Freedom of Speech, Shielding Children, and Transcending Balancing

The government has a strong interest in shielding children from unsuitable -- because sexually explicit or (perhaps) profane -- speech. So says the Court, and so say even many who generally frown on the regulation of sexually explicit material. 1 At the same time, the Court has held, much speech of this sort is constitutionally valuable. How can these strong competing claims, the government interest and the constitutional right, be reconciled?

The Court's official answer is strict scrutiny: Speech to adults may be restricted to serve the compelling interest in shielding children, but only if the restriction is the least restrictive means of doing so. 2 In Reno v ACLU, 3 the Court applied this framework to strike down the Communications Decency Act, a statute which pretty much banned not-for-pay online distribution of material containing "patently offensive" speech about "sexual or excretory activities." The decision was widely considered a great victory for free speech, 4 and I agree that it reached the right result.

Nonetheless, I believe that the logic of the ACLU opinion is deeply flawed, and that the flaws in the opinion reveal serious problems with the strict scrutiny framework. The opinion, I will argue, rests on a factually incorrect assertion. It fails to confront the critical normative judgment about the real sacrifice that free speech demands. The strict scrutiny framework that ACLU applies ultimately underprotects speech. And by following this framework, the opinion misses an opportunity to synthesize the cases into a different framework that's more accurate and more useful.

Below, I briefly describe the ACLU case (Part I), explain my criticism of the opinion (Parts II and III), present several alternative approaches to dealing with the problem (Part IV), and suggest how both the criticism and the alternatives might be generalized to other areas of free speech law (Part V).

I. Reno v ACLU

A. The Communications Decency Act

In 1996, Congress passed the Communications Decency Act (CDA). Its most controversial provision -- the only one I will focus on here -- provided that

(d) Whoever

(1) in interstate or foreign communications knowingly . . .

(B) uses any interactive computer service to display in a manner available to a person under 18 years of age, any comment, request, suggestion, proposal, image, or other communication that, in context, depicts or describes, in terms patently offensive as measured by contemporary community standards, sexual or excretory activities or organs . . .

shall be fined under Title 18, or imprisoned not more than two years, or both. 5

Most places on the Internet are generally open to everyone, child or adult; there's no way to check readers' ages, short of the expensive (and imperfect) proxy of demanding and verifying their credit card numbers. Therefore, the CDA would have essentially banned material that "depicts or describes, in terms patently offensive as measured by contemporary community standards, sexual or excretory activities or organs" -- which I will call "indecent" material for short 6 -- from all parts of the Internet except those that charge people for access using credit cards. 7 Because most of the Internet is now available for free, and because this free access is widely considered one of the Internet's great strengths, this naturally struck many as a very broad restraint.

Moreover, as Justice Stevens's majority opinion 8 pointed out, the restraint was made broader by its vagueness. "Could a speaker confidently assume," the opinion asked, "that a serious discussion about birth control practices, homosexuality, the First Amendment issues raised by the Appendix to [FCC v Pacifica Foundation, the `Seven Dirty Words' case], or the consequences of prison rape would not violate the CDA?" Justice Stevens was correct in thinking that the answer to this is "no"; while one might hope that prosecutors and juries wouldn't read the law this broadly, the text gives no such assurance. The statutory definition was potentially broad enough to cover such speech, with no safe harbor for speech that has substantial value, or for speech that doesn't appeal to prurient interests.

Thus, the Court faced a speech restriction that would have at least deterred, and quite possibly punished, a considerable amount of generally presumptively protected speech. At the same time, the restriction was said to be justified by a government interest to which the Court had paid considerable respect -- the interest in shielding children from offensive material. Either the right or the interest had to at least partly yield.

B. The Court's Choice of Level of Scrutiny

The Court could have avoided the full force of this conflict by fitting the CDA into one of the boxes where speech restrictions are more freely permitted than in other areas.

Low-value speech: If the CDA could have been seen as limited to "low-value" speech, the Court could have let the government prevail while theoretically imposing little sacrifice of free speech, because the burdened speech would be (by hypothesis) not very valuable. This wouldn't have entirely eliminated the free speech sacrifice -- low value does not equal no value -- but it would have made it seem less momentous.

But though Justice Stevens had in the past been the Court's leading proponent of this approach, 9 his opinion didn't even mention it. This might have been because the CDA covered much material that would be hard to call low value, 10 or because many of the other Justices are uncomfortable with the idea of treating "indecent" speech as low value. 11 In either event, the Court declined to take this escape route.

Secondary effects: The government urged the Court to treat the CDA as an essentially content-neutral law, justified not by the content of "indecent" speech, but by its supposed "secondary effects." 12 The Court correctly rejected this argument. Whatever the possible merits of the "secondary effects" doctrine in other contexts, the danger that speech will corrupt its listeners is a classic primary effect. "Listeners' reaction to speech is not a content-neutral basis for regulation"; 13 a law "justified by the [government's] desire to prevent the psychological damage it felt was associated with viewing [indecency] [must be analyzed] as a content-based statute." 14

The broadcasting analogy: The Court could also have concluded that online speech is entitled to the lesser constitutional protection afforded broadcast radio and television. For three decades, the Court has formally treated restrictions on broadcasting quite differently from similar restrictions on speech in other media. 15 The ACLU Court, though, refused to extend the broadcast test to the Internet. The Internet, the Court held, shares none of the "special justifications for regulation" that led to diminished protection for broadcast speakers: There is no "history of extensive government regulation of the . . . medium," no "scarcity of available frequencies at [the medium's] inception," and no specially "`invasive´ nature" to the medium that would make it easy for people to encounter offensive material by accident. 16 This holding was important but not surprising, because the broadcasting cases have generally had rather little gravitational force; in Turner Broadcasting v FCC (1994), 17 for example, the Court refused to extend them even to cable television. 18

C. Reconciling the Interest and the Right

Thus, the Court acknowledged that the CDA imposed a heavy burden on free speech: It restricted a great deal of presumptively fully protected speech based on its content, and did so in a fully protected medium. 19 But this cannot be the end of the inquiry, because orthodox free speech doctrine holds that even such a heavy burden may sometimes be imposed in the pursuit of a "compelling governmental interest." Given the choice between sacrificing free speech and sacrificing the compelling interest, sometimes free speech will have to yield. 20

One possible solution, of course, would be to say that shielding children from "patently offensive" descriptions of sex and excretion is not a compelling interest. Maybe such materials are somewhat harmful to children's upbringing; maybe children have no constitutional right to receive such material, and adults have no right to communicate it to them; but, the argument would go, the harm is not great enough to justify restraints on communication among adults. Better to sacrifice the shielding of children than to restrict speech. 21

The Court, however, has not taken this view. In Sable Communications v FCC, the Court noted that "there is a compelling interest in protecting the physical and psychological well-being of minors . . . [by] . . . shielding minors from the influence of literature that is not obscene by adult standards," 22 and ACLU did not contradict this. 23 One might still argue, of course, that this interest extends not to all patently offensive descriptions of sex and excretion, but only to extremely explicit ones, or only to those that appeal to prurient interests, or to some other narrow category; but ACLU did not rest on such an argument. 24

So in some situations free speech to adults may be restricted in order to shield minors. But, the Court said, not here. Why?

D. Applying Strict Scrutiny

The CDA is invalid, the Court said, because it is possible to protect speech in this context without any sacrifice of shielding of children. The CDA is simply insufficiently "carefully drafted," simply lacks "precision." 25 The conflict that might require hard trade-offs between a precious constitutional right and a compelling government interest was, the Court said, in fact absent.

The Court explained that the burden on free speech "is unacceptable if less restrictive alternatives would be at least as effective in achieving the legitimate purpose that the statute was enacted to serve." Though the Court has "repeatedly recognized the governmental interest in protecting children from harmful materials," that interest "does not justify an unnecessarily broad suppression of speech addressed to adults." Congress must "desig[n] its statute to accomplish its purpose `without imposing an unnecessarily great restriction on speech.´" The government bears a "heavy burden to explain why a less restrictive provision would not be as effective as the CDA." And given the "possible alternatives, such as requiring that indecent material be `tagged´ in a way that facilitates parental control of material coming into their homes, making exceptions for messages with artistic or educational value, providing some tolerance for parental choice, and regulating some portions of the Internet -- such as commercial Web sites -- differently than others, such as chat rooms," the government hasn't discharged this burden. 26

If the four phrases I italicized in the previous paragraph are correct, then Congress might have been able to have its cake and eat it too: It could have restricted speech less without sacrificing any shielding of children. The alternatives the Court identifies would be "at least as effective," and thus make the broader restraint imposed by the CDA "unnecessarily broad" and not "carefully drafted." (If they would not have been as effective, then the CDA would have been a "[]necessarily broad suppression" and a "[]necessarily great restriction," because it would have been necessary in order for Congress to fully accomplish its purpose. 27)

Tastes great and less filling! If the Court is right, then there really was no excuse at all for the CDA being passed. If the Court is right.

II. The Court's Error

A. No Equally Effective Alternatives

But the Court is wrong. None of the Court's proposed alternatives to the CDA -- or any other alternatives I can imagine -- would have been as effective as the CDA's more or less total ban.

