Schering Corporation v. Pfizer Inc. And Ucb Pharma, Inc.

This appeal invites us to revisit an increasingly important issue in the law of evidence, and one that has confused many courts. Today we clarify the circumstances under which scientifically conducted surveys may be admitted into evidence over a hearsay objection. The case comes to us on appeal from an order of the United States District Court for the Southern District of New York (McKenna, J.) denying a motion for a preliminary injunction made by plaintiff-appellant Schering Corporation (“Schering”) against defendants-appellees Pfizer Inc. (“Pfizer”) and UCB Pharma, Inc. (“UCB”). The primary evidence Schering offered in support of its motion were the results of five surveys in which physicians were asked to relate their memories, and sometimes their impressions, of communications that defendants’ sales representatives had made in the course of their promotional campaigns. Schering offered these surveys to support the more general proposition that defendants’ representatives were engaging in widespread false promotional activity in violation of both the Lanham Trade-Mark Act, 15 U.S.C. § 1051 et seq. (1994) (the “Lanham Act”), and a settlement agreement between the parties. The district court excluded all five surveys on hearsay grounds and then ruled that there was insufficient evidence to support a preliminary injunction.

We hold that two of the surveys in this case, which polled physicians’ impressions of the communications at issue, should have been admitted under the present state of mind exception to the hearsay rule, see FedR.Evid. 803(3), for the limited purpose of establishing a pattern of implied falsehood. We also find that in determining whether to admit the five surveys to establish literal falsehoods under the residual hearsay rule, see Fed.R.Evid. 807, the district court relied on an erroneous per se rulé against memory surveys offered to prove the facts remembered. The court thereby abused its discretion, and we remand to the court to determine the surveys’ trustworthiness on the basis of their methodological strengths and their relative susceptibilities to the risks of faulty memory and perception. Finally, we hold that the district court should have admitted one of the surveys, which Pfizer itself commissioned, as well as Pfizer’s internal analysis of this survey as party admissions. See Fed.R.Evid. 801(d)(2).

Because the survey evidence was integral to Schering’s case, we vacate the district court’s denial of a preliminary injunction and remand for reconsideration.

BACKGROUND

Schering is a pharmaceutical corporation that produces Claritin, a leading prescription antihistamine. A factor that strongly contributes to success in the antihistamine market is whether a drug can deliver effective relief from hay fever and other allergy symptoms without causing drowsiness. The first generation of antihistamines, introduced in the 1940s, caused drowsiness in a high percentage of users, and modern over-the-counter antihistamines have this same effect. Claritin, by contrast, is a new, second-generation prescription antihistamine, launched in 1993, which causes no more sedation than placebos in clinical tests. This feature has been important to Claritin’s market success.

UCB is a European pharmaceutical company that has developed a competing second-generation prescription antihistamine called “Zyrtec.” Because of UCB’s limited *222 presence in the United States, UCB licensed Pfizer, a Delaware corporation, to co-proraote the product domestically. Regulations promulgated by the United States Food and Drug Administration (“FDA”) required Pfizer to perform several controlled clinical tests on Zyrtec before the drug could be registered for domestic use, however, and these tests revealed that Zyrtec causes approximately twice as much sedation as a placebo. The FDA therefore required Pfizer to caution both physicians and consumers, through proper labeling and warning instructions, that Zyrtec has these sedating qualities. The FDA also warned Pfizer that it would be misleading to advertise Zyrtec by focusing on the rate at which customers discontinued using the drug because of somnolence (ie., 1%) rather than the drug’s somnolence level itself (ie., 11-14%).

In early 1996, Pfizer and UCB began selling Zyrtec in the United States. To promote the product, Pfizer used a method that is common in the pharmaceutical industry: it employed a team of approximately 1200 sales representatives to visit physicians across the nation and emphasize the product’s qualities in one-on-one informational meetings called “detailings.” In these meetings, Pfizer representatives were to promote Zyrtec, persuade doctors to prescribe it more often and respond to any questions or concerns that doctors might have about the drug. No records were kept of these meetings.

