Petro Stopping Centers, L.P. ("Petro Stopping") sued James River Petroleum, Inc. ("James River"), alleging federal and common law trademark infringement, federal and common law unfair competition, federal dilution, and Virginia false advertising claims. After a bench trial, the district court entered judgment for James River on all counts. The court found Petro Stopping's marks to be fairly weak and concluded that the remaining factors did not support a likelihood of confusion. We now affirm.

Petro Stopping operates and franchises over forty full service automobile and truck travel centers in twenty-seven states, including one each in Ruther Glen, Virginia and Fort Chiswell, Virginia. The company has been in business since 1975. Most of Petro Stopping's facilities operate under the marks PETRO STOPPING CENTER or PETRO TRAVEL PLAZA.

Petro Stopping has a total of fourteen registered federal service marks that contain the term PETRO. Examples include PETRO PRIDE, PETRO:LUBE, PETRO:2, PETRO:PLUS, and PETRO:TREAD. No registration is for the word PETRO alone, although the company has registered PETRO with background colors as a composite mark for truck stop services. The PETRO mark is almost always displayed in white letters with a green rectangular background, placed above a red-orange strip. Petro Stopping's first registered mark, PETRO STOPPING CENTER, was placed on the Principal Register of the Patent and Trademark Office ("PTO") in August 1982. The company registered its eighth service mark, the PETRO with background colors composite mark, in July 1990. At least eight of Petro Stopping's marks, including the registration for PETRO with back ground colors, have become incontestable pursuant to 15 U.S.C. § 1065.

Petro Stopping's truck stop facilities are located adjacent to interstate highways at the exits. They offer a wide variety of goods and services, including separate fueling areas for cars and trucks, truck maintenance, truck weighing scales, truck washes, restaurants, retail stores, fax machines, ATM machines, showers, laundry, quiet rooms, game rooms, television rooms, movie theaters, and barber shops. These travel facilities are staffed twenty-four hours a day. Customers purchasing fuel or other goods and services at Petro Stopping's sites can pay with cash, checks, or credit cards, including a PETRO credit card. The company advertises its facilities in radio spots, in publications directed to truckers, in truck stop directories, and on billboards and Virginia Department of Transportation ("VDOT") exit signs. The company claims that over the past five years it has sold in excess of $2 billion of goods and services and has spent over $6.25 million in promoting its truck stop services.

James River is engaged in petroleum distribution. In 1992, the company expanded into the "card lock" business. In this enterprise, James River generally sells its services to commercial owners of car or truck fleets. These fleet owners purchase accounts that allow their drivers or authorized agents to use payment cards at James River's unmanned, self-service filling stations. James River participates in the Commercial Fueling Network ("CFN"), and is therefore part of a network of independent card lock fuel facilities. Before trial, the company had five Richmond area stations and planned to open a sixth. James River operates its card lock facilities under the name JAMES RIVER PETRO CARD. The James River mark uses a white (or sometimes gray) background and contains one stylized, red "JR" logo, in large font. Next to the logo, in a much smaller font, appear the words "James River" in blue upper and lower case letters. Below that, in an even smaller font, the words "PETRO CARD" appear in slanted red upper case letters with horizontal lines drawn to give the effect that the words are moving.1

James River's card lock facilities are not located adjacent to interstate highways. Although one of the company's fuel stations is located near a separately-owned convenience store and another near a separately-owned restaurant, the company's card lock facilities do not offer any services comparable to Petro Stopping's truck stops except for fueling. James River's stations are not staffed. Fuel can be purchased only through the use of a card issued by James River or another CFN member. The company has run limited advertisements in small local trade newsletters, but does not utilize any billboards or VDOT signs. The company's sites are also listed in CFN directories. James River focuses its marketing, however, on the direct solicitation of commercial accounts.

Petro Stopping filed this suit against James River on July 1, 1996. The district court held a bench trial on December 2 and 3, 1996, and ruled for James River in all respects. On the trademark infringement and unfair competition claims, the district court found no likelihood of confusion, noting specifically that PETRO is "merely descriptive and a fairly weak mark." Petro Stopping now appeals.