1. Compulsory Tagging.

Consider the Court's first proposed alternative -- "requiring that indecent material be `tagged´ in a way that facilitates parental control of material coming into their homes." This is actually a pretty good alternative: Under it, parents who use special "filter" software can make their computers block access to any material that's tagged as indecent. 28 And it could be made better still if parents didn't have to spend time or money finding and buying the best filters, for instance, if the government made the software available for free, and service providers such as America Online had an option that easily turned it on. 29 Some parents still wouldn't use it, but perhaps one might say that they've chosen to let their children have unlimited access, so the government interest in shielding their children would then become less than compelling. (This latter view is controversial, but let's assume it for now.)

But even with this, compulsory tagging isn't "at least as effective" as the CDA in serving the CDA's "important purpose of protecting children from exposure to sexually explicit material." Filters work only on those computers on which they are installed and activated, and parents have little control over the computers used by their children's friends.

True, unusually conscientious parents may ask their children's friends' parents whether they have shielding software installed. But even conscientious parents often don't know every home that their teenagers might visit; and even for younger kids, what can one do when the other parent says, "Shielding software? Yeah, I think I have that option on, maybe"? Should one come over to check? What if the other parent says, "Yes, I definitely have the shielding software turned on," but turns out to be wrong? What if, unbeknownst to the other parent, his child has found some instructions for disarming the filter software, put on the World Wide Web by a helpful "anti-censorship" activist? 30

And so long as even a significant minority of homes in a particular social circle don't use shielding software -- whether intentionally or carelessly -- most kids in the circle will be able to get access to indecent material. They'll be able fairly quickly to find out who has the unshielded computer, and then come over to see what they want to see. 31 With online materials (unlike, say, with dial-a-porn), the kids don't even have to come over; a child who uses an unshielded computer can pull down the material, remove the tag, and forward it to others. 32

Congress might try to fight even this by (1) requiring tagging, (2) making it illegal for people -- including children -- to forward material with removed tags, and (3) making it illegal for people to let their unshielded computers be accessed by others' children. In theory, this might outlaw as much exposure of children to indecent material as the CDA would have.

But in practice, trying to deter Web site operators is much more effective than trying to deter kids from forwarding material to other kids, or trying to make parents police who is using their kids' unshielded computers. 33 Children are much less likely to know the law or to follow it, especially since the chances of a 12-year-old being prosecuted for e-mailing an indecent picture to another 12-year-old seem quite slim. And holding insufficiently watchful parents liable for access by their children's friends seems hardly fair or effective. 34

Compulsory tagging would thus provide considerably less shielding than the CDA would. It may be fairly effective; it may be as effective as you can get without very greatly burdening speech; but it's not "at least as effective" as the CDA. 35

2. Other Suggested Alternatives.

What about the second alternative, "making exceptions for messages with artistic or educational value"? Well, if the government's

interest in "protecting children from exposure to sexually explicit material" is sufficiently strong only when the material is artistically and educationally valueless, then this alternative would be as effective as the CDA at serving this limited interest. But the Court nowhere says that this is the relevant government interest, and nowhere explains why this would be so. Rather, the Court acknowledges the government interest in "protecting children from exposure to sexually explicit material" (or even "indecent material") with no qualifiers. 36 Maybe a CDA with an artistic/educational value exemption would not be dramatically less effective at serving this broader interest; but it certainly wouldn't be "at least as effective" as the CDA itself.

How about "providing some tolerance for parental choice"? Well, this sounds fine, but how exactly could this "tolerance" be implemented? If the Court is saying only that Congress must provide an exception for parents sending material directly to their children -- and if the government interest is limited to protecting children from indecent material when their parents think the material is harmful -- then that would indeed be a less restrictive and equally effective alternative. But if that's all the Court is saying, then the decision has been almost entirely for nought; Congress could tomorrow reenact a ban that's as broad as the CDA, and one that interferes as much with people's communications to consenting adults, subject to this one small exception. Surely that can't be all the Court is saying: Among other things, if it were, the statute couldn't be facially invalidated as substantially overbroad. 37

More likely, the Court might be suggesting that Congress "provide some tolerance for parental choice" by giving parents some means by which their children can freely surf the Web with no continuous intervention by parents -- that parents should have some option by which they can free their children of the Act's burden. This assumes that the government has no sufficient interest in shielding children whose parents don't want the shielding, itself a contested proposition. 38 But even if the interest is so limited, no such option could be as effective as the Act in shielding the children whose parents do want the shielding. Any "electronic note from my mother" system is just too easy to evade.

Finally, "regulating some portions of the Internet -- such as commercial Web sites -- differently than others, such as chat rooms" would likewise be less effective than the CDA at shielding children. Wherever the regulations are less restrictive, they'll also be less effective. Again, they might be fairly effective, reasonably effective, not ineffective. But they won't be "at least as effective."

3. Fact Findings Below.

Perhaps the Court's error would have been understandable if the Court had been led into it by erroneous fact findings at trial. But the district court never evaluated the supposed effectiveness of the Court's proposed alternatives.

The chief alternative dealt with at trial was filtering software: programs that, when run on a personal computer, block access from that computer to a list of "dirty" Internet locations. The list is maintained and frequently updated by the software manufacturers; the software can also block access to a range of locations (e.g., all Web pages at http://www.playboy.com/) or to materials that contain certain forbidden words. These filter programs could be modified to accommodate a mandatory tagging scheme, but the trial court findings focused only on filters as such, not on filters plus tagging. 39

And the three-judge district court was careful not to overstate the effectiveness of those alternatives that it considered. The court never found that any alternative would be "at least as effective" as the CDA; it found only that "a reasonably effective method . . . will soon be widely available" 40 -- surely the word "reasonably" ought to be a clue that the alternative might not be "equally effective." None of the three separate opinions suggested that there were any equally effective alternatives. The three-judge district court in Shea v Reno, another case that struck down the CDA, and that was pending before the Court when ACLU was decided, likewise never said there were any equally effective alternatives. 41 Rather, Shea avoided reaching this question by holding that "[e]ven if . . . nothing short of a total ban on indecent communication could be as effective," the law would still be unconstitutional. 42

The briefs before the Supreme Court likewise didn't focus on compulsory tagging as an alternative, probably because the party that had the most to gain from raising the matter -- the ACLU -- didn't want to be seen as endorsing compulsory tagging, which is itself a speech restriction (albeit one milder than the CDA). 43 Still, the government's opening brief did generally argue that "There Are No Alternatives That Would Be Equally Effective in Advancing the Government's Interests." 44 The government's reply brief likewise claimed that shielding software was "not effective," and that while it "could provide part of the answer to the problem of indecency of the Internet, it could not provide the full answer. The district court did not conclude otherwise." 45

Three of the four amicus briefs filed on the government's side explicitly pointed out that filters alone would not be "as effective" as filters plus the CDA because "children have access to many computers which will not employ software filtering devices such as in . . . neighbors' homes." 46 Even the ACLU's own brief said that the inquiry into whether "there are no `equally effective´ alternatives . . . misstates the relevant legal test. It is always true that only an `absolute ban´ on adult speech `can offer certain protection against assault by a determined child.´" 47 Wherever the Court got the erroneous notion that the alternatives would be equally effective, it wasn't from the findings below or from any concessions by the government.

B. The Problems Caused by the Error

The Court's error is more than just a harmless misstatement. To begin with, it's unfair to Congress. A Congress that restricts speech even though there are equally effective but less restrictive alternatives available -- that implements a genuinely "unnecessary" restriction -- is either incompetent or flagrantly unconcerned with the First Amendment: It's depriving us of free speech without getting us any benefit in return. In the Court's words, "[t]he CDA's burden on protected speech cannot be justified if it could be avoided by a more carefully drafted statute." 48 When the Court then says that the burden is indeed unjustified, it must be because Congress was the opposite of "careful" -- either careless or uncaring.

In reality, Congress wasn't acting that badly. It did sacrifice some free speech, but this sacrifice was necessary to shield children as well as possible. Such a trade-off may be unconstitutional, and the Court's trade-off -- sacrificing some shielding of children in order to more thoroughly protect free speech -- may be better. But this assertion is much less damning than a claim that Congress actually bought nothing with the trade-off it made.

Furthermore, as Part III.B.3 discusses in more detail, the Court's error makes ACLU a much less useful benchmark for future cases. The Court's First Amendment doctrine is "given meaning through the evolutionary process of common law adjudication," 49 in which the facts of a case can be compared and contrasted with the facts of earlier ones. If the facts in the original case are incorrectly stated or analyzed, then the case will be of extremely limited precedential value.

But most importantly, the Court's stress on equal effectiveness risks dramatically underprotecting speech in future cases. The pregnant negative in the Court's reasoning is that, had there really been no equally effective alternatives (as in fact there are not), the CDA should have been upheld.

