144 676 FEDERAL REPORTER, 3d SERIES - … · provides no guidance for our analysis here. ... The United States District Court ... 100,000 impressions over six years,

144 676 FEDERAL REPORTER, 3d SERIES

dent on a § 924(c) violation, and thereforeprovides no guidance for our analysis here.

Based on our reading of the statutoryscheme, we conclude that Congress intend-ed a defendant who violates subsection (c)to be subject to enhanced sentences byvirtue of the consecutive sentence man-date. A defendant who violates subsection(j) by definition violates subsection (c), andtherefore is subject to the mandate, re-gardless of whether § 924(j) constitutes adiscrete criminal offense from § 924(c).And when Congress required proof of a§ 924(c) violation before imposing the pen-alties listed under § 924(j), it intended toinclude a subsection (j) penalty within thescope of those sentences ‘‘imposed under’’subsection (c). Finding that Congressclearly intended to impose cumulative pun-ishment for a violation of subsection (j) andany other offense, see Albernaz, 450 U.S.at 344, 101 S.Ct. 1137, we reject Berrios’sdouble jeopardy challenge accordingly.

IV.

For the foregoing reasons, we will af-firm the judgments of conviction and sen-tence.

,

ROSETTA STONE LTD.,Plaintiff–Appellant,

v.

GOOGLE, INCORPORATED,Defendant–Appellee.

The UK Intellectual Property LawSociety, Amicus Curiae,

Eric Goldman; Public Citizen; MartinSchwimmer, Limited Intervenors.

International Trademark Association;Blues Destiny Records, LLC; Carfax,Incorporated; Ford Motor Company;Harmon International Industries, In-corporated; The Media Institute; Vi-acom, Inc.; Burlington Coat FactoryWarehouse Corporation; BusinessSoftware Alliance; Chanel, Incorpo-rated; Coach, Incorporated; Govern-ment Employees Insurance Company;Harrah’s Entertainment, Incorporat-ed; Longchamp USA, Incorporated;National Football League; Oakley,Incorporated; Professional Golfers’Association of America, Incorporat-ed; Rolls–Royce North America, In-corporated; S.A.S. Jean Cassegrain;Sunkist Growers, Incorporated;Swarovski North America, Ltd.; TheAssociation for Competitive Technol-ogy; The Sunrider Corporation;Tivo, Incorporated; Tiffany & Com-pany; Tumi, Incorporated; UnitedContinental Holdings, Incorporated;1–800 Contacts, Incorporated; Conva-Tec, Incorporated; Guru Denim, In-corporated; Monster Cable Products,Incorporated; PetMed Express, Inc.;Volunteers of America, Amici Sup-porting Appellant,

Public Citizen; Public Knowledge;Electronic Frontier Foundation;eBay Incorporated; Yahoo! Incorpo-rated, Amici Supporting Appellee.

No. 10–2007.

United States Court of Appeals,Fourth Circuit.

Argued: Sept. 22, 2011.

Decided: April 9, 2012.

Background: Provider of technology-based language learning products and ser-vices brought action alleging that internet

145ROSETTA STONE LTD. v. GOOGLE, INC.Cite as 676 F.3d 144 (4th Cir. 2012)

search engine operator actively assistedthird party advertisers to mislead consum-ers and misappropriated its trademarks byusing trademarks as keyword triggers forpaid advertisements and within title andtext of paid advertisements on operator’swebsite. The United States District Courtfor the Eastern District of Virginia, GeraldBruce Lee, J., dismissed unjust enrich-ment claim, 732 F.Supp.2d 628, and en-tered summary judgment in operator’s fa-vor, 730 F.Supp.2d 531. Provider appealed.

Holdings: The Court of Appeals, Traxler,Chief Judge, held that:

(1) summary judgment on provider’strademark infringement claim was notwarranted;

(2) operator’s use of marks was not pro-tected by functionality doctrine;

(3) summary judgment on provider’s con-tributory infringement claim was notwarranted;

(4) operator was not liable for vicarioustrademark infringement;

(5) operator was not unjustly enriched byits auctioning of trademarks; and

(6) summary judgment on provider’strademark dilution claim was not war-ranted.

Affirmed in part, vacated in part, and re-manded.

1. Trademarks O1421To establish trademark infringement

under Lanham Act, plaintiff must provethat: (1) it owns valid mark; (2) defendantused mark in commerce and without plain-tiff’s authorization; (3) defendant usedmark or imitation of it in connection withsale, offering for sale, distribution, or ad-vertising of goods or services; and (4)defendant’s use of mark is likely to confuseconsumers. Lanham Act, § 32(a), 15U.S.C.A. § 1114(a).

2. Federal Courts O893

Trademarks O1696

District court’s failure to consider allnine traditional likelihood-of-confusion fac-tors is not reversible error in ruling ontrademark infringement claim, especiallywhen offending use of plaintiff’s trademarkis referential or nominative in nature; how-ever, district court opting not to addressgiven factor or group of factors shouldprovide at least brief explanation of itsreasons. Lanham Act, § 32(a), 15U.S.C.A. § 1114(a).

3. Trademarks O1092

When defendant creates associationbetween its goods or services and plain-tiff’s mark, strength of mark is relevantconsideration in ruling on trademark in-fringement claim, since encroachmentupon strong mark is more likely to causeconfusion. Lanham Act, § 32(a), 15U.S.C.A. § 1114(a).

4. Federal Civil Procedure O2493

Genuine issues of material fact as towhether internet search engine operator’sauctioning of marks owned by provider oftechnology-based language learning prod-ucts and services as keyword triggers forlinks sponsored by third party advertiserswas intended to produce confusion inminds of consumers about origin of goodsor services in question, and whether it infact did so, precluded summary judgmenton provider’s trademark infringementclaim against operator. Lanham Act,§ 32(a), 15 U.S.C.A. § 1114(a).

5. Federal Civil Procedure O2493, 2545

Anecdotal customer testimony, inter-net search engine operator’s in-house stud-ies, and expert’s report were admissible asevidence of actual confusion for summaryjudgment purposes in trademark infringe-ment action brought by provider of tech-nology-based language learning products


and services against operator that auc-tioned its marks to third party advertisers,even though customers were aware thatthey were not purchasing directly fromprovider, and there were only five instanc-es of actual confusion out of more than100,000 impressions over six years, whereprovider claimed that customers were con-fused as to goods’ sponsorship, districtcourt permitted only five ‘‘actual confu-sion’’ depositions, and there was evidenceof more than 260 other instances of actualconfusion. Lanham Act, § 32(a), 15U.S.C.A. § 1114(a).

6. Trademarks O1610, 1611

Presumption of likelihood of consumerconfusion arises from intentional copyingof plaintiff’s trade dress or trademark bydefendant.

7. Trademarks O1065(3)

Internet search engine operator’s useof marks owned by provider of technology-based language learning products and ser-vices as keyword triggers for links spon-sored by third party advertisers was notprotected by functionality doctrine fromimposition of liability for trademark in-fringement under Lanham Act, even ifmarks made operator’s product more use-ful; provider’s use of its own mark was notessential for functioning of its language-learning products, which would operate nodifferently if they were branded with an-other name. Lanham Act, §§ 2(e)(5),32(a), 33(b)(8), 15 U.S.C.A. §§ 1052(e)(5),1114(a), 1115(b)(8).

8. Trademarks O1064

Product feature is ‘‘functional,’’ andthus is not protected by Lanham Act, if itis reason device works, or it constitutesactual benefit that customer wishes to pur-chase, as distinguished from assurancethat particular entity made, sponsored, or

endorsed product. Lanham Act, § 2(e)(5),15 U.S.C.A. § 1052(e)(5).

See publication Words and Phras-es for other judicial constructionsand definitions.

9. Trademarks O1566‘‘Contributory infringement’’ is judi-

cially created doctrine that derives fromcommon law of torts, under which liabilitymay be imposed upon those who facilitateor encourage trademark infringement.Lanham Act, § 32(a), 15 U.S.C.A.§ 1114(a).


10. Trademarks O1566To establish contributory trademark

infringement claim, it is not enough tohave general knowledge that some per-centage of purchasers of product or serviceis using it to engage in infringing activi-ties; rather, defendant must supply itsproduct or service to identified individualsthat it knows or has reason to know areengaging in trademark infringement.Lanham Act, § 32(a), 15 U.S.C.A.§ 1114(a).

11. Trademarks O1566For there to be liability for contribu-

tory trademark infringement, plaintiffmust establish underlying direct infringe-ment. Lanham Act, § 32(a), 15 U.S.C.A.§ 1114(a).

12. Federal Civil Procedure O2493Genuine issue of material fact as to

whether internet search engine operatorcontinued to supply its keyword advertis-ing services to known trademark infring-ers precluded summary judgment on con-tributory infringement claim broughtagainst operator by provider of technolo-gy-based language learning products andservices. Lanham Act, § 32(a), 15U.S.C.A. § 1114(a).


13. Trademarks O1565

Liability for vicarious trademark in-fringement requires finding that defendantand infringer have apparent or actual part-nership, have authority to bind one anoth-er in transactions with third parties orexercise joint ownership or control overinfringing product. Lanham Act, § 32(a),15 U.S.C.A. § 1114(a).


Internet search engine operator thatauctioned keyword triggers for links spon-sored to third party advertisers was notliable for vicarious trademark infringementunder Lanham Act based on advertisers’sales of counterfeit products, where therewas no evidence that operator acted jointlywith any advertiser to control counterfeitproducts. Lanham Act, § 32(a), 15U.S.C.A. § 1114(a).

15. Implied and Constructive ContractsO3

Under Virginia law, cause of action forunjust enrichment rests upon doctrine thata man shall not be allowed to enrich him-self unjustly at expense of another.


Under Virginia law, to avoid unjustenrichment, equity will effect contract im-plied in law requiring one who accepts andreceives services of another to make rea-sonable compensation for those services.


Under Virginia law, plaintiff assertingunjust enrichment must demonstrate that:(1) he conferred benefit on defendant; (2)defendant knew of benefit and should rea-sonably have expected to repay plaintiff;and (3) defendant accepted or retainedbenefit without paying for its value.

18. Contracts O27Under Virginia law, ‘‘implied-in-fact

contract’’ is actual contract that was notreduced to writing, but court infers exis-tence of contract from parties’ conduct.


19. Contracts O27Under Virginia law, to recover under

contract implied-in-fact, plaintiff must al-lege facts to raise implication that defen-dant promised to pay plaintiff for suchbenefit.


Under Virginia law, concept of im-plied-in-law contract, or quasi contract, ap-plies only when there is no actual contractor meeting of minds.


Under Virginia law, internet searchengine operator was not unjustly enrichedby its auctioning of trademarks belongingto provider of technology-based languagelearning products and services as keywordtriggers for links sponsored by third partyadvertisers, where there was no evidencethat operator should reasonably have ex-pected to pay for use of marks in itskeyword query process.

22. Trademarks O1461‘‘Trademark dilution’’ is whittling

away of established trademark’s sellingpower and value through its unauthorizeduse by others. Lanham Act, § 43(c)(1), 15U.S.C.A. § 1125(c)(1).


23. Trademarks O1523(2)Fair use essentially amounts to affir-

mative defense against claim of trademark


dilution. Lanham Act, § 43(c)(3)(A), 15U.S.C.A. § 1125(c)(3)(A).

24. Trademarks O1459To state prima facie dilution claim

under Federal Trademark Dilution Act(FTDA), plaintiff must show that: (1) plain-tiff owns famous mark that is distinctive;(2) defendant has commenced using markin commerce that allegedly is diluting fa-mous mark; (3) similarity between defen-dant’s mark and famous mark gives rise toassociation between marks; and (4) associ-ation is likely to impair famous mark’sdistinctiveness or likely to harm famousmark’s reputation. Lanham Act,§ 43(c)(1), 15 U.S.C.A. § 1125(c)(1).

