The Scotts Company sued Pursell Industries and United Industries (together, the "defendants"), claiming that the packaging of the defendants' "Vigoro" brand crabgrass-control product conveyed certain false messages to consumers. The district court granted a preliminary injunction in favor of Scotts. This court granted the defendants' motion for a stay of the injunction pending appeal and expedited the appeal. We now vacate the district court's grant of a preliminary injunction and remand for further proceedings on Scotts' request for a permanent injunction.

Generally speaking, chemical crabgrass-control products are most effective when applied to "pre-emergent" crabgrass — crabgrass that has yet to sprout. The products are also effective on early "post-emergent" crabgrass, but are ineffective when applied to mature crabgrass plants. Vigoro, the defendants' product, and Scotts' product ("Halts") offer some level of post-emergence control, if applied no later than about four weeks after germination. Scotts, however, contends that the Vigoro package misleads consumers into believing that Vigoro can kill mature crabgrass.

In 1999, Scotts filed a Lanham Act false advertising claim against Pursell based on the Vigoro package. At that time, the Vigoro package included what the parties refer to as the "crabgrass buster" logo — a cartoon of a crabgrass plant with a sinister face, surrounded by the familiar red circle with a slash through it. This graphic was located directly above a text block that stated (in all capital letters) "pre & post emergence control of crabgrass." The parties settled the 1999 action and entered into a settlement agreement under which Pursell admitted no liability but agreed not to use the crabgrass buster logo or a similar logo in close proximity to any claim that Vigoro controls or suppresses post-emergent crabgrass.

In March 2000, Scotts filed another action against Pursell. In that action, Scotts contended that Pursell breached the 1999 settlement agreement by including the crabgrass buster logo in its promotional materials. Scotts also asserted a new false advertising claim based on a time-line graphic included in those promotional materials which suggested that Vigoro was the only product with any post-emergence effect on crabgrass. The district court concluded that Pursell had breached the 1999 settlement agreement by using the crabgrass buster logo and that the time-line graphic was false because it wrongly suggested that Scotts' product had no post-emergence effect. The court therefore entered a preliminary injunction requiring Pursell to remove the offending logo and time-line from the promotional materials. The parties entered into another settlement agreement under which Pursell again admitted no liability.

In December 2001, United Industries purchased the Vigoro product line from Pursell. Pursuant to their agreement, Pursell continues to manufacture the product, and United distributes it exclusively to Home Depot stores.

For the 2002 spring selling season, the defendants introduced new packaging for the Vigoro products. A picture of lush green lawn occupies approximately a third of the bag, and the product name appears in the middle of that picture as follows (with some effort being made to approximate the relative font sizes):

ULTRA TURF PRE-EMERGENT Crabgrass Control plus Fertilizer 30-3-4

Centered on a bright blue background below this text and the lush-lawn picture is the following (again, with some attempt to show the relative font size):

Promotes Thick, Green Growth and Stops Crabgrass and Many Other Grassy & Broadleaf Weeds Before They Start.

To the left of this text block is the graphic at issue in this case: a realistic rendering of a mature crabgrass plant positioned directly above the following text:

Prevents Crabgrass up to 4 WEEKS After Germination

A small white cross by the crabgrass illustration leads to a disclaimer on the bottom of the front of the package, printed in dark blue or black ink and in very small type, which states, "Crabgrass image for illustration only as this product is for pre- and early post emergent control and does not control mature plants."

The crabgrass illustration on the Vigoro package is not unlike that used by Scotts on its Halts packages, and the Halts packages likewise claim some post-emergent effect. Scotts' post-emergent claim ("Delivers pre- and early post-emergent crabgrass control") is the first of three "bullet points" in a vertical column on the lower left side of the package front; the crabgrass illustration is placed just beside this column, in the middle of the lower part of the package. A disclaimer on the back of the Scotts' product states that "This product controls crabgrass in its early stages of growth. This product does not control mature crabgrass and other existing weeds."

Not satisfied by the revisions to the Vigoro packaging, Scotts filed this action in the Eastern District of Virginia, where both the 1999 and the 2000 actions had been filed. Scotts asserted false advertising claims under the Lanham Act, see 15 U.S.C.A. § 1125(a)(1)(B) (West 1998), claiming that the revamped packaging, by using the illustration of a mature crabgrass plant, falsely suggested that Vigoro could kill mature crabgrass.1

Scotts moved for a preliminary injunction and supported its motion with evidence of consumer confusion derived from two focus group discussions and face-to-face interviews with 40 shoppers selected from a shopping mall in Richmond. Although the district court questioned the persuasiveness of some of Scotts' evidence, the court concluded that Scotts had shown some evidence of consumer confusion, that Scotts had presented at least a triable issue on its false advertising claim, and that the balance of the harms favored Scotts. On July 2, 2002, the district court granted Scotts' request for a preliminary injunction and ordered the defendants to change the Vigoro packaging and remove existing product from store shelves or place a sticker covering the offending crabgrass graphic. The defendants thereafter sought a stay of the injunction from the district court. At the hearing on that motion, the district court stated that the question of whether to issue the injunction had been a very close one and that the question of staying the injunction was even closer. On July 15, 2002, however, the district court denied the motion to stay the injunction, although the court did modify the injunction so that it would take effect on August 30, 2002. The defendants appealed, and, as previously noted, we granted the defendants' motion to stay the injunction pending resolution of this expedited appeal.

In this circuit, the entry of a preliminary injunction is governed by the four-part test set forth in Blackwelder Furniture Co. of Statesville, Inc. v. Seilig Mfg. Co., Inc., 550 F.2d 189 (4th Cir.1977), which requires a court to consider "(1) the likelihood of irreparable harm to the plaintiff if the preliminary injunction is denied, (2) the likelihood of harm to the defendant if the requested relief is granted, (3) the likelihood that the plaintiff will succeed on the merits, and (4) the public interest." Direx Israel, Ltd. v. Breakthrough Med. Corp., 952 F.2d 802, 812 (4th Cir.1991) (internal quotation marks omitted); see Blackwelder, 550 F.2d at 193-95.

