Jellibeans, Incorporated, a Georgia Corporation, Cross-Appellant v. Skating Clubs of Georgia, Inc., a Georgia Corporation, Cross-Appellee

TJOFLAT, Circuit Judge: Skating Clubs of Georgia, Inc. appeals from a district court order enjoining it from using “Lollipops” as a name for one of its roller skating rinks because that name is too easily confused by consumers with the service mark 1 used by Jellibeans, Inc. to identify its roller rink, “Jellibeans.” Jellibeans, Inc. cross-appeals the district court’s denial of attorney fees. We affirm on both points. I. In November 1979, Jellibeans, Inc. opened a roller rink in Atlanta, Georgia, named “Jellibeans.” It advertised Jellibeans extensively, spending over $50,000 in November and December alone, and over $100,000 during the first year of the rink’s operation. Jellibeans, Inc. spent most of this money on radio advertisements, and the rest on newspaper advertisements and local promotional activities. Jellibeans, Inc.’s extensive advertising, and its use of a candy name for its rink, were unique among Atlanta roller rinks. The rink was a success from the outset. Skating Clubs of Georgia, Inc., operates six roller rinks in the Atlanta area. Skating Clubs established its first rink in 1959, another in 1971, two in 1973, and a fifth rink in 1975. It called each of these rinks *836 “Skating Clubs” and named them after the communities in which they were located, e.g., South Cobb Skating Club. In early 1980, a few months after Jellibeans opened, Skating Clubs decided to build its sixth rink. It purchased land for this new rink approximately 8-9 miles from Jellibeans and immediately began construction. It placed a sign on the site announcing the future opening of “Lollipops Disco Roller Rink.” The sign also stated, in small print, that Skating Clubs owned the rink. Skating Clubs later replaced this sign with one that omitted the word “Disco” from the name of the rink, and stated in much larger letters that Skating Clubs owned Lollipops, and that Albert Couey owned Skating Clubs. Several patients and a business acquaintance of Dr. Ronald Feinman, a principal shareholder of Jellibeans, Inc., saw the Lollipops sign. 2 They thought the name Lollipops was very similar to the name Jellibeans and therefore called Dr. Feinman to inquire whether he was opening a new rink. Dr. Feinman replied that he was not. He later became concerned that others, especially Jellibeans’ present and future customers, might also become confused by the similarity of the names and thus referred the matter to Jellibeans, Inc.’s counsel. Counsel concluded that Skating Clubs’ use of the name Lollipops infringed upon the company’s service mark, Jellibeans. Accordingly, on May 15,1980, counsel requested that Skating Clubs stop using the name Lollipops. Skating Clubs replied that its use of the name did not infringe on Jellibeans, Inc.’s service mark and therefore refused to comply with the request. On September 26, 1980, Jellibeans, Inc. brought a four-count complaint in the district court alleging that Skating Clubs’ use of the name Lollipops: (1) constituted a false description or representation of Skating Clubs’ services, in violation of the Lanham Act, 15 U.S.C. § 1125(a) (1976); 3 (2) infringed Jellibeans, Inc.’s service mark in violation of Georgia trademark law, 106 Ga. Code Ann. § 111; 4 (3) constituted a deceptive trade practice in violation of the Georgia Uniform Deceptive Trade Practices Act, 106 Ga.Code Ann. § 702; 5 and (4) constituted unfair competition under the common *837 law of Georgia. 6 Jellibeans, Inc. sought a declaratory judgment that Skating Clubs’ conduct was unlawful as alleged, and an injunction. It also sought attorney fees under 15 U.S.C. § 1117 (1976) and 106 Ga. Code Ann. 703(b). 7 Skating Clubs denied these allegations, and the case proceeded to trial. At trial, Jellibeans, Inc. called two patients and a business acquaintance of Dr. Feinman to testify. They stated that they saw Lollipops’ sign, believed that Jellibeans, Inc. was opening a new rink because of the similarity of the names, and called Dr. Feinman to inquire whether he was opening a new rink. Jellibeans, Inc. also produced a survey that showed 96% of those interviewed had heard of Jellibeans and 48% of them believed that the Lollipops rink was related to it. Skating Clubs objected to the introduction of this survey in evidence, contending that it was so flawed that it lacked probative value. 8 The court admitted the survey, however, ruling that the deficiencies impacted the weight to be accorded the survey and not its admissibility, a ruling Skating Clubs challenges on appeal. In its defense, Skating Clubs presented the testimony of owner, Albert Couey, and his nephew Jerry Taylor, who brokered the sale of the real estate where Lollipops was being built and who served as Skating Clubs’ insurance agent. Both Couey and Taylor testified that Taylor suggested the name Lollipops. Both denied that they had considered the similarity of the names Lollipops and Jellibeans. They admitted, however, that they had been aware of Jellibeans. Couey testified that the original sign announcing the construction of Lollipops stood only two days before it was burned by a fire of unknown origin. He testified that Skating Clubs replaced it with an identical sign, except for the word “Disco” which was omitted from the name of the rink. Jellibeans, Inc. materially impeached this testimony, however, by introducing pictures of the two signs that showed the substantial differences between the two, particularly with regard to the ownership of the rink. Finally, Couey testified that he promoted, and would in the future promote, the Lollipops rink only by circulating brochures to local schools and PTA’s, giving luncheons and teas for community groups, and newspaper advertisements. 9 He stated that he had promoted his other rinks in this manner and intended to do the same with Lollipops. On May 15, 1981, the district court rendered its decision. It dismissed the state law trademark infringement count because Jellibeans, Inc. had not registered its service mark in Georgia. 10 The court then held that Jellibeans, Inc.’s, claims under the Lanham Act, the Georgia Uniform Deceptive Trade Practices Act, and the Georgia law of unfair competition turned on the same question — whether rollerskaters, the purchasers of Jellibeans’ and Lollipops’ services, were likely to confuse Lollipops *838 with Jellibeans. 