"1st Choice of Doctors", in a blue ribbon on a product's packaging, conveys the message that more physicians prefer this product than any of its rivals. Does (must?) this phrase mean something more--for example, that a majority of all physicians prefer the product, or that the preference is strong or based on particular grounds? The phrase appears on the packaging of Similac, an infant formula made by the Ross Pediatrics division of Abbott Laboratories. More than a score of surveys show that pediatricians prefer Similac over Enfamil, the second-place formula (made by Mead Johnson), with all othercompetitors far behind. Many of these surveys show that Similac attracts majority support; most show that two physicians prefer Similac for every one who chooses Enfamil. But the district court nonetheless held that "1st Choice of Doctors" violates sec.43(a)(1) of the Lanham Act, 15 U.S.C. sec.1125(a)(1), because it implies to consumers that a majority of physicians strongly prefer the product for strictly professional reasons. 41 F. Supp. 2d 879 (S.D. Ind. 1999). All of the surveys that show majority support are inadequate, the judge concluded, because they were designed to elicit either weak preferences or those based on grounds other than medical judgment about quality. Other surveys, designed to eliminate slight or non-medical preferences, show that Similac enjoys only plurality support among physicians. A regular 2-to-1 margin is not enough to permit Abbott to make the "1st Choice" claim, the court held, and issued a preliminary injunction. In politics this would be a landslide: Bill Clinton was the "1st Choice of Voters" at the 1992 and 1996 Presidential elections even though he received less than half of the popular vote (43% in 1992, 49% in 1996). But in marketing, according to the district court, a product must have majority support to be "first."

In English, "first" is ordinal. It denotes rank in a series. A runner who crosses the finish line ahead of all others is "first" even if the race is slow and ends in a photo finish. A TV series ranks first in its time slot if it has a larger audience than any other series, even though there are so many networks, independent stations, and cable channels that no sitcom or drama attracts an absolute majority of viewers. A political candidate who receives more votes than the next- most-popular candidate finishes first, and for most offices a first-place finish is enough for election. Similac therefore is the "1st Choice of Doctors" according to ordinary usage. Perhaps a truthful claim of this kind could be misleading, and therefore actionable under sec.43(a)(1), if both absolute and relative levels of preference were small. Suppose 1.1% of pediatricians preferred Similac, 1% preferred Enfamil, 0.9% preferred some other formula, and 97% thought that all of the infant formulas were functionally identical. But absolute and relative preferences for Similac are substantial. Even if, like the district court, we throw out the surveys finding that a majority of medical professionals recommend Similac, the remaining surveys find that between 25% and 48% of those questioned rank Similac first, while Enfamil is the preference of between 10% and 40% of the respondents and never beats Similac. Surveys designed to elicit weaker preferences show that Similac receives between 51% and 64% support and Enfamil from 29% to 37%, so the roughly two-to-one ratio is not sensitive to methodology. Pediatricians may believe (as Abbott contends) that Similac is better tolerated by infants (i.e., less likely to induce unpleasant side effects such as gas, fussiness, and loose stools) and therefore is better in practice, even though clinical tests do not find nutritional differences and experts agree that both products are of high quality. No matter. When the absolute level of preference for the leading product is high, and the difference in support from the medical profession substantial, it is all but impossible to call the claim of "first choice" misleading.

Unless the meaning of language is itself a question of fact, to be determined by survey evidence. And this is what the district court concluded after a three-day hearing on the request for interlocutory relief. Mead Johnson commissioned a survey, which was administered at malls in 16 cities across the United States to women who either had recently purchased infant formula or contemplated doing so during the next year. The district court's opinion describes this survey in detail, 41 F. Supp. 2d at 887-92, so we summarize. Interviewers asked two sequences of questions. The first was open-ended. After identifying "1st Choice of Doctors" as the subject, interviewers asked consumers "what you understand this part of the labelto communicate to you." Most of the answers were treated as variations on "higher quality," if not medical superiority. The second sequence opened with the query: "in order for [the "1st Choice"] statement to be true, what percent of those doctors would have had to say that this product was their first choice?" More than 80% of the participants replied that at least a majority would have to prefer the product. The district court concluded from this survey that the claim "1st Choice of Doctors" conveyed to consumers the message that at least a majority of physicians prefer Similac on grounds of qualitative superiority. Then the judge scrutinized the surveys of physicians to see whether Abbott's claim, thus understood, had been substantiated. The judge found not, because surveys finding that a majority of surveyed physicians prefer Similac were not designed to ensure that the preference was based on Similac's superiority, while surveys that limited the grounds of preference did not show majority support for Similac.

