# Can Disclaimers of Affiliation Dispel Trademark Confusion? Evidence From Two Randomized Experiments

**Authors:** Barton Beebe, Roy Germano, Joel Steckel
**Citation:** "Can Disclaimers of Affiliation Dispel Trademark Confusion? Evidence From Two Randomized Experiments," *J. Empirical Legal Stud.* (2026) (with Roy Germano and Joel Steckel)
**Source:** https://doi.org/10.1111/jels.70029

## | Introduction

*p. 1*
Early on in the case Pennsylvania State University v. Vintage Brand LLC, 1 Chief Judge Matthew W. Brann queried whether the remedy for the defendant's allegedly infringing conduct should be limited to an injunction requiring that the defendant use a disclaimer of affiliation. 2 The main facts of the case are that Vintage Brand, a clothing company, was selling merchandise online bearing old Penn State logos that Penn State itself no longer uses and which Vintage Brand argues are in the public domain (Santoni 2024). 3 In his opinion denying Penn State's motion to dismiss various counterclaims, Judge Brann recognized a long and consistent line of Supreme Court dicta extending as far back as Kellogg Co. v. National Biscuit Co. 4 These dicta have suggested that in light of both First Amendment and competition concerns, disclaimers of affiliation may be a less speech-restrictive, more pro-competitive form of relief than an outright prohibition of the defendant's speech. 5 The preferred remedy, in other words, may sometimes be "more speech" 6 and also more competition.

*p. 1*
Judge Brann considered the possibility that consumers may be confused as to the true source of Vintage Brand merchandise bearing old Penn State insignia because of "a belief that appears to have come from what [consumers] thought the law required," namely that all such merchandise must be approved by Penn State. 7 But as Judge Brann pointed out, consumers' beliefs about what the law requires should not necessarily form the basis of what the law does in fact require:

*p. 1*
The circularity is apparent: the law only offers protection if there's belief, yet the belief comes from consumers' (mis)conception about the law. It would seem perverse to award market exclusivity based on a fake-it-until-you-make-it approach. If consumers' confusion stems from their incorrect belief that goods bearing Penn State's emblem must be licensed, shouldn't that belief be corrected, not perpetuated? 8 decide the law in fact requires. As trademark commentary has long observed, this circularity is vicious because it tends toward ever broader trademark rights, and as these rights broaden, they come into considerable tension with our deepest societal commitments to free speech and free market competition (Beebe 2005, 2066-69). 9 But as Judge Brann recognized, a way to stop and even reverse this circularity is by means of disclaimers of affiliation that teach consumers not to be confused as to source or sponsorship.

*p. 2*
Yet notwithstanding Supreme Court pronouncements and the occasional judge willing to consider less-restrictive injunctions of speech, the habitual view in trademark case law and commentary has been that disclaimers of affiliation are ineffective (Heymann 2010, 395). 10 The empirical foundations of this view can be traced back to a study that was published by Jacob Jacoby and Robert Raskopf in 1986 and cited by the Second Circuit in two prominent decisions the following year. 11 A recurring theme of this and later studies is that disclaimers are ineffective because consumers tend not to notice them, and even when they do, they tend not to read them or pay attention to their content (Jacoby and Raskopf 1986, 55).

*p. 2*
Jacoby and Raskopf, and the Second Circuit judges who relied on their research, were writing years before retail commerce began to move onto the internet. While on physical products it may be necessary to relegate disclaimers to the fine print to save room for other information, there is a great deal of flexibility in how product information can be presented online. For one, disclaimers can be displayed more prominently on product webpages than on physical products. Disclaimers can also be displayed at multiple points in the purchasing process and designed to require consumers to interact with them. For example, an online disclaimer may appear next to a product, and then again at checkout. At checkout, the consumer may be required to acknowledge that they have seen and read the disclaimer before they can complete their purchase.

*p. 2*
In this article, we revisit the debate over disclaimers and confusion with new evidence from two randomized experiments that tested the effectiveness of disclaimers in an online shopping context. In our first experiment, some participants saw a product webpage that included a standard fine-print disclaimer of affiliation, while others saw a version of the page with no disclaimer. Of the participants who saw the disclaimer, some saw the disclaimer again after viewing the product webpage and were required to retype its contents into a text box before they could continue with the study. Consistent with prior research, we found that, on its own, the disclaimer did not dispel confusion, although it reduced it significantly. But we also found that coupling the disclaimer with the typing exercise reduced the probability of confusion to a level that most courts would consider negligible.

*p. 2*
Because it is improbable that any merchant would ask its customers to retype a disclaimer, we designed our second experiment to test a more realistic acknowledgment task. We also used this opportunity to test the effect of a shorter, more prominently displayed disclaimer. Specifically, some Study 2 participants saw a version of the product webpage with the same fine-print disclaimer that Study 1 participants saw, some saw a version with the more prominent disclaimer, and some saw a version with no disclaimer. Of those who saw the fine-print disclaimer, some participants were asked to register their recognition of it in one of two ways: either by typing it out, as Study 1 participants did, or by attesting that they had read and understood the disclaimer by clicking an acknowledgment box.

*p. 2*
As we found in Study 1, the typing exercise reduced confusion to negligible levels. Moreover, we found that requiring participants to acknowledge the disclaimer by simply clicking a box was almost as effective in dispelling confusion as the more burdensome task of typing the disclaimer out. Finally, we found that a brief, ostensibly more noticeable disclaimer was slightly less effective in reducing confusion than a less noticeable but arguably more informative fine-print disclaimer.

*p. 2*
In sum, the disclaimers we tested, even the more prominently displayed disclaimer, were unable to dispel consumer confusion on their own. The effectiveness of the fine-print disclaimers, however, was greatly enhanced when consumers saw them a second time and were required to register acknowledgment of them. Requiring participants to click a box affirming that they read the disclaimer reduced the probability of confusion from 73% in the baseline condition down to 9% and eliminated all confusion as to the product's source.