When deciding whether to grant a preliminary injunction, the court must first determine whether the plaintiff has made a strong showing of irreparable harm if the injunction is denied; if such a showing is made, the court must then balance the likelihood of harm to the plaintiff against the likelihood of harm to the defendant. See Safety-Kleen, Inc. (Pinewood) v. Wyche, 274 F.3d 846, 859 (4th Cir.2001); Direx, 952 F.2d at 812. If the balance of the hardships "tips decidedly in favor of the plaintiff," Rum Creek Coal Sales, Inc. v. Caperton, 926 F.2d 353, 359 (4th Cir.1991) (internal quotation marks omitted), then typically it will "be enough that the plaintiff has raised questions going to the merits so serious, substantial, difficult and doubtful, as to make them fair ground for litigation and thus for more deliberate investigation," Blackwelder, 550 F.2d at 195 (internal quotation marks omitted). But if the balance of hardships is substantially equal as between the plaintiff and defendant, then "the probability of success begins to assume real significance, and interim relief is more likely to require a clear showing of a likelihood of success." Direx, 952 F.2d at 808 (internal quotation marks omitted).2

"We review the grant or denial of a preliminary injunction for abuse of discretion, recognizing that preliminary injunctions are extraordinary remedies involving the exercise of very far-reaching power to be granted only sparingly and in limited circumstances." MicroStrategy Inc. v. Motorola, Inc., 245 F.3d 335, 339 (4th Cir.2001) (internal quotation marks omitted). This standard "is not a rule of perfunctory appellate review but one of careful scrutiny." Direx, 952 F.2d at 815.

We begin by considering whether the district court erred when concluding that Scotts would suffer irreparable harm if an injunction were not issued. In this case, however, that inquiry is bound up with an inquiry into the likelihood that Scotts will prevail on its false advertising claims. By virtue of certain intricacies in the law governing Lanham Act claims, some courts apply a presumption of irreparable harm if the plaintiff has made a threshold showing as to the merits of his Lanham Act claim, and the district court here relied on such a presumption when concluding that Scotts would be irreparably harmed if the injunction were denied. See J.A. 698 ("Scotts need only demonstrate that the advertising at issue confuses consumers in order to satisfy the irreparable harm prong of the analysis."). Thus, because the balance-of-the-hardship question is intertwined with questions about the merits, our analysis of the balance-of-the-hardship question will first require a detour into the substance of Scotts' Lanham Act claims.

The Lanham Act prohibits the "false or misleading description of fact, or false or misleading representation of fact, which ... in commercial advertising or promotion, misrepresents the nature, characteristics, qualities, or geographic origin of his or her or another person's goods, services, or commercial activities." 15 U.S.C.A. § 1125(a)(1)(B) (West 1998). Thus, a plaintiff asserting a false advertising claim under the Lanham Act must establish that:

(1) the defendant made a false or misleading description of fact or representation of fact in a commercial advertisement about his own or another's product; (2) the misrepresentation is material, in that it is likely to influence the purchasing decision; (3) the misrepresentation actually deceives or has the tendency to deceive a substantial segment of its audience; (4) the defendant placed the false or misleading statement in interstate commerce; and (5) the plaintiff has been or is likely to be injured as a result of the misrepresentation, either by direct diversion of sales or by a lessening of goodwill associated with its products.

Cashmere & Camel Hair Mfrs. Inst. v. Saks Fifth Ave., 284 F.3d 302, 310-11 (1st Cir.), cert. denied, ___ U.S. ___, 123 S. Ct. 485, 154 L. Ed. 2d 396 (2002); accord IQ Prods. Co. v. Pennzoil Prods. Co., 305 F.3d 368, 375 (5th Cir.2002); United Indus. Corp. v. Clorox Co., 140 F.3d 1175, 1180 (8th Cir.1998). The only elements in dispute in this case are the first and third.

For liability to arise under the false advertising provisions of the Lanham Act, "the contested statement or representation must be either false on its face or, although literally true, likely to mislead and to confuse consumers given the merchandising context." C.B. Fleet Co. v. SmithKline Beecham Consumer Healthcare, L.P., 131 F.3d 430, 434 (4th Cir.1997) (internal quotation marks omitted). "Where the advertisement is literally false, a violation may be established without evidence of consumer deception." Cashmere & Camel Hair Mfrs., 284 F.3d at 311; accord Novartis Consumer Health, Inc. v. Johnson & Johnson-Merck Consumer Pharm. Co., 290 F.3d 578, 586 (3d Cir.2002); C.B. Fleet, 131 F.3d at 434. But if "a plaintiff's theory of recovery is premised upon a claim of implied falsehood, a plaintiff must demonstrate, by extrinsic evidence, that the challenged [advertisements] tend to mislead or confuse consumers." Johnson & Johnson * Merck Consumer Pharm. Co. v. Smithkline Beecham Corp., 960 F.2d 294, 297 (2d Cir.1992); see Abbott-Labs. v. Mead Johnson & Co., 971 F.2d 6, 14 (7th Cir.1992) ("[A] court may find on its own that a statement is literally false, but, absent a literal falsehood, may find that a statement is impliedly misleading only if presented with evidence of actual consumer deception.").

In Lanham Act cases involving trademark infringement, a presumption of irreparable injury is generally applied once the plaintiff has demonstrated a likelihood of confusion, the key element in an infringement case. See, e.g., Eli Lilly & Co. v. Natural Answers, Inc., 233 F.3d 456, 469 (7th Cir.2000) ("Irreparable harm is generally presumed in cases of trademark infringement and dilution."); Genesee Brewing Co. v. Stroh Brewing Co., 124 F.3d 137, 142 (2nd Cir.1997) ("In the context of trademark and unfair competition injunctions, the requirement of irreparable harm carries no independent weight, as we have held that a showing of likelihood of confusion (a requirement of both trademark infringement and unfair competition claims) establishes irreparable harm."). A similar presumption also appears in many Lanham Act false advertising cases, although there seems to be some disagreement among the courts as to when the presumption should be applied. Compare, e.g., United Indus., 140 F.3d at 1183 (suggesting that the presumption applies in all Lanham Act cases where the plaintiff has established a tendency to deceive), with Ortho Pharm. Corp. v. Cosprophar, Inc., 32 F.3d 690, 696 (2nd Cir.1994) (explaining that the presumption of irreparable harm is generally limited to cases involving false comparative advertising).