To prove trademark infringement, a plaintiff must show both that it has a valid, protectable trademark and that the defendant's use of a "reproduction, counterfeit, copy, or colorable imitation," 15 U.S.C. § 1114(1), creates a likelihood of confusion. Lone Star Steakhouse & Saloon, Inc. v. Alpha of Va., Inc., 43 F.3d 922, 930 (4th Cir.1995). Although Petro Stopping sometimes conflates these two distinct elements in its arguments to this court, only the second requirement--likelihood of confusion--is at issue in this appeal.2 In determining whether a plaintiff has demonstrated likelihood of confusion, courts of this circuit are guided by the factors announced in Pizzeria Uno Corp. v. Temple, 747 F.2d 1522, 1527 (4th Cir.1984). A court should consider (1) the strength or distinctiveness of the plaintiff's mark, (2) the similarity of the two parties' marks, (3) the similarity of the goods and services the marks identify, (4) the similarity of the facilities the two parties use in their businesses, (5) the similarity of advertising used by the two parties, (6) the defendant's intent, and (7) actual confusion. These factors are not of equal importance or equal relevance in every case. Id.

This circuit reviews district court determinations regarding likelihood of confusion under a clearly erroneous standard. Id. at 1526. We may find a district court's conclusions clearly erroneous only when there is no evidence in the record to support those conclusions or when, upon review of the record ourselves, we are left with a definite and firm conviction that a mistake has been committed. Id. We have previously recognized that likelihood of confusion is an inherently factual issue. Lone Star Steakhouse, 43 F.3d at 933. In this appeal, where only this type of factual determination is contested and the decision was rendered after a two-day bench trial, the standard of review is a deferential one requiring appellate respect for the trial court's findings.

The district court found that the first Pizzeria Uno factor--the strength of the plaintiff's mark--favored a finding of no likelihood of confusion because PETRO is descriptive and weak. Petro Stopping first contends that because eight of its PETRO marks are incontestable, the district court was legally precluded from reaching its conclusion. The company claims that the Supreme Court's decision in Park 'N Fly, Inc. v. Dollar Park & Fly, Inc., 469 U.S. 189, 105 S. Ct. 658, 83 L. Ed. 2d 582 (1985), forecloses any argument by James River that PETRO is descriptive.

We disagree. Petro Stopping misapprehends the Park 'N Fly holding and confuses the issue of a trademark's validity with the separate inquiry into a mark's strength for purposes of the likelihood of confusion determination. In Park 'N Fly, the Supreme Court held that a defendant in an infringement suit cannot claim that an incontestable mark is merely descriptive and therefore invalid and undeserving of any protection. The Court did not hold, however, that the descriptive nature of a mark may not be considered in the separate likelihood of confusion inquiry. This characterization of the Court's holding is confirmed by the Trademark Law Revision Act of 1988, which amended one of the Lanham Act's incontestability provisions to clarify that "[s]uch conclusive evidence of the right to use the registered mark shall be subject to proof of infringement." Pub.L. No. 100-667, § 128(b)(1), 102 Stat. 3935, 3945 (codified at 15 U.S.C. § 1115(b)) (emphasis added).

We recognized this crucial distinction in Lone Star Steakhouse. There the district court had concluded that the incontestability of the mark in question was dispositive of the likelihood of confusion inquiry. After considering Park 'N Fly, the holdings of other circuits, and the opinions of the commentators on this issue, we stated:

We agree with the above reasoning that incontestability affects the validity of the trademark but does not establish the likelihood of confusion necessary to warrant protection from infringement. Likelihood of consumer confusion remains an independent requirement for trademark infringement.... [We] hold that we are free to address whether Plaintiffs' incontestable trademark is descriptive or suggestive in determining whether the likelihood of consumer confusion exists in this case.