Suspicious that Zyrtec representatives might be misrepresenting the product’s somnolence levels in the detailings, Scher-ing hired IMS America, a commercial vendor of market surveys, to conduct a survey of physicians to determine whether any such misrepresentations were being made. The survey results led Schering to believe that they were. In February 1996, Scher-ing thus filed suit against Pfizer for violating Section 43(a)(2) of the Lanham Act, 15 U.S.C. § 1125(a)(2) (1994), by false advertising. See Schering Corp. v. Pfizer, Inc., No. 96 Civ. 1462(LMM) (S.D.N.Y. Feb. 29, 1996). On April 4, 1996, this dispute ended in a settlement agreement (the “Settlement Agreement”), under which Pfizer and UCB agreed not to permit their sales representatives to state either ‘expressly’ that Zyrtec was ‘lowsedating’ or ‘expressly or by implication’ that Zyrtec was ‘nonsedat-ing’ or ‘essentially nonsedating.’ These restrictions applied to all forms of advertising, including “verbal statements made to doctors.”

In order to monitor Pfizer’s and UCB’s compliance with the Settlement Agreement, Schering hired DTW in early 1998 to perform another survey of physicians concerning statements being made in the detailings. This survey polled 78 physicians who had been detailed between March 30 and June 5, 1998, and who were allegedly representative of a panel of over 6000 physicians nationwide who prescribe high volumes of antihistamines. 1 The doctors were asked the following question on the same day as the detailings:

What did the sales representatives tell you about [Zyrtec]? Please be as specific and complete as you can in describing the message or information that was conveyed to you about [Zyrtec].

The survey results suggested that approximately 30% of the Zyrtec agents were representing that the drug was either low sedating or nonsedating.

Schering subsequently commissioned a second survey from DTW, which was customized to be representative of a panel of over 6000 physicians nationwide who prescribe antihistamines in any amount. The survey polled 98 physicians shortly after they were detailed and asked more specifically: “What did the representative say about the product and sedation?” Approximately half of the physicians polled re *223 ported a representative describing Zyrtec as low sedating or nonsedating.

Sehering subsequently wrote a letter to Pfizer dated June 19, 1998, complaining of pervasive violations of the Settlement Agreement based on the results of these surveys. Pfizer denied the violations but assured Sehering that it would remind its sales force to comply with the mandates of the agreement. Sehering then continued to monitor Pfizer’s compliance by commissioning another survey, this time from the firm Clarke, Matire and Bartolomeo (“CMB”). This survey polled 200 physicians, who were selected randomly from those who prescribe antihistamines in above-minimal quantities nationwide. The interviews took place within a day of the detailings, between July 23 and September 17, 1998, and included the following question: “During (today’s/yesterday’s) Zyrtec detail, did the rep say anything about Zyrtec tablets and sedation?” The survey indicated that 47% of Zyrtec representatives were calling the product either low sedating or nonsedating.

On October 5, 1998, Sehering brought suit in the United States District Court for the Southern District of New York seeking a preliminary injunction against alleged ongoing and pervasive breaches of the Settlement Agreement and Section 43(a)(2) of the Lanham Act. In the course of discovery, Sehering learned that soon after receiving Schering’s original letter of complaint, Pfizer had commissioned Market Measures Inc. (“MMI”) to perform a “Fas-Tape Survey,” a common type of market survey that probes the main messages conveyed in an advertising campaign, “to understand the key messages and competitive claims concerning the sedation side effect pertaining to Zyrtec.” This survey involved a random sample of 74 physicians, who were allegedly representative of a panel of over 20,000 doctors prescribing antihistamines nationwide. The physicians were detailed between July 7 and July 12, 1998, and were instructed to call an 800 number immediately after receiving a sales presentation. They were then asked: “In one or two sentences, what was the main message of the presentation?” The FasTape Survey stated that “Zyrtec representatives appear to be focusing almost equally on the drug’s indications ... and efficacy, folloioed by its loio/no sedation.” (Emphasis added). During discovery, Sehering also uncovered an analysis of this FasTape Survey, prepared by Lakshmi Gengler, a market research executive at Pfizer. This internal analysis stated that “physicians report one-third of Zyrtec sales representatives mentioning low/non sedation with respect to Zyrtec’s side effect profile and mechanism of action.”

