In his article, “A Consumer Decision-Making Theory of Trademark Law,”1 Professor Mark McKenna makes two significant claims. The first is that the dominant Law and Economics theory of trademark law—the search-costs theory of the Chicago School—is in some way connected to recent undesirable expansions of trademark rights. The second is that a preferable theory of trademark law—one that would result in more tightly circumscribed and socially beneficial notions of trademark rights—would take consumer decision making, rather than search costs, as its guiding principle. I find myself sympathetic to these arguments, and yet I believe they are subject to criticism on grounds that they violate Hume’s Law—that is, they confuse the descriptive with the normative, the is with the ought.2 In this response I will suggest that a careful untangling of the descriptive and normative elements of McKenna’s project, while perhaps undermining some of his claims, lends considerable support to his overall ambition of organizing trademark doctrine by reference to a theory other than search-costs theory.
Take, for example, McKenna’s first major argument—that search-costs theory is causally connected to undesirable expansions of trademark rights. McKenna claims that modern trademark doctrine rests on a fallacious syllogism akin to the one that led Ionesco’s logician to conclude that Socrates was a cat,3 or that might give my young son difficulty as he tries to understand the relationship between the category “rectangles” and the category “squares.” In classical formal terms, it is the illicit major premise fallacy in the argument that proceeds as follows:
1. All A are B.
2. All A are C.
3. Therefore all B are C.4
In McKenna’s telling, our thinking about trademark law relies on a fallacious argument of equivalent form:
1. All consumer deception raises search costs.
2. All consumer deception is undesirable.
3. Therefore everything that raises search costs is undesirable.
McKenna seems to accept both premises of this argument—he concedes that consumer deception raises “a certain kind of search costs,”5 and much of his article is devoted to defending the principle that consumer deception is undesirable.6 But of course, as a matter of logical form an argument that attempts to derive the conclusion “everything that raises search costs is undesirable” from these two premises is invalid.7 Not content to stop here, McKenna further argues that the conclusion “everything that raises search costs is undesirable” is also contradicted by empirical evidence. Specifically, he would like to persuade us that not all increases in search costs are undesirable, at least not to all people.8
This is all fine and good so far as it goes. But the next move in the argument invites more scrutiny. McKenna would like to convince us that all of the problematic innovations in trademark doctrine over the past half-century—innovations that have expanded the scope of trademark rights far beyond what is necessary to prevent consumer deception—are somehow causally connected to courts’ failure to reject this syllogism, and specifically their equation of search costs with confusion.9 In his words: “search costs theory is complicit in trademark law’s expansion to the extent it encourages courts to equate confusion with search costs and to ignore the relationship between search costs and consumer decision making.”10
I have three objections to this argument. The first is that I am not sure what causal mechanism or relationship McKenna intends to suggest with the word “complicit.” If it is merely that search-costs theory has failed to rein in or roll back the doctrinal developments of the past half-century, then we have our first violation of Hume’s Law. Such an accusation of complicity is less a descriptive claim about a causal relationship than a normative claim about the proper scope of trademark rights. A theory’s failure to reverse doctrinal developments would count in its favor as a descriptive theory of the doctrine in question; such failure is only a fault if one thinks the theory ought to generate different doctrine.
If McKenna’s accusation of complicity is indeed intended as a descriptive claim that search-costs theory somehow generates the doctrines of which he complains—perhaps through courts’ alleged equation of “search costs” with “confusion”—then we come to my second objection. To wit: I don’t believe this claim is consistent with the cases McKenna cites in support of it. Of course, I concede that the Supreme Court has invoked search-costs theory as a general justification for legal regulation of trademarks,11 and that Judge Posner and his colleagues on the Seventh Circuit rely heavily on search-costs theory in trademark cases. But tellingly, the most obvious example McKenna cites—Judge Posner’s opinion in Ty v. Perryman—is about dilution, not confusion, and moreover demonstrates the fundamental silliness of attempting to retroactively justify many modern trademark doctrines in terms of search costs, when their purposes and effects clearly lie elsewhere.12
Moreover, many of the confusion-based decisions McKenna objects to are not informed by search-costs theory in any obvious way. Take, for example, the sponsorship and affiliation confusion cases that McKenna knows all too well.13 Not one of the cases he discusses in the section of his article dedicated to critiquing this doctrine says anything about search costs.14 On initial-interest confusion, McKenna essentially concedes that misappropriation theory, not search-costs theory, is driving the expansion of trademark rights.15 On post-sale confusion, again, it is not search-costs theory that McKenna objects to, but the fact that courts are “always able to fall back in these cases on broad references to the problem of consumer confusion.”16 If the judges in these cases are being led astray by search-costs theory, they are leaving no evidence of it in the text of their opinions.