*p. 2*
In the first section of this article, we discuss how courts have considered disclaimers as a remedy for trademark infringement in recent decades. Next, we provide an overview of the empirical studies that have most contributed to the view among judges and legal scholars that disclaimers are not effective. We then discuss why it is important to update our understanding of disclaimers for an era in which many consumers do their shopping online. Subsequently, we describe and report the results of our experiments. The final section considers the implications of our findings for trademark practice and policy. We focus in particular on the potential for disclaimers of affiliation to reverse trademark law's vicious circularity and allow for judicial remedies short of absolute injunctions on defendant conduct.

## | Legal Background

*p. 2*
From the late 1970s through the 1990s, disclaimers of affiliation received significant judicial and scholarly attention in trademark case law and commentary. The Supreme Court's long championing of disclaimer relief appears to have had some impact. But by the end of this period, opinions congealed around the view that disclaimers were generally not effective. Little has changed since then. We turn first to the case law.

*p. 2*
The high point for disclaimer relief in the trademark case law came almost a half-century ago in the 1978 Second Circuit case Consumers Union v. General Signal Corp. 12 The defendant was planning to run two television commercials for its REGINA vacuum cleaners that quoted the plaintiff's magazine Consumer Reports. 13 The commercials would include the disclaimer "Consumer Reports is not affiliated with Regina and does not endorse Regina products or any other products." 14 The district court issued a preliminary injunction prohibiting the airing of the commercials. 15 The Second Circuit reversed. The Second Circuit court was keenly aware of the First Amendment implications of blanket prohibitions on the defendant's speech. 16 In short, the Second Circuit argued that when effective, disclaimers were the least restrictive means of ensuring that consumers would not be misled. 17 Acting on this principle, Second Circuit courts approved of disclaimers of affiliation in numerous cases in the early 1980s (Palladino 1992). 18 But the Second Circuit's enthusiasm for disclaimers of affiliation did not last. Four years after Consumers Union, the Second Circuit changed course. On the same day in 1987, three separate Second Circuit panels issued their respective opinions in a trio of trademark infringement cases. 19 Each of these cases involved a disclaimer of affiliation. Each opinion cited the other two. 20 None of them cited Consumers Union.

*p. 3*
The most prominent of these cases was Home Box Office v. Showtime/The Movie Channel (Palladino 1992, 203). Mainly at issue were various Showtime slogans that Showtime featured in promotional materials that referenced HBO-for example, "SHOWTIME & HBO. It's Not Either/Or Anymore." 21 The district court opinion in Home Box Office had quoted at length from Consumers Union and ruled that the disclaimers Showtime used in these materials were adequate to dispel confusion. 22 The Second Circuit took a different view. It noted that in some of the materials the "disclaiming information does not appear in sufficiently close proximity to the infringing statements." 23 More broadly, it expressed concern about the effectiveness of disclaimers in general, citing case law from other circuits 24 as well as then-recent academic commentary. 25 The Second Circuit invited Showtime to revise their materials and seek approval from the district court "on the basis of its use of disclaimers or otherwise." 26 But the court added a kicker, which was that "[u]pon such an application there would be a heavy burden on Showtime to come forward with evidence sufficient to demonstrate that any proposed materials would significantly reduce the likelihood of consumer confusion." 27 In the Second Circuit, if a court found that the defendant's conduct created a likelihood of confusion as to source, the defendant would face a "heavy burden" to prove that any disclaimer would "significantly reduce" that confusion.

*p. 3*
The other two opinions that the Second Circuit issued along with Home Box Office similarly represented a retreat from Consumers Union. In Charles of the Ritz Group v. Quality King Distributors, the Second Circuit found that the defendant "failed to introduce empirical evidence that the disclaimer actually lessens consumer confusion as required to overcome such a previous finding [of confusion]." 28 In Soltex Polymer Corp. v. Fortex Industries, the Second Circuit actually found the disclaimer of affiliation to be effective. But the Soltex court substantially narrowed the impact of its holding. It emphasized that the case involved good faith conduct by the defendant and a "minimal or moderate amount of potential confusion" which could be cured by a disclaimer. 29 In the decades since Home Box Office, other circuits have adopted the Second Circuit's burden shifting framework 30 and courts have repeatedly expressed doubt about the efficacy of disclaimers of affiliation (McCarthy 2025, § 23:51). They have even suggested that in certain situations such disclaimers may increase the likelihood of confusion. 31 At least one court held that "[t]he voluntary act of placing a disclaimer on one's packaging constitutes a tacit admission that the use of plaintiff's trademarks, in the entire context of the packaging, is likely to confuse consumers" 32 -though this court's finding of a likelihood of confusion was reversed on appeal. 33 Defendants have meanwhile done the reputation of disclaimers no favors. Numerous disclaimer cases have presented courts with facts in which bad faith defendants presented their disclaimers in small print, in areas the consumer would not likely see, or in language that would not likely dispel confusion. 34 In rare instances, courts have prescribed disclaimer relief or otherwise cited a defendant's disclaimer of affiliation in support of a finding of no likelihood of confusion. 35 But the majority of trademark infringement case law that has addressed disclaimers has judged them to be ineffective unless empirical evidence shows their effectiveness.

## | Empirical Research on Trademark Disclaimers

*p. 3*
The most influential work of empirical scholarship on disclaimers of affiliation remains a 1986 article by Jacob Jacoby and Robert Raskopf entitled "Disclaimers in Trademark Infringement Litigation: More Trouble Than They Are Worth." Home Box Office, Charles of the Ritz, and many of the other disclaimer cases that parted from Consumers Union cited Jacoby and Raskopf's study in support of their conclusion that disclaimers were not effective. 36 Home Box Office went so far as to quote the article's recommendation that "[w]henever disclaimers are considered, empirical studies should be used to evaluate their likely impact." 37 Jacoby and Raskopf's study reports the results of a survey that Jacoby had conducted in connection with litigation between the National Football League and a manufacturer of "replica" NFL jerseys in National Football League v. Wichita Falls Sportswear. 38 The survey sought to determine if a disclaimer of affiliation appearing on the jerseys would dispel consumer confusion as to the source of the jerseys. The survey presented 1471 in-home respondents with various football jerseys bearing NFL designs and city and player names. Some of these jerseys carried a disclaimer sewn into the neckline stating, "Not authorized or sponsored by the N.F.L." Jacoby and Raskopf emphasized that, "[t]o insure readability, a larger print size was used for the disclaimer than for virtually all other lettering appearing on the label" (Jacoby and Raskopf 1986, 50-51).