This court has never spoken directly to the applicability of such a presumption of harm in Lanham Act cases.3 Not surprisingly, Scotts contends that we should endorse the use of the presumption in false advertising cases and that we should adopt the presumption in its broadest formulation. For their part, the defendants suggest that we should follow the approach of the Second Circuit and conclude that a presumption of irreparable harm applies, if at all, only to cases involving false comparative advertising. We need not, however, decide whether and under what circumstances a presumption of irreparable harm should be applied in false advertising cases. As we will explain below, Scotts has failed to make even a prima facie showing of consumer confusion, thus precluding application of the presumption, regardless of how it is formulated.

As previously noted, the evidence required of a false-advertising plaintiff is dependent upon whether the case involves advertising that is literally false or advertising that is only impliedly false. If the advertising is literally false, no evidence of consumer confusion is required. But if the advertising is impliedly false, the plaintiff must present extrinsic evidence of consumer confusion.

"In analyzing whether an advertisement... is literally false, a court must determine, first, the unambiguous claims made by the advertisement ..., and second, whether those claims are false." Novartis, 290 F.3d at 586. "A literally false message may be either explicit or conveyed by necessary implication when, considering the advertisement in its entirety, the audience would recognize the claim as readily as if it had been explicitly stated." Id. at 586-87 (internal quotation marks omitted).

Here, the only claim explicitly made on the Vigoro package is that the product prevents crabgrass up to four weeks after germination, and Scotts concedes that this message, standing alone, is true. Scotts, however, contends that the falsity springs not from the text standing alone, but from the juxtaposition of the mature crabgrass illustration directly above this text, which creates what Scotts describes as a "unitary graphic." According to Scotts, the placement of the mature crabgrass illustration so close to the phrase "prevents crabgrass up to 4 weeks after germination" necessarily leads consumers to believe that the picture shows what crabgrass looks like four weeks after germination. That message, in turn, leads consumers to believe that if they applied Vigoro to crabgrass that looked like the illustration, the product would kill the crabgrass, or at least prevent it from spreading. At four weeks after germination, however, crabgrass does not look like the illustration, but instead appears as small shoots. Thus, Scotts contends that the unitary graphic conveys the literally false message that Vigoro is capable of killing mature crabgrass.

The district court concluded that the Vigoro packaging made no literally false claims, a factual finding that is entitled to deference. See C.B. Fleet, 131 F.3d at 434 (explaining that literal falsity of an advertisement is a factual question); Fed.R.Civ.P. 52(a) (requiring district court to make specific findings of fact and conclusions of law when granting or denying preliminary injunctions and stating that such factual findings "shall not be set aside unless clearly erroneous"). Scotts does not even acknowledge the district court's factual finding, much less articulate why that finding is erroneous. Our own review of the record confirms the correctness of the district court's decision on this point.

Preliminarily, the sheer complexity of Scotts' explanation as to why this case involves a literal falsity seems to undercut the argument. See United Indus., 140 F.3d at 1181 ("The greater the degree to which a message relies upon the viewer or consumer to integrate its components and draw the apparent conclusion, however, the less likely it is that a finding of literal falsity will be supported."). Moreover, the illustration is teamed with a claim that Vigoro "prevents" crabgrass, a word that generally is used in the stopping-from-ever-starting sense. See, e.g., American Heritage College Dictionary 1085 (3d ed.1997) (giving "[t]o keep from happening" as the first definition of "prevent"). As the district court recognized, Vigoro can in fact prevent mature crabgrass by stopping the weed before it starts, which makes it difficult to see how the message conveyed by the Vigoro packaging can be literally false.

Scotts, however, contends that even though the stopping-crabgrass-before-it-starts message would be unobjectionable standing alone, the packaging conveys another message that is literally false. Scotts points out that while "prevent" is often used in the sense of stopping something from happening, it can also mean hindering, impeding, or eliminating something, or stopping something from existing. And according to Scotts, the defendants have exploited this secondary meaning of the word "[b]y juxtaposing a picture of mature crabgrass with the words `prevents after germination,'" which suggests that Vigoro "can stop, retard, and eliminate mature crabgrass once growth has occurred." Brief of Appellee at 25. That is, "[t]o say that the product is capable of preventing crabgrass after crabgrass has occurred is to say that the product eliminates crabgrass after the fact and after it has reached maturity." Brief of Appellee at 26.

We agree that "prevent" does carry the secondary message to which Scotts refers and that the Vigoro packaging conveys the message that Vigoro can stop or eliminate crabgrass once growth has occurred. But the packaging states that it can prevent (i.e., eliminate) crabgrass up to four weeks after it has emerged, a message which is true and which is a far cry from Scotts' assertion that the packaging claims that Vigoro eliminates crabgrass after maturity. Scotts' literal falsity argument is thus dependent on the claim that the placement of the crabgrass illustration so close to the "4 weeks after germination" text necessarily conveys to consumers that the illustration is of a crabgrass plant four weeks after germination. That argument, however, is untenable when the graphic is viewed in context, as it must be. See United Indus., 140 F.3d at 1180 (When "assessing whether an advertisement is literally false, a court must analyze the message conveyed within its full context.").