Thus, McKenna seems to be relying on a fallacy of his own when he argues that search-costs theory is somehow generating an undesirable expansion in trademark rights: the straw man. McKenna claims that courts’ reasoning can be encapsulated in three related propositions: (1) Courts “take it as a given that confusion necessarily harms consumers[, because] (2) [c]onfusion increases search costs, and (3) search costs are bad. End of story.”17 But propositions 2 and 3 do not seem to be playing any significant role in the expansion of doctrine McKenna decries. Rather, all the work appears to be done by proposition 1: the proposition that confusion harms consumers.
If this proposition is the fundamental flaw in modern trademark doctrine, it is difficult to attribute it to the Chicago School. Likelihood of confusion has been the standard of federal trademark infringement since well before the Law and Economics movement found its legs, let alone trained its sights on trademark law.18 To the contrary, McKenna, like myself and many other trademark commentators, has invoked principles of welfare economics from which search-costs theory is derived to critique recent expansions of trademark rights as theoretically unsound.19 We would not be able to make such arguments if trademark doctrine was consistent with those principles. Thus, I think McKenna’s descriptive claim that search-costs theory is somehow causally linked to an undesirable expansion in trademark rights is unpersuasive.20
This is not, I think, as serious a fault as it might seem for McKenna’s project, because of my third objection to his argument about the causal relationship between search-costs theory and expansive trademark rights: it is irrelevant. Even if I’m correct that McKenna’s descriptive claim fails to persuade, this does not settle the normative question whether reducing consumer search costs ought to be the goal of trademark law. On that normative question, I think McKenna has the better of the argument. The empirical evidence he reviews provides strong support for his normative claim that lowering search costs ought not to be our organizing principle in trademark cases. This is simply another way of saying that the conclusion of the flawed syllogism I noted at the outset is false. Search-costs theory need not be responsible for the expansion of trademark rights for us to conclude that those rights are overbroad. And McKenna has given us ample independent evidence of such overbreadth. So I view his claim that courts equate search costs with confusion, while not entirely correct as a descriptive matter, as superfluous to the normative thrust of his project—his defense of a model of trademark law based on consumer decision making—which I find appealing independent of any comparison to search-costs theory. But this leaves us with two important questions that I think have related answers. The first is a descriptive question: what is driving the expansion of trademark rights, if not search-costs theory? And the second is a normative question: what ought trademark law be trying to do (if not lower search costs) and why?
On the first question, McKenna’s article suggests some possible culprits. One is the misappropriation theory that he recognizes is behind initial-interest confusion doctrine and which has been the target of some (well-earned, in my view) derision in trademark law scholarship.21 But a more intriguing possibility is suggested by McKenna’s assertion that courts treat search costs and confusion interchangeably. This claim appears to be the basis of his argument for a causal link between search-costs theory and the expansion of trademark rights.22 But if I’m correct that search-costs theory is not playing a role in the cases in which courts expand trademark rights,23 then McKenna’s assertion suggests a far deeper problem. That is, I think his complaint is directed not against search-costs theory, but against the likelihood of confusion standard for infringement itself.
McKenna’s own historical scholarship clearly demonstrates that trademark liability was not always directed at consumer confusion—to the contrary, historically it was directed at fraud that resulted in diversion of trade.24 The Trademark Act of 1905, for example, made likelihood of confusion a barrier to registration, but provided a much narrower definition of actionable infringement.25 Not until the 1920 Act does something approaching confusion enter the statutory language defining a federal unfair competition claim,26 and only with the 1946 passage of the Lanham Act does likelihood of confusion become central to trademark infringement liability.27 It is the 1962 amendments to the Lanham Act that expand the range of actionable confusion beyond confusion of “purchasers”—a development that commentators (including McKenna himself) have tied to many of the doctrinal innovations he discusses.28
McKenna seems to suggest that a causal relationship between search-costs theory and expanding trademark rights is implied by the fact that both arise around the same time,29 but clearly these statutory changes also line up with the expansionary trend in the cases. So it seems to me that the historical record suggests at least as strong a case for a causal relationship between the likelihood of confusion standard and the expansion of trademark rights (though we might question which way the causation runs). In fact, I would suggest that this causal relationship is even more plausible than McKenna’s claimed link between search-costs and expanding rights, insofar as the expansionary cases uniformly cite the likelihood of confusion standard of the Lanham Act, but rarely discuss search costs at all.
What, then, might be the causal mechanism through which the likelihood of confusion standard and the expansion of trademark rights could be related? I think McKenna’s article suggests a highly plausible explanation: “confusion” is an extremely broad term that can encompass any number of mental states.30 Thus, as likelihood of confusion came to replace more tightly circumscribed statutory tests for infringement, the scope of trademark rights might well be expected to expand to include those mental states (or vice versa, perhaps). I believe that this insight is implicit in McKenna’s discussion of the development of trademark doctrine. But here again we run into Hume’s Law. McKenna’s closing argument on this point runs a descriptive claim as to what the likelihood of confusion test does together with a normative claim about the appropriate targets of trademark law: “it is obvious on a moment’s reflection,” he claims, “that trademark law cannot, and should not, respond to all forms of confusion or even to all confusion in the marketplace.”31 It is worth teasing these two claims apart.
