The Paddington Corporation ("Paddington") appeals from a February 18, 1992 judgment of the United States District Court for the Eastern District of New York (Thomas C. Platt, Jr., Chief Judge ) dismissing, pursuant to Fed.R.Civ.P. 52(c), Paddington's trademark and trade dress infringement action brought under § 43(a) of the Lanham Act, 15 U.S.C. § 1125(a), together with state claims, against a rival in the importation and distribution of the Greek liqueur ouzo.

Paddington argues on appeal that the district court's reliance on lack of secondary meaning in its trade dress analysis is now faulty in light of the Supreme Court's ruling in Two Pesos, Inc. v. Taco Cabana, Inc., --- U.S. ----, 112 S. Ct. 2753, 120 L. Ed. 2d 615 (1992), handed down after the judgment ordering dismissal was entered in the district court. Paddington also contends that the district court erred in finding no likelihood of confusion on either the trademark or the trade dress claim.BACKGROUND

Paddington is the United States importer and distributor of No. 12 Ouzo, a brand of the anise-based drink ouzo produced by Kaloyannis Bros., S.A. ("Kaloyannis"). Ouzo is the most popular alcoholic beverage in Greece, and No. 12 Ouzo, with a market share of more than forty percent, is the undisputed market leader in that country. Kaloyannis introduced its ouzo to the United States market with the unregistered mark "No. 12 Ouzo" and the current trade dress in 1969.

Attiki is the United States importer and distributor for Cavino S.A. ("Cavino"), a Greek ouzo producer that seeks, through Attiki, to bring an ouzo branded "#1 Ouzo" to the United States market. Attiki used to be the exclusive United States importer and distributor of No. 12 Ouzo, but in August, 1989, Kaloyannis dropped Attiki in favor of Paddington. Undaunted, in 1990, Attiki held discussions with Cavino about the prospect of introducing a rival ouzo brand into the United States. Cavino showed Attiki three potential labels, from which Attiki selected the #1 Ouzo label at issue in this case. Attiki and Cavino then entered an agreement under which Attiki would hold #1 Ouzo's United States trademark and trade dress rights.

Because the asserted trade dress similarity lies at the heart of this case, it is necessary to describe in some detail the physical appearances of the two ouzo bottles and their labels (see also Exhibit A).

No. 12 Ouzo is sold in a clear glass bottle with a large label on its side and a second, smaller label on its neck. Both labels are red, white, and black. The large label consists of a white background, inside of which is a rectangle consisting of a white top half and a red lower half, bordered by a thin black line. In the white half, there is a small "No" and a very large "12" in black block letters made to resemble a stencilled number on a crate or barrel. In the red lower half the word "ouzo" appears in white block letters with black shadowing, under which the word "Kaloyannis" is found in small, thin black letters, along with some other information in fine print. The neck label mirrors the large label in layout, except that a circle is used in place of the rectangle. The overall appearance of the bottle is simple, even stark.

The #1 Ouzo bottle also is clear, of a similar shape, and displays a large main label and a smaller neck label in the same tones of red, white, and black as found on the No. 12 Ouzo bottle. The large label is rectangular, with a white background and a red border. In the center of the label is a large ellipse, separated into a white top half and a red bottom half, and bordered by a thin black line. In the white half there is a small "#" and a large "1" in black block letters. In the red half of the circle the word "ouzo" appears in plain white block letters. Above the ellipse is the word "LIQUEUR" in small black letters, and below it is miscellaneous fine print in black. The neck label contains a circle with a configuration identical to the ellipse on the bottle's large label. The overall appearance of the #1 Ouzo dress, like the No. 12 Ouzo bottle, is simple and clean.

Paddington brought suit against Attiki on June 7, 1991, claiming trademark and trade dress infringement in violation of § 43(a) of the Lanham Act, and under New York law for common law unfair competition, dilution of a distinctive mark in violation of N.Y. General Business Law § 368-d, and deceptive acts in violation of N.Y. General Business Law § 349. It sought temporary and permanent injunctions against the use of the name "#1 Ouzo" and against the use of any trade dress that resembles that of No. 12 Ouzo. Paddington moved for a preliminary injunction on July 2, 1991, but this motion was withdrawn after Attiki agreed not to distribute #1 Ouzo pending the resolution of this action.