25. Trademarks O1612Once famous mark’s owner establishes

prima facie case of trademark dilution byblurring or tarnishment, it falls to defen-dant to demonstrate that its use constitut-ed fair use other than as designation ofsource for defendant’s own goods or ser-vices. Lanham Act, § 43(c)(3)(A), 15U.S.C.A. § 1125(c)(3)(A).

26. Trademarks O1523(2)Defendant in trademark dilution ac-

tion must show that its use of mark was ingood faith in order to establish fair usedefense under Federal Trademark Dilu-tion Act (FTDA). Lanham Act,§ 43(c)(3)(A), 15 U.S.C.A. § 1125(c)(3)(A).

27. Federal Civil Procedure O2493Genuine issues of material fact as to

whether internet search engine operator’suse of marks owned by provider of tech-nology-based language learning productsand services as keyword triggers for linkssponsored by third party advertisers dilut-ed provider’s marks, and whether provid-er’s marks were famous when operatorfirst permitted their use as keyword trig-gers precluded summary judgment on pro-vider’s trademark dilution claim against

operator. Lanham Act, § 43(c)(1), 15U.S.C.A. § 1125(c)(1).


Defendant’s first diluting use of fa-mous mark fixes time by which famous-ness is to be measured for purposes ofFederal Trademark Dilution Act (FTDA).Lanham Act, § 43(c)(1), 15 U.S.C.A.§ 1125(c)(1).

Trademarks O1800

ROSETTA STONE.

Trademarks O1800

ROSETTA STONE LANGUAGELEARNING SUCCESS.

Trademarks O1800

ROSETTA WORLD.

Trademarks O1800

ROSETTASTONE.COM.

ARGUED: Clifford M. Sloan, Skadden,Arps, Slate, Meagher & Flom, LLP,Washington, D.C., for Appellant. MargretMary Caruso, Quinn, Emanuel, Urquhart& Sullivan, LLP, Redwood Shores, Cali-fornia, for Appellee. ON BRIEF: Mitch-ell S. Ettinger, Jennifer L. Spaziano,Skadden, Arps, Slate, Meagher & Flom,LLP, Washington, D.C., for Appellant.Cheryl A. Galvin, Henry Lien, Austin D.Tarango, Quinn, Emanuel, Urquhart &Sullivan, LLP, Redwood Shores, Califor-nia; Jonathan D. Frieden, Odin, Feldman& Pittleman, PC, Fairfax, Virginia, for Ap-pellee. Charles Lee Thomason, Spalding& Thomason, Bardstown, Kentucky, forThe UK Intellectual Property Law Soci-ety, Amicus Curiae. David H. Bernstein,Debevoise & Plimpton LLP, New York,New York; Kurt E. Anderson, Giordano,Halleran & Ciesla, PC, Red Bank, New


Jersey; A. Justin Ourso, III, Jones, Walk-er, Waechter, Poitevent, Carrere & De-negre LLP, Baton Rouge, Louisiana, forInternational Trademark Association, Ami-cus Supporting Appellant. Marcia B.Paul, Kevan Choset, Davis Wright Tre-maine LLP, New York, New York, DanielP. Reing, Davis Wright Tremaine LLP,Washington, D.C., for Blues Destiny Rec-ords, LLC, Carfax, Incorporated, HarmonInternational Industries, Incorporated,The Media Institute, and Viacom, Inc.;Mark S. Sparschu, Brooks Kushman PC,Southfield, Michigan, for Ford MotorCompany, Amici Supporting Appellant.Randall K. Miller, Arnold & Porter, Mc-Lean, Virginia, Roberta L. Horton, TriciaA. Cross, Brent S. LaBarge, Arnold &Porter LLP, Washington, D.C., for Bur-lington Coat Factory Warehouse Corpora-tion, Business Software Alliance, Chanel,Incorporated, Coach, Incorporated, Gov-ernment Employees Insurance Company,Harrah’s Entertainment, Incorporated,Longchamp USA, Incorporated, NationalFootball League, Oakley, Incorporated,Professional Golfers’ Association of Amer-ica, Incorporated, Rolls–Royce NorthAmerica, Incorporated, S.A.S. Jean Casse-grain, Sunkist Growers, Incorporated,Swarovski North America, Ltd., The Asso-ciation for Competitive Technology, TheSunrider Corporation, TiVo, Incorporated,Tiffany & Company, Tumi, Incorporated,and United Continental Holdings, Incorpo-rated, Amici Supporting Appellant. BradR. Newberg, Reed Smith LLP, FallsChurch, Virginia, for 1–800 Contacts, In-corporated, ConvaTec, Incorporated, GuruDenim, Incorporated, Monster CableProducts, Incorporated, and PetMed Ex-press, Inc., Amici Supporting Appellant.Thomas G. Southard, Karl Wm. Means,Alan B. Sternstein, Shulman, Rogers, Gan-dal, Pordy & Ecker, PA, Potomac, Mary-land, for Volunteers of America, AmicusSupporting Appellant. Paul Alan Levy,

Public Citizen Litigation Group, Washing-ton, D.C., for Public Citizen, Amicus Sup-porting Appellee. Harold Feld, JohnBergmayer, Rashmi Rangnath, PublicKnowledge, Washington, D.C.; CorynneMcSherry, Electronic Frontier Founda-tion, San Francisco, California, for PublicKnowledge and Electronic Frontier Foun-dation, Amici Supporting Appellee. R.Bruce Rich, Jonathan Bloom, Mark J.Fiore, Weil, Gotshal & Manges LLP, NewYork, New York, Michael Lyle, Weil,Gotshal & Manges LLP, Washington,D.C., for Yahoo! Incorporated and eBay,Incorporated, Amici Supporting Appellee.

Before TRAXLER, Chief Judge,KEENAN, Circuit Judge, andHAMILTON, Senior Circuit Judge.

Affirmed in part, vacated in part, andremanded by published opinion. ChiefJudge TRAXLER wrote the opinion, inwhich Judge KEENAN and Senior JudgeHAMILTON joined.

OPINION

TRAXLER, Chief Judge:

Appellant Rosetta Stone Ltd. appealsfrom an order, see Rosetta Stone Ltd. v.Google Inc., 730 F.Supp.2d 531 (E.D.Va.2010), granting summary judgmentagainst Rosetta Stone on its claims againstAppellee Google Inc. for trademark in-fringement, see 15 U.S.C. § 1114(1)(a);contributory and vicarious trademark in-fringement; and trademark dilution, see15 U.S.C. § 1125(c)(1). Rosetta Stone alsoappeals from an order dismissing its un-just enrichment claim under Virginia Law.See Rosetta Stone Ltd. v. Google Inc., 732F.Supp.2d 628 (E.D.Va.2010). For thereasons that follow, we affirm the districtcourt’s order with respect to the vicariousinfringement and unjust enrichmentclaims; however, we vacate the district


court’s order with respect to the directinfringement, contributory infringementand dilution claims and remand theseclaims for further proceedings.

I. Background

In conducting a de novo review of thedistrict court’s order granting summaryjudgment in favor of Google, ‘‘we view thefacts and draw all reasonable inferencestherefrom in the light most favorable to[Rosetta Stone], as the nonmoving party.’’Georgia Pac. Consumer Prods., LP v. VonDrehle Corp., 618 F.3d 441, 445 (4th Cir.2010). Bearing this standard in mind, wereview the underlying facts briefly.

Rosetta Stone began in 1992 as a small,family-owned business that marketed itslanguage-learning software under thebrand name ‘‘Rosetta Stone.’’ 1 By 2006,Rosetta Stone had become an industryleader in technology-based language-learn-ing products and online services, and, byJanuary 2010, it had become a publiclytraded corporation with 1,738 employeesand gross revenues of approximately $252million. Its products consist of ‘‘software,online services and audio practice tools’’available in over thirty languages. J.A.203.

Rosetta Stone owns and uses severalregistered marks in connection with itsproducts and services: ROSETTA

STONE, ROSETTA STONE LAN-GUAGE LEARNING SUCCESS, RO-SETTASTONE.COM, and ROSETTAWORLD. Using this family of registeredmarks, Rosetta Stone markets its brandthrough various types of media, includingthe Internet, television, radio, magazinesand other print media, and kiosks in publicvenues. From 2003 through 2009, RosettaStone spent approximately $57 million fortelevision and radio advertising, $40 mil-lion for print media marketing, and $12.5million to advertise on the Internet. In2009, Rosetta Stone’s marks enjoyed thehighest level of brand recognition by far inthe domestic language-learning market.2

Rosetta Stone has achieved internationalsuccess as well, with its products in use inover 150 countries.

Rosetta Stone began advertising in con-nection with Google’s website and onlineservices in 2002 and has continued to doso since that time. Google operates oneof the world’s most popular Internetsearch engines—programs that enable in-dividuals to find websites and online con-tent, generally through the use of a ‘‘key-word’’ search. See Retail Servs., Inc. v.Freebies Publ’g, 364 F.3d 535, 541 n. 1(4th Cir.2004). When an Internet userenters a word or phrase—the keyword orkeywords—into Google’s search engine,Google returns a results list of links to

1. The actual Rosetta Stone, discovered in1799, is a granite stele bearing a royal Egyp-tian decree etched in three languages: Greek,hieroglyphic, and demotic. The discovery ofthis stone became the ‘‘key to the decipheringof Egyptian hieroglyphics.’’ Barbara Green,Cracking the Code: Interpreting and Enforcingthe Appellate Court’s Decision and Mandate,32 Stet. L.Rev. 393, 393 (2003) (internal quo-tation marks omitted). The term ‘‘RosettaStone’’ has become somewhat of a commonmetaphor for anything that provides themeans for solving a difficult problem or un-derstanding a code.

2. Rosetta Stone conducted a brand equitystudy in February 2009 showing a substantialgap in actual recognition of the Rosetta Stonemark and the closest competing brand.When asked to identify without prompting‘‘all brand names that come to mind whenyou think of language learning,’’ almost 45%of the respondents were able to recall ‘‘Roset-ta Stone,’’ while only about 6% thought of‘‘Berlitz,’’ the second-place finisher. J.A.2288. When prompted, 74% indicated theyhad heard of Rosetta Stone language prod-ucts. Berlitz, again the closest competitor,was familiar to only 23% of the respondentswhen prompted.


websites that the search engine has deter-mined to be relevant based on a proprie-tary algorithm.

In addition to the natural list of resultsproduced by the keyword search, Google’ssearch engine also displays paid advertise-ments known as ‘‘Sponsored Links’’ withthe natural results of an Internet search.Google’s AdWords advertising platformpermits a sponsor to ‘‘purchase’’ keywordsthat trigger the appearance of the spon-sor’s advertisement and link when the key-word is entered as a search term. Inother words, an advertiser purchases theright to have his ad and accompanying linkdisplayed with the search results for akeyword or combination of words relevantto the advertiser’s business. Most spon-sors advertising with Google pay on a‘‘cost-per-click’’ basis, meaning that the ad-vertiser pays whenever a user of Google’ssearch engine clicks on the sponsored link.

Google displays up to three sponsoredlinks in a highlighted box immediatelyabove the natural search results, and italso displays sponsored links to the rightof the search results, but separated by avertical line. As this suggests, more thanone sponsor can purchase the same key-word and have a link displayed when asearch for that keyword is conducted.Would-be advertisers purchase their de-sired keywords through an auction whereadvertisers bid competitively against eachother for page position on the search re-sults page. Generally speaking, users ofthe Internet are apparently more likely toclick on ads that appear higher up on thesearch results page. Accordingly, an ad-vertiser will try to outbid its competitorsfor the top positions in order to maximizethe number of clicks on the advertiser’stext ads. For the advertiser, more clicksyield increased web traffic, which meansmore potential website sales. Google, inturn, benefits by placing the most relevant

ads in the most desirable locations, whichincreases the likelihood of a high click-through rate and leads to increased adver-tising revenue.