Other cases have in fact stated this pregnant negative as a positive: Speech restrictions, these cases say, are valid if they are "necessary to serve a compelling state interest," 50 and alternatives that "fall short of serving [the] compelling interest[]" 51 aren't enough to rebut the claim of necessity. After all, there are always alternatives that are less restrictive but ineffective, for instance, having no government action at all, or perhaps just having the government urge people to go along voluntarily. But only the equally effective alternatives show that the restriction is in fact unnecessary -- not needed to accomplish the interest. 52 The Court has said similar things in other strict scrutiny contexts, 53 and lower court free speech decisions echo this view. 54

As Part III will explain, these cases are somewhat ambiguous about just how effective the alternatives must be. One could read them as saying that a speech restriction is constitutional if there are no equally effective alternatives, or one could read them (and I think they are probably best read this way) as saying that the restriction is constitutional if there are no more or less equally effective alternatives.

But this ambiguity only makes ACLU's emphasis on the alternatives being "at least as effective" particularly dangerous. A lower court, a legislator, or an executive official can easily read ACLU -- coupled with the other cases -- as choosing the "equally effective alternative" test: If none of the alternatives is "at least as effective" at serving the compelling interest, then the alternatives all "fall short of serving" the interest, the restriction is "necessary to serve" the interest, and therefore survives First Amendment scrutiny.

If this pregnant negative is accepted, then presumably Congress could just reenact the CDA whenever it gets enough evidence -- which, for the reasons described above, should not be hard to do -- that the alternatives would not be equally effective. Of course, in practice the Court may be unlikely to revisit the CDA's constitutionality simply in the face of more factual findings (though why shouldn't it, if constitutionality ultimately turns on a fact question?), but what about future bans, in cyberspace and out? They would be constitutional so long as there was no equally effective alternative for shielding children, a factual predicate that would almost always be met. As the Supreme Court has recognized, and as even the CDA's opponents acknowledged, only an "absolute ban" "can offer certain protection against assault by a determined child." 55 And yet surely this would be the wrong result, one that's inconsistent with the result in ACLU and with other cases, such as Butler v Michigan (which I discuss more below). 56

III. Harmless Hyperbole?

A. The "Pretty Much Equally Effective Alternative" Test

I think it's not too much to ask that the Court's factual assertions be literally accurate, 57 but I recognize that many might fault me for being a bit too persnickety here. Maybe Justice Stevens was just engaging in harmless hyperbole: Maybe he meant that the burden on speech is unacceptable if "less restrictive alternatives would be at least pretty much equally effective"; that Congress hadn't proven that "a less restrictive provision would not be more or less as effective as the CDA."

In fact, the Court has seemed to suggest this sort of rule in some earlier child shielding cases, including Sable Communications v FCC and Denver Area Educational Telecommunications Consortium v FCC; 58 though the cases aren't entirely clear, this is probably the best way of putting the strict scrutiny test. 59 Maybe the Court meant to say only that some alternatives -- perhaps the compelled tagging system -- would probably be almost as effective as the CDA.

B. The Problems with Reading Reno v ACLU This Way

1. Fostering Confusion among Lower Courts and Government Officials.

There are, however, serious difficulties even with this reading of ACLU. To begin with, hyperbole makes bad caselaw. Lower courts might well think that they ought to read Supreme Court opinions literally, and that "at least as effective" means "at least as effective." 60 Quite likely some courts will read ACLU this way, while others impose a "pretty much equally effective alternative" standard; likewise for legislators and executive branch officials. This is hardly a recipe for coherent decision making.

2. Hiding the Normative Judgment.

More importantly, the Court's factually erroneous claim that there are "equally effective" alternatives hides a significant and potentially controversial normative decision. By saying there are equally effective alternatives, alternatives that involve no loss of shielding for children, the Court could claim as a descriptive matter that the alternatives would be a win-win or at least a win-draw situation. But a requirement that the alternatives be merely "fairly effective," "almost as effective," or "pretty much equally effective" would mean some shielding of children would indeed be sacrificed. If the Court had to frankly admit that the alternatives would be a win for one side's concerns (free speech) at the expense of the other's (shielding children), then the Court would have had to give a normative explanation of why the losing side had to bear this loss.

The Justices would thus have had to say something like: "Any alternative to the CDA would sacrifice some shielding of children, and shielding children is a compelling government interest. But we think the benefits of protecting adults' access to free speech, even indecent speech, outweigh these costs. We realize Congress may have reached a different normative judgment, and concluded that even a small sacrifice of shielding was too high a price. But we disagree with Congress's normative judgment, for the following reasons. . . ."

There would have been nothing illegitimate about this frank substitution of the Court's own normative judgment for Congress's; I argue in Part IV that the Court should have done something quite like this. But the Court should acknowledge this normative disagreement, rather than trying to hide behind incorrect empirical claims.

3. Denying Lower Courts an Important Benchmark.

The Court's decisions are supposed to give guidance to government officials and lower courts. Subjective standards that turn on differences in degree, such as a "pretty much equally effective alternative" standard, provide little constraint on their face.

To make such a standard useful, the Court must use each case to set a benchmark. Where "the content of the rule is not revealed simply by its literal text," it must be "given meaning through the evolutionary process of common law adjudication"; 61 this requires the Court to confront the facts in each case, and explain why any particular alternative is or is not "pretty much equally effective."

Consider the position in which Congress now finds itself. ACLU suggested that the CDA's goals might be accomplished by a compulsory tagging scheme, which "requir[es] that indecent material be `tagged´ in a way that facilitates parental control of material coming into their homes." 62 Some legislators have been considering such an alternative. 63 Should the legislators (and eventually the President and the courts) conclude that the proposal is constitutional?

The Court suggested that compulsory tagging might be a "less restrictive alternative" to the CDA, but didn't say it was a constitutional alternative. It is, after all, either a content-based speech restriction (you may not post material that's indecent but not tagged), compelled speech (you must tag), or a content-based burden on speech (indecent material is specially burdened by the tagging requirement). Under any of these views, the tagging requirement would be subject to strict scrutiny, 64 and would itself be unconstitutional if there's a still less restrictive alternative that would be pretty much equally effective.

The clearest candidate for a still less restrictive alternative is a pure filtering scheme: Parents would use filters on their computers -- assume again that the government would provide them for free, or require service providers (such as America Online) to do so -- and the filter manufacturers would monitor the Net, find indecent material, and program the filters to block access to this matter. Filters can indeed screen out a great deal of indecent material; but filters are less effective than filters coupled with compulsory tagging, because they rely exclusively on the filter manufacturers' thoroughness. Even the best filter producer can't check every Web page as soon as it's put up, and of course the producers can't screen individual newsgroup posts, chatroom conversations, and discussion list messages. 65

So, given this, are filters a "pretty much equally effective alternative"? If ACLU had actually adopted this test, the Court would have had to describe the difference in effectiveness between the CDA and compulsory tagging (and the other alternatives), and to explain why this difference was so small as to be constitutionally insignificant. Then a legislator, President, or judge assessing the constitutionality of a compulsory tagging scheme could have used ACLU as a benchmark for determining whether the difference between tagging and pure filtering was likewise constitutionally insignificant.

But ACLU never acknowledged that the alternatives involved any sacrifice of effectiveness, so the Court didn't have to explain how much sacrifice of effectiveness was acceptable and how much would be too much. As a benchmark for judgments of "pretty much equal effectiveness," ACLU is largely useless. If "pretty much equal effectiveness" is the right test, then ACLU is an odd precedent for it: It neither properly sets forth the test, nor applies it in a way that provides a comparison point for future cases.

4. Neglecting to Show That the Alternatives Really Are Pretty Much as Effective.

So far, I've argued that it was a bad idea for the Court to speak of "equally effective" if it meant "pretty much equally effective." But is the ACLU result right even under a "pretty much equally effective alternative" test?

The most credible of the alternatives that the Court suggested is tagging. If indecent material is tagged, then a computer with properly functioning filter software can block the tagged material. From the computer user's perspective, the material will be as inaccessible as if it had never been posted online. And compliance with the tagging requirement should, if anything, be at least as high as compliance with the CDA.

But, as Part II.A.1 shows, tagging doesn't stop children from seeing indecent material at the homes of friends whose parents don't use shielding software. So long as there's one such child in a social circle, the others can use his computer, or even have him e-mail them the indecent material. Of course, if he does the latter, he might be violating the law (assuming the compulsory tagging statute applies to private forwarders), but that's not going to much deter many children.

This is a pretty big loophole, considerably bigger than the one found tolerable in the dial-a-porn case (Sable). The proposed less restrictive alternatives in Sable -- such as a credit card requirement or a requirement that a householder specifically ask the phone company to enable area code 900 phone calls 66 -- would have been quite effective. Though children might get their hands on their parents' credit cards, or use a phone at the home of someone who has enabled 900 calls, they would be deterred by the fact that the parents or the phone subscriber would see the unauthorized charge and hold the child responsible for it. If parents did let their child have access, or if the child had his own credit card, he'd still be less likely to invite other children to use the card or the phone, because this would cost more money. A child who gets access to dial-a-porn at least can't forward it to many friends at the touch of a button. 67 Even if the alternatives in Sable were "extremely effective [so that] only a few of the most enterprising and disobedient young people would manage to secure access to such messages," 68 the same is harder to say about the alternatives to the CDA.