A bench trial was held by the district court on November 7, 1991. At the close of Paddington's case, Attiki moved for dismissal under former Fed.R.Civ.P. 41(b). The court granted the motion and entered judgment for Attiki on February 18, 1992. We note that on December 1, 1991, Rule 41(b)'s procedure for involuntary dismissal at the close of a plaintiff's case during a bench trial was replaced with Rule 52(c), which for purposes of this case is substantively identical. The district court should have converted the motion to a Rule 52(c) motion, and we will refer to it as such. The court concluded that Paddington had failed to prove that its trade dress and trademark had acquired secondary meaning, that is, that they had come to be identified by the public as Kaloyannis's product, as required by Second Circuit case law at the time. Also, applying the analysis set forth in Polaroid Corp. v. Polarad Electronics Corp., 287 F.2d 492 (2d Cir.), cert. denied, 368 U.S. 820, 82 S. Ct. 36, 7 L. Ed. 2d 25 (1961), the court found that there was no likelihood of confusion between either the trade dresses or the trademarks. The court then dismissed Paddington's pendent state claims for lack of an independent basis of federal subject matter jurisdiction.

This appeal followed.

Unregistered trademarks and trade dresses are protectable under § 43(a) of the Lanham Act, 15 U.S.C. § 1125(a), which provides a private cause of action against any person who "in connection with any goods ... or any container for goods, uses in commerce any word, term, name, symbol, or device, or any combination thereof ... which ... is likely to cause confusion, or to cause mistake, or to deceive ... as to the origin, sponsorship, or approval of his or her goods ... by another person." 15 U.S.C. § 1125(a).

To prevail in an action for infringement of a trademark or trade dress under Lanham Act § 43(a), a plaintiff must prove (1) that its mark is distinctive and (2) that a likelihood of confusion exists between its product and the defendant's. Two Pesos, --- U.S. ----, ----, 112 S. Ct. 2753, 2758; Bristol-Myers Squibb Co. v. McNeil-P.P.C., Inc., 973 F.2d 1033, 1038-39 (2d Cir.1992). Additionally, functional packaging and product design are unprotected, and functionality may be raised as a defense to an action for trade dress infringement. LeSportsac, Inc. v. K Mart Corp., 754 F.2d 71, 75-76 (2d Cir.1985).

A. Distinctiveness of the No. 12 Ouzo Trade Dress

A product's trade dress is its "total image ... includ[ing] features such as size, shape, color or color combinations, texture, [or] graphics." Id. at 75 (quoting John H. Harland Co. v. Clarke Checks, Inc., 711 F.2d 966, 980 (11th Cir.1983) (second alteration in original)).

Prior to the Supreme Court's decision in Two Pesos, the Second Circuit required a plaintiff seeking § 43(a) protection of a trade dress to establish distinctiveness by proving secondary meaning. See, e.g., Stormy Clime Ltd. v. ProGroup, Inc., 809 F.2d 971, 974 (2d Cir.1987). If a plaintiff proved secondary meaning, then the plaintiff could recover if it could establish a likelihood of confusion between the two trade dresses. This stood in contrast to our treatment of trademark infringement under § 43(a), where a plaintiff can reach the likelihood-of-confusion issue if the mark either is inherently distinctive or has acquired secondary meaning.

The district court below, relying on our precedent, held that Paddington's action must fail because it had not established that its trade dress had acquired secondary meaning.

In Two Pesos, the Supreme Court specifically rejected the Second Circuit approach, reasoning that there is "no basis for requiring secondary meaning for inherently distinctive trade dress protection under § 43(a) but not for other distinctive words, symbols, or devices capable of identifying a producer's product." --- U.S. at ----, 112 S. Ct. at 2760. A secondary meaning requirement, the Court reasoned, would disfavor new market entrants who possessed a unique trade dress that had not become widely known, while favoring those who had enjoyed market success, without any basis for such a distinction in the Lanham Act. Id. at ----, ----, 112 S. Ct. at 2759, 2761. What is critical for Lanham Act § 43(a) analysis, the Court noted, is not whether a trade dress has in fact come to identify a specific producer, but whether it is "capable of identifying a particular source of the product." Id. at ----, 112 S.Ct. at 2759 (emphasis added).