I don’t agree that trademark law cannot, as a descriptive matter, respond to all forms of confusion in the marketplace resulting from unauthorized use of a trademark. Indeed, many of the doctrines McKenna criticizes strike me as examples of judges attempting to do precisely that in service to the Lanham Act’s broad command. I argued above that this statutory breadth, rather than reliance on search-costs theory, is probably what accounts for the expansion in trademark rights documented in the first half of McKenna’s article.
But if I’m correct, my descriptive claim makes McKenna’s second claim—his normative claim—all the more important. If the term “confusion” is as broad as it appears to be, then the “likelihood of confusion” test for infringement becomes something of an empty vessel that can be filled with any number of normative claims regarding the appropriate scope of trademark rights, any one of which might be highly contestable. When such contestable claims end up embodied in positive law, we may well have cause to complain about the result, but it is incumbent on us to explain through normative argument what makes the result undesirable. Even if the likelihood of confusion standard, as applied by the courts, could eliminate all forms of consumer confusion from the marketplace, the question remains whether we would want it to do so.
McKenna claims the answer to this question must be no, and part of his support for this claim is his empirical evidence on the psychological and welfare effects of search costs. But it seems to me that this evidence, standing alone, can at best rule out reduction of search costs as a comprehensive normative goal for trademark law. While this is not an insignificant achievement, and stands as a major contribution of McKenna’s article, it is not the same thing as an affirmative normative account of trademark law, which is what I think McKenna hopes to provide. Thus, we return to the second question I identified earlier: what ought trademark law be trying to do, and why?
McKenna’s answer to this question takes up the second half of his article, and can be summarized in the simple directive: “prevent deception that influences consumer purchasing decisions.” His reasons in support of this position are bound up in a particular theory of consumer autonomy, one which holds that consumers are the best judges of their own preferences,32 that they are entitled to determine and pursue those preferences without interference by others,33 and that they need not (or perhaps ought not) be protected from others’ efforts to shape those preferences in the absence of deception.34 This vision of consumer autonomy strikes me as the main contribution of McKenna’s article, and it bears unpacking.
I am interested in identifying the source of the normative content that informs McKenna’s proposal, because I believe that here as well there is an is/ought problem buried beneath the surface. Ostensibly, McKenna derives his model of consumer autonomy from false advertising law and commercial speech doctrine.35 But while the argument from authority is not necessarily fallacious in the common-law system, it provides no normative content of its own. The cases McKenna cites suggest that normative values inherent in the First Amendment underlie his model of autonomy,36 but his own analysis is remarkably quiet as to what those values are.
One possible normative value is improving outcomes: perhaps McKenna’s model of consumer autonomy will lead to the best results as measured by some relevant standard. Recall that McKenna’s argument for rejecting search-costs theory as a foundation for trademark law rests on his descriptive claim that not all reductions in search costs increase consumer welfare. A concern over consumer welfare is shot through his critique of search-costs theory,37 and it implies a normative criterion for evaluating that theory and the doctrine it purports to justify. This criterion is that increases in aggregate welfare are desirable, and decreases in aggregate welfare are undesirable—the welfare-maximization principle at the heart of the economic analysis of law.
I do not believe that McKenna’s model of consumer autonomy is entirely consistent with this normative principle, which creates an odd disconnect between the descriptive project of the first half of his article and the normative project of the second half. If McKenna’s problem with search-costs theory is that it generates doctrine that leads to decreases in consumer welfare, one would expect his alternative normative framework to avoid this failing. But it seems to me that his theory does not require all decreases in consumer welfare that result from trademark-related practices be avoided. And importantly, it is not clear to me whether McKenna’s reason for tolerating decreases in consumer welfare is grounded in a descriptive claim that the alternative to such toleration invites an even larger decrease in aggregate social welfare or a normative claim that values distinct from and incommensurate with consumer welfare—First Amendment values, perhaps—must take precedence.
Take McKenna’s discussion of the distinction between deception and persuasion—a distinction on which some of his proposals for doctrinal reform rest.38 One the one hand, he suggests that the administrative costs of attempting to regulate persuasion are too high—that it would be impracticable and “messy” to do so, and error in the attempt could lead to reduction in the availability of socially valuable information.39 This is a descriptive claim: the welfare losses that result from consumer susceptibility to persuasive messages are less than the welfare losses that would result from attempting to regulate such messages. On the other hand, McKenna suggests that attempts to regulate persuasive messaging are undesirable regardless of their effect on consumer welfare: “due regard for consumer autonomy requires us to live with [consumers’] decisions even if they are bad.”40 This is a normative argument, and it cannot be supported by reference to calculations of consumer welfare. To the contrary, it implies that even if regulation of persuasive messages would increase aggregate welfare, we should still refrain from such regulation.