11 The court found that rollerskaters were likely to confuse the two. It therefore declared Skating Clubs’ use of the name Lollipops unlawful and permanently enjoined its further use. The court denied Jellibeans, Inc.’s request for attorney fees under 15 U.S.C. § 1117, however, holding that Skating Clubs did not possess the type of intent that warranted a fee award. It did not rule on Jellibeans, Inc.’s request for attorney fees under 106 Ga.Code Ann. § 703(b). Skating Clubs appeals. It contends that the district court was clearly erroneous in finding that rollerskaters are likely to confuse Lollipops with Jellibeans. 12 Jellibeans, Inc. cross-appealing, contends that the district court abused its discretion in denying it attorney fees under 15 U.S.C. § 1117. Jellibeans, Inc. does not question, however, the district court’s failure to rule on its right to attorney fees under 106 Ga.Code Ann. § 703(b). II. We note subject matter jurisdiction in this case. Jurisdiction of Jellibeans, Inc.’s Lanham Act claim is conferred by the Lanham Act itself, 15 U.S.C. § 1121. 13 This statutory grant of jurisdiction, however, may not exceed the constitutional limitations on the power of Congress to regulate service marks; rather, it reaches and is coincident with the constitutional boundary embodied in the commerce clause. The Lanham Act proscribes the false description or representation of services that “enter into commerce.” 15 U.S.C. § 1125(a). The use of “[t]he word ‘commerce’ [in the Act] means all commerce which may lawfully be regulated by Congress.” 15 U.S.C. § 1127. This grant of jurisdiction has been construed to be “at least as broad as the definition of commerce employed in any other federal statute.” Shatel Corp. v. Mao Ta Lumber & Yacht Corp., 697 F.2d 1352, 1356 n. 3 (11th Cir.1983); Bulova Watch Co. v. Steele, 194 F.2d 567, 570 n. 11 (5th Cir. 1952), aff’d, 344 U.S. 280, 73 S.Ct. 252, 97 L.Ed. 252 (1952). Nonetheless, purely intrastate disputes do not fall within the commerce clause and therefore are not subject to the Lanham Act’s regulation. Iding v. Anaston, 266 F.Supp. 1015, 1019 (N.D.Ill. 1967). We are, accordingly, required to determine in this case whether: (1) the district court’s finding of fact as to the amount of interstate contact can be accepted as not clearly erroneous; and (2) whether as a matter of law the interstate contacts are sufficient to satisfy the jurisdictional requirements of the Lanham Act, 15 U.S.C. §§ 1121, 1125. The district court found that Jellibeans drew patrons from an area that was within 60 miles of Atlanta, and included parts of Alabama, as well as receiving some out-of-state convention business. In addition, the trial court found that Jellibeans received some free publicity in national magazines. None of the evidence *839 on which these conclusions were based was rebutted at trial. Therefore, we accept the district court’s findings as not clearly erroneous. These facts establish that Jellibeans used its mark in interstate commerce sufficiently to invoke the .court’s jurisdiction. There was no requirement that Skating Clubs’ business be shown to be interstate as well. “ ‘[I]n commerce’ ... does not mean that an infringer is immune from prosecution under the statute so long as he keeps his infringement entirely within the confines of a state.” Coca-Cola Co. v. Stewart, 621 F.2d 287, 290 (8th Cir.1980); World Carpets, Inc. v. Dick Littrell’s New World Carpets, 438 F.2d 482, 488 (5th Cir.1971). III. A. The district court held that the Georgia deceptive trade practices and unfair competition counts involved the same dispositive question as the federal Lanham Act count. This is a question of state law that the parties do not challenge on appeal. Therefore, we treat this holding as correct 14 and merely determine whether the district court properly decided the Lanham Act count. If we determine that the district court decided the Lanham Act count properly, we will also affirm its decision on the Georgia deceptive trade practices and unfair competition counts. B. The law of trademark and service mark infringement is part of “the rather swampy area of unfair competition.” John H. Harland Co. v. Clarke Checks, Inc., 711 F.2d 966 (11th Cir.1983) (quoting B.H. Bunn Co. v. AAA Replacement Parts Co., 451 F.2d 1254, 1258 (5th Cir.1971) (Goldberg, J.)). A service mark is a form of property and has property rights associated with it. The definition of the property right, where it begins and ends, is the correlative of the question of whether a given act infringes on the property right. Determining whether an infringement has taken place is but the obverse of determining whether the service mark owner’s property right extends into a given area. The property right entailed in a service mark has an amorphous, multi-dimensional character that is subject to change with circumstances. The factual setting surrounding the mark, including how it is created, where it is located, and how it is putatively infringed is crucial to determining whether such infringement has taken place. There is no simple, or for that matter complicated, test by which the question of infringement is resolved. The test for a service mark infringement under 15 U.S.C. § 1125 has often been expressed as whether or not the offending mark is “likely to cause confusion.” 15 Harland, 711 F.2d at 961; Amstar Corp. v. Domino's Pizza, Inc., 615 F.2d 252, 258 (5th Cir.1980), cert. denied, 449 U.S. 899, 101 S.Ct. 268, 66 L.Ed.2d 129 (1980); Boston Professional Hockey Ass’n v. Dallas Cap & Emblem Mfg., Inc., 510 F.2d 1004, 1013 (5th Cir.1975), cert. denied, 423 U.S. 868, 96 S.Ct. 132, 46 L.Ed.2d 98 (1975). Thus the dispositive issue in the court below was whether rollerskaters — the purchasers of Jellibeans’ and Lollipops’ services — are likely to confuse Lollipops’ services as being rendered, authorized, sponsored, or endorsed by Jellibeans. The standard of review applicable to the district court’s finding of a likelihood of confusion is dependent upon whether this is a question of law, or of fact, or a hybrid embodying both. This circuit has held that it is a question of fact, Harland, 711 F.2d at 972-973, and can be set aside only if clearly *840 erroneous. Safeway Stores, Inc. v. Safeway Discount Drugs, Inc., 675 F.2d 1160, 1163 (llth Cir.1982); Amstar, 615 F.2d at 257-58; see also Fed.R.Civ.P. 52(a). 