, Inc. v Prodigy Services Co., 1995 WL 323710 (NY Sup, May 24, 1995); Norman Redlich and David R. Lurie, First Amendment Issues Presented by the "Information Superhighway", 25 Seton Hall L Rev 1446, 1457-58 (1995) (suggesting a "legislative solution that would limit the defamation tort remedy to situations where a system operator exercises specific types of editorial control"); Jeffrey M. Taylor, Liability of Usenet Moderators for Defamation Published by Others: Flinging the Law of Defamation into Cyberspace, 47 Florida L Rev 247, 282-83 (1995).

 83.     New York Times v Sullivan, 376 US 254 (1964).

 84.     Gertz v Robert Welch, Inc., 418 US 323 (1974).

 85.     Dun & Bradstreet, Inc. v Greenmoss Builders, Inc., 472 US 749 (1985).

 86.     See Restatement (Second) of Torts § 580B (rejecting strict liability); Nelson v Lapeyrouse Grain Corp., 534 S2d 1085, 1091 (Ala 1988); Schneider v Pay'n Save Corp., 723 P2d 619, 625 (Alaska 1986); Savage v Pacific Gas & Electric Co, 26 Cal Rptr 2d 305, 313, 21 Cal App 4th 434 (1993); Miles v Perry, 11 Conn App 584, 529 A2d 199, 203, 207 (1987); Bagwell v Peninsula Regional Medical Center, 106 Md App 420, 665 A2d 297, 317 (1995); Great Coastal Express, Inc. v Ellington, 230 Va 142, 151-52, 334 SE2d 846 (1985); Ross v Bricker, 26 VI 314, 770 F Supp 1038, 1042 (1991). But see Sleem v Yale University, 843 F Supp 57, 62 (MD NC 1993) (holding that North Carolina law imposes strict liability).

 87.     Rodney A. Smolla, Law of Defamation § 3.02[3] (Clark Boardman Callaghan, 1991).

 88.     Restatement (Second) of Torts § 581, comment d (1977). See also Spence v Flynt, 647 F Supp 1266, 1273 (D Wyo 1986) (holding that magazine seller could be liable because it knew of the allegedly libelous statements).

 89.     If it does make a difference if a state does indeed impose strict liability on people who nonnegligently make false defamatory statements on matters of private concern. I believe that conference operators should indeed be treated as distributors. As I describe below, a conference operator is not much different than a bookstore. Both exercise some editorial control; the bookstore is in fact much more selective about the books it chooses to sell than is the typical conference operator about the posts it chooses to let through. Neither, though, can reasonably be expected to carefully read (or often even read at all) every item they choose to distribute.

 90.     1995 WL 323710 (NY Sup, May 24, 1995).

 91.     Id at *5.

 92.     Id.

 93.     Id at *4.

 94.     Stratton Oakmont, 1995 WL 323710 at *5.

 95.     Id at *4.

 96.     Id.

 97.     Id at *5.

 98.     Stratton Oakmont, 1995 WL 323710 at *5 (also describing Prodigy's conduct as "censorship"). See also Symposium, First Amendment and the Media: Regulating Interactive Communications on the Information Superhighway, 5 Fordham Intellectual Prop, Media & Enter L J 235, 265-66 (1995) (remarks of Jacob Zamansky, Stratton Oakmont's lawyer) (arguing that "the flip side" of editorial control should be increased defamation liability).

 99.     There are many levels of liability that the law could impose. It could, for instance, impose no liability at all on conference operators on the theory that any such liability will lead them to screen out some materials that might in fact not be defamatory, and that any duty to investigate even once they are on actual notice that the material may be false will be unduly chilling. Or it could say that, once a complaint is made about a particular message or even about a particular person who habitually posts allegedly defamatory messages, the only reasonable thing for the operator to do is to investigate. (In theory, even bookstores are supposed to do this, though practically they are rarely sued on these grounds.) It could even require that those conferences that have the economic ability to screen must do so, on the theory that they "should know" the speech that they are transmitting.