I don’t think either of these arguments is flawed in principle, but I would like to know which one McKenna would like us to rely on in evaluating his proposals. The main reason for my curiosity is that the descriptive argument is subject to empirical testing and (potentially) falsification, while the normative argument is not. This suggests a potential for conflicts between the two arguments, which would require us to decide which of them carries more weight.
I think the clearest example of such conflict lies in the potential for paternalist intervention that increases aggregate welfare. McKenna uses the term “paternalist” as a broadside, invoking all the most pejorative connotations of the term.41 “[T]rademark law,” he insists, “should not coddle consumers.”42 But of course there is a lively debate across a number of disciplines (including law) as to whether, and under what circumstances, paternalist intervention might be permissible or even desirable.43
For example, I have written in the past that trademarks supported by advertising can bias consumers.44 That is, marketing practices can lead consumers to hold beliefs about the objective qualities of products that are both false and resistant to correction by exposure to empirical evidence, without making claims about such objective qualities. I think that McKenna would agree that a consumer whose purchasing decisions are biased in the sense I have described is likely to make decisions that lead to lower aggregate welfare than would be the case in the absence of the biasing marketing practice. What I don’t know is whether McKenna would countenance regulation to curb such biasing marketing practices.
Another conflict arises in the case of unauthorized uses of trademarks on goods of equivalent quality to the plaintiff’s goods. Suppose a defendant applied a plaintiff’s trademark without authorization to a product that was in every measurable way precisely identical to the plaintiff’s products, but was offered by the defendant at a lower price. A consumer who purchased the defendant’s goods in reliance on the trademark would be deceived in a way that affected their decision to purchase, but would be better off for that purchase than she would have been in the absence of the deception. As I discuss in forthcoming work, extant doctrine suggests that such a defendant’s conduct would constitute trademark infringement,45 even though this result clearly causes a reduction in aggregate welfare, at least as a matter of first-order effects. Again, it is not clear to me whether McKenna’s model would endorse or reject that result.
These two examples are obviously outside the scope of McKenna’s article, but they reveal the tension at the heart of his project. In the first case, legal intervention might prevent a decrease in consumer welfare attributable to non-deceptive, non-persuasive uses of trademarks by their owners. In the second, legal intervention causes a decrease in consumer welfare in preventing a deception that actually makes consumers better off. To the extent McKenna’s rejection of search-costs theory in favor of a consumer decision-making theory is premised on the welfare effects of search costs, we might conclude that a concern for consumer welfare would lead McKenna to allow intervention in the first example and to reject it in the second. But to the extent McKenna’s embrace of a consumer decision-making theory is premised on other values that are inherent in a particular model of consumer autonomy, we might expect him to reject legal intervention in the first example and endorse it in the second on grounds that the law ought to leave it to consumers to negotiate the mix of information available in the marketplace even if the consequences reduce aggregate welfare.46
In sum, it is not clear to me whether McKenna’s consumer decision-making theory is grounded in the normative principles of welfare economics or in some alternative normative principle. If the latter, it is not clear to me what the alternative normative principle is (beyond the ipse dixit that autonomy is a value in itself). There are suggestions in his article that such alternative principles might be found in the justifications for the First Amendment, and on this issue perhaps the work of my co-respondent, Professor Laura Heymann, is most instructive.47 But even if this is correct, the argument by analogy to First Amendment principles requires us to explore the normative values that underlie those principles, and to evaluate their merits against those of the welfare-maximization principle.
For my part, I think McKenna’s consumer decision-making model will be more persuasive if it explicitly disavows welfare maximization as the sole—or even the most important—normative value at stake in trademark cases. Reliance on welfare maximization reduces too many close and contested policy debates to empirical questions in which the relevant data is necessarily absent.48 Frequently, we must take recourse to some other values that can avoid the empirical impasse. In some areas trademark law already recognizes such values—particularly with respect to expressive uses of trademarks.49 An interesting question, then, is whether we can build a more complete and coherent model of the values at stake in trademark law. I have some ideas about this,50 and I think they are largely consistent with McKenna’s consumer decision-making model. I would encourage him to flesh out these hidden foundations of his model—to accentuate the normative—in the hopes of escaping the is/ought problems that unnecessarily detract from the force of his otherwise very appealing claims.
Preferred citation: Jeremy N. Sheff, Accentuate The Normative: A Response to Professor McKenna, 98 Va. L. Rev. In Brief 48 (2012), http://www.virginialawreview.org/inbrief/2012/04/14/Sheff_Accentuate.pdf.
[Back to this issue]