An advertiser must register for a GoogleAdWords account before bidding on a key-word. Under AdWords’ boilerplate termsand conditions, the account holder mustagree to assume responsibility for its se-lected keywords, for all advertising con-tent, and for ‘‘ensuring that [its] use of thekeywords does not violate any applicablelaws.’’ J.A. 4081. Account holders mustalso agree to refrain from ‘‘advertis[ing]anything illegal or engag[ing] in any illegalor fraudulent business practice.’’ J.A.2382.

Prior to 2004, Google’s policy precludedboth the use of trademarks in the text ofan advertisement and the use of trade-marks as keywords upon request of thetrademark owner. In 2004, Google loosen-ed its trademark usage policy to allow theuse of third-party trademarks as keywordseven over the objection of the trademarkowner. Google later even introduced atrademark-specific keyword tool that sug-gested relevant trademarks for Google’sadvertising clients to bid on as keywords.Google, however, continued to block theuse of trademarks in the actual advertise-ment text at the request of a trademarkowner. At that time, Google’s internalstudies suggested the unrestricted use oftrademarks in the text of an advertisementmight confuse Internet users.

Finally, in 2009, Google changed its poli-cy to permit the limited use of trademarksin advertising text in four situations: (1)the sponsor is a reseller of a genuinetrademarked product; (2) the sponsormakes or sells component parts for atrademarked product; (3) the sponsor of-fers compatible parts or goods for use withthe trademarked product; or (4) the spon-sor provides information about or reviews


a trademarked product. Google’s policyshift came after it developed the technolo-gy to automatically check the linked web-sites to determine if the sponsor’s use ofthe trademark in the ad text was legiti-mate.3

Rosetta Stone contends that Google’spolicies concerning the use of trademarksas keywords and in ad text created notonly a likelihood of confusion but also actu-al confusion as well, misleading Internetusers into purchasing counterfeit ROSET-TA STONE software. Moreover, RosettaStone alleges that it has been plagued withcounterfeiters since Google announced itspolicy shift in 2009. According to RosettaStone, between September 3, 2009, andMarch 1, 2010, it was forced to report 190instances to Google in which one of Goo-gle’s sponsored links was marketing coun-terfeit ROSETTA STONE products.

Rosetta Stone filed this action againstGoogle, asserting several claims: directtrademark infringement under the Lan-ham Act, see 15 U.S.C. § 1114(1)(a); con-tributory trademark infringement; (3) vi-carious trademark infringement; (4)trademark dilution, see 15 U.S.C.§ 1125(c)(1); and (5) unjust enrichment.Google filed a motion for summary judg-ment as to all claims except unjust enrich-ment. As to that claim, Google moved todismiss. The district court granted Goo-gle’s motion for summary judgment on allclaims and granted the motion to dismissthe unjust enrichment claim. The districtcourt denied Rosetta Stone’s cross-motionfor partial summary judgment.

II. Direct Infringement

The district court entered summaryjudgment against Rosetta Stone as to its

direct trademark infringement claim, con-cluding (A) that there is not a genuineissue of fact as to whether Google’s use ofROSETTA STONE created a likelihood ofconfusion; and (B) that the ‘‘functionalitydoctrine’’ shielded Google from liability inany event. We conclude that neitherground can sustain the summary judgmentorder as to this claim. Accordingly, wevacate the district court’s order as it per-tains to the direct infringement claim andremand for further proceedings.

A. Likelihood of Confusion

[1] To establish trademark infringe-ment under the Lanham Act, a plaintiffmust prove: (1) that it owns a valid mark;(2) that the defendant used the mark ‘‘incommerce’’ and without plaintiff’s authori-zation; (3) that the defendant used themark (or an imitation of it) ‘‘in connectionwith the sale, offering for sale, distribu-tion, or advertising’’ of goods or services;and (4) that the defendant’s use of themark is likely to confuse consumers. 15U.S.C. § 1114(a); see Louis Vuitton Mal-letier S.A. v. Haute Diggity Dog, LLC, 507F.3d 252, 259 (4th Cir.2007); People forthe Ethical Treatment of Animals v.Doughney, 263 F.3d 359, 364 (4th Cir.2001).

According to the district court, Googledid not dispute that Rosetta Stone wasable to surmount the summary judgmentbarrier on all of the infringement elementsexcept the likelihood of confusion element.See Rosetta Stone, 730 F.Supp.2d at 540–41. On appeal, Google does not take issuewith this statement.4 Thus, we assume for

3. This automated tool checks the ‘‘landingpage’’—i.e., the page linked to the ad refer-ring to the trademark—and determineswhether the page uses the trademark promi-nently; whether the page contains commer-

cial information suggesting the sponsor is areseller; and whether the landing page is areview site.

4. We note, however, that Google, in its mem-orandum filed in support of its motion for


purposes of this appeal that Google’s policypermitting advertisers to use RosettaStone’s marks as keywords in the Ad-Words program and to use Rosetta Stone’smarks in the text of advertisements consti-tuted an unauthorized use ‘‘in commerce’’and ‘‘in connection with the sale, offeringfor sale, distribution, or advertising of anygoods or services.’’ 15 U.S.C.§ 1114(1)(a). The only question for us onRosetta Stone’s direct trademark infringe-ment claim is whether there is sufficientevidence for a finder of fact to concludethat Google’s ‘‘use’’ of the mark in itsAdWords program is ‘‘likely to produceconfusion in the minds of consumers aboutthe origin of the goods or services in ques-tion.’’ CareFirst of Md., Inc. v. FirstCare, P.C., 434 F.3d 263, 267 (4th Cir.2006)(internal quotation marks omitted).

This court has articulated at least ninefactors that generally are relevant to the‘‘likelihood of confusion’’ inquiry:

(1) the strength or distinctiveness of theplaintiff’s mark as actually used in themarketplace; (2) the similarity of thetwo marks to consumers; (3) the simi-larity of the goods or services that themarks identify; (4) the similarity of thefacilities used by the markholders; (5)the similarity of advertising used by themarkholders; (6) the defendant’s intent;(7) actual confusion; (8) the quality ofthe defendant’s product; and (9) the so-phistication of the consuming public.

George & Co., LLC v. Imagination Entm’tLtd., 575 F.3d 383, 393 (4th Cir.2009). Al-though summary judgment on the likeli-

hood of confusion issue is certainly permis-sible in appropriate cases, we have notedthis is ‘‘an inherently factual issue thatdepends on the facts and circumstances ineach case.’’ Lone Star Steakhouse & Sa-loon, Inc. v. Alpha of Va., Inc., 43 F.3d922, 933 (4th Cir.1995) (internal quotationmarks omitted).

The district court indicated that ‘‘onlythree of the nine confusion factors are indispute: (1) defendant’s intent; (2) actualconfusion; and (3) the consuming public’ssophistication.’’ Rosetta Stone, 730F.Supp.2d at 541. Weighing both RosettaStone’s evidence and Google’s rebuttal evi-dence, the district court concluded that allthree ‘‘disputed’’ factors favored Google.The district court then stated that it had‘‘[b]alanc[ed] all of the disputed likelihoodof confusion factors, TTT [and] conclude[d]that Google’s use of the Rosetta StoneMarks d[id] not amount to direct trade-mark infringement.’’ Id. at 545. On ap-peal, Rosetta Stone argues that the districtcourt failed to consider the effect of theother ‘‘undisputed’’ confusion factors, sug-gesting that all of these factors favor Ro-setta Stone. Rosetta Stone also contendsthat there was sufficient evidence to createa genuine issue of fact as to whether thethree ‘‘disputed’’ confusion factors favoredGoogle or Rosetta Stone. We addressthese arguments in turn.

1. Failure to Address All Factors

[2] Rosetta Stone contends that thedistrict court’s failure to consider all nineof the traditional likelihood-of-confusion

summary judgment, argued that it had not‘‘used’’ Rosetta Stone’s marks as contemplat-ed by 15 U.S.C. § 1114(a), but rather hadmerely sold advertising space to others whowere ‘‘using’’ the mark. J.A. 4103. And, wesee nothing in the hearing transcript suggest-ing that Google conceded that it ‘‘used’’ themark ‘‘in commerce’’ and ‘‘in connectionwith the sale, offering for sale, distribution, or

advertising of any goods or services.’’ 15U.S.C. § 1114(1)(a). Since it is not an issuein this appeal, we express no opinion today asto whether Google ‘‘used’’ these marks ascontemplated by the Lanham Act. See, e.g.,Rescuecom Corp. v. Google Inc., 562 F.3d 123,129–31 (2d Cir.2009) (holding that Google’sauctioning of trademarks qualifies as a ‘‘usein commerce’’).


factors was reversible error. We cannotagree. This judicially created list of fac-tors is not intended to be exhaustive ormandatory. See Pizzeria Uno Corp. v.Temple, 747 F.2d 1522, 1527 (4th Cir.1984)(setting forth factors one through seven);see also Sara Lee Corp. v. Kayser–RothCorp., 81 F.3d 455, 463–64 (4th Cir.1996)(identifying factors eight and nine). These‘‘factors are not always weighted equally,and not all factors are relevant in everycase.’’ Louis Vuitton, 507 F.3d at 259–60.In fact, ‘‘there is no need for each factor tosupport [the plaintiff’s] position on thelikelihood of confusion issue.’’ SynergisticInt’l, LLC v. Korman, 470 F.3d 162, 171(4th Cir.2006). Rather, the confusion ‘‘fac-tors are only a guide—a catalog of variousconsiderations that may be relevant in de-termining the ultimate statutory questionof likelihood of confusion.’’ Anheuser–Busch, Inc. v. L & L Wings, Inc., 962 F.2d316, 320 (4th Cir.1992). Accordingly, thereis no hard and fast rule that obligates thedistrict court to discuss each non-mandato-ry factor.

This is especially true when the offend-ing use of the plaintiff’s trademark is re-ferential or nominative in nature. SeeCentury 21 Real Estate Corp. v. Lending-tree, Inc., 425 F.3d 211, 217 (3d Cir.2005).Unlike the typical infringement fact-pat-tern wherein the defendant ‘‘passe[s] offanother’s mark as its own’’ and ‘‘confus[es]the public as to precisely whose goods arebeing sold,’’ id., a nominative use is one inwhich the defendant uses the plaintiff’strademark to identify the plaintiff’s owngoods, see Tiffany (NJ) Inc. v. eBay Inc.,600 F.3d 93, 102 (2d Cir.2010), and ‘‘makesit clear to consumers that the plaintiff, notthe defendant, is the source of the trade-marked product or service,’’ Century 21,425 F.3d at 220; see Tiffany, 600 F.3d at102 (explaining that a ‘‘nominative fairuse’’ does not create ‘‘confusion about thesource of [the] defendant’s product’’ (inter-

nal quotation marks omitted)). An exam-ple of this type of use would be where anautomobile repair shop specializing in for-eign vehicles runs an advertisement usingthe trademarked names of various makesand models to highlight the kind of cars itrepairs. See New Kids On The Block v.News Am. Publ’g, Inc., 971 F.2d 302, 306–07 (9th Cir.1992).

In the context of a referential or nomi-native type of use, the application of thetraditional multi-factor test is difficult be-cause often many of the factors ‘‘are eitherunworkable or not suited or helpful asindicators of confusion in this context.’’Century 21, 425 F.3d at 224; see PlayboyEnters., Inc. v. Welles, 279 F.3d 796, 801(9th Cir.2002). For example, the first twofactors in our list—the similarity of themarks and the strength of the plaintiff’smark—are clearly of limited value for as-sessing the kind of use at issue here. Con-sideration of the similarity of the markswill always suggest the presence of con-sumer confusion—the mark used will al-ways be identical ‘‘because, by definition,nominative use involves the use of anoth-er’s trademark in order to describe thetrademark owner’s own product.’’ Centu-ry 21, 425 F.3d at 224. The similarityfactor does not account for context and‘‘lead[s] to the incorrect conclusion thatvirtually all nominative uses are confus-ing.’’ Playboy Enters., 279 F.3d at 801.