*p. 3*
The survey initially asked respondents various questions probing for top-of-mind awareness and unaided and aided association (pp. 51-52). Jacoby and Raskopf reported that "[e]ven with the disclaimer present, more than half the respondents provided an NFL related association to each and every one of the unauthorized shirts to which they had been exposed" (p. 52). Their conclusion was that "the disclaimer label exerted a trivial impact on the number of NFL related associations."

*p. 3*
Turning to the question of consumer confusion, the survey then asked "Did the company that made the shirt have to get authorization or sponsorship, that is, permission, to make it?" For respondents who answered "no" to this question, the survey asked: "If you saw someone walking down the street, wearing a shirt like this, would you think that the company that made the shirt received authorization or sponsorship, that is, permission, from the National Football League or one of its teams?" (pp. 52-53). Jacoby and Raskopf reported that the disclaimer "exerted virtually no corrective impact on confusion:" 58% of respondents shown a jersey without a disclaimer responded that authorization was necessary while 59.1% of those shown a jersey with a disclaimer did so (p. 53).

*p. 4*
Jacoby further challenged the effectiveness of disclaimers in a 1994 study with coauthor George J. Szybillo. Jacoby and Szybillo reported the results of a survey Jacoby conducted in connection with Schering Corp. v. Schering Aktiengesellschaft, 39 a 1987 litigation in which the American and West German firms asserted competing claims to use the mark SCHERING in the United States. The survey sought to determine if a disclaimer of affiliation used by an American subsidiary of Schering AG in its marketing materials dispelled confusion as to any affiliation between the subsidiary and Schering Corporation.

*p. 4*
The Schering survey sampled three hundred physicians and three hundred pharmacists, with one-third of each group randomly exposed to one of three different twelve-page promotional brochures (Jacoby and Szybillo 1994, 228-29). One set of 100 physicians and 100 pharmacists were presented with a brochure for the subsidiary's products that contained a disclaimer in a footnote on page three stating: "Schering AG, West Germany, is not connected with Schering-Plough Corporation or Schering Corporation, Kenilworth, New Jersey" (pp. 229-30). A second set of physicians and pharmacists were exposed to the same brochure except that it did not include the disclaimer in the footnote (p. 230). The third set was exposed to the same brochure except that it included a "claimer" in the footnote that matched the disclaimer's language except for the omission of the word "not" (p. 230). The survey asked each set of physicians and pharmacists the same series of questions. Question 1 asked: "First, what is the name of the company that puts out the products mentioned in the brochure?" (p. 231). Question 2 asked: "Just from what you read in the brochure, is any other company involved in putting out the specific brands of oral contraceptives mentioned in the brochure?" (p. 232). The remaining questions were essentially reading comprehension questions. For example, Question 3: "As best you can remember, did the brochure refer to any other company or companies in any other context?" (p. 233). Question 6a: "Did the brochure state that any of the companies it mentioned were not connected or not related to one another in any way?" (pp. 234-35).

*p. 4*
Overall, Jacoby and Szybillo reported no significant differences in responses among those exposed to brochures containing either the disclaimer, no disclaimer, or the claimer (p. 237). They concluded that "disclaimers relying on brief negator words such as 'no' and 'not' are not likely to be effective." A sensible interpretation of the results is that a high proportion of those exposed to a twelve-page brochure simply did not notice a disclaimer in a footnote on the third page.

*p. 4*
That disclaimers, often relegated to fine print, go unnoticed is indeed the most obvious reason why they are ineffective (Jacoby and Morrin 1998). But drawing attention to the disclaimer is only part of the challenge. Consumers may still ignore disclaimers they notice. As Jacoby and coauthor Maureen Morrin put it in a 1998 review article, "the perceiver might not consider the information in a disclaimer sufficiently relevant to his or her goals, with the result that, no matter how prominent the disclaimer or how often it is repeated, he or she might choose simply to ignore or discount its contents" (p. 41). Studies on mutual fund disclaimers (Mercer et al. 2010), disclaimer footnotes in advertisements (Foxman et al. 1988), and disclaimers on images in fashion magazines (Tiggemann et al. 2017) confirm that more prominently displayed disclaimers are not necessarily more effective. These studies, like Jacoby's studies of trademark disclaimers, are generally skeptical that disclaimers can change minds and behaviors in the ways intended.

## | Updating Our Understanding of Disclaimers for the Internet

*p. 4*
It is surprising that what continues to be the most influential empirical studies and judicial opinions on trademark disclaimers were written many years before the internet became a fixture in Americans' daily lives. We propose an update to our understanding of the effectiveness of disclaimers to account for important changes brought about by the web and online shopping.

*p. 4*
One change is with regard to space. Space is finite on a physical product label. Even when companies are acting in good faith, limited space on packaging may necessitate that disclaimers appear in very small print. Websites do not have the same restrictions. Although there are good reasons to limit the amount of information that appears on a product webpage, the web offers a great deal of flexibility in terms of how much information a merchant can present and how prominently some information is displayed. The web also allows certain information to be repeated at different points in the purchasing process.