16 The “likely to confuse” test is not as unidimensional as its appellation would imply. While there are no subparts to this test in the sense of necessary and sufficient conditions there are a number of factors which may bear on the question of likelihood of confusion. In Safeway, this court enunciated a number of such factors: (1) How distinctive is the plaintiff’s mark? Do consumers strongly identify plaintiff’s service mark with his services? This is important because the more distinctive the plaintiff’s mark, the stronger it is considered, and the more protection it is accorded from confusingly similar marks; (2) How similar are the designs of the plaintiff’s and defendant’s marks? Obviously, the greater the similarity, the greater the likelihood that purchasers will confuse the plaintiff’s and defendant’s services. This principle also applies to subsidiary facts 3, 4, and 5; (3) How similar are the services the plaintiff’s and defendant’s marks represent; (4) How similar are the plaintiff’s and defendant’s retail outlets and their customers; (5) How similar is the advertising the plaintiff and defendant use; (6) Whether the defendant intended that purchasers would confuse his service mark with the plaintiff’s mark; and (7) Whether people were actually confused by the similarity of the marks. See Safeway, 675 F.2d at 1164 (and cases cited therein). The court evaluates and weighs these subsidiary findings to determine, as a matter of fact, whether consumers are likely to confuse the defendant’s services with the plaintiff’s services. 17 Since it is undisputed that the district court identified this analysis as the correct standard for deciding Jellibeans’ Lanham Act claim, our function on review is merely to determine whether the district court properly applied this factual analysis. More specifically, we must determine whether the district court was clearly erroneous in finding the subsidiary facts listed above, and in finding the ultimate fact— whether consumers are likely to confuse Lollipops with Jellibeans. See id. at 1164 n. 5; see Exxon Corp. v. Texas Motor Exchange, 628 F.2d 500, 506 (5th Cir.1980); Armstrong Cork Co. v. World Carpets, Inc., 597 F.2d 496, 501-02 (5th Cir.1979), cert. denied, 444 U.S. 932, 100 S.Ct. 277, 62 L.Ed.2d 190 (1979). We can reverse the district court’s decision only if after reviewing all the evidence, we are “left with the definite and firm conviction that a mistake has been committed,” Safeway, 675 at 1164, quoting United States v. United States Gypsum Co., 333 U.S. 364, 395, 68 S.Ct. 525, 541-42, 92 L.Ed. 746 (1948), by the district court in finding the ultimate fact issue. We can reach this decision in two ways. First, we can determine that the court clearly erred in finding one or more of the subsidiary facts, and that this mistake caused it clearly to err in its decision on the ultimate fact issue. In addition, we can determine that the district court did not *841 clearly err in its findings on any of the subsidiary facts, but nonetheless clearly erred in its decision on the ultimate fact issue. We now examine the evidence to determine whether the district court clearly erred in either of these ways. With regard to how distinctive a mark Jellibeans is, the court noted that Jellibeans is an arbitrary mark, 18 the type of service mark that is generally considered to be inherently distinctive. 19 See Sun Banks v. Sun Federal Savings & Loan, 651 F.2d 311, 315 (5th Cir.1981). The court also noted that Jellibeans is the only roller rink in the United States using that mark or a similar candy name. 20 Finally, the court found that because of Jellibeans, Inc.’s extensive advertising, the service mark Jellibeans was readily identifiable by the relevant market of young skaters in the Atlanta area. Based on these facts, the district court found the Jellibeans was a strong and distinctive service mark “entitled to broad protection in this geographic market as applied to another roller skating rink.” We have no difficulty upholding this finding. Jellibeans clearly is an arbitrary mark as applied to roller rinks. Survey evidence introduced by Jellibeans, Inc. showed that 96% of those interviewed had heard of Jellibeans, indicating strong recognition of the mark. Finally, there is ample evidence in the record that Jellibeans is the only roller rink in the United States using a candy name. In light of this evidence, we hold that the district court’s finding that Jellibeans is a very distinctive mark entitled to broad protection is not clearly erroneous. The second subsidiary fact that the district court examined is the similarity be *842 tween the design of Jellibeans, Inc.’s service mark, and the design of Skating Clubs’ service mark. This fact, the court held, is determined by considering the overall impression created by' the marks. See Amstar, 615 F.2d at 260-61; Armstrong Cork Co., 597 F.2d at 502. The district court found a visual and aural similarity to exist between these two marks. The court found Lollipops visually similar to Jellibeans because Skating Clubs spelled it with an “i” rather than a “y” — an equally acceptable spelling — and used the plural form. In finding an aural similarity the court noted that each name has three syllables, with an sound dividing the first and second syllables, and an “e” sound dividing the second and third syllables, and that each name ends in a plural “s.” The court found this aural similarity most important because both rinks advertised primarily either by word of mouth or radio. Finally, the court found that the names create a similar general impression because each is a similar sounding candy name used to signify a roller rink. Based on these considerations, the district court found that Lollipops and Jellibeans possess similar designs. After examining the names visually and aurally, we are not left with the definite and firm conviction that the district court made a mistake in finding them similar. Moreover, we cannot say that the district court erred in finding that the two names evoke the same general impression. 