If the illustration were positioned over text claiming that Vigoro kills crabgrass, we might be more sympathetic to Scotts' claim. But the graphic is simply one part of a package that focuses primarily on controlling pre-emergent crabgrass. The most obvious part of the package is the lush-lawn photo upon which the product is referred to (in very large letters) as "Ultra Turf Pre-Emergent Crabgrass Control." And beside the crabgrass graphic, in the center of the bottom third of the package front, is the large-letter statement that Vigoro "Stops Crabgrass and Many Other Grassy & Broadleaf Weeds Before They Start." Under these circumstances, we simply cannot agree with Scotts that the meaning it attaches to the unitary graphic is the one necessarily conveyed by the graphic. While it is possible that some consumers would interpret the graphic in the manner Scotts suggests, it is likewise possible that consumers would understand the illustration as showing mature crabgrass for identification purposes only, so that they can determine whether the weed that is vexing them is crabgrass or some other backyard pest. Because the graphic can reasonably be understood as conveying different messages, Scotts' literal falsity argument must fail. See Novartis, 290 F.3d at 587 ("[O]nly an unambiguous message can be literally false."); United Indus., 140 F.3d at 1181 ("Commercial claims that are implicit, attenuated, or merely suggestive usually cannot fairly be characterized as literally false.").4

Because we cannot conclude that the Vigoro packaging makes any literally false claims, the question is whether Scotts can establish that the packaging makes any impliedly false claims. See Clorox Co. Puerto Rico v. Proctor & Gamble Commercial Co., 228 F.3d 24, 35 (1st Cir.2000) ("[A] factfinder might conclude that the message conveyed by a particular advertisement remains so balanced between several plausible meanings that the claim made by the advertisement is too uncertain to serve as the basis for a literal falsity claim, though even in that case it could still form the basis of a claim that the advertising is misleading."). As discussed above, a plaintiff asserting an implied falsity claim generally must present extrinsic evidence of consumer confusion. See id. at 36 ("Unlike the requirements of a claim of literal falsity, the plaintiff alleging a misleading advertisement has the burden of proving that a substantial portion of the audience for that advertisement was actually misled."); Johnson & Johnson * Merck, 960 F.2d at 297 (explaining that if "a plaintiff's theory of recovery is premised upon a claim of implied falsehood, a plaintiff must demonstrate, by extrinsic evidence, that the challenged [advertisements] tend to mislead or confuse consumers").

Consumer confusion "is most often proved by consumer survey data," Clorox Co., 228 F.3d at 36, but "full-blown consumer surveys ... are not an absolute prerequisite" at the preliminary injunction stage, United Indus., 140 F.3d at 1183. Nevertheless, "the movant — even at the preliminary injunction stage — must present evidence of deception." Johnson & Johnson Vision Care, Inc. v. 1-800 Contacts, Inc., 299 F.3d 1242, 1247 (11th Cir.2002). Accordingly, we now turn to the defendants' challenges to Scotts' evidence of consumer confusion.

To show consumer confusion, Scotts presented evidence of two focus group discussions and the results of a survey of forty consumers. We consider each type of evidence separately.

Scotts' focus groups consisted of a small number of Richmond-area consumers who owned their homes and performed their own lawn care. A moderator led each of the group discussions. Dr. Kiecker, Scotts' marketing expert, summarized the results of the focus group sessions and concluded that "[m]ost participants felt that [the] purpose of the image of the plant on the front of the Vigoro bag was to show consumers the type of crabgrass the product would prevent." J.A. 102. Although none of the participants noticed the disclaimer on his own, Kiecker stated that once the disclaimer was pointed out to them, the participants believed that the purpose of the disclaimer was to "reverse the impression created by the plant image." J.A. 103.

Scotts asserts that this focus group evidence satisfies its burden of showing consumer confusion through extrinsic evidence. The defendants, however, contend that focus group evidence is inherently unscientific and thus inadmissible under Rule 702 of the Federal Rules of Evidence or Daubert v. Merrell Dow Pharmaceuticals, Inc., 509 U.S. 579, 113 S. Ct. 2786, 125 L. Ed. 2d 469 (1993), as evidence of consumer confusion in a false advertising case. In the alternative, the defendants contend that even if focus group evidence is generally admissible, the focus group evidence in this case is insufficient to support Scotts' claims.

We note that the very nature of a focus group seems, to some extent, to limit its ability to identify the message an advertisement conveys to an individual consumer. As the defendants' expert explained, focus groups "are a form of brainstorming where the participants are encouraged to `build on' the thoughts of others." J.A. 208. Because the participants in a focus group freely voice their opinions, the opinion of a participant can be shaped by those of the others. Thus, a participant who may have derived no false message from an advertisement viewed outside the context of the focus group might well change his opinion about the message conveyed by the advertisement after considering the views expressed by the other participants. Nevertheless, we need not decide the broader question of the general admissibility of focus group evidence, because we agree with the defendants that, in this case, the focus group evidence was unreliable and cannot be considered probative on the question of whether the Vigoro packaging was likely to mislead consumers.

In our view, the manner in which the focus group discussions were conducted prevents the results of those discussions from being considered as a reliable indicator of whether the Vigoro packaging conveyed misleading messages to consumers. This lack of reliability stems from the fact that the moderators of the groups channeled the discussions and led the participants into giving responses favorable to Scotts. See Novartis, 290 F.3d at 591 ("A survey is not credible if it relies on leading questions which are inherently suggestive and invite guessing by those who did not get any clear message at all." (internal quotation marks omitted)). The moderators asked highly leading questions, often ignored responses that were inconsistent with the view that the Vigoro conveyed the message that it could kill mature crabgrass, and typically explored in detail only the responses that were consistent with this hypothesis. For example, in one of the focus groups, a participant named John stated that he did not know what the purpose was of the crabgrass illustration on the Vigoro packaging. The moderator then asked, "if it says prevents crabgrass up to four weeks after germination do you think it will kill crabgrass that looks like that?" J.A. 613. Some discussion among the participants followed, none of which unambiguously supported Scotts' hypothesis. The moderator then asked if any participant knew what crabgrass looked like four weeks after germination, to which John responded, "I would assume it's just little shoots coming up." J.A. 619. The moderator responded, apparently to John and in reference to his earlier comment: "If you don't know the purpose, again, let's go back to the plant." J.A. 619. John immediately changed his tune and said, "This is what it would look like up to this point." J.A. 619.