 100.     Some have suggested that heightened defamation liability is more appropriate on an edited list because the very assertion that the list is edited makes the list more credible, and thus makes libels on it more dangerous. This premise, though, strikes me as factually incorrect. My sense is that most people who hear that a list is edited simply assume that the editor will take pains to keep discussions civil and on topic (if that). It would be unrealistic for them to assume that the editor will check the facts that are being asserted, and I doubt that people in fact assume this.

 101.     Pub L No 104-104, 110 Stat 113 (1996) amending 47 USC § 233.

 102.     47 USC § 230 (c).

 103.     See, for example, Hurley, 115 S Ct 2338 (parade on public streets); United Auto Workers v Gaston Festivals, Inc., 43 F3d 902 (4th Cir 1995) (festival in public park).

 104.     Rendell-Baker v Kohn, 457 US 830 (1982) (decision by school to fire an employee based on her speech not state action even though school was almost entirely government funded); Weaver v AIDS Services of Austin, Inc., 835 SW2d 798 (Tex App 1992) (decision by organization holding AIDS education classes to bar a person from its workshops not state action even though workshops were government funded); Sinn v The Daily Nebraskan, 829 F2d 662 (8th Cir 1987) (decision by public university student newspaper to exclude certain advertisements not state action even though newspaper was government funded). See also Gannett Satellite Information Network, Inc. v Berger, 894 F2d 61 (3d Cir 1990) (decision by lessee to exclude newspaper rack from its space not state action even though space was leased from government-owned airport); National Broadcasting Co. v Communications Workers of America, 860 F2d 1022 (11th Cir 1988) (decision by union to exclude broadcaster from its convention not state action even though convention was held in a place leased from a city government).

 105.     Cornelius v NAACP Legal Defense and Educational Fund, Inc., 473 US 788 (1985).

 106.     United States v Classic, 313 US 299, 326 (1941), quoted in Polk County v Dodson, 454 US 312, 317-18 (1981). See also United States v Price, 383 US 787, 794 n 7 (1966) (test for state action is the same as the test for when someone is acting under color of state law); Rendell-Baker, 457 US at 838.

 107.     Polk County, 454 US at 319. See also Jojola v Chavez, 55 F3d 488, 493 (10th Cir 1995) (government employee's assault on plaintiff not state action because it was not "committed on account of the authority vested in the employee by the state"); Mark v Borough of Hatboro, 51 F3d 1137, 1150-51 (3d Cir 1995) (same); Edwards v Wallace Community College, 49 F3d 1517, 1523 (11th Cir 1995) (same); Woodward v City of Worland, 977 F2d 1392, 1400-01 (10th Cir 1992) (same; "The mere fact that all [of the defendants] were state employees or that the offending acts occurred during working hours is not enough.").

 108.     See Lamb's Chapel v Center Moriches Union Free School District, 113 S Ct 2141 (1993).

 109.     Of course, restrictions on true threats, extortion, and so on should be as permissible online as they are offline.

 110.     FCC v Pacifica Foundation, 438 US 726, 766 (1978) (Brennan dissenting).

 111.     See Texas v Johnson, 491 US 397 (1989) (holding that public flagburning is protected); Erznoznik v City of Jacksonville, 422 US 205 (1975) (holding that publicly visible depictions of nudity are protected); Cohen v California, 403 US 15 (1971) (holding that public profanity is protected); Kunz v New York, 340 US 290 (1951) (holding that public anti-Catholic and anti-Semitic statements are protected). But see Pacifica Foundation, 438 US at 726 (holding that profanity on radio is unprotected). The Court has at times suggested that offensive speech may be restricted when a "captive audience" is present, but this is an extremely narrow, and in my view troubling, exception from the general rule. See Eugene Volokh, Comment, Freedom of Speech and Workplace Harassment, 39 UCLA L Rev 1791, 1832-43 (1992); Alan E. Brownstein, Rules of Engagement for Cultural Wars: Regulating Conduct, Unprotected Speech, and Protected Expression in Anti-Abortion Protests-Section II, 29 UC Davis L Rev 1163, 1178 n 38 (1996).

 112.     Frisby v Schultz, 487 US 474, 485 (1988) (emphasis added). See also id ("[t]he type of picketers banned by the [ordinance] generally do not seek to disseminate a message to the general public, but to intrude upon the targeted resident"); Rowan v United States Post Office Department, 397 US 728 (1970).