So tagging would probably provide considerably less shielding of children than the CDA would. How much less is impossible to measure, but there's reason to think it would be quite a bit less. According to some estimates, over 6 million children have access to the Internet today. 69 If even only half the children's parents would like them shielded, and if even only 20% of those will get access to indecent material under tagging but not under the CDA, the result is 600,000 children whose parents want shielding but who remain unshielded. It's not clear how, given this, compulsory tagging would be "pretty much equally effective." 70

5. Failing to Explain Butler v Michigan.

The final flaw with a "pretty much equally effective" standard is that it's inconsistent with the Court's first, and in some ways most important, ruling in this area: The 1957 holding in Butler v Michigan. 71

Butler struck down a law that banned all distribution, to anyone, in any medium, of material deemed unsuitable for minors: The government, the Court held, may not "reduce the adult population to reading only what is fit for children." 72 I think this is right, for the reasons given by the Court in the various cases holding that sexually themed material -- outside the narrow category of obscenity -- is entitled to constitutional protection.

But under a "pretty much equally effective" alternative standard -- "the Government may . . . regulate the content of constitutionally protected speech in order to promote a compelling interest [in shielding children] if it chooses the least restrictive means" 73 that are still "almost equally effective" 74 -- the law in Butler would have been valid. It serves the same interest that the Court seemed to accept in ACLU, and did accept in Sable: shielding children from the supposedly harmful effects of indecent material. And no less restrictive alternative is even close to equally effective.

The obvious alternative to the total ban is a ban on distribution of the material to children, 75 but of course that's much less effective than a total prohibition. Once the material is allowed to adults, some of it will inevitably fall into the hands of minors. An underage teenager might stumble across material owned by an adult friend, or by a friend's parents; adults might easily sell it or give it to the teenager in private transactions that will be hard to discover.

If the material were totally banned, the government could seize it as contraband at the border, at the manufacturer, or at any place in the chain of distribution -- a powerful tool for ensuring that the material will be kept away from minors. But so long as the material may lawfully be distributed to adults, it can be intercepted only when the government has concrete evidence that it's being distributed to children, evidence the government will only have for a tiny fraction of those materials that are actually falling into children's hands.

Allowing the distribution of indecent material to adults thus requires the government to make a significant sacrifice of its supposedly compelling interest in shielding children. Butler correctly concluded that it's better to pay this price than to pay the price of depriving adults of all access to indecent speech. But this is a judgment that the "pretty much equally effective alternatives" test doesn't accommodate. 76

ACLU at times seemed to acknowledge Butler. "[T]he Government," the Court said (indirectly quoting Butler), "may not `reduc[e] the adult population . . . to . . . only what is fit for children.´" "`Regardless of the strength of the government's interest´ in protecting children, `[t]he level of discourse reaching a mailbox simply cannot be limited to that which would be suitable for a sandbox.´" 77

But given this occasional acknowledgment, why the assertion in Sable that the government may indeed "regulate the content of constitutionally protected speech in order to promote a compelling interest [in shielding children] if it chooses the least restrictive means to further the . . . interest"? 78 Why the seeming pregnant negative in ACLU that the absence of equally effective alternatives would make the speech restriction valid?

IV. The "Balancing" Metaphor and Five Concrete Doctrinal Options

A. The "Balancing" Metaphor

As I mentioned above, in most free speech controversies the question is about trade-offs. How much free speech should we be willing to sacrifice in order to shield children, or to achieve any other government interest? Conversely, how much shielding of children -- or how much of any other important value -- must we sacrifice in order to protect free speech? Behind every framework for scrutiny of speech regulations lurks a judgment about the trade-offs that must be paid.

One cliché response is that the Court must reach this judgment by "balancing," and in a certain (largely tautological) sense this is true: The judgment by definition requires deciding when the sacrifice of one value is "weightier" than the sacrifice of another. In a perfect world, we would "weigh" the value that would be lost by the burden on speech against the value that would be lost by the burden on the competing government interest. 79

But this sort of "balancing" is not an answer; it's just a way of reframing the question. Balancing sounds manageable because the metaphor conjures up a familiar real-life device: a balance scale used for weighing two physical objects. The balance scale, though, works only because it uses a reliable physical process that unerringly compares a single, easily commensurable, attribute of two items. 80 No physical device can tell us whether some lump of government interest "weighs" more -- is of greater "constitutional gravity" -- than some chunk of free speech right. The statement "courts should balance" thus simply invites the question "How?"

Referring the matter to one's unarticulated intuitions -- the scale in each judge's conscience -- is, even if legitimate, simply impractical. The Supreme Court makes law for tens of thousands of legislative, executive, and judicial decision makers in American government. "Trust your instincts" is not a useful legal rule in such a system, especially (as the Court has repeatedly held) in free speech cases. 81

Because of this, the Court has done its free speech "balancing" by creating relatively concrete doctrinal structures that help decide when a competing government interest "outweighs" the free speech value. Thus, for instance, under Brandenburg v Ohio, the need to prevent violence "outweighs" the right to advocate violence only when the speech is intended to lead to imminent injury, and is in fact likely to do so; otherwise, the speech right "outweighs" the government interest. 82 The Court had to reconcile -- one might say balance -- two competing concerns, but in doing so it produced a rule that says more than simply "Balance!"

The Court's task in child-shielding cases should likewise be to set forth a rule that reconciles the competing concerns in a way that other decision makers can adequately implement -- to draw a more concretely applicable line that will still generally produce the right results. 83 There are five basic approaches to this problem.

B. Five Concrete Doctrinal Options

1. The Current Official Approach: "Compelling Interest Trumps"

The official strict scrutiny approach, set forth in Sable, is that the First Amendment tolerates the sacrifice of as much speech as it takes to shield children from indecency. True, the sacrifice must be genuinely "necessary" -- there must be no alternatives that exact a lower free speech price without significantly sacrificing the government interest. But if the only way to really satisfy the interest is to restrict speech, then the government may do it. 84

I call this the "compelling interest trumps" approach, 85 and as I argued above in Part III.B.5, I believe it is unsound. As Butler correctly holds, the government may not reduce adults to reading only what is fit for children. This is true even though letting adults access indecent material would necessarily sacrifice a great deal of shielding of children -- even though a total ban would genuinely be the only means to effectively further the interest. For Butler and the result in ACLU to be right, the First Amendment must sometimes demand significant sacrifice of the government interest. But the "compelling interest trumps" approach does not reflect this principle.

It may seem odd to characterize this approach as government-friendly, because the conventional wisdom is that strict scrutiny in free speech cases is "fatal in fact." 86 But the test says that when there is no other comparably effective way of serving a compelling interest, the speech restriction should be upheld. 87 And the test includes no inquiry into the magnitude of the burden on speech; even serious burdens might thus pass strict scrutiny. 88

And the Court has used strict scrutiny to uphold speech restrictions. In Buckley v Valeo, the Court upheld a restriction on campaign contributions, and a ban on speech (and not just contributions) that is coordinated with a candidate, advocates a candidate's election, and costs more than $1,000. 89 In Austin v Michigan Chamber of Commerce, the Court upheld a ban on speech by corporations in support or opposition to candidates. 90 A plurality in Burson v Freeman upheld a ban on campaign-related speech within 100 feet of a polling place. 91 In Riley v National Federation for the Blind, the Court said that a requirement that charity fundraisers make certain statements passed strict scrutiny. 92 And though the Court in Sable struck down a dial-a-porn ban, it did so on the grounds that other (presumably permissible) speech restrictions would be less burdensome. 93

Lower courts have followed suit. Two courts of appeals upheld the post-Sable dial-a-porn restrictions. 94 The Washington Supreme Court upheld an injunction banning "the use of the words `murder,´ `kill,´ and their derivatives" in abortion clinic picketing, on the theory that the ban was narrowly tailored to the "compelling State interest in preventing the [physical, emotional and psychological harm arising] when such words are heard by children." 95 Other cases have used strict scrutiny to uphold speech compulsions aimed at better informing would-be charitable contributors; 96 bans on anonymous political speech, aimed at better informing voters; 97 bans on all political speech within 600 feet of polls, including speech not related to items on the ballot; 98 restrictions on speech by judicial candidates, aimed at "protecting public confidence in the integrity of the judiciary"; 99 and bans on public display of "obscene-as-to-minors" material. 100 Whether these decisions are right or wrong, they suggest that courts are willing to apply strict scrutiny in a way that upholds many restrictions.

As I have argued in another article, strict scrutiny still seems demanding largely because courts are often correctly unwilling to live by it. 101 ACLU is good evidence for that proposition.

2. The "Substantial Burden Is Unconstitutional" Approach

What might substitute for the "compelling interest trumps" framework? One option is what I call the "substantial burden is unconstitutional" approach: If the law imposes a substantial burden on generally protected speech, then it is per se impermissible, even if this means we must sacrifice a significant amount of shielding of children.