Lone Star Steakhouse, 43 F.3d at 935. Similarly, in Shakespeare Co. v. Silstar Corp. of America, Inc., 110 F.3d 234, 238-40 (4th Cir.), petition for cert. filed, 66 U.S.L.W. 3283 (U.S. Oct. 1, 1997), we held that although a district court cannot cancel an incontestable trademark on grounds of functionality or descriptiveness, it can and should consider those grounds when determining whether likelihood of confusion has been established. Thus, our precedent squarely forecloses the argument Petro Stopping seeks to raise in this appeal.

The company next claims that the district court was precluded from finding PETRO to be descriptive and weak because the PTO had already found its marks to be at least suggestive. The PTO requires applicants with marks found to be descriptive to prove that the marks have secondary meaning before granting registration. 15 U.S.C. § 1052(f). Because the PTO did not require such proof with respect to Petro Stopping's applications for registration, it can safely be assumed that the PTO believed the marks were at least suggestive. Lone Star Steakhouse, 43 F.3d at 934. Relying on RFE Indus., Inc. v. SPM Corp., 105 F.3d 923, 926 (4th Cir.), cert. denied, --- U.S. ----, 117 S. Ct. 2512, 138 L. Ed. 2d 1015 (1997), Petro Stopping argues that a district court is not free to substitute its own finding for that of the PTO. Despite the PTO's apparent conclusions, we find no clear error in the district court's determination that PETRO is a descriptive, weak mark.

Although courts should accord deference to the PTO's findings when assessing the strength of a mark under the likelihood of confusion test, see Pizzeria Uno, 747 F.2d at 1534, the PTO's "determination is not conclusive." Id. The fact that the PTO does not view a mark as descriptive serves only as prima facie evidence that the mark is suggestive. Accordingly, the presumption may be rebutted by the defendant. Id.; see also Champions Golf Club, Inc. v. The Champions Golf Club, Inc., 78 F.3d 1111, 1118 (6th Cir.1996) (remanding to district court "to determine whether the mark is descriptive or suggestive" despite noting that PTO's registration created presumption that mark was not descriptive).

More significantly, the placement of a mark in either the suggestive or descriptive category is merely the first step in assessing the strength of a mark for purposes of the likelihood of confusion test. Lang v. Retirement Living Publishing Co., 949 F.2d 576, 581 (2d Cir.1991); Western Publishing Co. v. Rose Art Indus., Inc., 910 F.2d 57, 61 (2d Cir.1990); Miss World (UK) Ltd. v. Mrs. America Pageants, Inc., 856 F.2d 1445, 1449 (9th Cir.1988). Even a mark held to be suggestive may be found weak under the first likelihood of confusion factor. See Estee Lauder Inc. v. The Gap, Inc., 108 F.3d 1503, 1509-11 (2d Cir.1997) (finding mark suggestive, but holding district court's conclusion that mark was moderately strong to be clearly erroneous). The strength of a mark " 'ultimately depends on the degree to which the designation is associated by prospective purchasers with a particular source.' " Id. at 1510 (quoting Restatement (Third) of Unfair Competition § 21 cmt. i (1995)). Thus, courts must examine, in addition to the mark's characterization as suggestive or descriptive, the extent of secondary meaning a mark has acquired in the eyes of consumers. Lang, 949 F.2d at 581; Western Publishing, 910 F.2d at 61; Miss World (UK), 856 F.2d at 1449. This secondary meaning evidence may also serve to rebut the presumption arising from PTO action that a mark is not descriptive.

The district court considered just such evidence of secondary meaning, customer confusion, and third-party use in finding PETRO to be descriptive and weak. The court specifically found that "[i]n the present context, PETRO does not have secondary meaning in the card lock business." It also concluded that the lack of customer confusion evidence confirmed the weakness of PETRO as a mark. Finally, the district court found that a "number of other companies commonly use PETRO to describe petroleum related goods and services." In reviewing the record on which the district court based its conclusion, we are not left with a definite and firm conviction that a mistake has been made.