[3] The strength of the plaintiff’s markis also of limited probative value as to theconfusion created by a nominative use.When a defendant creates an associationbetween its goods or services and plain-tiff’s mark, the strength of the mark isrelevant since encroachment upon a strongmark is more likely to cause confusion.See CareFirst of Md., 434 F.3d at 270 (‘‘Astrong trademark is one that is rarely usedby parties other than the owner of the


trademark, while a weak trademark is onethat is often used by other parties.’’ (inter-nal quotation marks omitted)). Of course,in the nominative use context, the defen-dant is not passing off its products underthe plaintiff’s mark but rather is usingplaintiff’s mark to refer to plaintiff’s ownproducts. The strength of the mark isoften not informative as to confusion inthis context. See Century 21, 425 F.3d at225.

The district court also did not addressthe two factors relating to the trade-marked goods—the similarity of the par-ties’ goods and services and the quality ofthe defendant’s goods. Because Googleoffers no products or services under Ro-setta Stone’s mark, these factors are irrel-evant in this context.

The final two factors not addressed bythe district court—the similarity of facili-ties and the similarity of advertising—arelikewise of no relevance here. When con-sidering the similarity of facilities, courtsare trying to determine if confusion islikely based on ‘‘how and to whom therespective goods of the parties are sold,’’and the key question is whether ‘‘bothproducts [are] sold in the same ‘channelsof trade.’ ’’ 4 J. Thomas McCarthy, McCar-thy on Trademarks and Unfair Competi-tion § 24:51 [hereinafter McCarthy onTrademarks]; see Sara Lee Corp., 81 F.3dat 466 (similarity of distribution channelsfavored confusion where the parties’ prod-ucts were sold, ‘‘often side-by-side,’’ in thesame mass merchandising outlets). AsGoogle distributes no respective productvia the Internet or other outlets, this fac-tor does not aid the likelihood-of-confusionanalysis in this case.

We hasten to add that we are not adopt-ing a position about the viability of thenominative fair-use doctrine as a defenseto trademark infringement or whether thisdoctrine should formally alter our likeli-

hood-of-confusion test in some way. Thatquestion has not been presented here andwe leave it for another day. We havemerely attempted to highlight the prob-lems inherent in the robotic application ofeach and every factor in a case involving areferential, nontrademark use. According-ly, the district court did not commit re-versible error in failing to address everyfactor. In the future, however, a districtcourt opting not to address a given factoror group of factors should provide at leasta brief explanation of its reasons.

2. Remaining ‘‘Disputed’’ Factors:Genuine Issues of Fact

[4] Nevertheless, we agree that sum-mary judgment should not have beengranted. As explained in the discussionthat follows, the district court did notproperly apply the summary judgmentstandard of review but instead viewed theevidence much as it would during a benchtrial.

(a) Intent

The district court concluded that no rea-sonable trier of fact could find that Googleintended to create confusion by permittingthe use of ROSETTA STONE in the textof sponsored links or as keywords in Goo-gle’s AdWords program. The court foundit especially significant that ‘‘there is noevidence that Google is attempting to passoff its goods or services as RosettaStone’s.’’ Id. at 541.

The record shows that prior to 2004,Google did not allow the use of trademarksas keyword search triggers for unautho-rized advertisers or in the body or title ofthe text of an advertisement. In 2004,Google loosened its restrictions on the useof trademarks as keywords to ‘‘[p]rovideusers with more choice and greater accessto relevant information.’’ J.A. 4264. Theunderlying reason was largely financial, as


Google’s research showed that ‘‘[a]bout 7%[of its] total revenue [was] driven by[trademark]ed keywords.’’ J.A. 4265.With the policy shift, Google understoodthat ‘‘[t]here [would be] a slight increase inrisk that we and our partners will be thesubject of lawsuits from unhappy trade-mark owners.’’ J.A. 4271. At that time,however, Google ‘‘continue[d] to preventadvertisers from using TTT trademarks intheir ad text or ad titles unless the adver-tiser is authorized to do so by the trade-mark owner.’’ J.A. 4263. Indeed, internalstudies performed by Google at this timesuggested that there was significant sourceconfusion among Internet searchers whentrademarks were included in the title orbody of the advertisements.

Nonetheless, Google shifted its policyagain in 2009, telling its customers andpotential customers that ‘‘we are adjustingour trademark policy TTT to allow someads to use trademarks in the ad text.Under certain criteria, you can use trade-mark terms in your ad text TTT even if youdon’t own that trademark or have explicitapproval from the trademark owner to useit.’’ J.A. 4383. Google expected a sub-stantial boost in revenue from the policychange as well as an uptick in litigationfrom trademark owners. The record doesnot contain further Google studies or anyother evidence suggesting that in 2009source confusion relating to the use oftrademarks in the body of an advertise-ment was any less significant than in 2004.Viewing the evidence and all reasonableinferences in a light most favorable toRosetta Stone, as we are required to do ona motion for summary judgment, we con-clude that a reasonable trier of fact couldfind that Google intended to cause confu-sion in that it acted with the knowledgethat confusion was very likely to resultfrom its use of the marks.

(b) Actual Confusion

(i) Actual Purchaser Confusion

Rosetta Stone presented both surveyand anecdotal evidence of actual confusionin connection with Google’s use of trade-marks in its AdWords program. SeeGeorge & Co., 575 F.3d at 398 (‘‘Actualconfusion can be demonstrated by bothanecdotal and survey evidence.’’). Bothtypes of evidence are relevant, and neithercategory is necessarily required to proveactual confusion. See Tools USA & Equip.Co. v. Champ Frame StraighteningEquip., Inc., 87 F.3d 654, 661 (4th Cir.1996).

First, the record includes the depositiontestimony of five consumers who attempt-ed to buy a ROSETTA STONE softwarepackage via the Internet in 2009 afterGoogle began permitting use of ROSETTASTONE and other trademarks in the textof the sponsored links. Each of thesewould-be customers purchased bogus RO-SETTA STONE software from a spon-sored link that they mistakenly believed tobe either affiliated with Rosetta Stone orauthorized by Rosetta Stone to resell ordistribute genuine software. In each in-stance, the customer received fake soft-ware that would not load onto his or hercomputer or was so faulty after loading asto be altogether useless. Each witnesstestified that he or she called RosettaStone directly, believing that RosettaStone would assist because it was a defec-tive genuine product or that Rosetta Stonehad empowered the reseller to offer itsproducts. Typical of this set of witnesseswas Steve Dubow, a college-educatedfounder and owner of a software company.Mr. Dubow testified that he wanted tolearn Spanish and, after conducting hisown research on the Internet, concludedthat the ROSETTA STONE brand wasbest for him. Mr. Dubow then describedhow he arrived at the decision to purchase


from ‘‘bossdisk.com,’’ one of the sponsoredlinks that was selling counterfeit ROSET-TA STONE products:

TTT At the time that you entered theterms TTT ‘‘Rosetta Stone’’ in the Googlesearch engine TTT in October 2009, doyou recall whether any advertisementsappeared on the first page?

TTT [W]hat do you mean by advertise-ments?

Q. Links that appear to you to be com-panies selling goods in response to yourquery.

A. YesTTTT There were quite a fewunder that description, yes.

Q. What do you recall seeing on thesearch page results when you enteredRosetta Stone in the Google search en-gine?

A. I saw a number of sites TTT adver-tising Rosetta Stone software for a num-ber of different discounted prices.What attracted us to this particular sitewas that they presumed to be a RosettaStone reseller reselling OEM or originalequipment manufactured product.

TTT

Q. What do you mean by reseller?

A. That they were a TTT sanctionedreseller of Rosetta Stone product.

J.A. 4614c–4615a. Once Mr. Dubow re-ceived the shipment from bossdisk.comand determined that the software ap-peared to need a key code to become fullyoperational, he called Rosetta Stone be-cause he ‘‘thought that since this companywas a representative perhaps they justforgot to put the welcome kit in this pack-age and they would have a key.’’ J.A.4620c.

[5] The district court dismissed thisanecdotal customer testimony as evidenceof actual confusion for several reasons.We agree with Rosetta Stone that none of

these reasons provide a proper basis forrejecting this testimony completely.

First, the district court concluded thatthe witnesses indicated they knew theywere not purchasing directly from RosettaStone’s site and, therefore, ‘‘none of theRosetta Stone witnesses were confusedabout the source of their purchase but onlyas to whether what they purchased wasgenuine or counterfeit.’’ Rosetta Stone,730 F.Supp.2d at 544. More than justsource confusion is at issue in an infringe-ment claim since ‘‘[t]he unauthorized useof a trademark infringes the trademarkholder’s rights if it is likely to confuse anordinary consumer as to the source orsponsorship of the goods.’’ Doughney, 263F.3d at 366 (emphasis added) (internalquotation marks omitted). ‘‘The confusionthat is remedied by trademark and unfaircompetition law is confusion not only as tosource, but also as to affiliation, connectionor sponsorship.’’ 4 McCarthy on Trade-marks § 23:8.

The district court also reasoned thatnone of the five witnesses were confusedby a sponsored link ‘‘that conformed toGoogle’s policies—i.e., used the RosettaStone Marks in connection with advertis-ing genuine goods.’’ Rosetta Stone, 730F.Supp.2d at 543. This is no basis, howev-er, for rejecting this testimony. Whetherthe sponsored link conforms to Google’spolicy is not an issue that bears uponwhether the consuming public, which is notprivy to these policies, is confused by theactual use of the trademarks in sponsoredlinks. What matters is whether ‘‘the de-fendant’s actual practice is likely to pro-duce confusion in the minds of consumersabout the origin of the goods or services inquestion.’’ CareFirst of Md., 434 F.3d at267 (emphasis added) (internal quotationmarks omitted).

Finally, the district court dismissed theanecdotal evidence as de minimis giventhat there were only five instances of actu-


al confusion out of more than ‘‘100,000impressions over six years.’’ RosettaStone, 730 F.Supp.2d at 543. And, indeed,‘‘[e]vidence of only a small number of in-stances of actual confusion may be dis-missed as de minimis ’’ where the numberof opportunities for confusion is great.George & Co., 575 F.3d at 398; see 4McCarthy § 23:14 (‘‘If there is a verylarge volume of contacts or transactionswhich could give rise to confusion andthere is only a handful of instances ofactual confusion, the evidence of actualconfusion may receive relatively littleweight.’’). Rosetta Stone presented thedeposition testimony of five individualswho had experienced actual confusion—themaximum number of ‘‘actual confusion’’ de-positions permitted by the district court inthis case. The record, however, containsother evidence of actual confusion. Roset-ta Stone presented evidence that fromApril 1, 2009, through December 9, 2009,Rosetta Stone’s customer care center re-ceived 123 complaints ‘‘from individualswho ha[d] purchased pirated/counterfeitsoftware believing the software to be genu-ine Rosetta Stone product,’’ J.A. 5427, andRosetta Stone received 139 additional com-plaints from December 9, 2009, throughMarch 8, 2010. Although this evidencedoes not indicate whether each customerlogging a complaint made the purchase viaa sponsored link, it is reasonable, for pur-poses of summary judgment, to infer thata great number of these individuals wereconfused by the apparent relationship be-tween Rosetta Stone and the sponsoredlink given that Google began allowingtrademarks to be displayed in the ad textin 2009 and in light of the evidence show-ing a substantial ‘‘proliferation of spon-sored links to pirate/counterfeit sites.’’ Id.