Courts often evaluate the inherent distinctiveness of trademarks according to the test set forth by Judge Friendly in Abercrombie & Fitch Co. v. Hunting World, Inc., 537 F.2d 4, 9 (2d Cir.1976). Two Pesos, --- U.S. at ----, 112 S. Ct. at 2757. This is the approach used in this circuit. Bristol-Myers Squibb, 973 F.2d at 1039. Under the Abercrombie test, marks are classified as either (1) generic; (2) descriptive; (3) suggestive; or (4) arbitrary or fanciful. Abercrombie, 537 F.2d at 9. While the Court in Two Pesos lists arbitrary and fanciful as two separate classifications, --- U.S. at ----, 112 S. Ct. at 2757, the substance of Two Pesos treats them as being equivalent, in accordance with Abercrombie. Suggestive and arbitrary or fanciful marks are considered to be inherently distinctive, and therefore always satisfy the first prong of the test for Lanham Act § 43(a) trademark protection. Id. at ----, 112 S.Ct. at 2757; Bristol-Myers Squibb, 973 F.2d at 1039. If a mark is descriptive, however, the plaintiff must establish that it has acquired secondary meaning in order to become distinctive and thereby satisfy the first prong of Lanham Act trademark analysis. Two Pesos, --- U.S. at ---- - ----, 112 S. Ct. at 2757-58. Generic marks are never protectable. See id. at ----, 112 S. Ct. at 2757.

In classifying a trademark according to the Abercrombie test, a court examines the context in which the words constituting the mark are used. Bristol-Myers Squibb, 973 F.2d at 1041; Abercrombie, 537 F.2d at 12. As one commentator explained, "the word 'apple' would be arbitrary when used on personal computers, suggestive when used in 'Apple-A-Day' on vitamin tablets, descriptive when used in 'Tomapple' for combination tomato-apple juice and generic when used on apples." 1 J.T. McCarthy, Trademarks and Unfair Competition § 11:22, at 498-99 (2d ed. 1984) (quoted in Bristol-Myers Squibb, 973 F.2d at 1041 (upholding finding that the "pm" in "Excedrin PM" was descriptive rather than suggestive)); see also Physicians Formula Cosmetics Inc. v. West Cabot Cosmetics, Inc., 857 F.2d 80, 82 (2d Cir.1988) ("Physicians Formula" is suggestive when used for skin creams and lotions); McGregor-Doniger Inc. v. Drizzle Inc., 599 F.2d 1126, 1132 (2d Cir.1979) (noting that district court found that registered trademark "Drizzler" is suggestive when applied to rain jackets, but only barely over the line between descriptive and suggestive).

While the applicability of the Abercrombie classifications to trade dress was not at issue in Two Pesos, the Court noted that the Fifth Circuit below had applied the Abercrombie classifications to the trade dress at issue and discussed them without disapproval. --- U.S. at ----, 112 S. Ct. at 2757; see Taco Cabana Int'l, Inc. v. Two Pesos, Inc., 932 F.2d 1113, 1120 & n. 8 (5th Cir.1991). While the Court stated that the sole issue before it for decision was whether secondary meaning must be proven for an inherently distinctive trade dress and stopped just short of expressly ruling on whether the Abercrombie classifications apply to trade dress, we agree with the Fifth Circuit in Two Pesos that they do.

In Chevron Chemical Co. v. Voluntary Purchasing Groups, Inc., 659 F.2d 695 (5th Cir.1981), cert. denied, 457 U.S. 1126, 102 S. Ct. 2947, 73 L. Ed. 2d 1342 (1982), relied on heavily by the Court in Two Pesos, the Fifth Circuit determined that the plaintiff's bottles and packaging were inherently distinctive, noting that "the possible varieties of advertising display and packaging are virtually endless." Id. at 703. Since the choices that a producer has for packaging its products are, as the Fifth Circuit noted, almost unlimited, typically a trade dress will be arbitrary or fanciful and thus inherently distinctive, and the only real question for the courts will be whether there is a likelihood of confusion between the products, see id., provided, of course, the trade dress is not functional.