Substantial evidence existed in the record showing the considerable number of third-party registrations and uses of the term PETRO. James River submitted evidence showing that over 2,700 businesses in the United States use PETRO as part of their name. The company also submitted evidence showing that 117 third-party federal registrations or applications for registration contain the word PETRO, sixty-three of which are for use in connection with petroleum or fuel-related goods and services. Even before Petro Stopping registered its first service mark for PETRO STOPPING CENTER in 1982, forty-two other marks including the term PETRO existed on the trademark register.

Petro Stopping argues, however, that evidence of third-party registrations alone is insufficient to conclude that a mark is weak. The company maintains that only proof that third parties actually use the term PETRO would be relevant. We disagree. The frequency with which a term is used in other trademark registrations is indeed relevant to the distinctiveness inquiry under the first likelihood of confusion factor. Pizzeria Uno, 747 F.2d at 1530-31. This is especially true when the number of third-party registrations is great. For example, in Amstar Corp. v. Domino's Pizza, Inc., 615 F.2d 252, 259-60 (5th Cir. 1980), the court found it impossible to dismiss evidence of seventy-two third-party registrations in the PTO of the appellant's mark. See also Estee Lauder, 108 F.3d at 1511 (weakness of mark demonstrated by fact that over seventy trademark registrations, pending applications for registration or renewal, or publications-for-opposition included the term used in plaintiff's mark). "Specifically, third-party registrations are 'relevant to prove that some segment of the composite marks which both contesting parties use has a normally understood and well recognized descriptive or suggestive meaning, leading to the conclusion that that segment is relatively weak.' " First Sav. Bank, F.S.B. v. First Bank Sys., Inc., 101 F.3d 645, 654 (10th Cir.1996) (quoting 1 J. Thomas McCarthy, McCarthy on Trademarks and Unfair Competition § 11.27[b] (3d ed.1995)). Therefore, the abundant evidence of third-party registration and use of the term PETRO solidly supports the district court's conclusion that PETRO is a fairly weak mark.

Finally, Petro Stopping's own representations to the PTO in 1981 undercut the company's current arguments to this court. At that time, the PTO objected to registration of Petro Stopping's service mark because of likelihood of confusion with previously registered trademarks that included the term PETRO. Petro Stopping contended then that, because of the wide use and registration of PETRO as part of composite trademarks, "the term 'Petro' is entitled to a very narrow scope of protection." The company further urged "that the term is extremely dilute and entitled to a weak scope of protection." Petro Stopping now asks this court to accept precisely the opposite proposition. The company's promotion of, and commerce under, its service marks may have strengthened them since the time of its trademark counsel's representations. However, in light of the considerable number of federally registered trademarks using the term PETRO, and the extent of its use by businesses generally, the strength of Petro Stopping's marks must be understood to be confined to the field for which they are registered, namely truck stop services. See Amstar Corp., 615 F.2d at 260 ("The third-party uses and registrations discussed above merely limit the protection to be accorded plaintiff's mark outside the uses to which plaintiff has already put its mark."). In Petro Stopping's own words, outside of truck stop services its mark is "entitled to a weak scope of protection." In short, the district court's conclusion that PETRO is a descriptive and weak mark is not a clearly erroneous one.

Petro Stopping next argues that the district court erred by finding the two companies' marks to be dissimilar. Petro Stopping contends that the district court should have focused on the similarities between the marks rather than emphasizing the dissimilarities. This argument fails to persuade. In assessing similarity of marks under the second Pizzeria Uno factor, courts generally focus on the dominant portions of parties' marks. Lone Star Steakhouse, 43 F.3d at 936; Pizzeria Uno, 747 F.2d at 1534-35. This practice does not mean that courts confine their scrutiny merely to similarities. Because every triable case of trademark infringement involves marks that are similar at some level, Petro Stopping's rule would obviate the need for courts ever to inquire into the second Pizzeria Uno factor.