During discovery, Sehering also learned that Pfizer had commissioned Seott-Levin, another well-known market research company, to conduct a survey to identify the main messages being conveyed in the de-tailings. This survey, unlike the others, asked only about messages concerning nonsedation and did not poll for messages concerning low sedation. Sehering obtained several pages of this survey, which indicated that Zyrtec representatives were suggesting at a rate of approximately 21% in June of 1998 and 15% in July of 1998 that Zyrtec was nonsedating. Because Sehering was unable to obtain the entire survey, Sehering commissioned its own Seott-Levin Survey, which polled approximately 150 physicians per month from March through August of 1998. The survey results suggested that on average, approximately 16% of the Zyrtec agents were communicating a message that Zyrtec was nonsedating.

At the hearing on the preliminary injunction motion, Sehering sought to introduce five surveys: the two DTW surveys, the CMB Survey, the FasTape Survey and the Scotb-Levin Survey. Sehering also sought to introduce Pfizer’s internal analysis of the FasTape Survey. Finally, Sehering called to the stand Dr. Bartolo-meo, an executive at CMB, who testified that it is “very rare” for so many surveys of these kinds to have such consistent re- *224 suits. Pfizer responded with a motion arguing that all of this evidence contained hearsay and was thus inadmissible under Rule 802 of the Federal Rules of Evidence. The district court agreed and issued a written opinion refusing to admit Scher-ing’s surveys for any purpose. See Schering Corp. v. Pfizer, Inc., No. 98 Civ. 7000(LMM), 1999 WL 144921 (S.D.N.Y. Mar.16, 1999).

The only other significant evidence that Schering offered in support of its motion was a series of manuals used by Pfizer to train its sales representatives. These manuals contained some instructions to call Zyrtec “low sedating” and to focus attention away from the drug’s somnolence level and toward the rate at which customers discontinued using the drug because of somnolence effects. The manuals also indicated that eetirizine, Zyrtec’s active ingredient, does “not cause sedation.” The district court found this evidence insufficient to warrant a preliminary injunction and denied Schering’s motion. This appeal followed.

DISCUSSION

I. Standards of Review

We review a district court’s evi-dentiary rulings for abuse of discretion. See, e.g., Phoenix Assocs. III v. Stone, 60 F.3d 95, 100 (2d Cir.1995). “Either an error of law or a clear error of fact may constitute an abuse of discretion.” Charette v. Town of Oyster Bay, 159 F.3d 749, 755 (2d Cir.1998); see also Pullman-Standard v. Swint, 456 U.S. 273, 287, 102 S.Ct. 1781, 72 L.Ed.2d 66 (1982) (holding that a factual finding may be set aside if it “rest[s] on an erroneous view of the law”). An evidentiary ruling that is an abuse of discretion is, however, only reversible if it also affects a party’s substantial rights. See Fed.R.Evid. 103(a). This occurs when, for example, a district court excludes a party’s primary evidence in support of a material fact, and failure to prove that fact defeats the party’s claim. See O’Neal v. Esty, 637 F.2d 846, 848 (2d Cir.1980)

Here, there is no dispute that the district court excluded Schering’s primary evidence in support of its motion for a preliminary injunction, and there is no dispute that Schering’s motion was denied for lack of evidence that the surveys may have cured. In particular, the surveys may have helped establish that Pfizer and UCB were engaging in ongoing non-trivial violations of the Settlement Agreement and Section 43(a)(2) of the Lanham Act, 15 U.S.C. 1125(a)(2) (1994), and that a preliminary injunction was thus warranted. In what follows, we therefore limit our inquiry to whether the district court’s exclusion of these surveys was an abuse of discretion.