We believe this example well illustrates the problems with the focus group evidence in this case. Initially, no message was conveyed to John by the crabgrass illustration on the Vigoro packaging. But in a span of time reflected by a mere six pages in the joint appendix, John had moved from having no opinion as to the purpose of or message conveyed by the illustration to being convinced that the illustration did indeed convey a particular and false message. Thus, the manner in which the focus groups were conducted allowed the moderator to shape the opinions of the participants to mirror Scotts' theory of the case and then report those shaped opinions as evidence to support Scotts' claim.

We therefore conclude that the manner in which the focus groups were conducted destroyed the objectivity of the discussions, rendering the results utterly unreliable on the question of whether the Vigoro packaging conveys a false message. See Novartis, 290 F.3d at 591 ("The evidentiary value of a survey depends on its underlying objectivity as determined through many factors, such as whether the survey is properly filtered to screen out those who got no message from the advertisement, whether the questions are directed to the real issues, and whether the questions are leading or suggestive." (internal quotation marks and alteration omitted)). We recognize, of course, that a district court's evidentiary rulings are entitled to deference. See, e.g., Supermarket of Marlinton, Inc. v. Meadow Gold Dairies, Inc., 71 F.3d 119, 126 (4th Cir.1995). But the deficiencies in the focus group evidence are so substantial that we are constrained to conclude that the district court abused its discretion by giving the evidence any weight. See, e.g., Direx, 952 F.2d at 815 (explaining that an appellate court may overturn the district court's factual findings if, after reviewing the entire record, the appellate court is "left with the definite and firm conviction that a mistake has been committed." (internal quotation marks omitted)).

Without the focus group evidence, Scotts' only evidence of consumer confusion comes from its survey of 40 do-it-yourself lawn care consumers. According to Scotts, the survey showed that 92.5% of the respondents were misled by the Vigoro packaging, a level of confusion that would be sufficient to support Scotts' false advertising claims. See, e.g., Johnson & Johnson-Merck Consumer Pharms. Co. v. Rhone-Poulenc Rorer Pharms., Inc., 19 F.3d 125, 129 (3d Cir.1994) (explaining that a false advertising plaintiff must show that "there is actual deception or at least a tendency to deceive a substantial portion of the intended audience"). The defendants, however, challenge the admissibility of the survey evidence, arguing that the survey is inadmissible because it does not satisfy the standards of reliability set forth by the Supreme Court in Daubert. The defendants also contend that if the survey is admissible, it does not establish any level of consumer confusion on the relevant question. Because we agree with the latter contention, we will not address the former.

As previously discussed, the express messages conveyed by the Vigoro packaging are literally true. Vigoro prevents crabgrass if it is applied up to four weeks after the crabgrass has emerged from the soil, a claim that Scotts does not dispute. And to the extent that Vigoro's use of the mature crabgrass plant illustration directly above the claim that Vigoro prevents crabgrass up to 4 weeks post-emergence sends the message that Vigoro prevents mature crabgrass, that message, as the district court recognized, is literally true. By preventing crabgrass from even starting, Vigoro does prevent mature crabgrass. The critical question in this case, then, is whether the Vigoro packaging conveys the message that Vigoro kills mature crabgrass, a message that would be false. The survey, however, simply did not address that question.

The interviewer showed the respondents an empty Vigoro bag folded in quarters so that the lower left quadrant (where the crabgrass illustration was located) was visible and asked: "Based on your review of this section of the bag, should this product prevent the growth of crabgrass that looks like the crabgrass pictured?" J.A. 119. Through the use of the word "prevent," this question suffers from the same ambiguity as does the Vigoro packaging itself. That is, we cannot tell from this question whether the respondents who answered yes believed that Vigoro could prevent mature crabgrass by stopping it before it started, or whether they believed that Vigoro could kill if applied to established, mature crabgrass.

Scotts, however, insists that the survey does address the critical issue. First, Scotts points out that the survey did not simply ask whether Vigoro would prevent crabgrass that looks like the illustration, but instead asked whether Vigoro would prevent "the growth" of crabgrass that looks like the illustration. According to Scotts, by focusing on preventing growth, the question "asks whether the product controls (e.g., stops) `the growth of' the pictured crabgrass. And that of course goes to one of the key issues in the case." Brief of Appellee at 41-42. We disagree.

As previously pointed out, the primary and most common meaning of "prevent" is "[t]o keep from happening." American Heritage College Dictionary 1085 (3d ed.1997). Contrary to Scotts' suggestion, the inclusion of the phrase "the growth of" does not obliterate or transform this meaning — Vigoro can keep the growth of mature crabgrass from happening by stopping crabgrass before it has a chance to mature. The ambiguity in the survey question comes from the use of "prevent," and that ambiguity simply is not cured by adding the phrase "the growth of." Therefore, even though more than 90% of the respondents answered yes when asked whether Vigoro should "prevent the growth of crabgrass that looks like the crabgrass pictured," J.A. 119, those responses shed no light on the question that is key to Scotts' false advertising claims — whether the Vigoro packaging conveys the message that it can kill mature crabgrass.

Scotts also argues that the responses to the follow-up "why or why not" question provide sufficient evidence of consumer confusion. Again we disagree. When asked to explain why they believed that Vigoro would prevent crabgrass that looks like the crabgrass pictured, many respondents said things like "because the bag says so," J.A. 121; "because that is crabgrass," J.A. 123; because "that is the type of [crabgrass] I am familiar with," J.A. 127; "because that is what it says it does," J.A. 137; "because it pictures that [crabgrass] so it should prevent it," J.A. 145; and "because that is [crabgrass] and this is a prevent[ative]," J.A. 155. These responses are just as ambiguous as the initial question, and, like the responses to the initial question, provide no evidence of consumer confusion on the critical issue.