The underlying principle is that the freedom of speech need not be just a presumption that can be rebutted by strong enough claims of countervailing government interest. The First Amendment may instead be seen as embodying a judgment that some speech must be protected even if it unavoidably causes harm. 102

Under this view, the interest in shielding children might be seen as strong enough to justify modest restrictions on speech that's unsuitable for minors; 103 not every restriction need be considered an unconstitutional abridgement. Even relatively slight content-based restrictions are usually presumptively unconstitutional, 104 but such small prices may be worth paying when the strong government interest in shielding children is implicated. (As I discuss below, even Ginsberg v New York's ban on distribution to children of obscene-as-to-minors material imposed some burden on protected speech among adults.) But if the proposed restriction imposes a substantial burden on the free speech right, then under this framework it cannot stand.

Of course, this view rests on a contested assumption -- that avoiding substantial burdens on sexually themed or profane speech among adults is more important than shielding minors from such speech -- but this is an assumption that the First Amendment itself supports. The First Amendment protects speech even when it causes significant harm, perhaps including eventual violence. The theory of the First Amendment is that restricting speech is in the long run presumptively more harmful than permitting the speech. 105 And all the speech that we're discussing here is not obscene as to adults, and is thus constitutionally valuable; as ACLU points out, some of it may well be related to significant political, artistic, and scientific matters. 106

First Amendment doctrine does sometimes allow even substantial burdens on constitutionally valuable speech that is seen as too immediately harmful to tolerate. For instance, the government may sometimes ban newspapers from publishing certain military secrets, even if the publication is valuable to national debate; 107 the government may in narrow circumstances ban advocacy of violent conduct, even though the advocacy might contribute to political discussion. 108 The risk of imminent violence or death will always weigh heavily, even against the strongest of constitutional guarantees. 109

But these provisions are very much exceptions (and narrow ones at that), and should require extremely powerful justifications, more powerful than the justifications needed for laws that don't substantially burden valuable speech. 110 They should not be lightly extended absent strong evidence of dramatic, imminent harm. While there is a broad intuitive consensus that sexually themed or profane speech may indeed be in some measure unsuitable for minors, such a general sense of long-term corruption strikes me as inadequate justification for allowing substantial burdens on valuable speech, especially when this supposed corruption is compared to the dangers that the First Amendment routinely demands that we run. 111

Substantial burden tests are well known in constitutional law. In Planned Parenthood v Casey, for instance, the Court concluded that the government has a strong interest in protecting potential human life, but the woman's right to a pre-viability abortion must nonetheless prevail: Any substantial burdens on the right are per se invalid, though lesser burdens are presumptively constitutional. 112 Other doctrinal frameworks, such as those for religious freedom before Employment Division v Smith, freedom of expressive association, and the right to marry also include a substantial burden inquiry, though they at least theoretically allow even substantial burdens on the right so long as the law passes some form of heightened scrutiny. 113

Even some of the Court's free speech tests fit well with a "substantial burden" framework, though they generally aren't explained this way. Libel law, for instance, rests on the notion that the important interest in protecting reputation justifies some speech restrictions, so long as they do not substantially restrain valuable speech. The Court has concluded that false statements of fact are generally constitutionally valueless, 114 so the impact of libel law on false statements is seen as a constitutionally insignificant burden; but the Court has recognized that libel rules can incidentally burden constitutionally valuable truthful speech, and has therefore often inquired into the magnitude of the burden created by various rules. 115 The "compelling interest trumps" approach is not used here: For instance, the actual malice test is a considerably less effective alternative to traditional strict liability when it comes to protecting public figures' reputations from false defamatory statements, but the Court has nonetheless rejected strict liability (and even a negligence test) because it would impose too great a burden on speech. 116

To make this approach work for restrictions on material that's supposedly unsuitable for minors, the Court would have to do two things: (a) Identify what speech lacks substantial constitutional value when communicated to minors. Even a total ban on distributing such speech to minors would thus not create a substantial burden, so long as it's really limited to distribution to minors. (b) Determine when a burden on valuable speech (to adults or minors) qualifies as insubstantial. 117

a) Identifying Speech That Lacks Value When Communicated to Minors

i) Indecent versus "obscene as to minors." As of 1975, the Court had concluded that (1) there is no constitutional value in communications to minors (except perhaps by or with the approval of their parents) that are "obscene as to minors" -- essentially communications that fit the three-prong obscenity test, but with a "for minors" qualifier on each prong 118 -- but (2) there is value in communications to minors that fall outside this category. 119 This makes sense under the logic of obscenity law. Assume communicating obscenity to another person is generally constitutionally worthless as speech, because it appeals to prurient interests, is patently offensive, and lacks serious value. 120 Then communicating to a minor material that appeals to the minor's prurient interests, is patently offensive when distributed to minors, and lacks serious value for that minor would likewise be generally worthless. Even a total ban on such communications to people whom the speaker knows to be minors wouldn't substantially burden worthwhile speech, because it would ban only the speech that is, by assumption, valueless. 121

In FCC v Pacifica Foundation, however, the Court went beyond this, holding that government may regulate speech even to shield minors from material that is merely indecent -- that "refer[s] to excretory or sexual activities or organs" in a "patently offensive" way. 122 This includes material that doesn't appeal to the prurient interest, and may even cover material that has significant artistic, literary, or scientific value, even for minors. Sable seemed to echo this, acknowledging that the government has a compelling interest in prohibiting "indecent" telephone communications to minors, or at least those that are sexually themed. 123 ACLU seemed to agree, though it was ambiguous on the point. 124

This leaves much uncertain. Could the government, for instance, prohibit all indecent communications to minors, such as selling or even giving a 16-year-old a copy of Carlin's Seven Dirty Words? Sable and Pacifica suggest this might be constitutional, 125 while Ginsberg and Erznoznik suggest otherwise. Likewise, even if tagging schemes are generally constitutional, it's not clear whether Congress could constitutionally require tagging of indecent material or only of obscene-as-to-minors material. At some point, the Court will have to decide on the right standard.

"Obscene as to minors" seems to be the better solution. By definition, material that's "indecent" but not "obscene as to minors" either has serious artistic, literary, political, or scientific value for minors, or does not appeal to minors' prurient interests. If it has value for minors, then it ought to be protected. If it doesn't appeal to minors' prurient interests, then it's not clear why it should be treated differently than equally nonprurient material that deals with offensive topics other than sex or excretion.

More concretely, as ACLU pointed out, a lot of speech that might be considered "patently offensive" -- "serious discussion about birth control practices, homosexuality, the First Amendment issues raised by [indecency itself], or the consequences of prison rape" 126 -- does have considerable value. The one item the Court has held "indecent," the Carlin "Seven Dirty Words" monologue, is no less valuable than most humorous social commentary; 127 and it's not clear why the government has a compelling interest in shielding children from it.

But even if one disagrees with my judgment, and concludes that the line should be drawn at indecency, my basic point remains: The Court must identify which speech is valueless as to minors, and use this definition in determining whether or not a restriction substantially burdens valuable speech.

ii) Rights of parents. The Court must also decide whether parents have a constitutional right to communicate -- and to authorize others to communicate -- "obscene-as-to-minors" material to their children. Ginsberg suggested as much, 128 as did ACLU. 129

Doctrinally, this is not a trivial question; Ginsberg also suggested that the government has an interest in shielding children against the harmful effects of certain material, independent of the interest in protecting parental decisions about child-rearing. 130 If "obscene-as-to-minors" speech is indeed valueless to minors, one can argue that it remains valueless when it is communicated or tolerated by parents, and that parents should have no more right to expose their children to such speech than they would have with regard to simple obscenity.

Nonetheless, it is probably sounder to leave parents, rather than the government, with the ultimate decision here. Children of the same age vary widely in maturity, and parents usually know their child's maturity better than do prosecutors, judges, or juries. If parents believe there's educational value in giving their children access to supposedly "obscene-as-to-minors" material, there's good reason to defer to that judgment. And the notion that parents should, absent some powerful reason to the contrary, have discretion about how to raise their children buttresses this view. 131

Of course, this deference to parents' views is only a presumption: Parents generally can't, for instance, make it legal for their 12-year-olds to have sex or to drop out of school, even if the parents think the children are mature enough to do so. But though exposure of children to "obscene-as-to-minors" material may be harmful, the harm seems considerably smaller than the possible harms of early sex or of lack of education. This relatively modest harm ought not be enough to rebut our normal deference to parental decisions, especially when free speech rights are also implicated. 132

b) Identifying Insubstantial Burdens on Valuable Communications

Butler holds that banning all distribution of materials that are unsuitable for children is too great a burden on the free speech rights of adults, and surely this is correct. Though anything less than a total ban will increase the chances that some of the materials will leak out to children, that's a price the First Amendment requires us to pay.

But other burdens may not be so troublesome. Consider the law upheld in Ginsberg; though the law purports to ban only "knowing[]" sales to minors, it does in fact burden sales to at least some adults. Because the law essentially requires sellers to make "a reasonable bona fide attempt to ascertain the true age of such minor," 133 many sellers are reluctant to sell to adults who look underage and who don't have proof of age handy. Those adults are therefore burdened in their access to constitutionally protected material.