 113.     Rowan, 397 US at 728.

 114.     See Martin H. Redish, The Value of Free Speech, 130 U Pa L Rev 591 (1982).

 115.     A different rule might possibly apply in some unusual cases, such as speakers trying to communicate with public officials. See, for example, Lewis v New Orleans, 408 US 913 (1972) (Powell concurring) (arguing that words that would constitute "fighting words" when directed towards an ordinary citizen should be treated differently when directed towards a police officer who is expected to exercise a higher degree of restraint); United States Postal Service v Hustler Magazine, Inc., 630 F Supp 867 (D DC 1986) (holding that, unlike normal householders, members of Congress may not demand that a mailer stop sending them sexually themed material, and suggesting that this conclusion flows from the right to petition the government for a redress of grievances).

 116.     Rowan, 397 US at 738.

 117.     See Volokh, 39 UCLA L Rev at 1863-66 (cited in note 111); see also Brownstein, 29 UC Davis L Rev at 1166-70 (cited in note 111), for a very thoughtful discussion on this matter, albeit one with which I do not entirely agree.

 118.     Watts v United States, 394 US 705, 707 (1969).

 119.     Cal Penal Code § 653m(b), (c) (West 1988 & Supp 1996) (limited to home or work); Fla Stat Ann § 365.16(1)(d) (West 1968 & Supp 1996) (not so limited).

 120.     Conn Gen Stat Ann § 53a-183 (West 1994), as amended by 1995 Conn Legis Serv 95-143 (West). See also 47 USC § 223(a)(1)(A) (1988 & Supp 1994), as amended by Communications Act of 1996, Pub L No 104-104, 110 Stat 133 (1996); Idaho Code § 18-6710(1)(Michie 1995).

 121.     Pub L No 104-104, 110 Stat at 133, amending 47 USC § 223. See also Idaho Code § 18-6710(1) (Michie 1995).

 122.     Conn Gen Stat Ann § 53a-183, as amended by 1995 Conn Legis Serv 95-143.

 123.     Id.

 124.     47 USC § 223(a)(1)(B) (1988 & Supp 1994).

 125.     Compare Gormley v Director, 632 F2d 938 (2d Cir 1980) (upholding Connecticut law cited in note 117); State v Anonymous, 34 Conn Supp 689, 389 A2d 1270, appeal denied as State v Gormley, 174 Conn Supp 803, 382 A2d 1332 (1978); State v Elder, 382 S2d 687 (Fla 1980); Constantino v State, 243 Ga 595, 255 SE2d 710 (1979); United States v Lampley, 573 F2d 783 (3d Cir 1978); People v Smith, 392 NYS2d 968 (1977) with People v Klick, 66 Ill 2d 269, 362 NE2d 329 (1977) (striking down law); State v Dronso, 90 Wis 2d 110, 279 NW2d 710 (Wis App 1979) (striking down law as overbroad). See generally Gormley v Director, 449 US 1023 (1980) (White dissenting from denial of certiorari).

 126.     Conn Gen Stat Ann § 53a-183 (West 1994), as amended by 1995 Conn Legis Serv 95-143 (West 1995).

 127.     See, for example, Cohen v California, 403 US 15 (1971); Terminiello v City of Chicago, 337 US 1 (1949).

 128.     See, for example, Harris v Forklift Systems, Inc., 510 US 17, 21 (1993) (race, sex, religion, national origin); Eggleston v South Bend Community School Corp., 858 F Supp 841, 847-48 (ND Ind 1994) (age and disability); Leibert v Transworld Systems, Inc., 32 Cal App 4th 1693, 39 Cal Rptr 2d 65, 67 (1995) (sexual orientation); Cal Govt Code § 12940(h)(1) (marital status); DC Code § 1-2512 (1981) (banning discrimination in "terms, conditions, or privileges of employment", a phrase which has been interpreted in the Title VII context as applying to hostile environment harassment based on "marital status, personal appearance, sexual orientation, family responsibilities, disability, matriculation, or political affiliation"); NY City Charter & Admin Code § 8-107(a) (Supp 1995) (barring discrimination in "terms, conditions, or privileges of employment" based on "alienage or citizenship status"); 38 USC § 4311 (1994) (barring discrimination based on a person performing or having performed service in the armed forces); 41 CFR § 60-250.4 (1995) (barring discrimination by federal contractors against Vietnam-era veterans); infra note 136 (describing veteran status harassment case).
              I discuss here only hostile-environment harassment; I don't purport to deal with quid pro quo sexual harassment, in which a supervisor demands sex in exchange for favorable treatment.