(ii) Google’s In–House Studies andGoogle’s Corporate Designees

The record also includes various in-house studies conducted by Google ‘‘to an-

alyze user confusion (if any) associatedwith ads using [trademark] terms.’’ J.A.4362. One of the studies showed that ‘‘thelikelihood of confusion remains high’’ whentrademark terms are used in the title orbody of a sponsored link appearing on asearch results page. J.A. 4366. Thestudy recommended ‘‘that the only effec-tive [trademark] policy TTT is: (1) [to][a]llow [trademark] usage for keywords;(2) [but] not allow [trademark] usage in adtext—title or body.’’ Id. And, in fact, Goo-gle’s official policy change in 2004 thatcontinued to prohibit trademark usage inad text was based, in part, on these inter-nal studies. The district court concludedthese studies were not evidence of actualconfusion because the studies did not testconsumer impressions of the ROSETTASTONE mark specifically, but of a broadcross-section of 16 different brand namesof varying strengths. We conclude thatthese studies, one of which reflected that‘‘94% of users were confused at leastonce,’’ are probative as to actual confusionin connection with Google’s use of trade-marks; indeed, Google determined thatthere was ‘‘[n]o difference between strongand weak trademarks’’ with respect to con-fusion. J.A. 4375.

Additionally, when testifying on behalfof Google as its Rule 30(b)(6) designees,two of Google’s in-house trademark attor-neys were shown a Google search resultspage for the keyword phrase ‘‘RosettaStone,’’ and they were unable to determinewithout more research which sponsoredlinks were authorized resellers of ROSET-TA STONE products. The district courtrejected this evidence as proof of actualconfusion because the testimony appearedto the district court to ‘‘reflect a mereuncertainty about the source of a productrather than actual confusion.’’ RosettaStone, 730 F.Supp.2d at 544. ‘‘[U]ncer-


tain[ty about] the origin’’ of a product,however, is quintessential actual confusionevidence. Sara Lee Corp., 81 F.3d at 466.The district court should have accepted itas evidence of actual confusion for sum-mary judgment purposes; whether it isentitled to enough weight to carry the dayon the ultimate issue is a matter for trial.

(iii) Dr. Kent Van Liere’s Report

Rosetta Stone also presented a consum-er confusion survey report from Dr. KentVan Liere. Dr. Van Liere is an expert inmarket analysis and consumer behavior,with ‘‘experience conducting and using fo-cus groups and surveys to measure con-sumer opinions TTT regarding productsand services,’’ J.A. 5448, and ‘‘design[ing]and review[ing] studies on the applicationof sampling and survey research methodsin litigation for a variety of matters includ-ing trademark/trade dress infringement,’’J.A. 5449. Dr. Van Liere ‘‘tested for actu-al confusion regarding the appearance ofsponsored links when consumers conduct-ed a Google search for ‘Rosetta Stone.’ ’’J.A. 5449. Based on this study, Dr. VanLiere concluded that

a significant portion of consumers in therelevant population are likely to be con-fused as to the origin, sponsorship orapproval of the ‘‘sponsored links’’ thatappear on the search results page aftera consumer has conducted a Googlesearch using a Rosetta Stone trademarkas a keyword and/or are likely to beconfused as to the affiliation, endorse-ment, or association of the websiteslinked to those ‘‘sponsored links’’ withRosetta Stone.

J.A. 5450. Specifically, Dr. Van Liere’ssurvey ‘‘yield[ed] a net confusion rate of 17percent’’—that is, ‘‘17 percent of consum-ers demonstrate actual confusion.’’ J.A.5459. This result is clear evidence of actu-al confusion for purposes of summary

judgment. Cf. Sara Lee Corp., 81 F.3d at467 n. 15 (suggesting that survey evidence‘‘clearly favors the defendant when it dem-onstrates a level of confusion much belowten percent’’ but noting caselaw that‘‘hold[s] that survey evidence indicatingten to twelve percent confusion was suffi-cient to demonstrate actual confusion’’).

The district court, however, concludedthat the survey report was ‘‘unreliable evi-dence of actual confusion because the re-sult contained a measure of whether re-spondents thought Google ‘endorsed’ aSponsored Link, a non-issue.’’ RosettaStone, 730 F.Supp.2d at 544. Thus, thecourt did not consider this survey evidenceto be viable proof of actual confusion formuch the same reason it rejected the de-position testimony of the five individualswho purchased counterfeit software. Aswe previously stated, however, trademarkinfringement creates a likelihood of ‘‘confu-sion not only as to source, but also as toaffiliation, connection or sponsorship.’’ 4McCarthy on Trademarks § 23:8. Ac-cordingly, this evidence should have beenadded to the other evidence of actual con-fusion to be considered in the light mostfavorable to Rosetta Stone.

(c) Sophistication of theConsuming Public

The district court concluded that theconsumer sophistication factor also favoreda finding that Google’s use of the marks isnot likely to create confusion. Noting thesubstantial cost of Rosetta Stone’s prod-ucts (‘‘approximately $259 for a single-levelpackage and $579 for a three-level bun-dle’’), as well as the time commitment re-quired to learn a foreign language, thedistrict court concluded that the relevantmarket of potential purchasers ‘‘is com-prised of well-educated consumers’’ who‘‘are more likely to spend time searchingand learning about Rosetta Stone’s prod-


ucts.’’ Rosetta Stone Ltd., 730 F.Supp.2dat 545. From there, the court inferredconsumer sophistication—consumers will-ing to pay Rosetta Stone’s prices and,presumably, make the required time com-mitment ‘‘would tend to demonstrate thatthey are able to distinguish between theSponsored Links and organic results dis-played on Google’s search results page.’’Id.

The district court drew this inferencerelying on Star Industries, Inc. v. Bacardi& Co. Ltd., 412 F.3d 373 (2d Cir.2005), inwhich the Second Circuit noted that acourt may ‘‘reach a conclusion about con-sumer sophistication based solely on thenature of the product or its price.’’ Id. at390. This is correct if, as in Star Indus-tries, the court is making findings of facton the likelihood of confusion issue follow-ing a bench trial. See id. at 379. In themore relevant context of a summary judg-ment motion, however, that is not the case,as ‘‘[c]redibility determinations, the weigh-ing of the evidence, and the drawing oflegitimate inferences from the facts arejury functions, not those of a judge TTT

ruling on a motion for summary judg-ment.’’ Anderson v. Liberty Lobby, Inc.,477 U.S. 242, 255, 106 S.Ct. 2505, 91L.Ed.2d 202 (1986).

We conclude that there is sufficient evi-dence in the record to create a question offact as to consumer sophistication that can-

not be resolved on summary judgment.The record includes deposition testimonyfrom Rosetta Stone customers who pur-chased counterfeit ROSETTA STONEsoftware from sponsored links that theybelieved to be either affiliated with or au-thorized by Rosetta Stone to sell genuinesoftware. The evidence also includes aninternal Google study reflecting that evenwell-educated, seasoned Internet consum-ers are confused by the nature of Google’ssponsored links and are sometimes evenunaware that sponsored links are, in actu-ality, advertisements. At the summaryjudgment stage, we cannot say on thisrecord that the consumer sophisticationfactor favors Google as a matter of law.There is enough evidence, if viewed in alight most favorable to Rosetta Stone, tofind that this factor suggests a likelihoodof confusion.

[6] In sum, we conclude that there issufficient evidence in the record to createa question of fact on each of the ‘‘disputed’’factors—intent, actual confusion, and con-sumer sophistication—to preclude sum-mary judgment. Because the districtcourt’s likelihood-of-confusion analysis waslimited only to these ‘‘disputed’’ factors,the likelihood-of-confusion issue cannot beresolved on summary judgment, and wevacate the district court’s order in thisregard.5

5. We reject Rosetta Stone’s contention that itis entitled to a presumption of confusion onthe infringement claim and that the districtcourt erred in failing to afford such a pre-sumption. In this circuit, ‘‘a presumption oflikelihood of consumer confusion’’ arisesfrom the ‘‘intentional copying’’ of plaintiff’strade dress or trademark by a defendant.See, e.g., Osem Food Indus. Ltd. v. SherwoodFoods, Inc., 917 F.2d 161, 164 (4th Cir.1990);Shakespeare Co. v. Silstar Corp. of Am., Inc.,110 F.3d 234, 239 (4th Cir.1997). The ‘‘pre-sumption arises only when the copier in-ten[ds] to exploit the good will created by analready registered trademark.’’ Shakespeare,

110 F.3d at 239 (internal quotation marksomitted). Thus, where ‘‘one produces coun-terfeit goods in an apparent attempt to capi-talize upon the popularity of, and demand for,another’s product, there is a presumption of alikelihood of confusion.’’ Polo Fashions, Inc.v. Craftex, Inc., 816 F.2d 145, 148 (4th Cir.1987). We apply such a presumption because‘‘one who tries to deceive the public shouldhardly be allowed to prove that the public hasnot in fact been deceived.’’ Shakespeare, 110F.3d at 239. Here, however, there is abso-lutely no evidence that Google intentionallycopied or adopted Rosetta Stone’s mark in an


B. Functionality

[7] As an alternate to its conclusionthat Rosetta Stone failed to forecast suffi-cient evidence to establish a likelihood ofconfusion, the district court held that theuse of the ROSETTA STONE marks askeywords was protected by the ‘‘function-ality doctrine’’ and, as such, was non-in-fringing as a matter of law. See RosettaStone, 730 F.Supp.2d at 545. Because thefunctionality doctrine does not apply inthese circumstances, however, we concludethat the district court erred in awardingsummary judgment to Google on this ba-sis.

The functionality doctrine developed asa common law rule prohibiting trade dressor trademark rights in the functional fea-tures of a product or its packaging. SeeWilhelm Pudenz, GmbH v. Littlefuse, Inc.,177 F.3d 1204, 1207 (11th Cir.1999); 1McCarthy § 7:63. The purpose of the doc-trine is to preserve the distinction betweenthe realms of trademark law and patentlaw:

The functionality doctrine preventstrademark law, which seeks to promotecompetition by protecting a firm’s repu-tation, from instead inhibiting legitimatecompetition by allowing a producer tocontrol a useful product feature. It isthe province of patent law, not trade-mark law, to encourage invention bygranting inventors a monopoly over newproduct designs or functions for a limit-ed time, after which competitors are freeto use the innovation. If a product’sfunctional features could be used astrademarks, however, a monopoly oversuch features could be obtained withoutregard to whether they qualify as pat-ents and could be extended forever (be-cause trademarks may be renewed inperpetuity).

Qualitex Co. v. Jacobson Prods. Co., 514U.S. 159, 164–65, 115 S.Ct. 1300, 131L.Ed.2d 248 (1995) (internal citation omit-ted); see Georgia–Pacific ConsumerProds., LP v. Kimberly–Clark Corp., 647F.3d 723, 727 (7th Cir.2011) (explainingthat ‘‘patent law alone protects useful de-signs from mimicry; the functionality doc-trine polices the division of responsibilitiesbetween patent and trademark law by in-validating marks on useful designs’’ (inter-nal quotation marks omitted)).

[8] In 1998, Congress adopted thefunctionality doctrine by explicitly prohib-iting trademark registration or protectionunder the Lanham Act for a functionalproduct feature, see 15 U.S.C. § 1052(e)(5)(prohibiting registration of a mark which‘‘comprises any matter that, as a whole, isfunctional’’), and by making functionality astatutory defense to an incontestably reg-istered mark, see 15 U.S.C. § 1115(b)(8);see generally 1 McCarthy § 7:63. Al-though the Lanham Act does not definethe term ‘‘functional,’’ see 15 U.S.C.§ 1127, the Supreme Court has explainedthat ‘‘a product feature is functional if it isessential to the use or purpose of thearticle or if it affects the cost or quality ofthe article.’’ Inwood Labs., Inc. v. IvesLabs., Inc., 456 U.S. 844, 850 n. 10, 102S.Ct. 2182, 72 L.Ed.2d 606 (1982); seeTrafFix Devices, Inc. v. Marketing Dis-plays, Inc., 532 U.S. 23, 32–33, 121 S.Ct.1255, 149 L.Ed.2d 164 (2001). Under In-wood’s traditional rule, a product feature isfunctional if it is ‘‘the reason the deviceworks,’’ Board of Supervisors v. SmackApparel Co., 550 F.3d 465, 486 (5th Cir.2008) (internal quotation marks omitted),or it ‘‘constitute[s] the actual benefit thatthe customer wishes to purchase, as distin-guished from an assurance that a particu-lar entity made, sponsored, or endorsed aproduct,’’ Clamp Mfg. Co. v. Enco Mfg.

effort to pass off its own goods or services under the ROSETTA STONE mark.