However, where it is the custom of an industry to package products in a particular manner, a trade dress in that style would be generic and therefore not inherently distinctive. For example, packaging lime-flavored soda in green twelve-ounce cans is so common in the soft drink industry that such packaging probably is not inherently distinctive, although without the industry practice green cans would be either suggestive or arbitrary and therefore inherently distinctive. Descriptive trade dresses also are not inherently distinctive. In Blau Plumbing, Inc. v. S.O.S. Fix-It, Inc., 781 F.2d 604, 609 (7th Cir.1986), the court found that an advertisement for a sewer service company containing four boxes labeled north, south, east, and west, each containing the locations and phone numbers of different branches of the company within that area of the city, was a trade dress descriptive of the company's ability to service the entire city. Similarly, a trade dress featuring an illustration of a shining car on a bottle of car wax likely would be descriptive.

Trade dresses often utilize commonly used lettering styles, geometric shapes, or colors, or incorporate descriptive elements, such as an illustration of the sun on a bottle of suntan lotion. While each of these elements individually would not be inherently distinctive, it is the combination of elements and the total impression that the dress gives to the observer that should be the focus of a court's analysis of distinctiveness. If the overall dress is arbitrary, fanciful, or suggestive, it is inherently distinctive despite its incorporation of generic or descriptive elements. Roulo v. Russ Berrie & Co., 886 F.2d 931, 936 (7th Cir.1989), cert. denied, 493 U.S. 1075, 110 S. Ct. 1124, 107 L. Ed. 2d 1030 (1990). Cf. LeSportsac, 754 F.2d at 76 (despite functionality of individual elements of sports bag, bag is nonfunctional "when viewed in its entirety"). One could no more deny protection to a trade dress for using commonly used elements than one could deny protection to a trademark because it consisted of a combination of commonly used letters of the alphabet.

The No. 12 Ouzo bottle is inherently distinctive. There is no evidence in the record of any industry practice of using a design like the one that appears on the bottle's labels. There is nothing descriptive about the bottle and label design that conveys anything about its particular contents, except for the use of the trademark "No. 12 Ouzo," which will be discussed below, and the fact that the bottle is of a style such that it indicates to the observer that it contains a liquid that probably is potable. The tone and layout of the colors, the style and size of the lettering, and, most important, the overall appearance of the bottle's labeling, are undeniably arbitrary. They were selected from an almost limitless supply of patterns, colors and designs. Since the No. 12 trade dress is arbitrary, and therefore protectable under the Lanham Act, the secondary meaning analysis is unnecessary, and we turn to likelihood of confusion to determine if it should be protected in this case.

In both trademark and trade dress cases, in evaluating likelihood of confusion we are, of course, guided by the factors set forth by Judge Friendly in Polaroid Corp. v. Polarad Electronics Corp., 287 F.2d 492, 495 (2d Cir.), cert. denied, 368 U.S. 820, 82 S. Ct. 36, 7 L. Ed. 2d 25 (1961). See Bristol-Myers Squibb, 973 F.2d at 1043 (trade dress); Hasbro, Inc. v. Lanard Toys, Ltd., 858 F.2d 70, 75 (2d Cir.1988) (trademark). The factors we examine are: (1) the strength of the plaintiff's mark or dress; (2) the similarity between the two marks or dresses; (3) the proximity of the products in the marketplace; (4) the likelihood that the prior owner will bridge the gap between the products; (5) evidence of actual confusion; (6) the defendant's bad faith; (7) the quality of the defendant's product; and (8) the sophistication of the relevant consumer group. Bristol-Myers Squibb, 973 F.2d at 1043; Hasbro, 858 F.2d at 75. The eight-factor list is not exclusive. Bristol-Myers Squibb, 973 F.2d at 1043. Furthermore, the evaluation of the Polaroid factors is not a mechanical process "where the party with the greatest number of factors weighing in its favor wins." Physicians Formula, 857 F.2d at 85. Rather, a court should focus on the ultimate question of whether consumers are likely to be confused. See Lang v. Retirement Living Publishing Co., 949 F.2d 576, 580 (2d Cir.1991).