Abbott insists that because it has been applying the phrase "1st Choice of Doctors" (or a variant) to Similac for a decade, Mead Johnson's suit is barred by laches. Yet because a product's promotion must be true and non- misleading at the time of sale, the packaging's history cannot preclude a challenge. Suppose that when Abbott adopted the phrase physicians preferred Similac, but by the time suit was filed a decade later Enfamil had gained the lead in surveys of the medical profession. Similac then could not be labeled as physicians' first choice. Laches would supply a defense if nothing had changed over time (and if the other elements of that defense, such as prejudice, obtained). Conopco, Inc. v. Campbell Soup Co., 95 F.3d 187, 191-94 (2d Cir. 1996). Moreover, long use of a phrase justifies placing on the challenger a burden to establish that something has changed, so that what was once true is now false (or at least misleading). The district court did not find that critical numbers have changed in the last decade, so perhaps Abbott has a sound laches defense, but we need not say given the weakness of Mead Johnson's position on the record assembled so far. If on remand Mead Johnson decides to press forward with more evidence in pursuit of permanent relief, then the district judge will have to decide whether it has demonstrated a change in the truthfulness of Abbott's promotion.

Everything in the district court's analysis depends on the survey of consumers about their understanding of the phrase "1st Choice of Doctors." It is a problematic exercise, for the survey assumes that "first" is a cardinal number- -that is, a count such as "246" or "55" or a ratio of two such numbers, rather than a place in a series. Only if "first" is a ratio of cardinal numbers does it make sense to ask whether its meaning is 90% or 51% or a plurality. Respondents in survey research are suggestible; the form of a question implies an answer, or at least the range of proper answers, and this survey ensured that the answers would be numbers rather than places in a series. Having told the respondents to treat "first" as cardinal, the survey was bound to produce a misleading if not meaningless answer. "First" does not mean 51%, or 90%, or any other ratio, so it is not surprising that the responses were all over the lot. The survey's opening set of questions is less troublesome, but it is hard to get from there to a conclusion that surveys of physicians must attempt to limit the grounds of choice. Many a medical choice depends on ease of use rather than therapeutic effect; think of the enduring debate between champions of oral versus injected polio vaccine. If consumers took away from "1st Choice of Doctors" the implication that Similac is a high-quality product, as the author of the survey concluded, then they were not deceived (this Mead Johnson concedes); we doubt that it is proper to draw more from this survey.

There is a deeper problem: the use of a survey in the first place. Surveys areaccepted ways to probe for things such as confusion about the source of goods, for confusion depends on the effect of a phrase or trade dress on the consumer. E.g., Reed-Union Corp. v. Turtle Wax, Inc., 77 F.3d 909, 912 (7th Cir. 1996); Indianapolis Colts, Inc. v. Metropolitan Baltimore Football Club L.P., 34 F.3d 410, 415-16 (7th Cir. 1994). So far as we can tell, however, never before has survey research been used to determine the meaning of words, or to set the standard to which objectively verifiable claims must be held. Dictionaries themselves are a form of survey; lexicographers determine how words have been used in both scholarly and popular texts. But philologists and others who contribute to dictionaries devote their lives to discovering usage and interpreting nuance. It would be a bad idea to replace the work of these professionals with the first impressions of people on the street, especially because consumers' sketchy understanding of science means that survey results are apt to present firms with unrealistic demands for verification.

Suppose a tube of toothpaste bears this phrase in large type: "Sodium Fluoride Anticavity Toothpaste" immediately above a box with the words "ADA Accepted". These claims could be verified by showing that the American Dental Association had authorized the use of its name and that the dentifrice had "anticavity" effects. To a dentist, the word "anticavity" means that the toothpaste reduces the number of cavities compared with some benchmark (perhaps toothpaste without fluoride, perhaps no toothpaste at all) by a statistically significant amount. Perhaps a group of 1,000 persons using the toothpaste with fluoride for two years had 1,000 cavities, while a control group suffered 1,200 cavities. If the difference satisfies normal tests of significance--meaning that the difference is replicable, rather than the effect of chance-- then the claim is true and properly may be made to distinguish this toothpaste from others that are less effective at controlling cavities. It is valuable information to consumers. Imagine, however, a survey administered to shoppers in the toothpaste aisles of drugstores. First the questioner asks consumers "what you understand 'ADA Accepted' to communicate to you." Many are likely to respond that this means that the toothpaste is the best on the market, solves (or at least addresses) all dental problems, or some variation--even though the ADA's rules permit multiple products in the same classification to be "Accepted" if each is medically useful, and the phrase "ADA Accepted" may be applied to a toothpaste that does not have any therapeutic effect against gum disease. See American Dental Association Acceptance Program Guidelines for Fluoride-Containing Dentifrices (May 1998). Perhaps the survey would continue with a question about what "anticavity" means to the shoppers: does it prevent, say, 90% of all cavities?, a majority of cavities?, and so on. Suppose most of those surveyed thought that "anticavity" means that its use will cut cavities in half. Combining that survey result with the district court's approach to verification would mean that the product could not carry the words "anticavity" or "ADA Accepted" even though these representations help shoppers distinguish toothpaste according to effects that they value. Consumers can't be made better off by removing these words from the product, because then buyers who understand their significance will be deprived of information. The market share of the toothpastes that are most effective in reducing cavities will fall, and the number of cavities will rise.