21 Therefore, we uphold the district court’s finding that the two marks are similar in design. We also uphold the district court’s findings on the third and fourth subsidiary facts. The district court was not clearly erroneous in finding that both Jellibeans and Lollipops are skating rinks with a clientele primarily composed of teenagers and young adults. Even if, as Skating Clubs alleges, the rinks differ physically and attract patrons of different ages, we find sufficient overlap in their services and clientele to uphold the district court’s findings. See Safeway Stores, 675 F.2d at 1166. The fifth fact the district court examined is the similarity of the advertising used by Jellibeans, Inc. and Skating Clubs. The court found that Jellibeans, Inc. primarily uses radio advertising, supplemented with school and charitable promotions. The court found that Skating Clubs uses, or will use, similar forms of advertising. More specifically, it accepted Skating Clubs’ claim that it advertised, and would advertise, 22 Lollipops by “neighborhood publications, personal contacts with PTA’s and church groups, and fund-raising promotional activities.” It rejected Skating Clubs’ claim that it would eschew radio advertising, noting that “advertising habits can change at will. Lollipops will be defendant’s largest and most elaborate roller skating rink and it is not unreasonable to imagine that defendant may employ more extensive advertising for it than it does for its other rinks.... ” Even if Skating Clubs would not use radio advertising, however, the court determined the Skating Clubs’ use of school and charitable promotions was sufficiently similar to Jellibeans, Inc.’s local promotions to find a *843 similarity in the advertising used by the parties. We cannot say that the district court was clearly erroneous in making this finding. The evidence of Jellibeans, Inc.’s advertising practices, which include school and charitable promotions, is uncontradicted. Skating Clubs’ use, and intent to use these forms of advertising is also uncontradicted. The only evidence that Lollipops’ advertising practices would differ from Jellibeans’ is Albert Couey’s testimony that Lollipops would not use radio advertising, which the district court rejected. Determining the credibility of this testimony is a task solely within the province of the trial court, which had the opportunity to hear the witness and view his demeanor while testifying. We do not think the trial court’s finding on the credibility or lack thereof to be accorded this testimony is clearly erroneous. See United States v. Reddoch, 467 F.2d 897, 898 (5th Cir.1972). Given Couey’s other testimony about the content of the two Lollipops signs, which Jellibeans, Inc. materially impeached, we are not left with the definite and firm conviction that the trial court erred in disbelieving Mr. Couey’s testimony about Lollipops’ advertising. Given this conclusion, we also cannot find that the trial court erred in finding that the parties advertise similarly. The district court next examined Skating Clubs’ intent, and found that Skating Clubs selected its service mark, Lollipops, with the intent of confusing consumers with Jellibeans, Inc.’s successful mark, Jellibeans. The court based this conclusion solely on circumstantial evidence, rejecting Couey’s and Taylor’s denials that they had selected the name Lollipops with no intent of benefitting from its similarity to Jellibeans. Intent, as the district court noted, must often be proved by circumstantial evidence. Certainly the record in this case — • which includes testimony that Skating Clubs’ other rinks were called “skating clubs” and named after their geographic location, and that both Couey and Taylor knew of Jellibeans and its success when naming Lollipops — provides sufficient evidence to support a finding that Skating Clubs intended to confuse rollerskaters when it adopted the name Lollipops. Moreover, the court also had more than adequate evidence to disbelieve Couey’s and Taylor’s denials of improper intent. As stated supra, Jellibeans, Inc. materially impeached Couey’s testimony by showing the substantial differences in the Lollipops’ signs that Couey described as identical. Jellibeans, Inc. also impeached Taylor’s testimony by showing his business and familial relationship with Couey. Given this evidence, we cannot hold that the district court clearly erred in finding that Skating Clubs intended to capitalize on Jellibeans’ reputation when it named its new rink Lollipops. 23 The final fact the district court examined is whether people were actually confused by the similarity of the marks. The court found that such confusion between the marks did exist. It based this finding on two types of evidence: first, the testimony of two patients and one business acquaintance of Dr. Feinman, one of Jellibeans, Inc.’s major stockholders, that they saw the Lollipops sign, thought that Feinman’s corporation was opening a new rink, and inquired of or congratulated him on the new rink; second, a survey introduced by Jellibeans, Inc. that showed 48% of those interviewed believed that Lollipops was connected to Jellibeans. The court placed minimal weight on the survey evidence, finding it “seriously flawed,” but found that it, taken together with the witnesses’ *844 testimony and in the absence of any countervailing proof, was sufficient to show actual confusion. Skating Clubs contends that this finding is clearly erroneous, arguing that the testimony of only three witnesses is insufficient to prove actual confusion and that the survey is so flawed that it is inadmissible. We cannot agree. Skating Clubs cites Sun Banks v. Sun Federal Savings & Loan, 651 F.2d 311 (5th Cir.1981), and Amstar, 615 F.2d 252, to support its argument that three reports of actual confusion are insufficient to support a finding of actual confusion. In Sun Banks, the court held that nineteen reports of actual confusion over a three-year period were insufficient to establish a finding of actual confusion. Similarly, in Amstar the court held that three instances over a fifteen-year period were insufficient proof of actual confusion. Skating Clubs argues that if the evidence offered in those cases was insufficient, the evidence in this case is also insufficient. We disagree. Skating Clubs’ argument presumes that the reviewing court should review the evidence on some sort of absolute scale, rather than examining the evidence in light of the circumstances of the particular case. It is this latter mode of analysis, however, that Sun Banks and Amstar prescribe. In both of those cases, the court specifically qualified its insufficiency finding to the facts of the case. In Sun Banks, for instance, the court stated that the nineteen incidents were insignificant given the extent of the parties’ advertising and the number of transactions they handled during the three-year period. In Amstar, the court likewise found that the three instances of actual confusion over fifteen years were insufficient given the parties’ extensive advertising. Thus, Sun Banks and Amstar emphasized that in determining the sufficiency of the evidence to prove actual confusion, a court should examine