 129.     EEOC v Sage Realty Corp., 507 F Supp 599, 609-10 (SD NY 1981); see also Murray v New York University College of Dentistry, 57 F3d 243 (2d Cir 1995) (applying hostile environment educational environment analysis to a dental student's claim of harassment by a patient).

 130.     Brown Transportation Corp. v Commonwealth, 578 A2d 555, 562 (Pa Commw 1990).

 131.     Hilsman v Runyon, 1995 WL 217486 *3 (EEOC).

 132.     Pakizegi v First National Bank of Boston, 831 F Supp 901, 908-09 (D Mass 1993) (dictum) (also describing this a "discriminatory, anti-Iranian conduct"), aff'd without opinion, 56 F3d 59 (1st Cir 1995).

 133.     Turner v Barr, 806 F Supp 1025, 1029 (D DC 1992).

 134.     Robinson v Jacksonville Shipyards, Inc., 760 F Supp 1486, 1542 (MD Fla 1991).

 135.     Snell v Suffolk County, 611 F Supp 521, 532 (ED NY 1985), aff'd, 782 F2d 1094 (2d Cir 1986).


Conciliation Agreement Between the U.S. Department of Labor, Office of Federal Contract Compliance Programs and The Ohio State University (Sept 14, 1992). The requirements involved in this case apply only to federal contractors, but another statute, 38 USC § 4311 (1994), generally bars discrimination by all employers against present or former service members in "any benefit of employment" (which, under the logic of the harassment cases, would include the work environment).

 137.     Tunis v Corning Glass Works, 747 F Supp 951, 959 (SD NY 1990) (dictum), aff'd without opinion, 930 F2d 910 (2d Cir 1991), discussed in more detail in Volokh, 39 UCLA L Rev at 1805-06 (cited in note 111). A Kentucky human rights agency has in fact gotten a company to change its "Men Working" signs (at a cost of $35,000) on the theory that the signs "perpetuat[e] a discriminatory work environment and could have been deemed unlawful under the Kentucky Civil Rights Act." Andrew Wolfson, All Worked Up . . . Phone Company Called to Task Over Gender-Biased Signs, Louisville Courier-J 1-B (Mar 3, 1994).

 138.     See Cardin v Via Tropical Fruits, Inc., No. 88-14201, 1993 US Dist LEXIS 16302 at *24-26 & n 4 (SD Fla July 9, 1993) (finding a hostile environment based largely (though not entirely) on "caricatures of naked men and women, animals with human genitalia, . . . a cartoon entitled `Highway Signs You Should Know' [which showed] twelve drawings of sexually graphic `road signs' ( . . . for example, `merge,' `road open,' etc.)," and so on). Though "[m]any of the sexual cartoons and jokes . . . depicted both men and women," the court concluded that "widespread verbal and visual sexual humor, particularly vulgar and degrading jokes and cartoons . . ., may tend to demean women." Id at *46. The court ultimately held that "every incident reported by [plaintiff]", the jokes as well as the other conduct, "involves sexual harassment." Id at *61. For details on more such incidents, see Eugene Volokh, Freedom of Speech and Appellate Review in Workplace Harassment Cases, 90 Nw U L Rev 1009, 1014 n 23 (1996); Eugene Volokh, What Speech Does "Hostile Environment" Harassment Law Restrict?, 85 Georgetown L J (forthcoming 1997).

 139.     Offended employees have successfully lodged harassment complaints based on, among other things, Francisco Goya's Naked Maja, and a graduate student's 5" x 7" desktop photograph of his wife in a bikini. See Volokh, 90 Nw U L Rev at 1016-17 (cited in note 138) (discussing these incidents in more detail, and citing a number of others); Eugene Volokh, Thinking Ahead About Freedom of Speech and Hostile Work Environment Harrassment, 17 Berkeley J Employment & Labor L 305 (1996) (discussing one such incident in detail).