*p. 4*
Online shopping furthermore creates opportunities for consumers to interact with product information. While it would be unreasonable to expect clerks at physical stores to ask every consumer if they read and understood a disclaimer of affiliation, there are simple and unobtrusive ways to do this on the web. It may be particularly useful to present disclaimers again at checkout and require consumers to register their recognition of them before they can complete a purchase. Doing so gives the merchant an opportunity to expose consumers to the disclaimer a second time and in a setting in which the consumer is not also considering multiple product details. Requiring consumers to register their recognition of the disclaimer before they can complete the purchase also increases the relevance of the disclaimer to the consumer. These interventions could make the disclaimer harder for the consumer to ignore.

*p. 4*
In the next section, we present experimental findings on the effectiveness of disclaimers in an online purchasing context. We test a standard disclaimer that appears in small print as well as modifications to the standard disclaimer, such as increasing the visibility of the disclaimer and requiring consumers to register recognition of it as a condition of purchase. Our first experiment took the form, in essence, of a proof of concept to test the viability of our hypothesis that requiring consumers to acknowledge the disclaimer can boost effectiveness. The second experiment evaluated the replicability of our findings and tested a more practicable implementation of the acknowledgment task.

## | Experimental Evidence

*p. 5*
This section reports findings from two randomized experiments we conducted to test the effectiveness of trademark disclaimers. We used the online research platform Prolific to recruit 302 participants for our first study and 498 participants for the second. Using Prolific's prescreening questions, we limited the pool of participants to residents of the United States whose primary language is English. In the second study, we further screened out any potential participants who said they never purchase clothing online, and anyone who participated in the first study. After qualifying and agreeing to participate, participants were directed from the Prolific website to an online survey hosted by Qualtrics.

*p. 5*
The 800 people who participated in these studies were given the same instructions and put in the same purchasing context. First, they were instructed to begin by imagining that they were an online shopper who was in the market for a new Tshirt. Next, they were shown a product webpage with a logo for a company called Nostalgia Brands and a T-shirt bearing an old logo of the University of Notre Dame that Notre Dame itself no longer uses. We modeled this purchasing context from the dispute between Vintage Brand and Penn State, but used a different university and a fictitious company name to prevent participants from finding information about the case on the internet.

*p. 5*
As we detail below, all participants saw the same product page and the same survey questions. We only varied the disclaimers that appeared on or below the product page. So that all participants would see the product page in roughly the same way, we optimized it for viewing on mobile devices and required participants to access the study on a smartphone.

## | Study 1 Design

*p. 5*
Study 1 participants were randomly assigned to one of three groups. Participants assigned to the No Disclaimer group (n = 100) saw the product webpage without any disclaimer of affiliation between Nostalgia Brands and Notre Dame, as shown in Figure 1a. Participants assigned to the Disclaimer Only group (n = 103) saw the same product webpage except that it displayed a standard disclaimer in fine print above the image of the T-shirt that stated, "Nostalgia Brands not affiliated with or sponsored by the University of Notre Dame." This product page and disclaimer are shown in Figure 1b. The size, placement, and content of this disclaimer are similar to the real disclaimer used by Vintage Brand on its Penn State merchandise.

*p. 5*
Participants assigned to the Typing Exercise group (n = 99) were presented with the same webpage and disclaimer that the Disclaimer Only group saw (see Figure 1b). After viewing the product page, participants in the Typing Exercise group were told to type into a text box the following statement: "I understand that this T-shirt is not put out by, affiliated or connected with, or approved or sponsored by the University of Notre Dame." 40 This stimulus is displayed in Figure 2a. Although they were instructed to type the statement, participants were not prevented from copying-and-pasting it into the text box. The purpose of this condition was to expose participants to the disclaimer a second time and require them to register their

## | Study 2 Design

*p. 6*
We realize that a disclaimer typing exercise is impractical and that no court would issue an injunction demanding such a remedy. But we developed the Typing Exercise condition as a preliminary test of the viability of our hypothesis that requiring consumers to register recognition of a disclaimer can dispel confusion. Following this initial proof of concept, we developed Study 2 to test the hypothesis under more realistic conditions. We also developed Study 2 to test whether the results of Study 1 could be reproduced beyond our initial sample at a different point in time 41 and to test the effect of a shorter, more prominent disclaimer design.

*p. 6*
Study 2 participants were thus randomly assigned to one of five groups. Participants in the No Disclaimer group (n = 100), the Disclaimer Only group (n = 100), and the Typing Exercise group (n = 100) saw the same stimuli as their counterparts in Study 1.

*p. 6*
We added the Click Box group (n = 99) to further test the hypothesis that requiring consumers to register their recognition of a disclaimer can reduce confusion. Participants assigned to this group saw the webpage that participants in the Disclaimer Only group saw (see Figure 1b), followed by instructions to read and acknowledge the following statement, shown in Figure 2b: "This T-shirt is not put out by, affiliated or connected with, or approved or sponsored by the University of Notre Dame." Just below the statement was a box that said, "By clicking this box, I confirm that I have read the statement." Participants were required to click the box in order to proceed with the survey. This condition forces consumers to acknowledge the disclaimer but without the burden of retyping it.

*p. 6*
We developed a fifth group, the Bold Disclaimer group (n = 99), to test whether confusion can be reduced with a shorter, more noticeable disclaimer that does not require participants to take the extra step of registering their recognition. Participants assigned to this group saw a version of the product page with a disclaimer that read, "NOTRE DAME LOGO USED WITHOUT PERMISSION." This disclaimer, shown in Figure 1c, was located just above the product image, displayed in bold, capital letters and highlighted in yellow to draw participants' attention, in the style of a warning label.

## | Survey Questions

*p. 6*
After viewing the product page and completing the typing or click box acknowledgment tasks when applicable, all participants encountered the same three survey questions, which we drew from the standard Eveready survey format for the likelihood of consumer confusion (McCarthy 2025, § 32:174). 42 The first question asked: "In your opinion, what company or organization puts out this T-shirt?" Participants were instructed to type out their answer in a text entry box that appeared below the question. The second question asked: "In your opinion, is this T-shirt affiliated or connected with any company or organization apart from the one that puts it out?" Participants were provided with three answer choices: Yes, No, and Don't know. If participants selected Yes, an additional question instantly appeared: "What is that other company or organization?" Participants were instructed to type out their answer in a text entry box. The third question asked: "In your opinion, is this T-shirt approved or sponsored by any company or organization apart from the one that puts it out?" Participants were again provided with Yes, No, and Don't know answer choices. Those who selected Yes were instructed to type the name of the company or organization in an accompanying text entry box.