the totality of the circumstances to determine how likely instances of actual confusion would be reported. This examination may include consideration of the time period in question and how extensively the product is advertised or made known to the public, see Sun Banks, 651 F.2d 311; Amstar, 615 F.2d 252, as well as the type of confusion that exists and who suffers the confusion. See Armstrong Cork Co., 597 F.2d 496; Holiday Inns, Inc. v. Holiday Out, 481 F.2d 445 (5th Cir.1973). Considering the testimony of the three witnesses in this manner, we have no difficulty according it probative value. When the witnesses saw the Lollipops sign, it had been up only a short time and was Skating Clubs’ only advertisement. Therefore, few reports of actual confusion could be expected. Cf. 3 R. Callman, The Law of Unfair Competition, Trademarks and Monopolies § 80.6 (3rd ed. 1969) (“when equitable relief [from the defendant’s use of a confusingly similar service mark] is sought with due promptitude, the use of defendant’s mark will have been of such duration that, even if actual confusion has occurred, proof thereof is virtually unattainable.”). Moreover, the reports of confusion that do exist appear genuine, given that the three witnesses independently and on their own initiative contacted Dr. Feinman to inquire about the new rink. Thus, this testimony alone might be sufficient to support a finding of actual confusion. We need not decide this question, though, because the record also contains survey evidence on this issue. Skating Clubs argues, however, that the district court should not have admitted this survey evidence because of its alleged technical deficiencies: (1) poor sampling; (2) inexperienced interviewers; (3) poorly designed questions; and (4) other errors in execution. These alleged technical deficiencies affect the survey’s weight, however, and not its admissibility. See Exxon Corp. v. Texas Motor Exchange, 628 F.2d 500, 507 (5th Cir.1980); Amstar, 615 F.2d at 264; Holiday Inns, Inc., 481 F.2d at 447 (“the district court properly admitted the survey evidence in this case, leaving the format of the questions and the manner of conducting the survey for consideration as to the weight of the evidence.”); see also C.A. May Marine Supply Co. v. Brunswick Corp., 649 F.2d 1049, 1055 n. 10 (5th Cir. 1981) (court excluded survey evidence be *845 cause irrelevant, but noted that “[i]f the inadequacies in the survey had been technical, such as the format of the question or the manner in which it was the survey [sic] was taken, those shortcomings would have borne on the weight of the evidence, not its admissibility.”) Therefore, the district court properly admitted the survey evidence, however flawed. The question before us then is one of weight — whether this survey evidence combined with the testimony of the three witnesses sufficiently supports a finding of actual confusion. In answering this question, we note that the quantum of evidence needed to show actual confusion is relatively small. Safeway Stores, 675 F.2d at 1167; Roto Rooter Corp. v. O’Neal, 513 F.2d 44, 47 (5th Cir.1975); World Carpets, Inc. v. Dick Littrell’s New World Carpets, 438 F.2d 482, 489 (5th Cir.1971). We believe that the evidence in this case satisfies this minimal requirement. As stated supra, the testimony of the witnesses is probative of actual confusion. Notwithstanding its technical deficiencies, the survey evidence is also probative. 24 Whether either type of evidence *846 would be sufficient alone to support a finding of actual confusion, we need not decide. It is enough that when this evidence is combined, and absent any countervailing evidence, we cannot say that we are left with the firm and definite conviction that the district court clearly erred in finding that actual confusion exists between the names. Having found the seven subsidiary findings not clearly erroneous, we now must determine whether the district court’s evaluation of these findings and its decision on the ultimate fact question — whether roller skaters are likely to confuse Lollipops with Jellibeans — is clearly erroneous. We hold that it is not. Jellibeans, Inc.’s service mark is distinctive and is entitled to broad protection; there is a great similarity between Lollipops’ and Jellibeans’ designs, services, retail outlets and purchasers, and advertising; and Skating Clubs intended to use these similarities to confuse others, which it did. Weighing these factors together, we conclude that there is a strong likelihood that roller skaters will confuse Lollipops with Jellibeans. We therefore affirm the district court’s decision that Skating Clubs violated section 1125 of the Lanham Act. Accordingly, we also affirm the district court’s decision that Skating Clubs violated the Georgia Uniform Deceptive Trade Practices Act and engaged in unfair competition under Georgia common law. IV. Jellibeans, Inc. cross appeals the district court’s denial of its claim for attorney fees. Having found that Skating Clubs intentionally adopted a mark confusingly similar to its own mark, Jellibeans, Inc. argues that the district court should have awarded it counsel fees. We disagree. Under the Lanham Act, a district court may award attorney fees to the prevailing party “in exceptional cases.” 15 U.S.C. § 1117 (1976). “Exceptional cases” are those that involve “acts which the courts have characterized as malicious, fraudulent, deliberate, and willful.” S.Rep. No. 93-1400, 93rd Cong. 2d Sess. (1974), reprinted in 1974 U.S.Code Cong. & Admin.News 7132, 7136. See Safeway Stores, 675 F.2d at 1169; Vermont Castings, Inc. v. Evans Products Co., No. 79-265 (D.Vt. Dec. 23, 1981); Clairol, Inc. v. Save-Way Industries, Inc., 210 U.S.P.Q. 459 (S.D.Fla.1980); O’Brien International, Inc. v. Mitch, 209 U.S.P.Q. 212 (N.D.Cal.1980); Salton, Inc. v. Cornwall Corp., 477 F.Supp. 975, 992 (D.N.J. 1979); RCA Records v. Kory Records, Inc., 197 U.S.P.Q. 908 (E.D.N.Y.1978). Even if the court determines the case to be exceptional, the decision to award attorney fees is still within the court’s discretion. See *847 Safeway Stores, 675 F.2d at 1169; Donsco, Inc. v. Gasper Corp., 587 F.2d 602, 608 (3d Cir.1978). Thus, we can reverse the district court’s denial of attorney fees only by holding that this is an exceptional case and that the district court abused its discretion in not awarding fees. We do not so hold. The district court denied Jellibeans, Inc. attorney fees because it found that Skating Clubs’ “intent amounted more to imitation