 140.     Compare, for example, Kingsley R. Browne, Title VII as Censorship: Hostile-Environment Harassment and the First Amendment, 52 Ohio State L J 481 (1991) (arguing that hostile work environment law is largely unconstitutional when applied to speech) with Volokh, 39 UCLA L Rev at 1791 (cited in note 111) (arguing that it's generally unconstitutional when applied to speech but constitutional when applied to directed, one-to-one speech) and Richard H. Fallon, Jr., Sexual Harassment, Content Neutrality, and the First Amendment Dog That Didn't Bark, 1994 S Ct Rev 1 (arguing that it's constitutional in some situations and unconstitutional in others) and Deborah Epstein, Can a "Dumb Ass Woman" Achieve Equality in the Workplace? Running the Gauntlet of Hostile Environment Harassing Speech, 84 Georgetown L J 399 (1996) (arguing that it's entirely constitutional). Compare also, for example, DeAngelis v El Paso Municipal Police Officers Association, 51 F3d 591, 596 (5th Cir 1995) (suggesting free speech defense may have merit) with Jacksonville Shipyards, 760 F Supp at 1534-37 (rejecting free speech defense).

 141.     See, for example, Harris v Forklift Systems, Inc., 510 US 17 (1993); Meritor Savings Bank, FSB v Vinson, 477 US 57 (1986). See generally Volokh, 39 UCLA L Rev at 1793-1807 (cited in note 111).

 142.     Mich Comp Laws Ann § 37.2103(i) (West Supp 1995); see also Minn Stat Ann § 363.01, subd 41 (West 1991); ND Cent Code § 14-02.4-01 (Supp 1995); Cook County, Illinois ord. no 93-0-13 art V(c); cf. Montana Human Rights Commission document at (asserting that Montana state discrimination law bans sexual harassment in public accommodations); New York City Commission on Human Rights document at (asserting that New York City human rights law bars harassment "on the basis of race, color, creed, age, national origin, alienage or citizenship status, gender, sexual orientation, disability, marital status . . . lawful occupation, retaliation, and record of conviction or arrest" in, among other things, public accommodations).

 143.     Bond v Michael's Family Restaurant, Wisconsin Labor & Indus Rev Commission, Case Nos 9150755, 9151204 (Mar 30, 1994). The case suggested that this theory may be limited only to speech by the restaurant owner, but a later case by the same agency made clear that the proprietor can be held liable for a hostile environment created by its patrons, so long as it is able to eject patrons but declines to do so. Neldaughter v Dickeyville Athletic Club, Wisconsin Labor & Indus Rev Commission, Case No 9132522 (May 24, 1994).
              See also Harris v American Airlines, Inc., 55 F3d 1472 (9th Cir 1995) (passenger sued airline based on racist statement made by a fellow passenger; court held that in the airline context such state claims are preempted by federal aviation law); Hodges v Washington Tennis Service Intl, Inc., 870 F Supp 386 (D DC 1994) (health club member sued club over racist statements made by employee; claim dismissed on procedural grounds); In re Totem Taxi, Inc. v New York State Human Rights Appeal Bd, 65 NY2d 300, 480 NE2d 1075, 491 NYS2d 293 (1985) (passengers sued taxi company over racist statements and threats made by taxi driver; claim dismissed because company had taken reasonable steps to prevent such conduct); compare King v Greyhound Lines, Inc., 61 Or App 197, 656 P2d 349 (1982) (customer sued bus company over racist statements made by an employee; court held for customer, but suggested that the law might not cover racist statements made by a fellow patron, and that employment harassment law might not be an apt analogy); Robert A. Sedler, The Unconstitutionality of Campus Bans on "Racist Speech": The View from Without and Within, 53 U Pitt L Rev 631, 673 (1992) (asserting, though not specifically in the cyberspace context, that racist statements by employees of places of public accommodation to customers violate laws that bar discrimination in public accommodation); Deborah M. Thompson, "The Woman in the Street": Reclaiming the Public Space from Sexual Harassment, 6 Yale J L & Feminism 313 (1994) (accepting the notion that public accommodation laws bar harassing speech and suggesting that they be extended to public parks).