II. Admissibility of Surveys in General

In the first half of this century, surveys were generally regarded as inherently untrustworthy because they contained hearsay, or out-of-court statements offered to prove the truth of the matters asserted. See, e.g., DuPont Cellophane Co. v. Waxed Prods., 6 F.Supp. 859, 884 (E.D.N.Y.1934), modified, 85 F.2d 75 (2d Cir.1936) (refusing to admit survey because court could not “see how plaintiff could even test the facts, as it had no opportunity for cross-examination of those who were supposed to have answered the questions”); Elgin Nat’l Watch Co. v. Elgin Clock Co., 26 F.2d 376, 376-77 (D.Del.1928) (refusing to admit survey, even as basis for expert opinion, because survey collected hearsay). Schering argues, however, that “the modern view is that the hearsay objection is without merit and that any technical deficiencies in survey methodology go to [a survey’s] weight as evidence, not to its admissibility.” (Pl.’s Br. at 33-34 (quoting 5 J. Thomas McCarthy, McCarthy on Trademarks and Unfair Competition § 32:169, at 32-248 (4th ed.1998) [hereinafter McCarthy on Trademarks ]).) Under the “modern view,” according to Schering, surveys should be admitted as a general *225 rule, and their weight should, be determined by whether:

(1) the “universe” was properly defined, (2) a representative sample of that universe was selected, (3) the questions to be asked of interviewees were framed in a clear, precise and non-leading manner, (4) sound interview procedures were followed by competent interviewers who had no knowledge of the litigation or the purpose for which the survey was conducted, (5) the data gathered was accurately reported, (6) the data was analyzed in accordance with accepted statistical principles and (7) the objectivity of the entire process was ensured.

Toys “R” Us, Inc. v. Canarsie Kiddie Shop, Inc., 559 F.Supp. 1189, 1205 (E.D.N.Y.1983) (paraphrasing Manual for Complex Litigation § 2.712 (5th ed.1981)); Hutchinson v. Essence Communications, Inc., 769 F.Supp. 541, 557 (S.D.N.Y.1991); see also Manual for Complex Litigation, Third § 21.493 (1995). These factors derive from accepted principles of survey methodology and help define when a survey has been properly conducted. See, e.g., McCarthy on Trademarks § 32:181, at 32-272.

There is no doubt that beginning with Judge Wilfred Feinberg’s seminal decision in Zippo Manufacturing Co. v. Rogers Imports, Inc., 216 F.Supp. 670 (S.D.N.Y.1963), the general trend has been toward the admission of surveys of various kinds. Surveys are, for example, routinely admitted in trademark and false advertising cases to show actual confusion, genericness of a name or secondary meaning, all of which depend on establishing that certain associations have been drawn in the public mind. See, e.g., Bristol-Myers Squibb Co. v. McNeil-P.P.C., Inc., 973 F.2d 1033, 1043 (2d Cir.1992) (secondary meaning); PPX Enters., Inc. v. Audiofidelity Enters., Inc., 818 F.2d 266, 271 (2d Cir.1987) (actual confusion); Nestle Co. v. Chester’s Market, Inc., 571 F.Supp. 763, 769-70, 773-75 (D.Conn.1983) (genericness of name). See generally Marcia B. Paul & Anthony F. Lo Cicero, Litigating Trademark, Section 43(a) and Unfair Competition Cases, P.L.I. Patents, Copyrights, Trademarks, and Literary Property Course Handbook Series No. G4-3925 (Oct.1994) (describing use of surveys in trademark and false advertising context). Surveys of other types have occasionally been admitted for other limited purposes. See, e.g., Keith v. Volpe, 858 F.2d 467, 479-81 (9th Cir.1988) (admitting survey to show statistics concerning respondents’ race, income and housing preferences); Debra P. v. Turlington, 730 F.2d 1405, 1408, 1412-14 (11th Cir.1984) (admitting survey to show “whether the teacherfs] [surveyed] had provided instruction during 1981-82 relating to the skills tested on the SSAT-II and if so, whether that instruction had been sufficient for a student to master the skills”).