In reviewing a district court's evaluation of the likelihood-of-confusion issue, we review the determination of each individual Polaroid factor under a clearly erroneous standard, but the ultimate evaluation of likelihood of confusion, which is based on a weighing of the factors, we review de novo. Bristol-Myers Squibb, 973 F.2d at 1043-44; Western Publishing Co. v. Rose Art Indus., 910 F.2d 57, 61 (2d Cir.1990). While we used to apply de novo review to the similarity-of-the-mark inquiry, on the theory that an appellate court is in as good a position to examine a mark or dress as a trial court, e.g., Spring Mills, Inc. v. Ultracashmere House, Ltd., 689 F.2d 1127, 1130 (2d Cir.1982), we held in Bristol-Myers Squibb, 973 F.2d at 1043-44, that this practice was inappropriate in light of a 1985 amendment to the Federal Rules requiring that findings of fact "whether based on oral or documentary evidence," should not be set aside unless clearly erroneous. Fed.R.Civ.P. 52(a); see 1985 Advisory Committee Note. While a few of our cases have applied the old de novo standard after the 1985 amendment's effective date, see Western Publishing, 910 F.2d at 62; Centaur Communications, Ltd. v. A/S/M Communications, Inc., 830 F.2d 1217, 1226 (2d Cir.1987); Oral-B Laboratories, Inc. v. Mi-Lor Corp., 810 F.2d 20, 23 (2d Cir.1987), these decisions did not discuss the 1985 amendment and cited only pre-amendment cases as authority. We therefore follow Bristol-Myers Squibb and apply a clearly erroneous standard, which adjusted the Second Circuit standard to bring it in line with the Federal Rules.

The district court correctly found that there was no evidence of actual confusion and no evidence of a disparity in the quality of the two ouzos. We turn, then, to the remaining six Polaroid factors.

The district court held that the No. 12 trade dress was weak based solely on Paddington's failure to establish secondary meaning. We believe this to have been clear error.

A mark's strength is its "distinctiveness ..., or more precisely, its tendency to identify the goods sold under the mark as emanating from a particular, although possibly anonymous, source." McGregor-Doniger, 599 F.2d at 1131. Such strength may result solely from the mark's inherent distinctiveness. If the mark or dress is such that it seems to the consumer uniquely intended to indicate a product's source, it will be strong whether or not the consumer is familiar with the mark or knows the source. Of course there is no bar to a court's consideration of secondary meaning in evaluating a mark's strength. Lang, 949 F.2d at 581; McGregor-Doniger, 599 F.2d at 1132-33. However, it is error for a court to hold that a mark cannot be strong absent proof of secondary meaning. W.W.W. Pharmaceutical Co. v. Gillette Co., 984 F.2d 567, 573 (2d Cir.1993); McGregor-Doniger, 599 F.2d at 1132. This rule is bolstered by Two Pesos, which is grounded in the notion that a newcomer to the market with an inherently distinctive trade dress may be entitled to protection. Two Pesos, --- U.S. at ---- - ----, 112 S. Ct. at 2759, 2761. While Two Pesos dealt with the inherent distinctiveness prong of trade dress protection, it plainly informs the entire analysis of Lanham Act protectibility. To not require secondary meaning as a threshold matter but then bring it in through the back door in the strength-of-the-mark inquiry would undermine the core holding of Two Pesos.

It was clear error for the district court to find that the No. 12 trade dress was weak based solely on the lack of secondary meaning. Based on the record before us, the No. 12 trade dress clearly is arbitrary and fanciful and would appear to a consumer to be intended to identify the origin of the product, and therefore it is a strong mark. See Charles of the Ritz Group, Ltd. v. Quality King Distribs., Inc., 832 F.2d 1317, 1321 (2d Cir.1987) ("Opium" perfume's trademark and trade dress are arbitrary rather than generic or descriptive "and thus are among the strongest and most highly protected class of trademarks"); Ambrit, Inc. v. Kraft, Inc., 812 F.2d 1531, 1539 (11th Cir.1986) (in analyzing the strength of a trade dress, "the scope of protection increases as the trade dress moves toward the arbitrary end of the spectrum"), cert. denied, 481 U.S. 1041, 107 S. Ct. 1983, 95 L. Ed. 2d 822 (1987).

The district court found that while there was some similarity between the two trade dresses, there were sufficient differences for this factor to weigh in favor of Attiki. We believe this to have been clear error.