Moreover, the law burdens the right of underage-looking adults to buy anonymously, because proof of age generally includes a person's name. The law also burdens sellers, who must take the time and effort to check the age of the young-looking adults; and it burdens publishers, who lose some adult potential readers as a result.

Finally, the Ginsberg statute burdens the rights of minors to receive material that's protected as to them, and the rights of writers, publishers, and sellers to communicate such material to them. The "obscene-as-to-minors" standard is vague, and as Reno v ACLU points out, even laws that are not vague enough to be unconstitutional tend to make speakers cautious. 134 A seller might well be reluctant to sell any sexually oriented material to a minor, even if a jury would ultimately conclude that this material wasn't "obscene as to minors." Minors might therefore find it hard to get material that they're theoretically entitled to read (though determined minors will probably find some store that will sell it to them). 135

I don't want to overstate these burdens: They do seem relatively slight. In a sense they may be seen as incidental to a ban on only supposedly unprotected speech (distribution to minors of obscene-as-to-minors material). But the law explicitly applies whenever the distributor "[has] reason to know . . . or [has] a belief or ground for belief which warrants further inspection or inquiry of . . . the character and content of [the] material [and] the age of the minor," 136 thus effectively requiring the distributor to investigate all borderline cases, including those where the material is ultimately found to be protected or the customer is found to be an adult. This is not merely a burden imposed by a general law that is not focused on speech (such as tax law or contract law); 137 the burden applies to a content-based category of speech, and necessarily interferes in some measure with the distribution of the speech to people who have a right to receive it.

Ginsberg thus shows that, while the most serious burdens on protected speech -- such as a total ban -- are unconstitutional, slighter burdens may not be. 138 We might hesitate to ask courts to draw lines between slight burdens and substantial burdens, and I believe the Court is generally correct in avoiding such line-drawing for most content-based speech restrictions. 139 But so long as we are committed to allowing some restrictions on speech that's unsuitable for children, some such line-drawing is inevitable there.

Most hard cases involve burdens that are in between Ginsberg and Butler. In Pacifica, for instance, adult listeners were deprived of the opportunity to hear material for free during the day and evening, 140 and speakers were deprived of the opportunity to reach the normal radio audience. In ACLU, adult users were deprived of the opportunity to see material for free in cyberspace, and speakers were deprived of the opportunity to reach cyberspace users who didn't want to pay for the speech. In the compelled rating scheme suggested by ACLU, adult users would not be appreciably burdened, but speakers would have to take the time and effort to rate their material, and might suffer a hard-to-measure loss to the effectiveness of their message caused by the compelled statement. 141

The Court's evaluation of these burdens must turn on its judgment as to (1) the value of the burdened speech, and (2) the significance of the interference with the size of the speaker's audience and with the adult listeners' ease of accessing the material. 142 It would be best if the Court could articulate some rules about this, but even if it can only make ad-hoc decisions, these decisions can be used as benchmarks by other decision makers, so long as the facts in each case are candidly described.

I would suggest that any restriction that substantially reduces a speaker's audience -- as the rules in Pacifica and ACLU would have -- or that makes it substantially harder for adults to get the material they want should be seen as an impermissible substantial burden. Justice Powell was right that the FCC's action in Pacifica "[did] not prevent willing adults from purchasing Carlin's record [or] from attending his performances." 143 Likewise, the CDA would have allowed indecent speech so long as it was posted on sites that charged users through credit cards. Nonetheless, the ACLU Court was correct to stress the breadth of the restriction and to suggest that it "threaten[ed] to torch a large segment of the Internet community." 144 The CDA -- like the regulation in Pacifica -- substantially burdened listeners by making it considerably more expensive (and, in Pacifica, more time consuming) for them to get this information. As importantly, the laws substantially burdened speakers by requiring them to sell their speech rather than distributing it for free, thus dramatically reducing the size of their audience. 145

Of course, others may want to draw the substantiality threshold higher. One court, for instance, has borrowed the "adequate alternative channels" test from the framework used for content-neutral time, place, or manner restrictions. 146 Under this view, burdens that leave open adequate alternative channels are deemed insubstantial, and therefore permissible. I don't agree: The adequate alternative channels test, at least as used by the Court, is generally quite tolerant of government regulation, even when the regulations significantly reduce the size of the audience and make it considerably harder for potential viewers to get the speech. This has been criticized even for content-neutral regulations, 147 but given the special concerns involved with content-based regulations, 148 it seems particularly inappropriate here.

In any event, though, even if some would draw the "substantial burden" line in a slightly different place than I would, I hope I have shown that the "substantial burden is unconstitutional" framework is generally the sounder one. Even when some restrictions on a certain category of speech (here speech that's unsuitable for minors) are allowed, the Court should recognize that some burdens are per se unconstitutional, even if they are genuinely necessary to maximally serve the compelling government interest.

c) Results the Rule Would Likely Yield

The "substantial burden is unconstitutional" framework would produce the following results:

i) Butler. The ban on all distribution of material that is unsuitable for minors would be unconstitutional, for a reason compatible with the Court's: The restriction greatly burdens constitutionally protected speech to adults.

ii) Ginsberg. The ban on sale of "obscene-as-to-minors" material to people the seller knows to be minors would be constitutional, for a reason again compatible with the Court's: The restriction bans only constitutionally valueless speech to minors, and imposes only a slight burden on valuable communications.

iii) Pacifica. The restriction on broadcasting "indecent" material would be unconstitutional, even if it is necessary to most fully serve the interest in shielding children, because it substantially burdens communications to adults. 149 And if profane speech has constitutional value even when the listener is a minor, then the restriction would also create a substantial burden on communications to children.

iv) Sable. The dial-a-porn ban would be unconstitutional, not because there are less restrictive but pretty much equally effective alternatives, but because the ban substantially burdens communications to adults.

v) Post-Sable dial-a-porn cases. 150 After Sable, Congress passed the Helms Amendment, which, as implemented by the FCC, prohibits access to dial-a-porn unless (a) the phone service subscriber specifically tells the phone company to allow area code 900 calls on his line, or (b) the dial-a-porn service asks for a credit card number. Such a law would be upheld -- as it in fact was by the courts of appeals, though on least restrictive alternative grounds -- because it doesn't substantially burden speakers or listeners. It does prevent dial-a-porn calls by people who don't have their own phone line and who don't have credit cards, but without a personal phone bill or a credit card, one generally lacks the mechanism to pay for the calls anyway. The only appreciable burden is a slight delay for those who don't have credit cards, and who thus have to wait until the phone company unblocks their phone lines; but dial-a-porn seems not to be a particularly time-sensitive medium, so this delay is not a substantial burden.

vi) ACLU. The CDA would be unconstitutional because -- as I discuss above -- banning not-for-pay communications substantially burdens speech to adults, even if the material is still available for sale. 151

vii) Hypothetical tagging requirement for "obscene-as-to-minors" material. Such a requirement would be constitutional because it does not impose a substantial burden. Its chief burdens are that it (a) requires people who put potentially obscene-as-to-minors material online to take the time and effort to rate it; (b) tempts people who put on borderline material to err on the safe side and rate it "obscene as to minors" even if it probably doesn't fit within the category; (c) risks depriving minors of such misrated-for-safety's-sake material; and (d) in some measure alters the speaker's message by compelling them to include the rating. The first three burdens, though, are no greater than what the Court upheld in Ginsberg, and they seem modest though not negligible. The fourth burden is harder to evaluate, but strikes me as likewise modest. 152

viii) Hypothetical tagging requirement for "indecent" material. Such a requirement would be unconstitutional if the government's compelling interest extends only to shielding children from "obscene-as-to-minors" matter. Even slight content-based burdens, such as a tagging requirement, should be unconstitutional in the absence of a compelling interest. 153 If the compelling interest extends only to obscene-as-to-minors matter, then compelled ratings may be applied only to such material, not to indecent speech, violent speech, or other categories.

ix) Ban on public display and unattended sales of "obscene-as-to-minors" material. Many states bar public display and unattended sales, in places where minors might be present, of "obscene-as-to-minors" materials. 154 These restrictions have generally been upheld, 155 but they would probably be unconstitutional under the "substantial burden" analysis. Bans on public display -- for instance, on murals, paintings hanging in restaurants, and the like -- do substantially burden the displayer's ability to communicate to the public. Bans on unattended sales (or unattended browsing) are a lesser burden, but probably still a substantial one, because forcing the publications out of a newsrack and into a bookstore or an attended newsstand will probably substantially decrease their audience.

As the above shows, applying the "substantial burden is unconstitutional" approach would involve some hard calls. Setting the "substantiality" threshold higher might uphold bans on public display of "obscene-as-to-minors" matter, and perhaps even the Pacifica regulation. Setting it lower might invalidate the hypothetical "obscene-as-to-minors" tagging requirement, and perhaps even the Helms Amendment and the law upheld in Ginsberg. Nonetheless, the rule creates a zone of fairly easy cases, 156 and in the hard cases focuses decision makers on the important question.