*p. 6*
Our three questions tested for three forms of consumer confusion: source confusion, affiliation confusion, and sponsorship confusion. Even if a consumer is not confused as to the true source of a product, that consumer may nevertheless believe that the producer of the product is commercially affiliated with another company or has received approval from another company to put out the product. To find infringement, courts generally require a showing that at least 20% to 30% of the relevant population is experiencing any of these three types of confusion (Beebe and Hemphill 2017; McCarthy 2025, § 32:188). 43 However, some courts have expressed skepticism of evidence showing only affiliation or sponsorship confusion (and not source confusion) because consumers may mistakenly assume affiliation or sponsorship for precisely the reasons Judge Brann identified in Vintage Brand. 44 We coded the open-ended responses generously. Responses were labeled "confused" if they explicitly referenced Notre Dame. This includes common misspellings, such as "Noter Dame," "Nota Dame," and "Notre Dam." We also counted as confused more general phrases that clearly alluded to the university whose logo was on the T-shirt even if Notre Dame was not explicitly mentioned in the response. This includes phrases such as "the university," "the school," "a college," "college team," "college sports team," and so on.

*p. 6*
After answering the survey questions about the source, affiliation, and sponsor of the T-shirt, participants were asked demographic questions and questions about their level of familiarity with the University of Notre Dame and how regularly they follow college sports. To measure our online participants' engagement, we concluded with an attention check that was disguised as a question about T-shirt size. 45

## | Participant Characteristics

*p. 7*
Tables 1 and 2 report characteristics of study participants by group assignment. Differences between the groups are generally small and not statistically significant, suggesting that the randomizations were successful. Overall, Study 1 participants were in their mid-thirties on average, and 48% identified themselves as female. Sixty-eight percent reported their race/ ethnicity as White, 11% as Hispanic or Latino, 11% as Asian, and 8% as Black or African American. In terms of education, about 40% of participants said they have a high school degree or less, 44% a college degree, and 15% a graduate or professional degree. Only 12% of participants said they are not familiar with the University of Notre Dame, while 64% said they are slightly or moderately familiar and 24% said they are very or extremely familiar. 46 Moreover, 40% of respondents said they do not follow college sports, while 27% said they follow college sports a little, 24% a moderate amount, and 8% a lot. An impressive 88% of Study 1 participants passed the attention check question. Participants who did not pass the attention check still appear to have completed the survey in good faith, providing sensical, on-topic responses to the survey's many open-ended questions.

*p. 7*
Study 2 participants, recruited six months later, were in their late thirties on average, and 55% reported that they are female. Seventy percent of Study 2 participants said they identify as White, 10% as Black or African American, 9% as Asian, and 9% as Hispanic or Latino. Thirty-two percent reported that they have a high school degree or less, 50% a college degree, and 18% a graduate or professional degree. As in Study 1, almost all Study 2 participants said they have at least some familiarity with the University of Notre Dame, and about two-thirds follow college sports at least a little. Eighty-four percent of Study 2 participants passed the attention check. Finally, Study 2 included an additional question that was used to screen out any potential participants who said they never purchase clothing online. Forty-nine percent said they buy clothing online once every few months,

## | Study 1 Results

*p. 8*
We used logistic regression to compare the causal effects of multiple treatments. We specified the following model:

*p. 8*
where Y i is a binary variable that identifies whether participant i was confused and β 1 and β 2 represent the causal effects of participant i's exposure to either the disclaimer alone or to the disclaimer and the typing exercise relative to the baseline condition of no disclaimer. X i is a vector of covariates that adjust for participant i's age, gender, education, interest in college sports, and self-reported familiarity with the University of Notre Dame. 47 We measured the dependent variable in three ways. First, we counted participants as confused if they mentioned Notre Dame in response to at least one of the three survey questions. This is our primary measure of confusion. Second, we counted participants as confused only if they said that Notre Dame puts out the T-shirt. This outcome measures source confusion. Third, we counted participants as confused only if they said Notre Dame The coefficients in Table 3 indicate that seeing the disclaimer on the product page and then being asked to type it out greatly reduced the odds that participant i was confused. The effect was large and statistically significant in all models. The disclaimer alone also reduced the odds of confusion relative to no disclaimer, but the effect was not nearly as large and was confined to confusion related to affiliation or sponsorship, not source. Coefficients and standard errors for the treatment variables were similar regardless of whether we included covariates (Models 1-3) or excluded them (Models 4-6). Consistent with the standard practice in trademark studies, the results shown in Table 3 do not include the 35 participants who failed the attention check question.

*p. 10*
Coefficients and standard errors, however, were similar when we reran the models with all participants included (see Table A1). 48 The stability of the estimates suggests that the treatment effects are robust and not significantly influenced by potential confounding variables.

*p. 10*
Figure 3 reports probabilities of confusion for each group. The first panel shows that participants who saw the product page with no disclaimer, the baseline condition, had a 71% probability of confusion. Participants who saw the disclaimer had a much lower probability of confusion at 29%. The disclaimer thus played a substantively and statistically significant role in lowering the likelihood of confusion (p < 0.001). But while it reduced confusion, the disclaimer did not dispel it. Many courts, in fact, would find a 29% probability of confusion unacceptably high. 49 These results are thus consistent with Jacoby's position that disclaimers do not dispel confusion.