There are, however, some explanations that do seem to support Scotts' view of the message conveyed by the Vigoro packaging. For example, one respondent explained that Vigoro should prevent the growth of crabgrass that looks like the illustration "because it says it prevents up to 4 [weeks] after germination which would look like this." J.A. 119. Other explanations included: "because that is c[rab]g[rass] (`pre-emergent')," J.A. 143; "says it prevents [crabgrass] up to 4 [weeks] and that looks like a picture of [crabgrass] up to 4 [weeks]," J.A. 157; "the picture shows what you are trying to kill," J.A. 161; and "photo use[d] should be representative," J.A. 175. But even some of these explanations, however, are as ambiguous as the initial question. For example, the explanation that "the picture shows what you are trying to kill" would be consistent with an understanding that the crabgrass illustration is for identification purposes only and that Vigoro prevents the pictured crabgrass by stopping it from starting. While a few of these explanations more unambiguously support Scotts' claim, we simply cannot conclude that the relatively small number of such explanations establish that a substantial number (or, a "not insubstantial number") of consumers are likely to be misled by the Vigoro packaging. See Rhone-Poulenc, 19 F.3d at 134 (noting that a Lanham Act plaintiff asserting an implied falsehood claim must establish that "the advertising tends to deceive or mislead a substantial portion of the intended audience" (internal quotation marks omitted)); Johnson & Johnson * Merck, 960 F.2d at 298 (explaining that a plaintiff raising a claim of implied falsity must establish "that a not insubstantial number of consumers" are likely to be confused or misled (internal quotation marks omitted)).

Moreover, the manner in which the survey was conducted creates significant questions about its relevance and reliability. The purpose of consumer surveys in false advertising cases is to determine the message actually conveyed to consumers. But the interviewers conducted the survey in this case in a way that effectively required the respondents to express a specific opinion, even if they did not have an opinion, by specifically not offering the respondents the opportunity to give "not sure" as a response. See J.A. 118-19. In addition, the survey respondents were given an empty Vigoro bag folded into quarters so that only the quadrant with the crabgrass illustration was visible, thus obscuring the parts of the Vigoro packaging that emphasized its focus on stopping crabgrass before it started, and the respondents were asked to review only that portion of the bag. The survey thus elicited information about the consumer's reaction to an isolated part of the packaging, when the relevant issue in a false advertising case is the consumer's reaction to the advertisement as a whole and in context. We believe that these deficiencies in the survey's design, particularly when considered in light of the survey's failure to focus on the critical question, weaken the relevance and credibility of the survey evidence to the point that it sheds no light on the critical question in this case. See Johnson & Johnson * Merck, 960 F.2d at 300 ("The evidentiary value of a survey's results rests upon the underlying objectivity of the survey itself. This objectivity, in turn, depends upon many factors, such as whether the survey is properly filtered to screen out those who got no message from the advertisement, whether the questions are directed to the real issues, and whether the questions are leading or suggestive." (internal quotation marks and alterations omitted)).

In its order granting the preliminary injunction, the district court expressed concerns about the survey evidence, noting that "the results of portions of the consumer studies are not entirely helpful or convincing." J.A. 700 n. 7. The court recognized that while 92.5% of the respondents

believed the product would "prevent" the crabgrass pictured in the graphic, this number does not necessarily demonstrate confusion. The product also indicates that it is a pre-emergent. As a result, if it is applied correctly, it will prevent mature crabgrass like the plant pictured on the package. The fact that the vast majority of respondents believe the product would "prevent" mature crabgrass, consequently, provides little insight.

J.A. 700-01 n. 7. Notwithstanding these concerns, the court relied on the survey evidence when concluding that Scotts had established a likelihood of confusion. As we have explained above, the survey evidence does not establish consumer confusion, primarily because it addresses issues that are not relevant to Scotts' false advertising claim and yet fails to address the single question that is relevant to Scotts' claims — whether Vigoro kills mature crabgrass. We therefore conclude that the district court abused its discretion by crediting the survey evidence.

Scotts, however, contends that any deficiencies in its consumer confusion evidence are irrelevant, because it established that the defendants intentionally set out to deceive consumers about Vigoro's effect on mature crabgrass. Although this court has not yet addressed the question, other circuits have held that, like a finding of literal falsity, a conclusion that a defendant intended to deceive triggers a presumption of consumer confusion that relieves a Lanham Act plaintiff of any obligation to present evidence of likely confusion. See Cashmere & Camel Hair Mfrs., 284 F.3d at 316 ("It is well established that if there is proof that a defendant intentionally set out to deceive or mislead consumers, a presumption arises that customers in fact have been deceived."); William H. Morris Co. v. Group W, Inc., 66 F.3d 255, 258 (9th Cir.1995) ("If [the defendant] intentionally misled consumers, we would presume consumers were in fact deceived and [the defendant] would have the burden of demonstrating otherwise."); Johnson & Johnson * Merck, 960 F.2d at 298 ("[W]here a plaintiff adequately demonstrates that a defendant has intentionally set out to deceive the public, and the defendant's deliberate conduct in this regard is of an egregious nature, a presumption arises that consumers are, in fact, being deceived." (internal quotation marks omitted)). Scotts contends this intent to deceive is evidenced by the defendants' "pattern and practice of misleading consumers," Brief of Appellee at 20, and by the inclusion of the disclaimer on the Vigoro packaging.

The disclaimer, which is linked to the graphic by a faint cross symbol, states that "crabgrass image for illustration only as this product is for pre- and early post emergent control and does not control mature plants." Scotts argues that "[i]f the graphic conveyed a truthful message, there would be no reason to `tell' consumers in a nearly-invisible footnote that the mature crabgrass depicted ... is shown for `illustration only.' ... At the very least, the disclaimer ... reveals a significant awareness of a risk that the current graphic conveys a misleading message." Brief of Appellee at 30-31. Scotts contends that counsel for United effectively admitted the intent to deceive at the hearing when he stated to the district court that the disclaimer was intended, in part, to forestall a legal challenge by Scotts. See J.A. at 457 ("The factual point of the disclaimer was trying to keep Scotts from suing us again."). With regard to the defendants' "pattern and practice of misleading consumers," Scotts states that the "Defendants have appeared before the same judge three separate times for using three strikingly similar versions of the same misleading graphic to deceive consumers into believing that Vigoro products can control mature crabgrass." Brief of Appellee at 20. Scotts contends that "[o]n this record of past deception, it would have defied common sense for [the district court] to have insisted upon Scotts providing mounds of extrinsic evidence to support its motion." Brief of Appellee at 22.5