3. The "Substantial Burden/Less Restrictive Alternative Hybrid" Approach

The "substantial burden is unconstitutional" test would generally uphold any restriction on material unsuitable for minors (whether that is defined as indecent material or obscene-as-to-minors material) if the restriction imposes an insubstantial burden on free speech. 157 One might, however, argue that even minor burdens should be unconstitutional if they provide only slight marginal gains to the government interest -- if there's a less restrictive but pretty much equally effective alternative. Even a small speech cost, the argument would go, can be justified only by a significant shielding benefit.

But whatever the merits of inquiries into "pretty much equally effective alternatives" in other strict scrutiny contexts, I doubt that such an inquiry would be particularly helpful here. Less restrictive alternatives will rarely be clearly pretty much as effective as more restrictive ones. This is particularly evident in ACLU, where few alternatives would work nearly as well as a total ban on free online material. Likewise, the alternative given in Butler to a total ban on the distribution of explicit material -- a ban on distributing such material to children -- is much less effective than the total ban. The alternatives to bans on broadcast indecency, such as restrictions on day and evening broadcasts, are likewise considerably less effective. 158

These alternatives are less restrictive than a total ban because, unlike the ban, they aim to segregate two categories of people: (1) adults (and minors whose parents choose not to shield them), who should have access to the materials, and (2) all other minors, who shouldn't have access. But any such attempted segregation will also be less effective, because it will necessarily lump some minors from the second category together with the people in the first. Compelled tagging of online material (as opposed to a ban on free online indecency) protects the access rights of people who don't have filters running on their computer, but this class of people will necessarily include some minors who are visiting friends whose computers are unshielded. Channeling broadcast indecency to 10 p.m. through 6 a.m. (as opposed to totally banning it) protects the access rights of adults who are awake between 10 and 6, but also gives access to some minors who are up during those times. Allowing the distribution of explicit materials to adults (as opposed to totally banning such distribution) protects the access rights of adults, but also gives access to some minors who get the material from those adults.

Thus, while I think the substantial burden/less restrictive alternative hybrid approach is plausible, I suspect it would usually (if properly applied) reach the same results as the "substantial burden is unconstitutional" framework: Less restrictive means will almost always be less effective. I also suspect that the extra inquiry into less restrictive alternatives would be complicated and error prone, tending to lead courts into the same kinds of empirical mistakes that the ACLU Court fell into. These two points, I think, counsel in favor of the simpler "substantial burden is unconstitutional" framework. On the other hand, because the hybrid approach might at least theoretically protect speech more than a pure substantial burden framework would -- without unduly sacrificing the government's ability to shield children -- many free speech maximalists may support it.

4. The "Cost-Benefit Weighing" Approach

As I mentioned above, in theory the most appealing approach may be a sliding-scale one: If there is value in shielding children, then as a restriction provides more shielding (as compared to the next best alternative), we should be willing to tolerate a greater burden on the countervailing value of free speech. Thus, consider two examples:

A ban on unattended coin-operated newsracks containing obscene-as-to-minors (not just indecent) material. Striking down the ban would allow minors easy access to some hardcore matter. Upholding the ban would substantially interfere with distribution of the material to adults: Many publications are sold through newsracks precisely because newsracks make it easier for people to buy the publications, and thus substantially increase the publication's audience; requiring the material to be sold through attended newsstands would probably substantially decrease the audience. Nonetheless, while the burden is substantial, it is not a total ban, and perhaps is less substantial than the burden imposed by the CDA. 159

Time restrictions on broadcast indecency. Here, at least before V-chip-like technology became available, there really were no less restrictive but pretty much equally effective alternatives to a total ban. 160 Striking down the ban would expose minors to whatever indecency people broadcast. Upholding the ban would substantially burden free speech, though not as much as the law in Butler, which applied to all media. Following some hints given by Pacifica, Congress and the FCC limited the indecency ban to 6 a.m. to 10 p.m., 161 which made it less burdensome but also less effective. Any even less burdensome restrictions would be less effective still.

My sense is that many people who generally favor strong free speech protections nonetheless support the newsrack ban and perhaps even some broadcast time restrictions: 162 The extra shielding of children seems enough to "outweigh" the substantial, but not very great, burden on free speech. Some lower court decisions dealing with newsrack restrictions, general restrictions on public display of obscene-as-to-minors material, and even the CDA itself in fact use something like this weighing approach. 163 Justice Scalia's Sable concurrence likewise argues that "the more pornographic" -- and thus presumably the more harmful -- the material, "the more reasonable it becomes to insist upon greater assurance of insulation from minors." 164

One might therefore try to directly implement a "cost-benefit weighing" approach, in which the government decision maker (legislator, executive official, or judge) "weighs" the magnitude of the burden on speech against the magnitude of the extra shielding that the law would provide. Thus, under this test, the ACLU Court would have "weighed" the marginal speech burden imposed by the CDA (compared to that imposed by alternatives, such as a tagging requirement) -- which is quite high, since a ban is much more burdensome than tagging -- against the level of marginal shielding that the CDA provides relative to the alternatives, which the Court thought was minimal but which in fact was probably quite high.

But for the reasons I discussed above in Part IV.A, I don't believe this approach can be practically administered by the many judicial, legislative, and executive officials throughout the country. The "compelling interest trumps" and "substantial burden is unconstitutional" tests provide some definiteness because they involve the comparison of something, for instance, the degree of burden on speech, against a fixed benchmark that is similar in kind, such as a threshold burden: How large is the burden here relative to the burdens recognized as substantial or insubstantial by prior cases? Cost-benefit weighing, on the other hand, in each case demands a comparison of two things (the burden and the benefit) that are very different, a much harder proposition.

The one free speech doctrine that explicitly tries to use cost-benefit weighing -- the Pickering test for speech restrictions imposed by the government as employer -- supports this skepticism. Pickering specifically calls for sliding-scale "balanc[ing]" of "the interests of the [employee], as a citizen, in commenting upon matters of public concern" against "the interest of the State, as an employer, in promoting the efficiency of the public services it performs through its employees." 165 The Court has acknowledged that "such particularized balancing is difficult," 166 and this seems an understatement. From all I've seen of the lower court decisions, the test is essentially indeterminate in all but the easiest cases, cases that could well be resolved through more categorical rules. This indeterminacy is troubling even when the government is acting as employer, a context in which the Court properly accepts greater speech restrictions, but would be especially improper when evaluating laws imposed by the government acting as sovereign. 167

Cost-benefit weighing in the child shielding context would also work only if the Court does what it has so far avoided doing: if it actually determines the extent of the harm caused by indecent or "obscene-as-to-minors" speech -- rather than just the presence of some nontrivial harm -- and thus the true benefit of shielding. There is something of a consensus that some such benefit exists, which is probably enough to justify trying to realize this benefit through less than substantially burdensome restrictions. But once one must "weigh" even substantial free speech burdens against the purported benefits of shielding, one can't avoid having to determine just how serious these benefits are. The relatively uncontroversial claim that the benefits exist will not be very helpful here: One will need some agreement on at least the rough magnitude of the benefits, compared to the magnitude of the speech burden, and such an agreement may be hard to achieve. 168

The weighing approach seems theoretically most satisfying, because it promises a full recognition and comparison of both the costs to free speech of regulation and the costs to minors of adult freedom. But for the reasons given above, it would probably be practically intractable, especially given the need for the Court to provide meaningful constraints to other government actors.

5. The "Multiple Speech Subclasses" Approach

We can avoid some of the administrative problems of a sliding-scale approach by trying to approximate it through a sort of graduated step-ladder -- what I call the "multiple speech subclasses" framework, in which the Court sets up different levels of tolerable burden for different kinds of speech, based on perceived differences in harm. Applying this framework doesn't require direct comparison of burden and benefit, but only of burden against threshold burden.

For instance, the Court might say that merely indecent speech is fairly harmless, that any loss of shielding against indecent speech is thus of rather little consequence, and that any burden beyond the mildest should therefore be impermissible. This would be a "substantial burden is unconstitutional" approach, with a low threshold of substantiality. On the other hand, the Court might conclude that obscene-as-to-minors speech is more harmful (because it appeals to prurient interests), that the loss of shielding against that kind of speech is thus more important, and that even fairly substantial -- but not very great -- burdens on the speech should therefore be allowed. This would also be a "substantial burden is unconstitutional" approach, but with a higher substantiality threshold. 169

This framework is less attentive than pure cost-benefit weighing to fine differences in weights on both sides, and is thus both theoretically less perfect and practically more administrable. It would require only a few up-front decisions about the price we should be willing to pay for the freedom to communicate various kinds of speech, rather than a fresh weighing in each case. A good analogy is the libel law framework, which sets up different tests for private concern speech, private figure/public concern speech, and public figure/public concern speech, each tolerating a different level of incidental burden on constitutionally protected expression. 170

I suspect even this framework would probably be too complex to be workable; the "substantial burden is unconstitutional" approach discussed in Part IV.B.2 would be practically superior. Nonetheless, even some version of the "multiple speech subclasses" framework is better than the orthodox strict scrutiny "compelling interest trumps" test.

6. Summary

I have argued that, where speech that's supposedly unsuitable for children is concerned, the Court should focus primarily on the burden that the restriction imposes on valuable speech. The Court should draw benchmarks that distinguish permissible insubstantial burdens from impermissible substantial ones, benchmarks that government decision makers can consult in the future.