of a successful mark in order to engender similar good will than to actual conversion of the mark.” This statement is a finding by the district court that Skating Clubs’ infringement of Jellibeans, Inc.’s mark was not malicious, fraudulent, deliberate, or willful, but rather an honest attempt lawfully to copy a successful mark, which unfortunately for Skating Clubs proved unlawful. We cannot say that the district court erred in this finding. Moreover, even if we could hold that the court erred in this finding, we certainly could not hold that the district court abused its discretion in denying the fees in this case. We, therefore, affirm the district court’s denial of attorney fees. AFFIRMED. Footnotes [1] A service mark is any word, name, symbol, device, or any combination thereof, or titles, character names and other distinctive features of radio or television programs “used in the sale or advertising of services to identify the services of one person and distinguish them from the services of others.” 15 U.S.C. § 1127 (1976). [2] It is not clear from the record whether these persons saw the first or second Lollipops sign. [3] Title 15 U.S.C. § 1125(a) provides: Any person who shall affix, apply, or annex, or use in connection with any ... services, ... any false description or representation, including words or other symbols tending falsely to describe or represent the same, and shall cause such ... services to enter into commerce, and any person who shall with knowledge of the falsity of such ... description or representation cause or procure the same to be ... used in commerce ... shall be liable to a civil action by any person ... who believes that he is or is likely to be damaged by the use of any such false description or representation. [4] Title 106 Ga.Code Ann. § 111 provides: [A]ny person who shall: (a) Use, without the consent of the registrant any ... colorable imitation of a ... service mark registered under this Chapter in connection with the sale, offering for sale, or advertising of any ... services on or in connection with which such use is likely to cause confusion or mistake ... or (b) ... colorably imitate any such ... service mark and apply such ... colorable imitation to ... signs ... or advertisements intended to be used upon or in connection with the sale or other distribution in this State of such ... services; shall be liable to a civil action by the owner of such registered . .. service mark for any or all of the remedies provided in section 106-112 .... [5] Title 106 Ga.Code Ann. § 702(a) provides: A person engages in a deceptive trade practice when, in the course of his business ... he: (1) Passes off ... services as those of another; (2) Causes likelihood of confusion or of misunderstanding as to the source, sponsorship, approval, or certification of ... services; (3) Causes likelihood of confusion or of misunderstanding as to affiliation, connection, or association with, or certification by, another; (5) Represents that ... services have sponsorship, approval ... that they do not have ... * * * * * * (12) Engages in any other conduct which similarly creates a likelihood of confusion or of misunderstanding. [6] Georgia statutory law specifically preserves the common law claim of unfair competition. Title 106 Ga.Code Ann. § 113 provides: “Nothing herein shall adversely affect the rights or the enforcement of rights in ... service marks, acquired in good faith at any time at common law.” Title 106 Ga.Code Ann. § 702(c) provides: “This section does not affect unfair trade practices otherwise actionable at common law or under other statutes of this State.” [7] Title 15 U.S.C. § 1117 provides in pertinent part: “The court in exceptional cases may award reasonable attorney fees to the prevailing party.” See infra text part IV. Title 106 Ga.Code Ann. § 703(b) provides in pertinent part: “The court (in its discretion) may award attorney’s fees to the prevailing party if ... the party charged with a deceptive trade practice has wilfully engaged in the trade practice knowing it to be deceptive. [8] For more information on the survey and its alleged deficiencies, see infra note 24. [9] At the time of trial, Skating Clubs had not begun advertising Lollipops extensively because the rink was still under construction. [10] See supra note 4. Skating Clubs does not appeal this ruling. The court also noted that federal registration of the service mark is not a prerequisite for bringing a Lanham Act claim under 15 U.S.C. § 1125(a). See Boston Professional Hockey Ass’n v. Dallas Cap & Emblem Mfg., 510 F.2d 1004, 1010 (5th Cir.1975), cert. denied, 423 U.S. 868, 96 S.Ct. 132, 46 L.Ed.2d 98 (1975). [11] To prevail on these claims, of course, Jellibeans, Inc. also had to allege and prove other elements. Skating Clubs does not appeal the district court’s rulings on any of these other elements and therefore we do not discuss them. [12] Skating Clubs also argues that the association of a candy name with a roller rink is a marketing concept that is not protected by the Lanham Act. This argument does not merit textual discussion. That the name Jellibeans is a valid service mark, albeit unregistered, cannot be disputed. That the district court did no more than enjoin Skating Clubs’ continued use of a name confusingly similar to that mark also cannot be disputed. Thus, in issuing the injunction, the district court merely protected a valid service mark and in no other way affected Skating Clubs’ marketing of its rink. [13] Title 15 U.S.C. § 1121 provides: The district and territorial courts of the United States shall have original jurisdiction and the courts of appeal of the United States shall have appellate jurisdiction, of all actions arising under [the Lanham Act], without regard to the amount in controversy or to diversity or lack of diversity of the citizenship of the parties. Jurisdiction of Jellibeans’ claim existed in the district court for the unfair competition count under 28 U.S.C. § 1338(b), and for the two Georgia statutory counts under the doctrine of pendent jurisdiction as diversity is lacking (both parties being Georgia corporations). [14] We note parenthetically that Amstar Corp. v. Domino’s Pizza, Inc., 615 F.2d 252, 258 (5th Cir.1980), cert. denied, 449 U.S. 899, 101 S.Ct. 268, 66 L.Ed.2d 129 (1980), appears to support the district court’s ruling. [15] There is a superficial distinction between this “likely to confuse” test and the test in registered trademark cases under 15 U.S.C. § 1114 in which the offending