The dresses of the two bottles bear substantial similarities, both in their details and their overall appearance. Both bottles are clear, with a large main label and a smaller neck label using identical shades of red, white, and black. The labels are both broken into two fields, a white top and a red bottom. Both use black block lettering, with the numbers in very large type and the "#" and "No" in small type. Both "Ouzo" 's are in white block lettering over red fields, the only difference being subtle black shadowing on the No. 12 labels. The neck labels of the two bottles are nearly identical to each other.

The differences are minimal: an ellipse versus a square on the large labels; a plain black block-lettered "#1" versus a stencilled-look black block-lettered "No 12" (and this distinction disappears if one takes a few steps back); a plain white block-lettered "ouzo" versus a shadowed white block-lettered "ouzo"; and a few other very minor distinctions such as subtle differences in bottle shape, the design of the label borders, and the cap design.

More significant than the striking similarities in various details are the dresses' similarity in overall appearance. See Perfect Fit Indus. v. Acme Quilting Co., 618 F.2d 950, 954-55 (2d Cir.1980) (construing New York law of unfair competition). Each label's lettering style, layout, and coloration, taken together, convey the same impression: a design that is simple, clean, and stark. In light of the marked similarity between the two bottles, the district court's determination that they were not similar was clearly erroneous.

Attiki argues that we should examine the gift boxes in which the ouzos are shipped rather than the bottles (see Exhibit B). Although less similar, they are sufficiently so that this Polaroid factor still weighs in favor of Paddington. Both use the same colors and block lettering style as found on the labels. Each of the four side panels on the No. 12 Ouzo box is broken into a white field in the upper third and a red lower field. The top field contains a "No 12" identical to that on the bottle label, and the bottom has the same "ouzo" in large letters with the same thin "Kaloyannis" underneath it as on the bottle.

The #1 Ouzo box contains two different kinds of side panels. One is a drawing or photograph of a #1 Ouzo bottle. The other has a white background with a circle in the middle that is identical to that on the main bottle label and neck label. Toward the top of the panel are the words "DRY GREEK APERITIF" in thin, medium-sized lettering.

Despite a few more differences in details than between the bottles, the boxes nonetheless are sufficiently similar in overall impression to suggest that they are made by the same manufacturer. The district court's decision does not specifically describe and compare the boxes, but it may have considered them in reaching a conclusion of non-similarity. To the extent that it did, this was clear error.

Competitive Proximity of the Two Products and Likelihood of Bridging the Gap

The district court correctly found that the two ouzos were similar products that would compete in the same market, a factor that, as the district court noted, weighs in favor of a finding of likelihood of confusion. Since the ouzos would compete in the same market, the district court correctly found that the likelihood-of-bridging-the-gap factor, which examines whether the prior user may wish to enter the defendant's market in the future, was irrelevant.

The district court found that Attiki did not copy the No. 12 dress in bad faith since, although Attiki had been the distributor of No. 12 Ouzo for eight years, and thus undoubtedly was fully familiar with the No. 12 Ouzo trade dress, Attiki had not created the #1 Ouzo trade dress itself, but rather chose it from among three trade dresses and trademarks designed by Cavino.

Where a second-comer acts in bad faith and intentionally copies a trademark or trade dress, a presumption arises that the copier has succeeded in causing confusion. See Warner Bros. Inc. v. American Broadcasting Cos., 720 F.2d 231, 246-47 (2d Cir.1983) (trademark); Perfect Fit, 618 F.2d at 954 (trade dress); see also Bristol-Myers Squibb, 973 F.2d at 1044-45; Charles of the Ritz, 832 F.2d at 1322 (intentional copying "bolsters a finding of consumer confusion"). In determining a defendant's intent, "actual or constructive knowledge" of the prior user's mark or dress may indicate bad faith. Mobil Oil Corp. v. Pegasus Petroleum Corp., 818 F.2d 254, 259 (2d Cir.1987). Where such prior knowledge is accompanied by similarities so strong that it seems plain that deliberate copying has occurred, we have upheld findings of bad faith. See id. at 258-59; Charles of the Ritz, 832 F.2d at 1318-20, 1322.