Co., 870 F.2d 512, 516 (9th Cir.1989) (inter-nal quotation marks omitted); see I.P.Lund Trading v. Kohler Co., 163 F.3d 27,37 n. 5 (1st Cir.1998). (‘‘[F]unctional fea-tures or designs should be defined as thosethat are driven by practical, engineering-type considerations such as making theproduct work more efficiently, with fewerparts and longer life, or with less dangerto operators, or be shaped so as to reduceexpenses of delivery or damage in ship-ping.’’ (internal quotation marks omitted)).6

The district court did not conclude, norcould it, that Rosetta Stone’s marks werefunctional product features or that RosettaStone’s own use of this phrase was some-how functional. Instead, the district courtconcluded that trademarked keywords—beit ROSETTA STONE or any othermark—are ‘‘functional’’ when entered intoGoogle’s AdWords program:

The keywords TTT have an essential in-dexing function because they enableGoogle to readily identify in its databas-es relevant information in response to aweb user’s query TTT [T]he keywordsalso serve an advertising function thatbenefits consumers who expend the timeand energy to locate particular informa-tion, goods, or services, and to compareprices.

Rosetta Stone, 730 F.Supp.2d at 546.

The functionality doctrine simply doesnot apply in these circumstances. Thefunctionality analysis below was focused onwhether Rosetta Stone’s mark made Goo-gle’s product more useful, neglecting toconsider whether the mark was functional

as Rosetta Stone used it. Rosetta Stoneuses its registered mark as a classic sourceidentifier in connection with its languagelearning products. Clearly, there is noth-ing functional about Rosetta Stone’s use ofits own mark; use of the words ‘‘RosettaStone’’ is not essential for the functioningof its language-learning products, whichwould operate no differently if RosettaStone had branded its product ‘‘SPHINX’’instead of ROSETTA STONE. See Play-boy Enters., Inc. v. Netscape Commc’nsCorp., 354 F.3d 1020, 1030–31 (9th Cir.2004) (‘‘Nothing about the marks used toidentify PEI’s products is a functional partof the design of those products’’ since‘‘PEI could easily have called its magazineand its models entirely different thingswithout losing any of their intended func-tion.’’). Once it is determined that theproduct feature—the word mark ROSET-TA STONE in this case—is not functional,then the functionality doctrine has no ap-plication, and it is irrelevant whether Goo-gle’s computer program functions betterby use of Rosetta Stone’s nonfunctionalmark. See id. at 1031 (concluding that‘‘[t]he fact that the [word] marks makedefendants’ computer program more func-tional is irrelevant’’ where plaintiff used itsword marks merely to identify its prod-ucts).

As the case progresses on remand, Goo-gle may well be able to establish that itsuse of Rosetta Stone’s marks in its Ad-Words program is not an infringing use ofsuch marks; however, Google will not beable to do so based on the functionality

6. Elaborating on the idea that the functionali-ty doctrine keeps trademark law from ‘‘inhib-iting legitimate competition by allowing aproducer to control a useful product feature,’’Qualitex Co. v. Jacobson Prods. Co., 514 U.S.159, 164, 115 S.Ct. 1300, 131 L.Ed.2d 248(1995), the Supreme Court noted that if afeature is functional, ‘‘exclusive use TTT wouldput competitors at a significant nonreputa-

tion-related disadvantage,’’ id. at 165, 115S.Ct. 1300. However, ‘‘[w]here the design isfunctional under the Inwood formulationthere is no need to proceed further to consid-er if there is a competitive necessity for thefeature.’’ TrafFix Devices, Inc. v. MarketingDisplays, Inc., 532 U.S. 23, 33, 121 S.Ct.1255, 149 L.Ed.2d 164 (2001).


doctrine. The doctrine does not applyhere, and we reject it as a possible affir-mative defense for Google.

III. Contributory Infringement

[9–11] Rosetta Stone next challengesthe district court’s grant of summaryjudgment in favor of Google on the con-tributory trademark infringement claim.Contributory infringement is a ‘‘judiciallycreated doctrine’’ that ‘‘derive[s] from thecommon law of torts,’’ Von Drehle, 618F.3d at 449, under which liability may beimposed upon those who facilitate or en-courage infringement, see 4 McCarthy onTrademarks § 25:17. The Supreme Courtexplained in Inwood Laboratories that

if a manufacturer or distributor inten-tionally induces another to infringe atrademark, or if it continues to supplyits product to one whom it knows or hasreason to know is engaging in trade-mark infringement, the manufacturer ordistributor is contributorily responsiblefor any harm done as a result of thedeceit.

456 U.S. at 854, 102 S.Ct. 2182. It is notenough to have general knowledge thatsome percentage of the purchasers of aproduct or service is using it to engage ininfringing activities; rather, the defendantmust supply its product or service to‘‘identified individuals’’ that it knows orhas reason to know are engaging in trade-mark infringement. See Sony Corp. ofAmerica v. Universal City Studios, Inc.,464 U.S. 417, 439 n. 19, 104 S.Ct. 774, 78L.Ed.2d 574 (1984) (contributory trade-mark infringement requires a showing thatthe defendant ‘‘intentionally induc[ed] itscustomers to make infringing uses’’ of themarks or ‘‘suppl[ied] its products to identi-fied individuals known by it to be engagingin continuing infringement’’ (internal quo-tation marks omitted)). Finally, for thereto be liability for contributory trademark

infringement, the plaintiff must establishunderlying direct infringement. See VonDrehle, 618 F.3d at 451. In other words,there must necessarily have been an in-fringing use of the plaintiff’s mark thatwas encouraged or facilitated by the defen-dant.

The district court recognized that Roset-ta Stone had come forward with evidencerelevant to its contributory infringementclaim. The most significant evidence inthis regard reflected Google’s purportedallowance of known infringers and counter-feiters to bid on the Rosetta Stone marksas keywords:

[The evidence included] a spreadsheetthat Google received which reflects thedates when Rosetta Stone advised Goo-gle that a Sponsored Link was fraudu-lent, the domain names associated witheach such Sponsored Link, the text ofeach Sponsored Link, and the date andsubstance of Google’s response. As doc-umented, from September 3, 2009through March 1, 2010, Rosetta Stonenotified Google of approximately 200 in-stances of Sponsored Links advertisingcounterfeit Rosetta Stone products. Ro-setta Stone contends that even after be-ing notified of these websites, Googlecontinued to allow Sponsored Links forother websites by these same advertis-ers to use the Rosetta Stone Marks askeyword triggers and in the text of theirSponsored Link advertisements. Forexample, between October 2009 to De-cember 2009, 110 different SponsoredLinks purportedly selling Rosetta Stoneproducts used ‘‘Rosetta Stone’’ as a key-word trigger, and most of the Linksincluded ‘‘Rosetta Stone’’ or ‘‘Rosetta-stone’’ in their display. Registered tothe same individual, these 110 Linkswere displayed on 356,675 differentsearch-results pages.


Rosetta Stone, 730 F.Supp.2d at 547 (inter-nal citations omitted).

Nevertheless, the district court indicatedit was ‘‘unpersuaded’’ by this evidence.Id. at 547. The district court’s conclusionwas based largely on Tiffany (NJ) Inc. v.eBay Inc., 600 F.3d 93 (2d Cir.2010), inwhich the Second Circuit rejected a con-tributory trademark infringement claimagainst an Internet auction site, eBay, by atrademark owner, Tiffany, whose markwas being used by jewelry counterfeiterson eBay’s site. The record at trial in thatcase contained evidence ‘‘demonstrat[ing]that eBay had generalized notice thatsome portion of the Tiffany goods sold onits website might be counterfeit,’’ id. at106, having received ‘‘thousands of [Noticeof Claimed Infringement Forms] [Tiffany]filed with eBay alleging TTT that certainlistings were counterfeit,’’ id. The SecondCircuit concluded that such evidence wasinsufficient to satisfy Inwood’s ‘‘knows orhas reason to know’’ requirement and thatTiffany ‘‘would have to show that eBayknew or had reason to know of specificinstances of actual infringement beyondthose that it addressed upon learning ofthem.’’ Id. at 107 (emphasis added; inter-nal quotation marks omitted). The SecondCircuit noted, however, that had therebeen evidence of willful blindness, thatwould have satisfied the Inwood standard.See id. at 109. ‘‘[C]ontributory liabilitymay arise where a defendant is (as waseBay here) made aware that there wasinfringement on its site but (unlike eBayhere) ignored that fact.’’ Id. at 110 n. 15.7

Applying Tiffany, the district court con-cluded that Rosetta Stone failed to estab-lish with the requisite specificity that Goo-

gle knew or should have known of theinfringing activity:

Comparing the evidence of knowledgeattributed to eBay to the roughly 200notices Google received of SponsoredLinks advertising counterfeit RosettaStone products on its search resultspages, the Court necessarily holds thatRosetta Stone has not met the burden ofshowing that summary judgment isproper as to its contributory trademarkinfringement claim.

See Rosetta Stone, 730 F.Supp.2d at 549(emphasis added). The court also notedthat Google did not turn a blind eye toRosetta Stone’s complaints about counter-feiters, explaining that ‘‘[t]here is littleGoogle can do beyond expressly prohibit-ing advertisements for counterfeit goods,taking down those advertisements when itlearns of their existence, and creating ateam dedicated to fighting advertisementsfor counterfeit goods.’’ Id. at 548.

[12] On appeal, Rosetta Stone arguesthat the district court misapplied the stan-dard of review and incorrectly awardedsummary judgment to Google where theevidence was sufficient to permit a trier offact to find contributory infringement.We agree. In granting summary judg-ment to Google because ‘‘Rosetta Stonehas not met the burden of showing thatsummary judgment is proper as to itscontributory trademark infringementclaim,’’ the district court turned the sum-mary judgment standard on its head.While it may very well be that RosettaStone was not entitled to summary judg-ment, that issue is not before us. The onlyquestion in this appeal is whether, viewingthe evidence and drawing all reasonable

7. eBay maintained a ‘‘Verified Rights Owner(‘VeRO’) Program,’’ which allowed trademarkowners to report potentially infringing itemsso that eBay could remove the associatedlistings. See Tiffany (NJ) Inc. v. eBay Inc.,

600 F.3d 93, 99 (2d Cir.2010). The districtcourt found that the trial evidence showedeBay promptly removed challenged listingsfrom its website. See id. at 106.


inferences from that evidence in a lightmost favorable to Rosetta Stone, a reason-able trier of fact could find in favor ofRosetta Stone, the nonmoving party. SeeVon Drehle, 618 F.3d at 445. Of course,the Tiffany court did not view the evidencethrough the lense of summary judgment;rather, Tiffany involved an appeal of judg-ment rendered after a lengthy bench trial.Because of its procedural posture, the dis-trict court in Tiffany appropriatelyweighed the evidence sitting as a trier offact. Accordingly, Tiffany is of limitedapplication in these circumstances, and thedistrict court’s heavy reliance on Tiffanywas misplaced. We conclude that the evi-dence recited by the district court is suffi-cient to establish a question of fact as towhether Google continued to supply itsservices to known infringers. Accordingly,we vacate the district court’s order to theextent it grants summary judgment in fa-vor of Google on Rosetta Stone’s contribu-tory infringement claim.