The use of surveys is, however, far more prevalent in trademark law than in most other areas. See Neal Miller, Facts, Expert Facts, and Statistics, 40 Rutgers L.Rev. 101, 137 (1987). Moreover, contrary to Schering’s contention, the case law does not support a general rule allowing surveys into evidence for all purposes. See, e.g., United States v. Pryba, 900 F.2d 748, 757 (4th Cir.1990) (excluding survey in obscenity case when offered “to demonstrate the community’s attitude, toleration and standards with regard to sexually explicit materials”); American Footwear Corp. v. General Footwear Co., 609 F.2d 655, 660 n. 4, 663 (2d Cir.1979) (holding that district court’s rejection of survey offered to show secondary meaning was not clearly erroneous); Ortho Pharm. Corp. v. Cosprophar, Inc., 828 F.Supp. 1114, 1122 (S.D.N.Y.1993) (excluding survey offered to show actual confusion).

A review of the case law suggests that there are, in fact, two ongoing controversies that tend to complicate the question of whether survey evidence should be admitted in a particular case. First, there is a dispute over the proper consequence of a finding of methodological error in a survey. While some courts in this Circuit *226 believe such flaws are proper grounds for exclusion, others view methodological errors as affecting only the weight of the evidence. Compare, e.g., Mobil Oil Corp. v. Pegasus Petroleum Corp., 818 F.2d 254 (2d Cir.1987) (“The district court properly admitted these surveys into evidence, despite claims of statistical imperfections by both sides, as those criticisms affected the weight accorded to the evidence rather than its admissibility.”), Bristol-Myers, 973 F.2d at 1043 (affirming district court’s decision to admit survey and allow methodological errors to affect the weight of the evidence), and Centaur Communications, Ltd. v. A/S/M Communications, Inc., 652 F.Supp. 1105, 1110 (S.D.N.Y.1987) (admitting survey and allowing methodological errors to affect only the weight of the evidence), with, e.g., Ortho Pharm., 828 F.Supp. at 1122 (“This court cannot conclude that the surveys were ‘properly conducted’ ... and therefore concludes ... that the surveys are inadmissible hearsay.”), and Toys “R” Us, 559 F.Supp. at 1205 (excluding survey because of methodological errors).

Indeed, this same division runs among, and sometimes within, other circuits. Compare, e.g., Indianapolis Colts, Inc. v. Metropolitan Baltimore Football Club Ltd. Partnership, 34 F.3d 410, 415-16 (7th Cir.1994) (finding that court was correct to draw inferences from survey, despite methodological problems), Brunswick Corp. v. Spinit Reel Co., 832 F.2d 513, 523 (10th Cir.1987) (“As to the technical and methodological deficiencies in the survey that Spinit charges, those relate not the survey’s admissibility but to the weight to be given such evidence.”), Levi Strauss & Co. v. Blue Bell, Inc., 778 F.2d 1352, 1358 (9th Cir.1985) (en banc) (affirming district court’s decision to admit surveys and determine their weight by examining methodology), Zatarains, Inc. v. Oak Grove Smokehouse, Inc., 698 F.2d 786, 795, 797 (5th Cir.1983) (finding that “survey evidence is the most direct and persuasive way to establish secondary meaning,” but that particular survey at issue was entitled to “little evidentiary weight” because of its methodological defects), and President and Trustees of Colby College v. Colby College-New Hampshire, 508 F.2d 804, 809 (1st Cir.1975) (finding it clearly erroneous to reject survey altogether, and giving survey some weight, despite methodological defects), with, e.g., Harolds Stores, Inc. v. Dillard Dep’t Stores, Inc., 82 F.3d 1533, 1544 (10th Cir.1996) (“We allow the admission of survey evidence as an exception to the hearsay rule if the survey is material, more probative on the issue than other evidence and if it has guarantees of trustworthiness.” (citations omitted) (internal quotations omitted)), C.A. May Marine Supply Co. v. Brunswick Corp., 649 F.2d 1049, 1054 (5th Cir.1981) (“Surveys and customer questionnaires are admissible, if they are pertinent to the inquiry, upon a showing that the poll is reliable and was compiled in accordance with accepted survey methods.”), Baumholser v. Amax Coal Co., 630 F.2d 550, 552 (7th Cir.1980) (“To qualify a study or opinion poll for admission into evidence, there must be a substantial showing of reliability.”), and Pittsburgh Press Club v. United States, 579 F.2d 751, 758 (3d Cir.1978) (holding that surveys are admissible only if they are “conducted with proper safeguards to insure accuracy and reliability”). Courts have, however, rarely articulated their reasons for falling onto one or another side of this divide.