A defendant's intent is a "quintessentially factual question." United States v. McKie, 951 F.2d 399, 403 (D.C.Cir.1991); see Warren v. Halstead Indus., Inc., 802 F.2d 746, 751 (4th Cir.1986) ("Issues of intent are included in the factual matters for the trier of fact to discern."). The district court, however, did not conclude that the defendants intended to deceive consumers. The court was, of course, aware of the prior actions brought by Scotts, and the court did consider the effect of the disclaimer when concluding that the packaging contained an implied falsehood. But nothing in the court's opinion suggests that the court concluded as a factual matter that the defendants' prior history, the disclaimer, or any other evidence demonstrated that the defendants intended to deceive consumers, and it is not the function of this court to supplement the factual findings made by the district court. Moreover, while there is some evidence at this point that could perhaps support a finding that the defendants intended to deceive consumers, the evidence is limited and far from overwhelming.

As to the disclaimer, we note that the mere fact that the defendants recognized the possibility that Scotts might sue does not inexorably lead to the conclusion that the defendants believed that the graphic was misleading and intended the graphic to deceive consumers. See J.A. at 457 (counsel for United explaining that "[s]ometimes, though, even when you haven't done anything wrong, you go ahead and you do a little bit more to try to prevent somebody from complaining"). While we do not rule out the possibility that a factfinder might attach some significance to a particular disclaimer and the circumstances surrounding its implementation when concluding that the defendant intended to deceive consumers, the district court did not do so in this case.

As to the defendants' prior conduct, we note that the prior actions did not involve the graphic at issue in this case, but instead centered on the cartoon "crabgrass buster" logo and a time-line suggesting that Scotts' products had no post-emergence effect. The district court in the second action did conclude that the time-line was false, but the court did not conclude in either the first action or the second action that the crabgrass buster logo conveyed a false or misleading message. Instead, through settlement agreements under which it admitted no liability, Pursell agreed not to use the crabgrass buster logo, but specifically reserved the right to use a graphic claiming post-emergence control. And in our view, the graphic at issue in this case — a realistic rendering of a mature crabgrass plant — bears little if any resemblance to the menacing crabgrass buster logo. Again, a fact-finder might draw negative inferences from this evidence, but the district court did not do so here.

Scotts is of course free to make these arguments to the district court at the trial on Scotts' request for a permanent injunction, and nothing in this opinion should be understood as suggesting any particular outcome should the issue be raised. But because the district court did not make any factual findings with regard to whether the defendants intended to deceive consumers, it would be improper for this court on this record to apply a presumption of consumer confusion so as to excuse the deficiencies we have identified in Scotts' evidence.

Which brings us to this point. We have concluded that the Vigoro packaging is not literally false and that Scotts can prevail on its implied falsity claim only by presenting extrinsic evidence of likely consumer confusion. And because we have rejected Scotts' evidence of consumer confusion (and its various arguments as to why no extrinsic evidence was required), it follows that the district court erred by applying the presumption of irreparable harm, a presumption that was dependent on Scotts' establishing consumer confusion. We therefore proceed to consider whether the preliminary injunction can be sustained without reference to the presumption.

As previously discussed, the critical issue in a preliminary injunction case involves the balancing of the harms likely to be suffered by the parties. If the plaintiff has made a strong showing that it will suffer irreparable harm if the injunction is denied, the court must balance the likelihood of that harm against the likelihood of harm that would be suffered by the defendant. The plaintiff must make a "clear showing of irreparable harm ..., and the required irreparable harm must be neither remote nor speculative, but actual and imminent." Direx, 952 F.2d at 812 (internal quotation marks omitted).

Scotts contends that it will suffer irreparable injury if a preliminary injunction is not issued because it faces the risk of a permanent loss of customers and goodwill. First, Scotts asserts that the misleading statements on the Vigoro packaging will cause customers to believe that Vigoro is the superior product and will thus decrease Scotts' sales and perhaps even tarnish its reputation. Second, Scotts contends that by duping customers into believing that there is a product that will rid their yards of mature crabgrass, the defendants "run the risk of driving consumers from the do-it-yourself lawn care market altogether," an event that would disproportionately affect Scotts, the market leader. Brief of Appellee at 34.

These arguments fall far short of establishing an actual and imminent injury in this case. The record establishes that between now and January 2003 (when the trial is scheduled on Scotts' claims), sales of crabgrass control products are virtually non-existent. Sales will begin increasing in February, peaking in March and April. Moreover, Vigoro products are sold exclusively at Home Depot and currently are not even on the shelves in most of those stores. Therefore, even if the Vigoro packaging were misleading, the denial of a preliminary injunction would cause very little harm to Scotts. A permanent injunction after the trial in January, should the district court conclude an injunction is warranted, would be more than sufficient to protect Scotts' market share and reputation, because the injunction would take effect before the crabgrass-preventative season begins in earnest in the spring. We therefore conclude that, without the presumption of irreparable harm that was applied by the district court, Scotts has failed to establish that, without a preliminary injunction, it will suffer irreparable harm that is "neither remote nor speculative, but actual and imminent." Direx, 952 F.2d at 812.

For their part, the defendants contend that they will suffer irreparable harm from the granting of the injunction, pointing to the expense of creating new packaging and covering over the offending graphic on existing stock, as well as the harm to their reputation that would be caused by a determination that they engaged in false advertising. The district court largely dismissed these concerns, stating that "[a]ny harm that would result from a preliminary injunction would only be the result of the Defendants' own allegedly deceptive advertising," J.A. 701, a fact that "should be considered in balancing the harms." J.A. 702.

We agree with the defendants that the district court gave impermissibly short shrift to the question of the harm that would be visited upon the defendants. The harm to a defendant, particularly a defendant in a false advertising case, could almost always be described as of the defendant's own making. If self-made harm is given substantially less weight, as it was by the district court in this case, then the balance of the harms will almost always favor the plaintiff, thus transforming a preliminary injunction from an extraordinary remedy into a routine occurrence. And when the purpose behind the requirement that the court balance the harms is recognized, it becomes apparent that it is error to dismiss as self-inflicted the harms that might be suffered by a defendant if an injunction were to issue.