 144.     In re Minnesota by McClure v Sports & Health Club, Inc., 370 NW2d 844, 872-73 n 40 (Minn 1985) (Peterson dissenting); id at 853 & n 16; id at 867 n 25 (Peterson dissenting). But see Haney v University of Illinois, No 1993SP0431, 1994 WL 880339 (Ill. Human Rights Commission) (holding that state public accommodations law does not bar the creation of a hostile environment, in large part because of free speech concerns).

 145.     See, for example, Murray v New York Univ. College of Dentistry, 57 F3d 243 (2d Cir 1995) (education); see also Honce v Vigil, 1 F3d 1085, 1090 (10th Cir 1993) (housing).

 146.     See note 49 and accompanying text.

 147.     Stuart Biegel, Hostile Connections, LA Daily J 7 (Aug 22, 1996).

 148.     The remarks were "anatomically explicit and sexually derogatory," but there was no allegation that they were threatening or otherwise generally outside the First Amendment's protections. Letter from John E. Palomino, Regional Civil Rights Director for United States Department of Education, Office of Civil Rights, to Dr. Robert F. Agrella, President of Santa Rosa Junior College 2 (June 23, 1994) (on file with author).

 149.     See id at 7. The message was posted on a men-only conference by student request, SOLO included men-only and women-only conferences as well as integrated ones, but this factor didn't affect the OCR's harassment analysis. It seems to me that, if the messages were posted on an integrated conference, especially one which the offended women read, this would have only exacerbated their harassing effect.

 150.     The Office concluded the college didn't violate Title IX, but only because it "took immediate action to remedy the harm . . . and to prevent sexual harassment from occurring in the future." Id at 8. See also CMU Disciplinary Charges, at, reporting various "harassment" claims based on online speech at Carnegie Mellon University.

 151.     Regarding proposed "hate speech" restrictions generally, and on campus in particular, see Nicholas Wolfson, Free Speech Theory and Hateful Words, 60 U Cinn L Rev 1 (1991); Nadine Strossen, Regulating Racist Speech on Campus: A Modest Proposal?, 1990 Duke L J 484, 498; Michael Kent Curtis, Critics of "Free Speech" and the Uses of the Past, 12 Const Comm 29 (1995); see also Alex Kozinski and Eugene Volokh, A Penumbra Too Far, 106 Harv L Rev 1639 (1993); Eugene Volokh, Freedom of Speech and the Constitutional Tension Method, 3 U Chi Roundtable 223 (1996); Henry Mark Holzer, ed, Speaking Freely: The Case Against Speech Codes (Second Thoughts Books, 1994). Regarding workplace harassment law, see note 140.

 152.     Hustler Magazine v Falwell, 485 US 46 (1988); Brandenburg v Ohio, 395 US 444 (1969) (per curiam); Cohen v California, 403 US 15 (1971); Cantwell v Connecticut, 310 US 296 (1940).

 153.     See Dambrot v Central Michigan University, 55 F3d 1177 (6th Cir 1995); UWM Post, Inc. v Board of Regents, 774 F Supp 1163 (E D Wis 1991); see also Iota Xi v George Mason University, 993 F2d 386 (4th Cir 1993) (striking down speech restriction that was not patterned directly on harassment law); Doe v University of Michigan, 721 F Supp 852 (ED Mich 1989) (same); compare Corry v Leland Stanford Jr. University, No 740309 (Cal Super Ct, Feb 27, 1995) (speech code at private university, patterned partly on harassment law, struck down under state law that extended Free Speech Clause protection to students at private universities).

 154.     See cases collected in Volokh, 90 Nw U L Rev at 1029 n 85 (cited in note 138).

 155.     Jacksonville Shipyards, 760 F Supp at 1535.

 156.     Recall that Pacifica Foundation, 438 US at 726, held that mere profanity can be described as "indecent," even if it doesn't appeal to the prurient interest.

 157.     The Occasional Screenful, We have over 1650 subscribers; to subscribe, send the message "subscribe occasional-screenful" (not followed by your name) to To submit poems, send them to

 158.     E-mail to author, October 19, 1995.

 159.     See cases cited in note 125.

 160.     See, for example, United States v Lampley, 573 F2d 783, 787 (3d Cir 1978).

 161.     See, for example, Gene Barton, Taking a Byte out of Crime: E-Mail Harassment and the Inefficacy of Existing Law, 70 Wash L Rev 465, 480-83 (1995).