The second dispute reflected in the case law involves the proper rationale for allowing admissible surveys into evidence. As Judge Feinberg explained in Zippo:

Some cases hold that surveys are not hearsay at all; other cases hold that surveys are hearsay but are admissible because they are within the recognized exception to the hearsay rule for statements of present state of mind, attitude, or belief.

216 F.Supp. at 682; see also Grotrian, Helfferich, Schulz, Th. Steinweg Nachf. v. Steinway & Sons, 523 F.2d 1331, 1341 (2d *227 Cir.1975) (‘We are mindful of the on-going dispute as to whether a survey should be admissible ... under the state of mind exception to the hearsay rule or based on the need for it plus adequate guarantees of trustworthiness.”). We have sometimes suggested that surveys are properly admitted under the residual hearsay rule, see, e.g., Starter Corp. v. Converse, Inc., 170 F.3d 286, 297 (2d Cir.1999) (citing residual hearsay rule as basis for admitting survey), and the Advisory Committee for the Federal Rules of Evidence has suggested that surveys are sometimes best admitted as bases of expert testimony pursuant to Rule 703, see, e.g., Fed.R.Evid. 703 advisory committee’s note (“[Rule 703] offers a more satisfactory basis for ruling upon the admissibility of public opinion poll evidence. Attention is directed to the validity of the techniques employed rather than to relatively fruitless inquiries whether hearsay is involved.”). Quite often, courts simply fail to cite a statutory basis for an admission. See Zippo, 216 F.Supp. at 682. These confusions in the case law have led at least one prominent commentator to conclude that

[a] skeptic would classify the survey cases into two categories: a survey is accepted and relied upon when the judge already has his or her mind made up in favor of the survey results; and a survey is rejected and torn apart when the judge subjectively disagrees with the survey results.

McCarthy on Trademarks § 32:196, at 32-298.

This skepticism is unwarranted. Because the Federal Rules of Evidence govern the admissibility of all evidence over a hearsay objection, the question of survey admissibility is ultimately a question of statutory interpretation. As shown in the following two sections, careful attention to the possible statutory grounds 'for admitting hearsay can help harmonize most of the case law. Whether the district court was correct to exclude the five surveys in this case thus merits detailed discussion, with particular attention to the types of statements contained in the surveys, the purposes for which they were offered and the various possible grounds for their admission.

III. The State of Mind Exception: Admissibility to Show What Was Implied

One of the two most common bases for admitting survey evidence is Rule 803(3), which creates an exception to the hearsay rule for statements that express a declarant’s state of mind at the time of the utterance. In particular, Rule 803(3) excepts any

statement of [a] declarant’s then existing state of mind, emotion, sensation, or physical condition (such as intent, plan, motive, design, mental feeling, pain, and bodily health), but not including a statement of memory or belief to prove the fact remembered or believed unless it relates to the execution, revocation, identification, or terms of declarant’s will.

Fed.R.Evid. 803(3). The great majority of surveys admitted in this Circuit, including those used in Lanham Act cases to establish actual confusion or secondary meaning, fall into this category: they poll individuals about their presently-existing states of mind to establish facts about the group’s mental impressions. See, e.g., Sterling Drug, Inc. v. Bayer AG, 14 F.3d 733, 741-42 (2d Cir.1994) (polling consumers for their then-existing perceptions to establish actual confusion); Mobil Oil Corp., 818 F.2d at 259 (same); Harlequin Enters. Ltd. v. Gulf & Western Corp., 644 F.2d 946, 949-50 (2d Cir.1981) (polling consumers for their then-existing states of mind to establish secondary meaning, or tendencies to associate certain product features with a particular corporate origin).