Laws that impose sufficiently substantial burdens, I've argued, are unconstitutional even if they're genuinely necessary to shield children; the "compelling interest trumps" framework is therefore unsound. And while it seems appealing to "weigh" the degree of the burden in each case against the benefit the law brings, such a "cost-benefit weighing" framework is practically unadministrable.

I've identified three alternatives to these frameworks: (a) "Substantial burden is unconstitutional," which strikes down all substantial burdens and allows all insubstantial ones. (b) "Substantial burden/less restrictive alternative hybrid," which strikes down all substantial burdens and those insubstantial burdens that are not genuinely necessary to shield children, and allows only those burdens that are both insubstantial and necessary. (c) "Multiple speech subclasses," which sets different levels of permissible burden for different subclasses of unsuitable-for-minors speech.

The "substantial burden is unconstitutional" framework is the simplest of the three, and I think the best. Though it's in some respects theoretically less satisfying than the others, because it seems to sometimes overprotect speech and sometimes underprotect it, it's probably practically more effective. "Less restrictive alternative" inquiries are, if done candidly, largely fruitless in this area; and defining one level of "unsuitability for minors" and one "substantiality" threshold seems hard enough that I doubt courts can effectively define multiple levels of unsuitability and multiple thresholds of substantiality.

Nonetheless, the choice between these three frameworks is a close call. They all recognize that some burdens are high enough to be impermissible even if the burdens are really necessary to shield children. They all accept, as the Court and much of the country seem to accept, that certain kinds of burdens on material that is unsuitable for minors are slight enough to be allowed. And because they require that burdens be compared against threshold burdens, rather than commanding an amorphous case-by-case weighing of burden on speech against increase of shielding, they are all relatively practically administrable. Each of them would have yielded a better -- more honest and more precedentially valuable -- opinion in ACLU than the Court produced using the orthodox "compelling interest trumps" framework.

V. Conclusion: Applications to Other Kinds of Speech Restrictions

This article has focused on one particular set of speech restrictions -- restrictions on material that's supposedly unsuitable for minors. I have argued that:

1. ACLU reached the right result but erred in its reasoning.

2. This error sheds light on a deeper problem: The Court's official test in this area -- strict scrutiny, with its "compelling interest trumps" approach -- is unsound, and if applied candidly would lead to wrong results.

3. Instead of compensating for the unsound test's errors by misapplying it, the Court should design a test that's better suited to the concerns raised by the interest in shielding children.

4. Directly "weighing" the burden of a restriction against its benefits is not feasible in a system in which the Court must create administrable rules for a myriad other government actors.

5. Rather, the best solution is one that tolerates certain less-than-substantial burdens on a narrow class of speech, but that categorically invalidates any burdens that are substantial.

Much of this, I believe, is also applicable to other free speech contexts. ACLU is just one of many cases in which, as I have argued in a recent article, 171 the Court has fudged the facts in order to reach the right results under its strict scrutiny framework. This felt need to fudge strongly suggests that the strict scrutiny framework is itself unsound in these cases: Speech must often be protected even though protecting it unavoidably causes substantial costs to compelling government interests -- even though there are no less restrictive but pretty much equally effective alternatives to speech suppression. 172 The line drawn by the "compelling interest trumps" test inadequately reconciles the values at stake in such cases.

I propose that instead the Court should return to the framework it generally followed as it was developing libel law, obscenity law, and various other major free speech doctrines:

1. A strong presumption that speech restrictions justified by the communicative impact of speech are unconstitutional. 173

2. A recognition that certain classes of restrictions are constitutional, either because they do not substantially burden valuable speech, or (in rare situations) because, despite their substantial burden on valuable speech, they are genuinely so crucial that we must sometimes substantially sacrifice speech in their name.

3. Development of individual bodies of law to deal with these classes of restrictions.

Thus, libel law evolved from a conclusion that some restrictions on false and defamatory speech would serve an important interest without substantially burdening valuable expression. 174 Obscenity law evolved from a (controversial) judgment that restrictions on some sexually explicit material would likewise not substantially burden valuable speech. Incitement law recognized that some burdens even on core political advocacy might be needed to preserve life and limb, though even that most important of interests must often yield to free speech rights. Similar rules should be developed for those areas (which, I hope, will be relatively few) where -- as with speech unsuitable for minors -- the Court believes some additional restrictions are proper.

I am not particularly happy about the Court fragmenting First Amendment doctrine this way, and there certainly is great appeal to a simple, consistent Grand Principle that explains all free speech law. Perhaps one day such a principle may evolve from the Court's experience with the individual subtests; 175 in particular, I think versions of the substantial burden approach outlined above have promise in some (though not all) areas. 176 But so far the Court has not found a principle that is both broad and accurate.

For now, I would be satisfied with the Court jettisoning its current, unsound, Grand Principle, and reawakening to the need to look for better (even if more particularistic) approaches; and there's reason to think the Court is indeed prepared to do so. Justice Stevens himself has often criticized the strict scrutiny framework. 177 Justice Kennedy has at times explicitly urged that it be rejected and replaced by a more categorical approach similar to the one described above. 178 A recent opinion by Justice Thomas, while formally using strict scrutiny, in effect urged a substantial-burden-like test, one that would reject certain laws even if they are necessary to accomplish certain interests that the Court has found compelling. 179

Believing as I do in rather strong free speech protection, I hope the Court will make the areas of permissible regulation as narrow as possible; and I recognize that discarding existing rules always poses a risk that the new rules will be worse. Nonetheless, the current situation is already not very good. The strict scrutiny test would on its face uphold certain restrictions that ought not be upheld; and though the Court has tended to avoid these results, it has done so only by creating precedents that, because of their lack of candor, make the law even murkier. Given this, there's much to be gained from rejecting strict scrutiny and confronting the important questions more openly and clearly.

1. See, for example, Sable Communications v FCC, 492 US 115 (1989); Reno v ACLU, 117 S Ct 2329 (1997); Ginsberg v New York, 390 US 629, 649-50 (1968) (Stewart concurring) (distinguishing children's rights to access speech from adults' rights); Paris Adult Theatre I v Slaton, 413 US 49, 113 (1973) (Brennan dissenting, joined by Stewart and Marshall) (arguing that obscenity is constitutionally protected "at least in the absence of distribution to juveniles"); Pope v Illinois, 481 US 497, 513, 517 (1987) (Stevens dissenting, joined by Brennan and Marshall) ("government may not constitutionally criminalize mere possession or sale of obscene literature, absent some connection to minors"); Sable Communications v FCC, 492 US 115, 134 (1989) (Brennan dissenting in part, joined by Marshall and Stevens) ("To be sure, the Government has a strong interest in protecting children against exposure to pornographic material that might be harmful to them."). But see Ginsberg, 390 US at 650 (Douglas, joined by Black, dissenting) (arguing against any such restrictions).

2. See cases cited in Part III.A.

3. 117 S Ct 2329 (1997).

4. For example: "Senator Patrick J. Leahy . . . who was an opponent of the legislation, said, `. . . This is a victory for the First Amendment.´" Supreme Court Strikes Down Communications Decency Act, Facts on File World News Digest 473 A1 (July 3, 1997). "Free speech scored an important victory last week, when the U.S. Supreme Court ruled that freedom of expression applies to the Internet." Editorial, Denver Post E4 (July 6, 1997). See also Michael Loftin, Victory for the First Amendment, Chattanooga Times A8 (July 14, 1997).

5. Pub L No 104-104, 110 Stat 56 (1996), codified at 47 USC § 223(d). Two other provisions barred communicating indecency "knowing that the recipient of the communication is under 18 years of age" and communicating indecency "to a specific person or persons under 18 years of age." 47 USC §§ 223(a)(1)(B), (d)(1)(A). If these provisions were interpreted narrowly, to cover only speech to a particular person whom the speaker knows to be a minor, they might well be constitutional, at least as applied to indecent speech that is also "harmful-to-minors"; Justice O'Connor so argued in her dissent. ACLU, 117 S Ct at 2354-55 (claiming that this was the best reading, and that even if the ban on indecent speech was overbroad in including more than just "harmful-to-minors" speech, it was not substantially overbroad); see Part IV.B.2.a.i (discussing distinction between "harmful-to-minors" and "indecent"). The majority struck down the provisions because it thought they could not be so narrowly read. Id at 2348-50.

6. See ACLU, 117 S Ct at 2345 ("assum[ing] arguendo" that "indecent" was synonymous with "patently offensive [depiction or description] of sexual or excretory activities or organs"); FCC v Pacifica Foundation, 438 US 726, 740-41 (1978) (accepting a similar FCC definition of "indecent").

7. Free sites can't practically use credit card numbers for verification because the verification costs money, apparently about $1 per transaction. ACLU v Reno, 929 F Supp 824, 846 finding of fact no. 99 (ED Pa 1996).

If there were a way in which a cyberspace speaker could, for free, check a would-be listener's age -- perhaps by checking some reliable "cyber ID" that the listener c