*p. 10*
Results for the Typing Exercise group show that the disclaimer's effectiveness was greatly enhanced when consumers were required to register their recognition of it. The probability of confusion among participants who retyped the disclaimer was 8%, which is significantly different from the 29% probability of confusion in the Disclaimer Only group (p < 0.001) and below the threshold that would typically trigger a finding of likelihood of confusion. The second panel of Figure 3 To summarize, our first study showed that participants who were not exposed to a disclaimer were overwhelmingly inclined to believe that Notre Dame was, at least to some degree, involved in the production of the T-shirt. Placing a fine-print disclaimer above the image of the T-shirt reduced confusion significantly, but the disclaimer did not dispel it. Requiring participants to register their recognition of the disclaimer by typing it out or copying and pasting it brought confusion down to negligible levels and almost eliminated source confusion.

## | Study 2 Results

*p. 10*
We used a similar specification in our analysis of data from Study 2, but with additional parameters for participants who were required to register their recognition of the disclaimer by clicking a box (β 3 ) and participants who were exposed to the shorter yet more prominent disclaimer (β 4 ). Additionally, the covariate adjusted models now include dummy variables that indicate the frequency with which participants said they purchase clothing online:

*p. 10*
Table 4 shows that all four treatments caused a statistically significant reduction in confusion relative to the baseline condition of no disclaimer, whether covariates were included (Models 1-3) or excluded (Models 4-6). 50 These effects hold for our general measure of confusion and the narrower measures of source confusion and confusion as to sponsorship and affiliation.

*p. 10*
Comparing the coefficients in Table 4 to those in used in Model 1. 51 The magnitude of the coefficients in Table 4 furthermore indicates that the click box acknowledgment reduced the odds of confusion nearly as much as the typing exercise did, whereas the odds of confusion among participants who saw the bold disclaimer were higher than those who only saw the fine-print disclaimer.

*p. 12*
We could not meaningfully estimate the source confusion models for Study 2 (Model 2 and Model 5) because no participants in the Typing Exercise or Click Box groups who passed the attention check said they believe the T-shirt is put out by Notre Dame. The absence of any source confusion in these groups is compelling evidence that requiring recognition of disclaimers enhances their effectiveness.

*p. 12*
Figure 4 reports probabilities of confusion for each group. Participants who did not see any disclaimer had a 73% probability of confusion, while participants who only saw the standard, fine-print disclaimer had a 28% probability of confusion (p < 0.001). These are similar to the corresponding probabilities of 71% and 29% observed in Study 1. So, as we found in Study 1, the standard disclaimer reduced confusion from the baseline condition, but not below the threshold at which most courts would find for likelihood of confusion. The second and third panels of Figure 4 demonstrate that most of the confusion is related to sponsorship or affiliation. In this study, the standard disclaimer reduced the probability of source confusion to 5% from a probability of 22% in the baseline condition (p < 0.001), and it reduced the probability of confusion regarding sponsorship or affiliation to 22% from 56% in the baseline condition (p < 0.001).

*p. 12*
We again find that the effectiveness of the standard disclaimer was greatly enhanced when participants were required to register recognition of it. Participants assigned to see the disclaimer a second time and complete the typing exercise had a 3% probability of confusion, which is significantly different from the Disclaimer Only and No Disclaimer groups (p < 0.001). The typing exercise furthermore eliminated all confusion as to the source of the T-shirt. 52 We further observe that the click box acknowledgment reduced the probability of confusion to nearly the same level as the more burdensome and time-consuming typing exercise. Participants who saw the disclaimer a second time and registered their acknowledgment of it by clicking the box had a 9% probability of confusion, which is below the threshold that typically leads to a finding of likelihood of confusion and significantly different from the Disclaimer Only group (p = 0.005). The click box also, like the typing exercise, completely eliminated source confusion in Study 2. Requiring purchasers to acknowledge the disclaimer by clicking a box therefore appears to dispel confusion almost as effectively as requiring them to retype the disclaimer, but with only a fraction of the effort required on the part of the consumer. 53 Finally, we find that there is no advantage to using a briefer yet more prominently displayed disclaimer over a more informative fine-print disclaimer. Participants who saw the bold disclaimer had a 36% probability of confusion, which is an improvement on the 73% probability in the No Disclaimer group (p < 0.001), but higher than the 28% probability associated with the standard disclaimer (p = 0.24). This finding is consistent with studies referenced earlier, which report that disclaimers that are more prominently displayed are not necessarily effective in dispelling confusion because people still tend to ignore them. In this case, we suspect that the bold disclaimer was less effective than the fine-print disclaimer because it did not provide any information about the relationship between Notre Dame and Nostalgia Brands.

## | Discussion

*p. 12*
Our experiments showed that requiring consumers to register acknowledgment of a disclaimer may reduce confusion more effectively than merely showing them a disclaimer without requiring acknowledgment. Participants in Study 1 and Study 2 had high baseline probabilities of confusion at 71% and 73%, respectively. Although the standard, fine-print disclaimer reduced confusion significantly, the probability of confusion remained high at 29% and 28%. But the probability of confusion fell to 9% and below when participants saw the disclaimer again and had to affirm that they read it, either by retyping it or clicking an acknowledgment box. The click box acknowledgment reduced confusion nearly as well as the burdensome typing exercise. Our findings were consistent across two studies conducted at different points in time, when covariates were included or removed from the models, and when the analyses were run on data generated by all study participants or only the most attentive participants.

*p. 12*
There are multiple reasons why the acknowledgment tasks could have been so effective. First, the tasks exposed participants to the disclaimer for a second time. For some, this repetition may have reinforced the message conveyed by the fine-print disclaimer that appeared on the product page. Others likely did not notice the fine-print disclaimer, so the acknowledgment tasks provided a second opportunity to observe it. The tasks also isolated the disclaimer from the rest of the product information, perhaps making it more memorable.

*p. 13*
The acknowledgment tasks moreover required participants to engage with the disclaimer before they could proceed with the rest of the survey. By actively engaging with the disclaimer, participants may have been more likely to read it and remember it. This seems especially true of the typing exercise, but also of the click box exercise if the task caused participants to stop and think about the disclaimer before affirming that they read and understood it. Lastly, the acknowledgment tasks increased the relevance of the disclaimer to participants because they could not complete the survey unless they registered their recognition of it.