Attiki, as the former distributor of No. 12 Ouzo, was of course intimately familiar with its trade dress. The similarities between the No. 12 and #1 Ouzo dresses are so striking that it is hard to imagine how Cavino and Attiki could have come up with the dress for #1 Ouzo without intentional copying. Intentional copying, of course, does not require identical copying. Where the copier references the prior dress in establishing her design with the apparent aim of securing the customers of the other based on confusion, intentional copying may be found. That is what occurred here. Attiki's bad faith is not diminished by the fact that it did not create the #1 Ouzo trade dress but rather selected it from three options presented to it by Cavino. Upon the full record, we hold the district court's finding of the absence of bad faith to be clearly erroneous.

The more sophisticated the consumers of a product are, "the less likely it is that similarities in trade dress or trade marks will result in confusion concerning the source or sponsorship of the product." Bristol-Myers Squibb, 973 F.2d at 1046. But see Centaur Communications, 830 F.2d at 1228 (noting that while customer sophistication typically weighs against a finding of likelihood of confusion, sophistication "might on occasion increase the likelihood of confusion, depending upon the circumstances of the market and the products").

The district court found that most ouzo drinkers in America are either Greek citizens or Greek-Americans, both of whom have a high level of brand awareness regarding the national drink of Greece. Therefore, the court found the potential purchasers of the two ouzos to be highly sophisticated. However, Two Pesos instructs that a user of a distinctive trade dress, even if a newcomer to a market or one whose product has not been successful, "should be able to maintain what competitive position it has and continue to seek wider identification among potential customers." --- U.S. at ----, 112 S. Ct. at 2759. While it may be true that today most consumers of the Greek national drink are Greek or of Greek descent, the Lanham Act protects Paddington's right to try to expand its market to include substantial numbers of people not of Greek origin and to rely on its distinctive trade dress to build brand loyalty. Thus it was error for the district court to confine the relevant market to persons of Greek origin in finding buyer sophistication, since a market of unsophisticated potential purchasers may exist as well.

Weighing the various Polaroid factors, we find, based on Paddington's evidence at trial, upon which the district court based its Rule 52(c) dismissal, that #1 Ouzo's trade dress creates a strong likelihood of confusion with No. 12 Ouzo's dress. The #1 Ouzo dress is so similar to No. 12 Ouzo's distinctive trade dress that it is likely that consumers would be confused, whether by believing that #1 Ouzo came from the producer of No. 12 Ouzo, or by confusing the two products outright. We note that the district court made errors of law in evaluating the strength of the mark based solely on secondary meaning, and in focusing its analysis on too limited a consumer group in the "sophistication" inquiry. While the district court's opinion does not reflect consideration of all of the factors it should have in its strength-of-the-mark analysis, based on the evidence Paddington presented at trial, it is plain that "No. 12 Ouzo" is a strong mark, and it was clear error to hold otherwise. As to the sophistication-of-the-consumer prong, the district court never addressed whether there was a broader, untapped market of ouzo drinkers who were unsophisticated. However, we need not remand this issue to the district court since even if the district court were to find that no such market existed and adhered to its finding of a sophisticated consumer group, the other Polaroid factors weigh so heavily in favor of a likelihood of confusion that our balancing of the factors would be unaffected.

Accordingly, the district court erred in dismissing Paddington's claim under Rule 52(c). We therefore reverse the district court's dismissal of the trade dress claim.

Paddington contends that Attiki infringed its unregistered trademarks "No. 12 Ouzo" and "World's No. 1 Ouzo," a phrase Paddington uses in advertising No. 12 Ouzo, by using a mark, "#1 Ouzo," that will cause confusion since it uses one of the digits of "12" and since No. 12 Ouzo, not #1 Ouzo, is the world's number-one seller.

Paddington's selection of the mark "No. 12 Ouzo" is arbitrary, and therefore is protectable under the Lanham Act without a showing of secondary meaning. The number "12" was selected from among an endless variety of names that could be given to an ouzo product. The district court found that at least four other brands of ouzo in Greece use a number designation for ouzo. This grew out of the former practice of ouzo producers in Greece to stamp barrels of ouzo with a number to distinguish their brand from another at a time when ouzo was sold in bulk. The district court correctly considered this common use of a number designation in evaluating the strength of the mark. We note that the common practice of using a number designation does not affect the arbitrary nature of the No. 12 Ouzo mark. While the decision to use a number designation for ouzo probably could not be considered arbitrary in light of the commonness of this practice in Greece, the choice of the number "12" was undeniably arbitrary.