IV. Vicarious Infringement

[13] Rosetta Stone next challenges thedistrict court’s rejection of its vicariousliability theory. ‘‘Vicarious liability’’ in thetrademark context is essentially the sameas in the tort context: the plaintiff seeksto impose liability based on the defendant’srelationship with a third party tortfeasor.Thus, liability for vicarious trademark in-fringement requires ‘‘a finding that thedefendant and the infringer have an appar-ent or actual partnership, have authorityto bind one another in transactions withthird parties or exercise joint ownership orcontrol over the infringing product.’’Hard Rock Cafe Licensing Corp. v. Con-cession Servs., Inc., 955 F.2d 1143, 1150(7th Cir.1992).

[14] Rosetta Stone argues that the evi-dence proffered was sufficient to create aquestion of fact regarding whether Google

jointly controls the appearance of the adsor sponsored links on Google’s search-en-gine results page. This is not evidence,however, that Google acts jointly with anyof the advertisers to control the counterfeitROSETTA STONE products. According-ly, we affirm the district court’s grant ofsummary judgment in favor of Google onRosetta Stone’s vicarious liability claim.

V. Unjust Enrichment

Rosetta Stone contends that the districtcourt improperly dismissed its claim forunjust enrichment under Virginia law.The district court dismissed this claim ontwo grounds, concluding that RosettaStone failed to allege facts sufficient tostate a claim of unjust enrichment, seeRosetta Stone, 732 F.Supp.2d at 631–32,and that the Communications Decency Act(CDA), see 47 U.S.C. § 230(c)(1), bars theunjust enrichment claim, see RosettaStone, 732 F.Supp.2d at 633. We concludethat Rosetta Stone failed to sufficientlyplead the elements of its unjust enrich-ment claim and therefore affirm, albeit onreasoning different than that of the districtcourt.

[15–17] A cause of action for unjustenrichment in Virginia ‘‘rests upon thedoctrine that a man shall not be allowed toenrich himself unjustly at the expense ofanother.’’ Kern v. Freed Co., 224 Va. 678,299 S.E.2d 363, 365 (1983) (internal quota-tion marks omitted); see Nossen v. Hoy,750 F.Supp. 740, 744 (E.D.Va.1990). ‘‘Toavoid unjust enrichment, equity will effecta ‘contract implied in law,’ ’’ i.e., a quasicontract, ‘‘requiring one who accepts andreceives the services of another to makereasonable compensation for those ser-vices.’’ Po River Water and Sewer Co. v.Indian Acres Club of Thornburg, Inc., 255Va. 108, 114, 495 S.E.2d 478 (1998). Aplaintiff asserting unjust enrichment mustdemonstrate the following three elements:


‘‘(1) he conferred a benefit on [the defen-dant]; (2) [the defendant] knew of thebenefit and should reasonably have expect-ed to repay [the plaintiff]; and (3) [thedefendant] accepted or retained the benefitwithout paying for its value.’’ Schmidt v.Household Finance Corp., 276 Va. 108, 661S.E.2d 834, 838 (2008).

[18, 19] The district court concludedthat Rosetta Stone failed to state a claimbecause it did not allege ‘‘facts which implythat [Google] promised to pay the plaintifffor the benefit received’’ or that there was‘‘an understanding by Google that it owedRosetta Stone revenue earned for paidadvertisements containing the RosettaStone Marks.’’ Rosetta Stone, 732F.Supp.2d at 631, 632. Failure to allegean implicit promise to pay, however, is notnecessarily fatal to an implied contracttheory. Virginia distinguishes betweentwo types of implied contracts: contractsthat are implied-in-fact and contracts thatare implied-in-law. An implied-in-fact con-tract is an actual contract that was notreduced to writing, but the court infers theexistence of the contract from the conductof the parties. See Nossen, 750 F.Supp. at744. To recover under a contract ‘‘im-plied-in-fact,’’ a plaintiff must allege ‘‘factsto raise an implication that the defendantpromised to pay the plaintiff for such ben-efit.’’ Nedrich v. Jones, 245 Va. 465, 429S.E.2d 201, 207 (1993) (internal quotationmarks omitted & emphasis added).

[20] By contrast, the concept of an im-plied-in-law contract, or quasi contract, ap-plies only when there is not an actualcontract or meeting of the minds. See id.We understand Rosetta Stone’s unjust en-richment claim to be an implied-in-law con-tract claim; thus, the failure to allege thatGoogle implicitly promised to pay is notfatal.

[21] Nonetheless, this court can affirmthe dismissal of the complaint ‘‘on anybasis fairly supported by the record.’’ Ei-senberg v. Wachovia Bank, N.A., 301 F.3d220, 222 (4th Cir.2002). We conclude thatRosetta Stone failed to allege facts show-ing that it ‘‘conferred a benefit’’ on Googlefor which Google ‘‘should reasonably haveexpected’’ to repay. According to RosettaStone, the keyword trigger auctions consti-tute the unauthorized sale of the ROSET-TA STONE marks. Rosetta Stone allegesthat through the auctions it conferred abenefit ‘‘involuntarily’’ on Google, and thatGoogle ‘‘is knowingly using the goodwillestablished in [the] trademarks to deriveTTT revenues.’’ J.A. 197. Rosetta Stone,however, has not alleged facts supportingits general assertion that Google ‘‘shouldreasonably have expected’’ to pay for theuse of marks in its keyword query process.Indeed, Rosetta Stone does not contend,and did not allege, that Google pays anyother mark holder for the right to use amark in its AdWords program. In ourview, these allegations are insufficient tosurmount even the minimal barrier pre-sented by a motion to dismiss.8

8. On appeal, Rosetta Stone clarified that itsunjust enrichment claim arises from Google’sbusiness practice of selling trademarks as key-words that trigger the display of sponsoredlinks rather than the content of the sponsoredlinks. In light of our conclusion that RosettaStone failed to state an unjust enrichmentclaim as to the use of its marks as keywords,we need not address the district court’s alter-native holding that, to the extent advertisersused Rosetta Stone’s marks in the text of their

ads, Google was entitled to ‘‘immunity’’ underthe Communications Decency Act ‘‘becauseGoogle is no more than an interactive com-puter service provider and cannot be liablefor the actions of third party advertisers.’’Rosetta Stone Ltd. v. Google Inc., 732F.Supp.2d 628, 632 (E.D.Va.2010) (footnoteomitted); see 47 U.S.C. § 230(c)(1) (‘‘No pro-vider or user of an interactive computer ser-vice shall be treated as the publisher or speak-


VI. Trademark Dilution

[22] Rosetta Stone next challenges thedistrict court’s summary judgment orderas to its trademark dilution claim. ‘‘Un-like traditional infringement law, the pro-hibitions against trademark dilution TTT

are not motivated by an interest in pro-tecting consumers.’’ Moseley v. V. SecretCatalogue, Inc., 537 U.S. 418, 429, 123S.Ct. 1115, 155 L.Ed.2d 1 (2003). Dilutionis not concerned with confusion in the mar-ketplace. Rather, dilution theory providesthat ‘‘if customers or prospective custom-ers see the plaintiff’s famous mark used byother persons in a non-confusing way toidentify other sources for many differentgoods and services, then the ability of thefamous mark to clearly identify and distin-guish only one source might be ‘diluted’ orweakened.’’ 4 McCarthy § 24:67. Thus,trademark dilution is ‘‘the whittling awayof the established trademark’s selling pow-er and value through its unauthorized useby others.’’ Tiffany, 600 F.3d at 111 (in-ternal quotation marks and alterationomitted).

Until 1996, trademark dilution wasbased entirely upon state law because fed-eral law did not recognize the dilutiondoctrine. The Federal Trademark Dilu-tion Act (FTDA) was passed in 1996, seePub.L. No. 104–98, 109 Stat. 985 (1996),and was amended substantially in 2006with the passage of the Trademark Dilu-tion Revision Act of 2006, see Pub.L. No.109–312, § 2, 120 Stat. 1730 (2006). TheFTDA currently provides:

[T]he owner of a famous mark TTT shallbe entitled to an injunction against an-other person who TTT commences use ofa mark or trade name in commerce thatis likely to cause dilution by blurring ordilution by tarnishment of the famousmark, regardless of the presence or ab-

sence of actual or likely confusion, ofcompetition, or of actual economic inju-ry.

15 U.S.C. § 1125(c)(1) (emphasis added).The statute defines ‘‘dilution by blurring’’as the ‘‘association arising from the simi-larity between a mark or trade name and afamous mark that impairs the distinctive-ness of the famous mark.’’ 15 U.S.C.§ 1125(c)(2)(B). ‘‘[D]ilution by tarnish-ment’’ is defined as the ‘‘association arisingfrom the similarity between a mark ortrade name and a famous mark that harmsthe reputation of the famous mark.’’ 15U.S.C. § 1125(c)(2)(C). Thus, blurring un-der the federal statute involves the classic‘‘whittling away’’ of the selling power andstrength of the famous mark. Tarnish-ment, by contrast, creates consumer aver-sion to the famous brand—e.g., when theplaintiff’s famous trademark is ‘‘linked toproducts of shoddy quality, or is portrayedin an unwholesome or unsavory context’’such that ‘‘the public will associate the lackof quality or lack of prestige in the defen-dant’s goods with the plaintiff’s unrelatedgoods.’’ Scott Fetzer Co. v. House of Vac-uums Inc., 381 F.3d 477, 489 (5th Cir.2004)(internal quotation marks omitted).

[23] Finally, the FTDA expressly ex-cludes from its reach ‘‘[a]ny fair use, in-cluding a nominative or descriptive fairuse, or facilitation of such fair use, of afamous mark by another person other thanas a designation of source for the person’sown goods or services.’’ 15 U.S.C.§ 1125(c)(3)(A). The statute specificallyprovides comparative advertising and par-ody as examples of non-dilutive fair uses.See 15 U.S.C. § 1125(c)(3)(A)(i) & (ii). Ac-cordingly, ‘‘fair use,’’ though not so labeledin the statute, essentially amounts to anaffirmative defense against a claim oftrademark dilution. Cf. KP PermanentMake–Up v. Lasting Impression I, Inc.,

er of any information provided by another information content provider.’’).


543 U.S. 111, 117–18, 125 S.Ct. 542, 160L.Ed.2d 440 (2004).

[24] To state a prima facie dilutionclaim under the FTDA, the plaintiff mustshow the following:

(1) that the plaintiff owns a famousmark that is distinctive;(2) that the defendant has commencedusing a mark in commerce that allegedlyis diluting the famous mark;(3) that a similarity between the defen-dant’s mark and the famous mark givesrise to an association between themarks; and(4) that the association is likely to im-pair the distinctiveness of the famousmark or likely to harm the reputation ofthe famous mark.

Louis Vuitton, 507 F.3d at 264–65.

The district court granted summaryjudgment for Google on the dilution claimon two bases. First, the district court heldthat Rosetta Stone was required but failedto present evidence that Google was‘‘us[ing] the Rosetta Stone Marks to iden-tify its own goods and services.’’ RosettaStone, 730 F.Supp.2d at 551. To supportits conclusion, the district court relied onthe text of the statutory ‘‘fair use’’ defensethat shields a person’s ‘‘fair use’’ of plain-tiff’s mark so long as such use is not as ‘‘adesignation of source for the person’s owngoods or services.’’ 15 U.S.C.§ 1125(c)(3)(A).

Second, the district court concluded thatRosetta Stone failed to show that Google’suse of the mark was likely to impair thedistinctiveness of or harm the reputationof the ROSETTA STONE marks. Specifi-cally, the district court indicated that therewas ‘‘no evidence of dilution by blurringwhen Rosetta Stone’s brand awareness hasonly increased since Google revised itstrademark policy in 2004,’’ and the courtnoted evidence that Rosetta Stone’s

‘‘brand awareness equity also increasedfrom 19% in 2005 to 95% in 2009.’’ Roset-ta Stone, 730 F.Supp.2d at 551. In sup-port of this conclusion, the district courtread our decision in Louis Vuitton to es-tablish the proposition that ‘‘no claim fordilution by blurring exists where a defen-dants’ product only increases public identi-fication of the plaintiffs’ marks.’’ Id.