We noted in Pizzeria Uno that an overall test under this factor is whether there exists a "similarity in appearance and sound which would result in confusion." Id. at 1534. Here, the companies' two marks present entirely different appearances. Petro Stopping's service mark is displayed in green, white, and red-orange. James River's mark utilizes red, blue, white, and sometimes gray. Petro Stopping's mark never contains the words "James River," never includes a JR logo, and never is portrayed with horizontal lines suggesting movement. To the extent that the James River mark has a dominant portion, it appears to be the large JR logo, not the term PETRO. Because of these marked differences in appearance, we do not find the district court's conclusion under this factor to be clearly erroneous.

Petro Stopping next maintains the district court erred by finding the third and fourth Pizzeria Uno factors--similarity of goods, services, and facilities--to support a finding of no likelihood of confusion. Besides the sale of fuel, the two parties' services and facilities differ in virtually every respect. Petro Stopping's travel center facilities offer, among other services, maintenance, weighing scales, truck washes, restaurants, retail stores, fax and ATM machines, showers, laundry, quiet rooms, game rooms, television rooms, movie theaters, and even barber shops. James River's facilities sell only fuel. Petro Stopping's travel center sites are located adjacent to interstate highways near the exits and usually occupy anywhere between fifteen to thirty acres of land. James River's card lock facilities are not visible from interstate highways, not located at interstate exits, and normally occupy only one acre or less of land. Even Petro Stopping's own Chief Executive Officer admitted that the distinctive design utilized by his company's facilities would suffice to distinguish them from a hypothetical truck stop using the term PETRO if opened by another company within twenty miles. Even if these stark differences did not suffice, it would be nearly impossible for a person seeking a Petro Stopping facility to be lured into purchasing fuel at a James River station because the latter accepts only cards issued by James River or other CFN members as a means of payment. The district court's findings under the third and fourth factors, therefore, were not clearly erroneous.

Petro Stopping next argues that the district court erred by finding that the fifth Pizzeria Uno factor--similarity of advertising--supported a finding of no likelihood of confusion. The company contends specifically that James River promotes its services in the CFN directories, in a manner similar to Petro Stopping's own promotion of its services in truck stop directories. We are not persuaded. It is true that the CFN directories, which are available only to members of the CFN network and not to the public generally, do list the James River stations as member companies. However, because truck drivers whose employers hold an account with a CFN member probably would frequent only those facilities at which they can purchase fuel through their account, the listing of James River's facilities in the CFN directories is unlikely to create confusion with Petro Stopping's travel centers. Furthermore, Petro Stopping ignores important differences between the types of advertising practiced by each company. Petro Stopping advertises in radio spots, in publications directed at truckers, on billboards, and on VDOT exit signs. The company emphasizes the diversity of services available at its facilities. James River has run only limited advertisements in small, local newsletters. The company promotes the advantages of the card lock system, not the variety of available services. We identify no clear error in the district court's finding under the similarity of advertising factor.

Finally, Petro Stopping claims that the district court erred under the last Pizzeria Uno factor--actual confusion. The district court found that there was little evidence of such confusion in the record. Petro Stopping maintains that courts normally accord significance to even one instance of actual confusion. This argument, however, sweeps too broadly. Petro Stopping submitted evidence showing it is a significant commercial actor in the truck stop industry. In light of its huge volume of commerce, Petro Stopping's meager evidence of actual confusion is at best de minimis. Universal Money Ctrs., Inc. v. American Tel. & Tel. Co., 22 F.3d 1527, 1535 (10th Cir.1994). At worst, the company's failure to uncover more than a few instances of actual confusion creates a "presumption against likelihood of confusion in the future." Amstar Corp., 615 F.2d at 263. We perceive no clear error in the district court's determination under this final factor.

The district court's findings on each of the Pizzeria Uno factors were firmly grounded in the evidence. We also hold that the district court's ultimate finding of no likelihood of confusion, based on its overall perception of the evidence adduced at trial, is not a clearly erroneous one. For the foregoing reasons, we affirm the judgment of the district court.

AFFIRMED.