Section 43(a)(1) forbids misleading as well as false claims, but interpreting "misleading" to include factual propositions that are susceptible to misunderstanding would make consumers as a whole worse off by suppressing truthful statements that will help many of them find superior products. A "misunderstood" statement is not the same as one designed to mislead. Reducing ads and packaging to meaningless puffery can't be the objective of the Lanham Act--though it is a logical (and likely)outcome of Mead Johnson's approach, given the normal level of confusion and misunderstanding reflected in consumer surveys. Asked at oral argument whether a seller of aspirin could label that drug as an anti- inflammatory useful for arthritis (a medically established property of aspirin) if a survey showed that consumers confused palliation of symptoms with a cure for the disease, counsel for Mead Johnson replied that the claim of anti- inflammatory properties would be misleading for the same reason "1st Choice of Doctors" is misleading. This consequence of Mead Johnson's view is so counterproductive that the basic position cannot be accepted. We are not comforted by Mead Johnson's assurance that a seller could overcome consumer misunderstanding and make the claim about anti-inflammatory (or anticavity) benefits if it delivered additional details about the nature and extent of these effects. Requirements along the lines of a package insert with medical details are the province of regulations issued by the Food and Drug Administration, not of litigation under the Lanham Act. What is more, adding details could be so costly and burdensome that sellers might choose to omit all of the information. See Morales v. Trans World Airlines, Inc., 504 U.S. 374, 389-90 (1992); cf. Todd v. Societe BIC, S.A., 9 F.3d 1216, 1218-19 (7th Cir. 1993) (en banc) (observing that compendious advice is not always more useful to consumers). Anyway, if consumers did not read (or understand) the medical details they would be none the wiser, and on Mead Johnson's view the claim should be enjoined anyway.

By using Mead Johnson's survey to define the meaning of the phrase "1st Choice of Doctors" and then insisting that verification meet the standards thus established, the district court committed a legal error. That is reason enough to reverse the preliminary injunction, so we need not address most of the other subjects on which the parties have locked horns. But we cannot close without expressing concern about the level of the bond, which the district judge set at $1 million under Fed. R. Civ. P. 65(c):

No restraining order or preliminary injunction shall issue except upon the giving of security by the applicant, in such sum as the court deems proper, for the payment of such costs and damages as may be incurred or suffered by any party who is found to have been wrongfully enjoined or restrained.

The district court calculated Abbott's "costs and damages" solely by reference to the costs of printing new labels and promotional literature. But this suit is not about out-of-pocket costs (which Abbott estimates at $5.8 million); Mead Johnson and Abbott care about the issue principally because it affects the sales of their products. Mead Johnson conceded in the district court that the loss of even 1% of market share as a result of the injunction would cost Abbott about $10 million. (Abbott's estimate is $16 million per point of market share.) The principal reason the district judge gave for ignoring this effect (and for disregarding most of the out-of- pocket costs) was that Abbott should have quantified its losses earlier, and the only earlier opportunity was at the preliminary injunction hearing. Nothing in the record suggests, however, that the district judge alerted Abbott to this requirement in advance of the hearing. A litigant naturally would suppose (in the absence of notice) that a hearing on a request for a preliminary injunction will be devoted to the merits of that request, rather than to fixing the amount of bond. The judge also called Abbott's estimate "soft"--which it was, given the need to come up with some number quickly. That might have justified using Mead Johnson's smaller estimate; it did not justify ignoring a cost that was sure to be large, even if the total was hard to determine on short notice. We trust that in the future the district court will notify the parties of the ground rules and endeavor to set bonds at levels reflecting full consequences.

When setting the amount of security, district courts should err on the high side. If the district judge had set the bond at $50 million, as Abbott requested, this would not have entitled Abbott to that sum; Abbott still would have had to prove its loss, converting the "soft" numbers to hard ones. An error in setting the bond too high thus is not serious. (The fee for a solvent firm such as Mead Johnson or its parent Bristol- Myers Squibb Co. to post a bond, a standby letter of credit, or equivalent security is a very small fraction of the sum involved.) See generally Note, Recovery for Wrongful Interlocutory Injunctions Under Rule 65(c), 99 Harv. L. Rev. 828 (1986). Unfortunately, an error in the other direction produces irreparable injury, because the damages for an erroneous preliminary injunction cannot exceed the amount of the bond. W.R. Grace & Co. v. Rubber Workers, 461 U.S. 757, 770 n.14 (1983); Russell v. Farley, 105 U.S. 433, 437-38 (1882); Coyne-Delany Co. v. Capital Development Board, 717 F.2d 385, 393-94 (7th Cir. 1983). Abbott now must swallow substantial losses as a result of the district court's decision.

Trademark suits, like much other commercial litigation, often are characterized by firms' desire to heap costs on their rivals, imposing marketplace losses out of proportion to the legal merits. That's why bonds must reflect full costs. Shifting back to the plaintiff the complete injury occasioned by the errors that sometimes occur when preliminary relief is issued after an abridged judicial inquiry will hold in check the incentive business rivals have to pursue relief that gives them a competitive edge even if, as in this case, they lose in the end.