*p. 13*
Although the acknowledgment tasks had strong and consistent effects on confusion in our experiments, their impact may be less pronounced in the real world. First, our experiments exposed participants to a novel intervention at a single point in time. We did not investigate whether or to what extent disclaimer acknowledgment tasks are effective after multiple exposures over long periods of time. If they were to become common features of retail websites, it is possible that consumers would become inured to them and click them reflexively without reading.

*p. 13*
Second, although we designed the product website to appear realistic and instructed participants to imagine they are shopping for a new T-shirt, participants were aware that they were not making a real purchase. For some, this may have lowered the stakes, causing them to pay less attention to the disclaimers than they would in a real-world purchasing context. Other participants may have paid more attention to the disclaimers because they were being compensated to participate in a study that instructed them to look over a product page. It is difficult to know to what extent participants may have erred in either of these directions; however, we attempted to account for variability in consumer attention by also analyzing and reporting data generated by all participants, not only the most attentive. We further expect that random assignment distributed different modes of attention evenly across groups.

*p. 13*
Finally, so that all participants would see the product page in a similar way, we optimized it for viewing on mobile devices and required participants to take the survey on smartphones. An estimated 76% of online retail website visits in the United States are made on smartphones, while 23% are made on desktop and laptop computers and 1% on tablets (Salesforce Research 2024).

*p. 13*
Consumers may notice disclaimers differently on the larger, horizontal screens of desktops and laptops than on the smaller, vertical screens used by participants in our study. Although the vast majority of online shoppers experience retail websites on smartphones, we cannot be sure that our findings generalize to the share of online shopping that takes place on non-mobile devices.

## | Legal Implications

*p. 13*
Too much disclaimer case law has involved defendants using fine-print disclaimers in such a way that consumers could not be expected to notice them, let alone appreciate their meaning. 54 Our studies demonstrate that disclaimers can be effective in dispelling consumer confusion as to source, affiliation, or sponsorship when consumers actually see them and pay enough attention to them. We have shown that these twin goals can be accomplished by requiring consumers to register their recognition of the disclaimer at the point of sale with something as simple as a click box. In this section, we consider how enhancing the effectiveness of disclaimers may reverse the vicious circularity of trademark law. We then consider a common objection to disclaimers of affiliation, that they do not mitigate post-sale confusion or trademark dilution.

## | Reversing the Vicious Circularity of Trademark Law

*p. 14*
Enhancing the effectiveness of disclaimers opens the door to a wide array of reforms of trademark law and policy. The vicious circularity that Judge Brann identified in Vintage Brand has long shaped trademark law and driven it mainly toward ever broader rights over the past half century. Trademark scholars have just as long insisted, mostly unsuccessfully, that the law should not merely passively react to consumer perceptions but actively seek to shape those perceptions-by allowing conduct that consumers will eventually realize is in fact, or should be, allowed under the law (Dinwoodie 2023)

## | The Problems of Post-Sale Confusion and Trademark Dilution

*p. 14*
Even if requirements such as our typing exercise or click box may successfully dispel consumer confusion at the point of purchase, others who later encounter the purchaser using or displaying the product will typically not benefit from the disclaimer of affiliation. As courts and commentators have recognized, disclaimers are usually not effective against the phenomenon of "post-sale confusion." 56

*p. 14*
Disclaimers are also ineffective in preventing trademark dilution by blurring or tarnishment, which by definition affect consumers who are not confused (McCarthy 2025, § 24:72). 57 To the extent that a disclaimer emphasizes that two different entities are using the same mark, the disclaimer may worsen the blurring of the link between that mark and its senior user.

*p. 14*
Nor will a disclaimer mitigate tarnishment, which involves one entity damaging by association the reputation of another entity even when consumers know that there are two different entities.

*p. 14*
As with blurring, in informing consumers that two entities are commercially unaffiliated, a disclaimer of affiliation may nevertheless reinforce consumers' association between the two entities and their reputations.

*p. 14*
Post-sale confusion and trademark dilution are trump cards that trademark owners try to play when they cannot prove point of sale confusion. 58 Both theories of harm are highly controversial in trademark commentary, largely because they grant an enormous expansion of exclusive rights in speech to prevent a harm that is rarely made clear (Tushnet 2008, 527-42). We do not seek here to engage the full debates on either theory of harm. It may be enough to point out, however, that to the extent disclaimers of affiliation, even when effective, will not mitigate either form of harm, that may be a strike not against disclaimers, but against the two theories of harm. Courts typically justify prohibitions against post-sale confusion on the ground not that they will protect consumers from confusion, but rather that they will preserve the rarity of the plaintiff's trademark. 59 Meanwhile, prohibitions against trademark dilution are, by definition, prohibitions on non-confusing, non-deceptive speech. No matter how effective disclaimers of affiliation are in informing consumers of the truth, that will usually not protect a defendant from liability for dilution and for most forms of post-sale confusion. This may be as good an indication as any that neither basis for an outright injunction of speech is consistent with the Free Speech Clause of the First Amendment.

## | Conclusion

*p. 14*
Legal scholars and the courts have long viewed disclaimers as an ineffective tool for dispelling trademark confusion. Our experiments showed, however, that simply requiring consumers to register acknowledgment of a disclaimer before completing their purchase can reduce confusion to negligible levels. In a preliminary test of this hypothesis, we asked participants to acknowledge a disclaimer by retyping it. This task reduced the overall probability of confusion to 8%, down from a probability of 71% among participants who did not see any disclaimer. The retyping task also nearly eliminated any confusion as to the source of the product. We furthermore found that a far less burdensome approach-requiring participants to acknowledge that they read the disclaimer by clicking a box-had nearly the same impact as requiring them to retype the disclaimer. Participants who were asked to click the box to acknowledge the disclaimer had only a 9% probability of confusion overall and experienced no confusion as to the source of the T-shirt.