It is important for district courts to recognize surveys of this type because their qualification for a traditional hearsay exception obviates the need to examine methodology before overruling a hearsay *228 objection. Regardless of the basis cited for admitting these surveys, errors in methodology thus properly go only to the weight of the evidence—subject, of course, to Rule 403’s more general prohibition against evidence that is less probative than prejudicial or confusing. See, e.g., Grotrian, 523 F.2d at 1341 (finding, in the context of a survey offered to show actual confusion, that while courts cite different bases for admitting survey evidence, “[t]he more relevant issue [is] the weight to be given to the surveys”); see also Starter, 170 F.3d at 297 (holding that survey evidence offered to show actual confusion was properly excludable under Rule 403 when it was so flawed that its probative value was outweighed by the risk of prejudice or confusion). The “modern view” urged by Schering is thus fully applicable to these kinds of surveys. See McCarthy on Trademarks § 32:168, at 32-246 (“[S]ince at least 1951 the cases are now unanimous that evidence of the state of mind of persons surveyed is not inadmissible as hearsay.” (footnote omitted)) (collecting cases). See generally Jack P. Lipton, Trademark Litigation: A New Look at the Use of Social Science Evidence in Trademark Litigation, 29 Ariz. L.Rev. 639 (1987). But cf. Section IV(i), infra (explaining that where residual hearsay rule is only possible basis for admission, methodological defects must be examined to determine not only weight but admissibility of survey evidence).

Schering argues that both the Fas-Tape and Scott Levin surveys fall into this category because they asked physicians to relate thé “main messages” conveyed by Zyrtee agents in the detailings. The surveys thereby asked physicians to relate not only what was said in the meetings but the impressions with which they were left. Such impressions are classic states of mind and, as such, fall under Rule 803(3). See, e.g., Zippo, 216 F.Supp. at 683-84; Telebrands Corp. v. E. Mishan & Sons, No. 97 Civ. 1414(RPP), 1997 WL 232595, at *20 (S.D.N.Y. May 7, 1997).

The district court did not address directly whether these surveys polled for then-existing states of mind. Instead, the court appears to have viewed evidence of these states as irrelevant to the present litigation. The court explained:

The Court does not fully concur with Schering’s suggestion that the Court need not conclude that the surveys report “what exact words were spoken.” While it may not need to know the exact words spoken, the Court must know the exact substance of what was said by the representatives to the physicians, because it is that substance that violates, or does not violate, the settlement agreement. To determine that substance requires an effort to determine, as closely as possible, the exact words that were spoken.

Schering Corp., 1999 WL 144921, at *4 n. 4 (emphasis added) (citation omitted). We disagree.

The Settlement Agreement states in relevant part that:

Pfizer and UCB Pharma hereby agree that, in connection with their advertising and promotion of ZYRTEC in the United States, they will not claim or allow those acting on their behalf to claim, either expressly or by implication that:

(1) ZYRTEC and/or its active ingredient cetirizine is nonsedating or essentially nonsedating;

(2) ZYRTEC and/or cetirizine is as non-sedating as CLARITIN .... or is comparable to CLARITIN ... or to the “second” generation of antihistamines ... in terms of its somnolence or sedation;

(3) ZYRTEC’s and/or cetirizine’s discontinuance rate due to somnolence means that ZYRTEC and/or cetirizine is non-sedating or essentially non-sedating;

(4) After a short period of time, ZYRTEC and/or cetirizine users *229 who experience sedation develop tolerance to its sedating effect.