 162.     I speak here only of speech that is generally protected by the Free Speech Clause; unprotected speech, such as threats, can and should be prohibited online as well as offline.

 163.     397 US at 728.

 164.     Id at 737.

 165.     Id at 738.

 166.     Id at 737.

 167.     47 USC § 223(d)(1)(B), as amended by Pub L No 104-104, 110 Stat at 133 (1996).

 168.     ACLU v Reno, 929 F Supp 824 (ED Pa 1996), cert granted, 65 USLW 3411 (US Dec 10, 1996) (No. 96-511).

 169.     See, for example, American Booksellers v Webb, 919 F2d 1493 (11th Cir 1990) (upholding Georgia law cited in note 137); Davis-Kidd Booksellers, Inc. v McWherter, 866 SW2d 520 (Tenn 1993).

 170.     Ga Code Ann § 16-12-103(e) (Michie 1992).

 171.     Ga Code Ann § 16-12-102(1) (Michie 1992). See also Ala Stat § § 13A-12-200.1(3), 13A-12-200.5 (1994) (probably prohibiting only display for sale); Ariz Rev Stat Ann § 13-3507 (West 1989) (prohibiting any display in any "place where minors are invited as part of the general public"); Fla Stat Ann § 847.0125 (West 1994) (prohibiting only display for sale); Ind Code Ann § 35-49-3-3(2) (West 1995) (prohibiting any display "in an area to which minors have visual, auditory, or physical access"); Kan Stat Ann § 21-4301c(a)(1) (1988) (prohibiting display in commercial establishments only); La Rev Stat Ann § 14:91.11 (West 1995) (prohibiting any display "at a newsstand or any other commercial establishment which is open to persons under the age of seventeen years"); Minn Stat Ann § 617.293, subd 2(a) (West 1987 & Supp 1996) (prohibiting commercial display); NM Stat Ann § 30-37-2.1 (1978 & Supp 1995) (prohibiting display only while offering for sale, "in a retail establishment open to the general public," and "in such a way that it is on open display to, or within the convenient reach of, minors who may frequent the retail establishment"); NC Gen Stat § 14-190.14(a) (1993) (prohibiting display in commercial establishments only); Okla Stat Ann § § 1040.75, 1040.76 (West 1983 & Supp 1996) (prohibiting all display, "including but not limited to . . . commercial establishment[s]"); Tenn Code Ann § 39-17-914(a) (1991) (prohibiting display for sale or rent); Tex Penal Code Ann § 43.24 (Vernon 1991) (prohibiting all display, whenever person is "reckless about whether a minor is present who will be offended or alarmed by the display"); 13 Vt Stat Ann § § 2801(8), 2804a (Equity 1971 & Supp 1995) (prohibiting display "for advertising purposes").

 172.     See, for example, Davis-Kidd Booksellers, 866 SW2d 520 (largely upholding a similar Tennessee statute).
                 I am suggesting only that a court may plausibly interpret the statute to cover online locations, not that the statutes should necessarily be interpreted this way. Calling cyberspace locations "places" is a metaphor, and one may certainly argue that the statutes ought not be read metaphorically. On the other hand, this metaphor seems both reasonable and widely accepted; cyberspace is in many respects (though, as the text points out, not in all respects) functionally similar to bookstores, theaters, and other tradtional "public places." Compare the discussion in note 49 about whether cyberspace locations can be "places of public accommodation."

 173.     492 US 115, 128-131 (1989).

 174.     Burson v Freeman, 504 US 191, 199 (1992) (plurality opinion). See also Buckley v Valeo, 424 US 1, 28-29 (1976).

 175.     352 US 380, 383 (1957).

 176.     I suggest elsewhere that strict scrutiny may be the wrong approach, both wrong normatively and wrong descriptively, for this analysis; Butler may in fact be read as mandating more protection of speech than the strict scrutiny test provides. See Eugene Volokh, Freedom of Speech, Permissible Tailoring, and Transcending Strict Scrutiny, 144 Penn L Rev 2417 (1996). For purposes of this discussion, though, I'll accept the validity of the orthodox strict scrutiny framework.