*p. 15*
To a large degree, our other results confirmed lessons from previous studies. On their own, we found that disclaimers of affiliation are not sufficiently effective in mitigating trademark confusion, at least when the baseline level of confusion is as high as it was in our studies. Participants who saw only the standard disclaimer, for example, had probabilities of confusion as high as 29%. The standard disclaimer therefore reduced confusion relative to a condition of no disclaimer, but not enough to preclude a finding of likelihood of confusion in most courts. A more prominent but less informative disclaimer was even less effective, with a 36% probability of confusion.

*p. 15*
Societal commitments to free speech and free market competition can often be forgotten in the hurly-burly of frontline trademark litigation. Also often forgotten is the basic First Amendment doctrine that when government regulates commercial speech, it must do so by means of laws that are "narrowly drawn" 60 and "extend only as far as the interest [they] serve" (Farley and Ramsey 2023). When effective, disclaimers of affiliation support these societal commitments and calibrate the constitutional test. They also promote both speech and competition. They are a means of educating consumers not to be confused and of reversing trademark law's vicious circularity. For decades, courts have habitually assumed that disclaimers are ineffective, and so we have not had to come to terms with the constitutional and policy issues that they raise. Our hope is that evidence that requiring consumers to register recognition of disclaimers can in fact be effective in dispelling confusion will restore these issues to the prominence they deserve. 40 Cf. Ayres (2012, 2076) (discussing as an example of a "thoughtrequiring altering rule" for displacing a contractual default the eighteenth-century requirement in French admiralty law that, to be enforceable, certain contractual provisions departing from the Ordinance de la Marine or the common law be written out by hand).

*p. 16*
41 Study 1 was administered on September 19, 2023. Study 2 was administered on April 2, 2024. 42 See Union Carbide Corp. v. Ever-Ready Inc., 531 F.2d 366 (7th Cir. 1976). 43 See Beebe and Hemphill (2017, 1348) ("There is a threshold proportion of confusion above which a defendant will be found liable for trademark infringement. Cases involving survey evidence suggest that this threshold is typically set at 20% to 30%."). See also McCarthy ( § 32:188) ("In the author's view, survey confusion numbers that go below 20% need to be carefully viewed against the background of other evidence weighing for and against a conclusion of likely confusion."). 44 See supra notes 2-4 and accompanying text. 45 The attention check question appeared to ask participants to select their T-shirt size from a list of common sizes. Participants who were reading carefully noticed buried in the question text that they were to select the "Other" option and enter a unique, randomly generated four-digit number in the text entry box. Participants who entered the correct four-digit number passed the attention check.

*p. 16*
46 Such a high degree of familiarity is not surprising given that Notre Dame is consistently among the best-selling brands in college sports (Smith 2013).

*p. 16*
47 Study 1 and Study 2 results were similar whether we included or excluded a variable for participant i's race or ethnicity. 48 Given that consumers pay varying levels of attention to product information, including less attentive participants in the analysis may provide a more practical test of the real-world impact of the disclaimer and disclaimer acknowledgment tasks. Such estimates are also based on the original random assignment and are the unbiased intent-to-treat estimates (Aronow et al. 2019; Montgomery et al. 2018; Varaine 2023). 49 See supra note 43 and accompanying text. 50 The analyses shown in Table 4 do not include the 82 participants who failed the attention check. Results, however, were largely the same when we reran the models on all participants (see Table A2). One difference is that the typing exercise had a somewhat more pronounced effect in this study when less attentive participants were excluded.

*p. 16*
51 Similarities between Study 1 and Study 2 are reassuring given justifiable concerns in the social sciences about replication of experimental findings (Korbmacher et al. 2023). 52 The effect of the typing exercise was less pronounced in analyses that included Study 2 participants who failed the attention check (Table A2). This group had an 8% probability of confusion overall and a 2% probability of source confusion. 53 The nonoverlapping confidence intervals in Figure 4 suggest that the typing exercise and click box affected confusion differently (p = 0.07).

*p. 16*
The effects of the typing exercise and the click box, however, were nearly identical in analyses that included participants who failed the attention check. In those models, shown in Table A2, participants in the Typing Exercise group had an 8% probability of confusion compared to 10% among participants in the Click Box group (p = 0.40). 54 See supra note 34. 55 Inc., 457 F.3d 1062, 1077-78 (9th Cir. 2006) ("Shorn of their disclaimercovered packaging, Auto Gold's products display no indication visible to the general public that the items are not associated with Audi or Volkswagen. The disclaimers do nothing to dispel post-purchase confusion.").

*p. 16*
57 "A given unauthorized use by defendant can cause confusion in some people's minds and in other people's minds cause dilution by blurring, but in no one person's mind can both perceptions occur at the same time." 58 See, for example, the recent case of adidas Am. Inc. v. Thom Browne Inc., No. 23-166, 2024 WL 1953594 (2d Cir. May 3, 2024), in which the plaintiff argued initial-interest and post-sale confusion. The jury was unpersuaded and found no likelihood of confusion.

## Footnotes

> 17401461, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jels.70029 by New York University, Wiley Online Library on [18/03/2026]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License

> Note: Standard errors in parentheses. ***p < 0.001, **p < 0.01, *p < 0.05.

> , we observe that the log-odds of confusion among Study 2 participants in the Disclaimer Only and Typing Exercise groups are similar to those in Study 1, at least for the general measure of confusionlogit P Y i = 1 = 0 + 1 Disclaimer Only i + 2 Typing Exercise i + 3 Click Box i + 4 Bold Disclaimer i + X i + iFIGURE 3 | Study 1: probability of confusion, with 95% confidence intervals, by group and type of confusion. [Color figure can be viewed at wileyonlinelibrary.com]

> 56 See, for example, Au-Tomotive Gold Inc. v. Volkswagen of America

> Note: Standard errors in parentheses. ***p < 0.001, **p < 0.01, *p < 0.05.17401461