Applying the first Polaroid factor, the district court found that the trademark "No. 12 Ouzo" was a weak mark. It based this determination largely on the fact that it is common in Greece to designate brands of ouzo by number. The district court also based its determination on the lack of secondary meaning of the No. 12 mark, which is a relevant consideration for determining a mark's strength, as discussed above. This determination was not clearly erroneous.

The district court also found that the marks were not similar. We agree. The "# 1 Ouzo" mark, while it shares the numeral "1" with "No. 12 Ouzo," cannot be said to be similar to it. The marks do not sound alike, nor do they look alike, except for the common use of the numeral "1". Furthermore, the use of the "#" symbol instead of "No." tends to reduce any similarity. We also note that in light of the common practice of ouzo producers of designating their ouzos by number, new market entrants likely will want to select a number for their ouzos. Were we to prevent them from using all of the integers already part of current ouzo producers' trademarks, or, as Paddington urges, prevent them from using the numbers surrounding the numbers of current brands, such as "# 11" or "# 13" Ouzo, the number of low-digit integers available would quickly disappear. Where there is no evidence of industry custom, and where larger numbers or numbers combined with letters are used for a trademark, courts have been more willing to find similarity between marks containing numbers. See Nabisco Brands, Inc. v. Kaye, 760 F. Supp. 25 (D.Conn.1991) ("A.2" mark for steak sauce infringed plaintiff's registered "A.1" steak sauce mark); Clorox Co. v. State Chemical Mfg. Co., 197 U.S.P.Q. 840 (T.T.A.B. 1977) (denying registration of "Formula 999" for degreasing compound due to similarity to opposer's registered mark "Formula 409" for degreaser). We also reject Paddington's argument that the two marks are similar because "No. 12 Ouzo" is the world's number-one seller. This connection is too attenuated to create a similarity between the marks. We therefore hold that the district court was not clearly erroneous in finding no similarity between "No. 12 Ouzo" and "# 1 Ouzo."As noted in our earlier discussion of trade dress, the district court properly found that the products would be in direct competition, which weighs toward a finding of likelihood of confusion, and that the likelihood of bridging the gap factor and the evidence of actual confusion factor are not applicable here.

Our holding that the district court erred in finding no bad faith in Attiki's adoption of the # 1 Ouzo trade dress might normally militate toward a conclusion that the court erred in finding no bad faith as to the adoption of the trademark. However, in light of the fact that the trademarks are dissimilar and the common practice in Greece of branding ouzos with numbers, any bad faith by Attiki in choosing the trademark "# 1 Ouzo" is of minimal significance in evaluating likelihood of confusion. See Warner Bros., 720 F.2d at 231 ("if comparison of the works reveals no fair jury issue concerning likelihood of confusion, then intent to copy, even if found from the proffered evidence, would not establish a Lanham Act violation.").

Finally, as discussed above, the district court erred in not considering whether there was a broader market of potential ouzo drinkers in the United States. However, balancing the various Polaroid factors, we hold that regardless of the existence or non-existence of such a market, there is not a likelihood of confusion between the two marks. We therefore affirm the district court's order dismissing Paddington's trademark claim under Rule 52(c) and its entry of judgment for Attiki on that claim.

The district court's Rule 52(c) dismissal of Paddington's trademark infringement claim and entry of judgment for Attiki on that claim is affirmed. The dismissal of Paddington's trade dress claim and its entry of judgment for Attiki is reversed. We remand the case to the district court for further proceedings, including Attiki's presentation of evidence. See Trans-Orient Marine Corp. v. Star Trading & Marine, Inc., 925 F.2d 566, 571 (2d Cir.1991) (remand is correct procedure upon reversing dismissal under 41(b), the predecessor to Rule 52(c)). Our reversal of the district court's dismissal of Paddington's trade dress carries with it the reinstatement of Paddington's common law and state statutory claims.

Affirmed in part, reversed in part, and remanded. Costs are awarded to Paddington.APPENDIX A

NOTE: OPINION CONTAINS TABLE OR OTHER DATA THAT IS NOT VIEWABLEAPPENDIX B

NOTE: OPINION CONTAINS TABLE OR OTHER DATA THAT IS NOT VIEWABLE