 177.     See Ginsberg v New York, 390 US 629 (1968).

 178.     See, for example, Davis-Kidd Booksellers, 866 SW2d at 528-29 (upholding a display statute in part because "compliance does not impose an impermissible burden on merchants subject to its provisions. . . . [T]he statute specifies a variety of methods of compliance allowing booksellers to choose the least burdensome means. . . . [A]dults are not completely denied access to the minimal amount of materials subject to the provisions of the statute.").

 179.     Phone conversation with salesman at Egghead Software in Los Angeles, October 30, 1995.

 180.     SurfWatch subscription information as of June 29, 1996 at

 181.     A filter distributor might avoid this problem to some extent by marking entire sites, rather than just single pages, as clean; it could, for instance (this is something of an oversimplification), check the current Web pages housed at http://www.a.b.c/joe.schmoe/, confirm that they're all clean, and conclude that any Web page whose address starts that way is clean. The cost of that, though, is that if Joe Schmoe adds a dirty page his entire Web site, including that page, will still be seen as clean until the filter distributors check the entire site again.

 182.     One can certainly argue that the parents want to have it both ways, they want to both have shielding (even legally enforced shielding) from sexually explicit material and at the same time not have too much shielding, but perhaps under Sable, Ginsburg, and the public display cases they might be entitled to have it both ways.

 183.     The closest analogy here is the plurality opinion in Burson v Freeman, which upheld a ban on electioneering around polling places on the grounds that it was the least restrictive means of serving a compelling interest in preventing interference with voting. 504 US 191 (1992). The dissent suggested that laws directly banning intimidation would be a less restrictive alternative, but the plurality disagreed. "[B]ecause law enforcement officers generally are barred from the vicinity of the polls to avoid any appearance of coercion in the electoral process, many acts of interference would go undetected." Id at 207 (internal citation omitted). Thus, the plurality suggests that even if allowing police officers near the polls to enforce the laws would have been a less restrictive alternative, it is not the constitutionally compelled alternative, because it would achieve the government's goals only at an unacceptable harm to other goals (the avoidance of any appearance of coercion).

 I disagree with the Burson plurality on other grounds, but the general notion that undesirable side effects of proposed alternatives must be considered seems to me to be sound.

 184.     See, for example, Sable Communications, 492 US at 131-32 (Scalia concurring).

 185.     Id at 126-31.

 186.     Burson, 504 US at 206-207 (plurality opinion). See also Buckley, 424 US at 27-28.

 187.     This system would shield children from "harmful to minors" material only if their parents so choose, but that's as it should be. See Ginsberg, 390 US at 639.

 188.     Given that foreign sites won't be bound by the law, and will presumably post all their pictures with no ratings, the assumption that an unrated picture may be dirty is probably the safest one. Of course, foreign sites might explicitly misrate "dirty" contents as "clean," but this sort of intentional action seems much less likely than a simple failure to attach a rating that isn't even required in one's own country.

 189.     Ginsberg, 390 US at 643-45. Requiring people to rate fully protected material, material that's not obscene even as to minors, might indeed raise serious constitutional problems. One might view it as an impermissible content-based burden on speech, or as an impermissible form of compelled speech. But see Meese v. Keene, 481 US 465 (1987) (upholding a labeling requirement, in my view incorrectly). Even requiring people to rate harmful-to-minors material is not without constitutional difficulty, but in light of the permitted restrictions on harmful-to-minors matter, a labeling requirement for such matter would probably be constitutional.

 190.     Some might dislike courts second-guessing legislative judgments on the relative effectiveness of various alternatives, but under current First Amendment law the courts have a constitutional duty to arrive at their own independent judgment on such matters. Sable Communications, 492 US at 129.

 191.     Sable Communications, 492 US at 124-26.

 192.     See I. Trotter Hardy, The Proper Legal Regime for "Cyberspace", 55 U Pitt L Rev 993, 1012-13 (1994).

 193.     Virginia Postrel, The Age of the Editor, in The State and Fate of Publishing: A Flair Symposium 25 (1995).