(Settlement Agreement of 4/4/1996, at 3-4 (emphasis added).) The Settlement Agreement thus prohibits not only explicit but also implicit falsehoods. In this sense, the agreement is reminiscent of the Act upon which it was based — i.e., the Lanham Act — which prohibits advertisements that are not only literally but also impliedly false. See L & F Prods. v. Procter & Gamble Co., 45 F.3d 709, 711 (2d Cir.1995) (noting that Lanham Act prohibits impliedly false advertising, or advertisements that “although literally true, [are] still likely to mislead or confuse consumers”); see also S.C. Johnson & Son, Inc. v. Carter-Wallace, Inc., 614 F.Supp. 1278, 1319 (S.D.N.Y.1985) (“There are two types of actionable false advertising: (1) advertising which makes claims which are literally false on their face, and (2) advertising which, although literally true on its face, is perceived by a significant proportion of the relevant market as making ‘subliminal’ or ‘implicit’ claims which are provably false. With regard to the second type of false advertising, the courts sometimes say that the advertising has a tendency to ‘mislead, confuse or deceive.’ ”). 2

Cases in the Lanham Act context demonstrate, moreover, that the mental impressions with which an audience is left can be relevant, and sometimes even necessary, to establish what a defendant is implying in a challenged representation. See, e.g., Cocar-Cola Co. v. Tropicana Prods., Inc., 690 F.2d 312, 317 (2nd Cir.1982) (“When a merchandising statement or representation is literally or explicitly false, the court may grant relief without reference to the advertisement’s impact on the buying public. When[, however,] the challenged rather than explicitly false, its tendency to violate the Lanham Act by misleading, confusing or deceiving should be tested by public reaction.” (emphasis added) (citations omitted)). In fact, although plaintiffs seeking to establish a literal falsehood must generally show the substance of what is conveyed, we have held that a “district court must rely on extrinsic evidence to support a finding of an implicitly false message.” See Johnson & Johnson * Merck Consumer Pharms. Co. v. Smithkline Beecham Corp., 960 F.2d 294, 297 (2d Cir.1992) (emphasis added). This is because plaintiffs alleging a literal falsehood are claiming that a statement, on its face, conflicts with reality, a claim that is best supported by comparing the statement itself with the reality it purports to describe. By contrast, plaintiffs alleging an implied falsehood are claiming that a statement, whatever its literal truth, has left an impression on the listener that conflicts with reality. This latter claim invites a comparison of the impression, rather than the statement, with the truth. See generally Jacob Jacoby et al, Survey Evidence in Deceptive Advertising Cases Under the Lanham Act: An Historical Review of Comments from the Bench, 954 PLI/Corp. 83, 87-88 (1996). Given this distinction in the false advertising context, we see no reason to believe that Schering’s allegations of breach of the Settlement Agreement by implied falsehood cannot similarly be tested by this type of evidence. Moreover, Sehering rested its argument for a preliminary injunction not only on alleged breaches of the Settlement Agreement but also on alleged violations of the Lanham Act.

*230 These considerations persuade us that the FasTape and Scott Levin surveys are relevant to show ongoing violations of both the Settlement Agreement and the Lan-ham Act, on a theory that Zyrtec agents were making statements that regardless of their veracity, left the physicians with false impressions. Although statements by out-of-court declarants—e.g., the surveyed physicians—-relating such impressions are hearsay where, as here, they are offered to establish the existence of the impressions themselves, 3 see Fed.R.Evid. 801(c), these declarations are independently admissible under Rule 803(3) as expressions of the declarant physicians’ then-existing states of mind. The district court should have thus admitted the FastTape and Scott Levin surveys for the limited purpose of establishing a pattern of implied falsehood.

This ruling concerning admissibility in no way suggests that the district court should give these surveys any particular weight on remand. The fact that the two surveys corroborate one another is certainly a point in their favor, as is the fact that Schering produced independent evidence of Pfizer training manuals, which appear to have recommended misleading promotional practices. Still, Pfizer has offered a series of criticisms concerning the surveys’ methodologies and the potentially leading or ambiguous nature of the questions they posed. Schering also has failed to bring even a small subset of the surveyed physicians into court, a process that would have allowed for their cross-examination and would have greatly assisted the court in determining what weight to give the surveys overall. Finally, Pfizer has argued that