mark must be “likely to cause confusion, or to cause mistake, or to deceive.” As applied, however, the case law does not reveal a meaningful distinction between these tests. See, e.g., Amstar, 615 F.2d at 258-59. [16] There is some older authority within this circuit for the contrary proposition that “likelihood of confusion” is a mixed law and fact question. See, e.g., B.H. Bunn, 451 F.2d at 1261. This proposition is adhered to by a number of the circuit courts; see, e.g., Fleischmann Distilling Corp. v. Maier Brewing Co., 314 F.2d 149, 159 (9th Cir.1963) cert. denied, 374 U.S. 830, 83 S.Ct. 1870, 10 L.Ed.2d 1053 (1963) (ultimate issue of likelihood of confusion is a question of law); Sears, Roebuck & Co. v. Johnson, 219 F.2d 590, 591 (3rd Cir.1955); Millworth Converting Corp. v. Slifka, 276 F.2d 443, 445-46 (2nd Cir.1960); McCormick & Co. v. B. Manischewitz Co., 206 F.2d 744 (6th Cir.1953); California Fruit Growers Exch. v. Sunkist Baking Co., 166 F.2d 971, 973 (7th Cir.1947). The recent cases in this circuit are in accord that this is a question of fact. We adopt this view. [17] We note that the district court should not determine whether a likelihood of confusion exists by merely computing whether a majority of the subsidiary facts indicates that such a likelihood exists. Rather, the district court must evaluate the weight to be accorded the individual subsidiary facts and then make its ultimate fact decision. Cf. Roto-Rooter Corp. v. O’Neal, 513 F.2d 44, 45-46 (5th Cir.1975) (“While proof of actual confusion is not required to sustain a claim of infringement, the view of this court is that it is the best evidence of likelihood of confusion.”). [18] Trademarks and service marks can be categorized as: (1) descriptive or generic, i.e., the mark describes the product or service itself; (2) suggestive, i.e., the mark describes or suggests a characteristic of the product or service; (3) arbitrary, i.e., the mark is a word in common use, but applied to a product or service unrelated to its meaning, so that the word neither describes nor suggests the product or service; and (4) coined, i.e., the mark is a word devised or invented for the purpose of identifying the product or service. Tisch Hotels, Inc. v. Americana Inn, Inc., 350 F.2d 609, 611 n. 2 (7th Cir.1965). [19] The great weight given to the category into which a mark falls (see supra note 18) and the relationship between the strength of a mark and its categorization tends to undercut the assertion that the standard for infringement is “likelihood of confusion.” If a mark is generic or descriptive it is considered inherently weak; on the other hand, if it is arbitrary or coined it is considered inherently strong. Thus by this reasoning, the first person to use the name “Roller Rink” has a less distinctive mark than the first person to name a roller rink “Jellibeans.” This assertion defies reason. A mark that is not only unique but conveys important information about the service offered is clearly more valuable than one that does not ceteris paribus. Thus, the unwillingness of the law to protect the use of generic terms in service marks (see, e.g., Harland, supra, 711 F.2d at 975 (anyone may use a generic word)) must have its explanation elsewhere. The obvious reason for giving less protection to generic terms in a service mark than to arbitrary terms is that to do otherwise would place an unreasonable burden on third parties. To exclude other providers of a good or service from the right to use generic terms once they have been appropriated by a provider of the good, would quickly result in a situation in which businesses were struggling to convey information about their business in their mark but without using any words that described their business, because those words would already be protected in the service mark of another firm. In such a world, there would be almost no possibility of confusing one firm’s service mark with another. Unfortunately the public would also be receiving very little information about the nature of a business’ service from its mark. Therefore it seems that the true legal standard is not simply the “likelihood of confusion” but rather the “likelihood of unreasonable confusion.” “Reasonable” confusion is generated by the legitimate efforts of a business to convey vital information to the public about the basic nature of one’s business, and to invite comparison with one’s competitors. Thus implicit in the legal standard is a balancing between the confusion created between the putatively offending service mark and the burden placed on its owner to find a mark which can convey information to the public about the nature of his service. [20] The fact that other local businesses used the same mark to identify unrelated products, the district court properly held, did not dilute the strength of Jellibeans, Inc.’s mark. See Safeway, 675 F.2d at 1165. [21] Skating Clubs cites several cases, Hancock v. American Steel & Wire Co., 203 F.2d 737, 739, 40 Cust. & Pat.App. 931 (1953); American Technical Industries, Inc. v. General Foam Plastics Corp., 200 U.S.P.Q. 244, 246-47 (S.D. N.Y.1978); General Foods Corp. v. Borden, Inc., 191 U.S.P.Q. 674 (N.D.Ill.1976), to support its argument that the district court was clearly erroneous in finding that the marks create the same general impression. The first two of these cases, Hancock and American Technical, involved marks that were synonyms, while General Foods involved marks spelled almost identically. Skating Clubs argues from these cases that the district court must find the marks synonymous or almost identically spelled before it can find a similar general impression. We disagree. Neither these cases, nor any other examined by this court state that, as a matter of law, a court can find that two marks convey the same general impression only if they are synonyms or spelled similarly. Rather, the cases state merely that this matter is a question of fact to be determined in each individual case, and reviewed under the clearly erroneous standard. [22] See supra note 10. [23] It is possible that Couey’s intent was to invite comparison between his skating rink and Jellibeans, or to convey the same imagery and symbolic message in his service mark. Neither intent is to be condemned and if actualized would not therefore be illegal. Both of these efforts can be realized by choosing a service mark close to “Jellibeans.” In doing so, however, even if Couey’s intent was innocent and he realized his purpose he ran the risk of creating