"[G]ranting a preliminary injunction requires that a district court, acting on an incomplete record, order a party to act, or refrain from acting, in a certain way. `The danger of a mistake in this setting is substantial.'" Hughes Network Sys., Inc. v. InterDigital Communications Corp., 17 F.3d 691, 693 (4th Cir.1994) (internal alteration omitted) (quoting American Hosp. Supply Corp. v. Hospital Prods. Ltd., 780 F.2d 589, 593 (7th Cir.1986)). The entire preliminary injunction inquiry, and particularly the requirement that the district court carefully balance the harms to the parties, is intended to ensure that the district court "choose[s] the course of action that will minimize the costs of being mistaken." American Hosp. Supply Corp. 780 F.2d at 593 (7th Cir.1986). Thus, while cases frequently speak in the shorthand of considering the harm to the plaintiff if the injunction is denied and the harm to the defendant if the injunction is granted, the real issue in this regard is the degree of harm that will be suffered by the plaintiff or the defendant if the injunction is improperly granted or denied:

If the judge grants the preliminary injunction to a plaintiff who it later turns out is not entitled to any judicial relief — whose legal rights have not been violated — the judge commits a mistake whose gravity is measured by the irreparable harm, if any, that the injunction causes to the defendant while it is in effect. If the judge denies the preliminary injunction to a plaintiff who it later turns out is entitled to judicial relief, the judge commits a mistake whose gravity is measured by the irreparable harm, if any, that the denial of the preliminary injunction does to the plaintiff.

American Hospital Supply Corp., 780 F.2d at 593. By dismissing outright or giving less weight to the harm that would be suffered by a defendant on the grounds that the harm was self-inflicted, a court is effectively considering the harms that would flow from a properly entered injunction (that is, an injunction entered against a defendant who would go on to lose) rather than considering the harms that would flow from an injunction entered in error (an injunction entered against a defendant who would go on to win). We therefore believe that it is error for a district court to conclude that any harm that would be suffered by a defendant was self-inflicted and thus entitled to lesser weight in the balancing-of-the-harms portion of the preliminary injunction calculus.

But in this case, any error by the district court in its consideration of the harm that would be suffered by the defendants is of little consequence, because we conclude that the defendants would suffer only minimal harm. While the defendants would incur the monetary costs of complying with the injunction by creating new packaging or placing stickers to cover the offending graphic on the existing packaging, they would be protected by the substantial bond posted by Scotts. See Hoechst Diafoil Co. v. Nan Ya Plastics Corp., 174 F.3d 411, 421 n. 3 (4th Cir.1999) ("In fixing the amount of an injunction bond, the district court should be guided by the purpose underlying Rule 65(c), which is to provide a mechanism for reimbursing an enjoined party for harm it suffers as a result of an improvidently issued injunction or restraining order."); Gateway Eastern Ry. Co. v. Terminal R.R. Ass'n of St. Louis, 35 F.3d 1134, 1140 (7th Cir.1994) ("In deciding whether to grant a preliminary injunction, the court must also consider any irreparable harm that the defendant might suffer from the injunction — harm that would not be either cured by the defendant's ultimately prevailing in the trial on the merits or fully compensated by the injunction bond ...." (internal quotation marks omitted)). And while the defendants' concern about the effect of an injunction is legitimate, we believe that this concern is significantly lessened by the fact that Home Depot is the defendants' only customer. If a preliminary injunction were entered and the defendants were exonerated after trial, the defendants' task in alleviating any concerns would be less onerous (and more likely to succeed) with only one customer in need of updating as to the outcome of the litigation.

We therefore conclude that relatively little harm would befall either side if the preliminary injunction issue were improperly decided against it. Because the balance of the harms does not tip toward either party, "the probability of success begins to assume real significance, and interim relief is more likely to require a clear showing of a likelihood of success." Direx, 952 F.2d at 808 (internal quotation marks omitted); see Blackwelder, 550 F.2d at 195 n. 3.

We have previously determined that the Vigoro packaging makes no literally false claims and that Scotts must therefore present evidence of consumer confusion to succeed on its false advertising claims. Because the evidence presented by Scotts is insufficient to show a likelihood of consumer confusion, Scotts has therefore failed to show a likelihood of success on the merits. This factor, then, counsels against the entry of a preliminary injunction.

The final Blackwelder factor to be considered is whether the public interest would be served by the injunction. As the district court noted, "there is a strong public interest in the prevention of misleading advertisements." Novartis, 290 F.3d at 597 (internal quotation marks and alteration omitted). But because we have determined that Scotts has not, at least at this juncture, established a likelihood of success on its false advertising claim, we cannot conclude that public interest weighs in favor of the issuance of an injunction.

Thus, after reviewing the record and considering the Blackwelder factors, we conclude that a preliminary injunction is not warranted in this case, a conclusion that flows directly from our determination that Scotts was required to present extrinsic evidence showing a likelihood of consumer confusion and that the focus group and survey evidence submitted by Scotts was not reliable or probative of the critical question in this case. The district court's error in accepting this evidence as sufficient led it to erroneously apply a presumption of irreparable harm, which in turn led to an erroneous consideration and application of the preliminary injunction standards. We therefore conclude that the district court abused its discretion by granting Scotts' request for a preliminary injunction. See, e.g., Quince Orchard Valley Citizens Ass'n, Inc. v. Hodel, 872 F.2d 75, 78 (4th Cir.1989) ("A district court abuses its discretion by applying an incorrect preliminary injunction standard, by resting its decision on a clearly erroneous finding of a material fact, or by misapprehending the law with respect to underlying issues in litigation." (internal quotation marks omitted)). Accordingly, we vacate the injunction and remand for proceedings on Scotts' request for a permanent injunction and other relief. Given the seasonal nature of the products at issue here, we encourage the district court to make every effort to resolve this case as expeditiously as possible.

VACATED AND REMANDED.