a likelihood of unreasonable confusion in the minds of the public which would have inured to his benefit and to the detriment of “Jellibeans” and the public. Thus, while intent is of some probative value, particularly if a bad intent is shown, it is neither a necessary nor sufficient condition for determining the ultimate legal fact of the “likelihood of confusion.” [24] Two hundred fourteen persons, aged approximately 12 to 24 and who had rollerskated in the last year, were interviewed at six shopping centers located near Jellibeans and Lollipops. They were told that a new roller rink named Lollipops would be opening soon and then asked whether they thought the Lollipops rink may be connected with any other rink. Forty-eight percent of those interviewed believed that Lollipops was connected to Jellibeans because of the similarity of the names. We find this level of responses probative of actual confusion between the names. The technical deficiencies alleged by Skating Clubs merely reduced the weight of this evidence. For instance, the fact that the survey interviews occurred only at six shopping centers and involved only roller skaters aged 12 to 24 may not provide a truly representative survey sample. It may be that the rollerskating universe is not limited to this age group and does not frequent these shopping centers. Even if this is so, the survey does provide some indication of the substantial confusion experienced by this segment of the population— which Jellibeans, Inc., claims primarily composes its clientele and which Skating Clubs has not contradicted. Therefore, notwithstanding this alleged deficiency, the survey is still probative. Similarly, the fact that the three interviewers may have been inexperienced does not wholly destroy the probativeness of the survey. These interviewers were college seniors majoring in marketing with top marks in their survey course. Moreover, they were given 90 minutes of instruction on how to conduct the survey. Finally, the interview itself merely required the interviewer to read the questionnaire to the interviewee and properly record his answer. This task does not appear beyond the competence of the “inexperienced” interviewers. Skating Clubs also argues that because four of the 214 questionnaires unexplainedly have an unknown person’s initials on them, there must have been a fourth untrained interviewer. Even if this is true, it does not trouble us. The remaining 210 properly recorded questionnaires still remain probative. The allegation that the survey questions may have been poorly designed also does not totally destroy the probativeness of the survey. Although the survey could have included pictures of Lollipops’ and Jellibeans’ logos, we believe that a comparison of the names alone deserves some weight. Skating Clubs disagrees, arguing that “word association” surveys are entitled to no weight. Skating Clubs cites Amstar, 615 F.2d 252, and Holiday Inns, Inc., 481 F.2d 445, to support this argument. Skating Clubs, however, reads these cases too broadly. In Holiday Inns, Inc., the Holiday Inns hotel company claimed that the name Holiday Out, when used by a campground company, was too easily confused with its service mark. To prove actual confusion, Holiday Inns introduced a survey that recorded people’s responses when shown a placard bearing the words “Holiday Out” and asked what other company it brought to mind. The Holiday Inns, Inc. court held that this survey merited little weight “because the format failed to account for the number of responses attributable to the word ‘Holiday’ as distinguished from the service mark Holiday Out.” 481 F.2d at 448. Thus, the Holiday Inns, Inc. court did not hold that all word association surveys are without weight. Rather, it held that if a service mark contains a word that independently may be associated with other names, the survey should distinguish between the confusion caused by the word, and the confusion caused by the mark that contains the word. If this distinction is not made, the survey is accorded little weight. In Amstar, that is exactly what happened. There the Amstar Corporation claimed that the “Domino’s Pizza” was too easily confused with its service mark, “Domino Sugar.” In its survey, Amstar Corporation showed persons the Domino’s Pizza mark and asked if it brought anything else to mind; it did not try to account for the number of responses attributable to the word “Domino’s.” For that reason, inter alia, the Amstar court gave little weight to the survey. *846 Thus, the mere fact that a survey involves a word association test does not mean that it deserves little weight. As long as no words in the mark blur the results of the test, the survey merits some weight. Since the marks involved in this case contain no such words, the survey is entitled to some weight. Certain alleged errors in the execution of the survey also do not destroy the survey’s probativeness. Skating Clubs alleges that the expert who conducted the survey, Doctor Edward W. Cundiff, did not (1) pretest the questionnaire to determine if the questions were confusing; (2) validate the survey by calling several interviewees and verifying their responses; and (3) initially compute the results correctly. However, Dr. Cundiff stated that he pretested the questionnaire by asking two teaching colleagues to fill them out; he felt further pretesting was unnecessary given the simplicity of the questionnaire. Dr. Cundiff also stated that he validated the survey by: (1) secretly observing the interviewers while they conducted some of the interviews; (2) asking the interviewers after the interviews whether they followed his instructions; and (3) randomly phoning several interviewees to confirm that they were interviewed; he did not attempt to confirm their answers because people often rethink their answers after the interview and offer different ones. Lastly, Dr. Cundiff stated that he recomputed the survey results accurately prior to trial. These uncontradicted statements by Dr. Cundiff show that certain safeguards existed to ensure the proper execution of the survey. Dr. Cundiff might have used better safeguards, as Skating Clubs alleges, but the ones he used are nevertheless sufficient to allow one to accord some weight to the survey results. Finally, we emphasize that we do not decide today precisely how much weight should be given the survey results. We only uphold the district court’s decision that the survey’s technical deficiencies do not strip it of all probative value.