A. Google’s Non–Trademark Useof Rosetta Stone’s Marks

We first consider the district court’sgrant of summary judgment based on thelack of evidence that Google used the RO-SETTA STONE marks ‘‘to identify itsown goods and services.’’ Id. The districtcourt held that Rosetta Stone could notestablish its dilution claim, specifically, thethird element, without showing that Goo-gle used the mark as a source identifier forits products and services. See id. at 550–51. In support of this conclusion, however,the district court relied upon the ‘‘fair use’’defense available under the FTDA. See 15U.S.C. § 1125(c)(3)(A) (‘‘Any fair use, in-cluding a nominative or descriptive fairuse, or facilitation of such fair use, of afamous mark by another person other thanas a designation of source for the person’sown goods or services’’ is not ‘‘actionableas dilution by blurring or dilution by tar-nishment.’’) Thus, the district court ap-parently concluded that Rosetta Stone wasrequired, as part of its prima facie showingof dilution under the FTDA, to demon-strate that Google was using the mark as asource identifier for Google’s own goods.

[25] We view § 1125(c)(3)(A) as afford-ing a fair use defense to defendants indilution actions. See Louis Vuitton, 507F.3d at 265–66. In our view, once theowner of a famous mark establishes a pri-ma facie case of dilution by blurring ortarnishment, it falls to the defendant todemonstrate that its use constituted a ‘‘fair


use TTT other than as a designation ofsource for the [defendant’s] own goods orservices,’’ 15 U.S.C. § 1125(c)(3)(A).Whether Google used the mark other thanas a source identifier and in good faith isan issue that Google, not Rosetta Stone, isobligated to establish. Thus, the districtcourt erroneously required Rosetta Stoneto demonstrate that Google was using theROSETTA STONE mark as a source iden-tifier for Google’s own products.

More importantly, the district courterred when it ruled that Google was notliable for dilution simply because there wasno evidence that Google uses the RosettaStone marks to identify Google’s owngoods and services. In essence, the dis-trict court made nontrademark use coex-tensive with the ‘‘fair use’’ defense underthe FTDA. The statute, however, requiresmore than showing that defendant’s usewas ‘‘other than as a designation ofsource’’—the defendant’s use must alsoqualify as a ‘‘fair use.’’ 15 U.S.C.§ 1125(c)(3)(A). Indeed, if the districtcourt’s analysis is correct—that is, if afederal trademark dilution claim is doomedsolely by the lack of proof showing that thedefendant used the famous mark as atrademark—then the term ‘‘fair use’’ asset forth in § 1125(c)(3)(A) would be su-perfluous.

The district court failed to determinewhether this was ‘‘fair use’’. Although theFTDA does not expressly define ‘‘fair use,’’the classic concept of ‘‘fair use’’ is well-established and incorporated as an affir-mative defense to a claim of trademarkinfringement. See 15 U.S.C. § 1115(b)(4).The contours of the fair-use defense in theinfringement context are therefore instruc-tive on the classic or descriptive fair-usedefense to a dilution claim. See Sullivanv. Stroop, 496 U.S. 478, 484, 110 S.Ct.2499, 110 L.Ed.2d 438 (1990) (‘‘[I]denticalwords used in different parts of the same

act are intended to have the same mean-ing.’’ (internal quotation marks omitted)).

[26] Descriptive, or classic, fair use ap-plies when the defendant is using a trade-mark ‘‘in its primary, descriptive sense’’ todescribe the defendant’s goods or services.Fortune Dynamic, Inc. v. Victoria’s SecretStores Brand Mgmt., Inc., 618 F.3d 1025,1031 (9th Cir.2010) (internal quotationmarks omitted); see 15 U.S.C.§ 1115(b)(4). The FTDA also expresslyincludes ‘‘nominative’’ fair use as a de-fense. See 15 U.S.C. § 1125(c)(3)(A).Typically, nominative fair use comes intoplay when the defendant uses the famousmark to identify or compare the trademarkowner’s product. See New Kids on theBlock, 971 F.2d at 308; 4 McCarthy§ 23.11. Regardless of the type of fair useclaimed by a defendant, a common compo-nent of fair use is good faith. See, e.g., JAApparel Corp. v. Abboud, 568 F.3d 390,401 (2d Cir.2009) (‘‘Assessment of this de-fense thus requires analysis of whether agiven use was (1) other than as a mark, (2)in a descriptive sense, and (3) in goodfaith.’’ (internal quotation marks omitted));Sands, Taylor & Wood Co. v. Quaker OatsCo., 978 F.2d 947, 951 (7th Cir.1992) (‘‘Toprevail on the fair use defense, the defen-dant must establish that it has used theplaintiff’s mark, in good faith, to describeits (defendant’s) product and otherwisethan as a trademark.’’ (internal quotationmarks omitted)). In this context, ‘‘the in-quiry into the defendant’s good faith ‘con-cerns the question whether the user of amark intended to create consumer confu-sion as to source or sponsorship.’ ’’ JAApparel Corp., 568 F.3d at 400; see alsoBd. of Supervisors v. Smack Apparel Co.,550 F.3d 465, 489 (5th Cir.2008) (explain-ing that ‘‘in order to avail [itself] of thenominative fair use defense[,] the defen-dant (1) may only use so much of the markas necessary to identify the product or


service and (2) may not do anything thatsuggests affiliation, sponsorship, or en-dorsement by the markholder.’’ (internalquotation marks omitted)).

In short, the court’s summary judgmentorder omitted this analysis, impermissiblyomitting the question of good faith andcollapsing the fair-use defense into onequestion—whether or not Google uses theROSETTA STONE mark as a source iden-tifier for its own products. Accordingly,we vacate the district court’s summaryjudgment order and remand for reconsid-eration of Rosetta Stone’s dilution claim.If the district court determines that Roset-ta Stone has made a prima facie showingunder the elements set forth in LouisVuitton, 507 F.3d at 264–65, it should re-examine the nominative fair-use defense inlight of this opinion.

B. Likelihood of Dilution

Alternatively, the district court held thatRosetta Stone failed to satisfy the fourthand final element of its trademark dilutionclaim requiring that the plaintiff show de-fendant’s use is ‘‘likely to impair the dis-tinctiveness of the famous mark or likelyto harm the reputation of the famousmark.’’ Id. at 265. The court based itsconclusion solely on the fact that ‘‘RosettaStone’s brand awareness ha[d] only in-creased since Google revised its trademarkpolicy in 2004.’’ Rosetta Stone, 730F.Supp.2d at 551. On the strength of thisevidence, the district court concluded that‘‘the distinctiveness of the Rosetta StoneMarks has not been impaired’’ and there-fore that ‘‘Rosetta Stone cannot show thatGoogle’s trademark policy likely caused di-lution by blurring.’’ Id.

To determine whether the defendant’suse is likely to impair the distinctiveness ofthe plaintiff’s famous mark, the FTDAenumerates a non-exhaustive list of six

factors that are to be considered by thecourts:

In determining whether a mark or tradename is likely to cause dilution by blur-ring, the court may consider all relevantfactors, including the following:

(i) The degree of similarity between themark or trade name and the famousmark.

(ii) The degree of inherent or acquireddistinctiveness of the famous mark.

(iii) The extent to which the owner ofthe famous mark is engaging in substan-tially exclusive use of the mark.

(iv) The degree of recognition of thefamous mark.

(v) Whether the user of the mark ortrade name intended to create an associ-ation with the famous mark.

(vi) Any actual association between themark or trade name and the famousmark.

15 U.S.C. § 1125(c)(2)(B). Although ‘‘[n]otevery factor will be relevant in every case,and not every blurring claim will requireextensive discussion of the factors[,] TTT atrial court must offer a sufficient indicationof which factors it has found persuasiveand explain why they are persuasive.’’Louis Vuitton, 507 F.3d at 266.

[27] The district court addressed onlyone factor—the degree of recognition ofRosetta Stone’s mark—and did not men-tion any other remaining statutory factor.The court’s reliance on Louis Vuitton forthe proposition that no claim for dilutionby blurring exists when there is evidencethat public recognition of the defendants’product increased was error. Louis Vuit-ton addressed a far different fact pattern,where the defendant’s fair use claim wasbased on parody, which Congress express-ly included as a protected fair use underthe FTDA so long as the mark beingparodied is not being ‘‘used as a designa-


tion of source for the person’s own goodsor services.’’ See 15 U.S.C.A.§ 1125(c)(3)(A)(ii). We concluded that asuccessful parody ‘‘might actually enhancethe famous mark’s distinctiveness by mak-ing it an icon. The brunt of the jokebecomes yet more famous.’’ Louis Vuit-ton, 507 F.3d at 267 (4th Cir.2007) (empha-sis added). We disagree, therefore, thedistrict court’s reading of Louis Vuitton.Under the FTDA, Rosetta Stone mustshow only a likelihood of dilution and neednot prove actual economic loss or reputa-tional injury. See id. at 264 n. 2. Thedecision below employed a truncated anal-ysis that placed a very heavy emphasisupon whether there had been any actualinjury suffered by Rosetta Stone’s brand.On remand, the court should addresswhichever additional factors might applyto inform its determination of whetherGoogle’s use is likely to impair the distinc-tiveness of Rosetta Stone’s mark. See 15U.S.C. § 1125(c)(2)(B).

C. When did Rosetta Stone’smarks become famous?

Under the FTDA, the owner of a famousmark may obtain injunctive relief againstany ‘‘person who, at any time after theowner’s mark has become famous, com-mences use of a mark TTT in commercethat is likely to cause dilution.’’ 15 U.S.C.§ 1125(c)(1) (emphasis added). A thresh-old issue, therefore, is whether the plain-tiff’s mark became famous, if at all, beforethe defendant began using the mark incommerce. Although the district courtheld that Rosetta Stone’s mark had be-come famous before Google began using it,‘‘we are not limited to evaluation of thegrounds offered by the district court tosupport its decision TTT [and] may affirmon any grounds apparent from the record.’’Pitt Cnty. v. Hotels.com, L.P., 553 F.3d308, 311 (4th Cir.2009) (internal quotationmarks omitted). Accordingly, we consider

Google’s argument that Rosetta Stone’smarks were not famous in 2004 when Goo-gle allegedly began using the mark in com-merce.

Under the statute, ‘‘a mark is famous ifit is widely recognized by the general con-suming public of the United States as adesignation of source of the goods or ser-vices of the mark’s owner.’’ 15 U.S.C.§ 1125(c)(2)(A). This is not an easy stan-dard to achieve. ‘‘[C]ourts agree that amark must be truly prominent and re-nowned to be granted the extraordinaryscope of exclusive rights created by theFederal Antidilution Act.’’ 4 McCarthy§ 24:104. ‘‘Because protection from dilu-tion comes close to being a ‘right in gross,’TTT the FTDA extends dilution protectiononly to those whose mark is a ‘householdname.’ ’’ Nissan Motor Co. v. Nissan Com-puter Corp., 378 F.3d 1002, 1011 (9th Cir.2004).

[28] Additionally, for § 1125(c)(1) toapply, the defendant must have ‘‘com-mence[d]’’ a diluting use of the plaintiff’smark after the point at which the markbecame famous. The policy basis for thisrule ‘‘reflects the fair and equitable princi-ple that one should not be liable for dilu-tion by the use of a mark which was legalwhen first used.’’ 4 McCarthy § 24:103.Professor McCarthy explains as follows:

[I]f at the time of first use, Zeta’s markdid not dilute Alpha’s mark because Al-pha’s mark was not then famous, Zeta’suse will not at some future time becomediluting and illegal solely because Al-pha’s mark later became ‘‘famous.’’That is, Alpha will not at some futuretime have a federal dilution claimagainst Zeta’s mark. Thus, the junioruser must be proven to have first usedits mark after the time that plaintiff’smark achieved fameTTTT


This rule is modeled after that appliedin traditional confusion cases where theplaintiff must prove secondary meaning.In those cases, the senior user mustprove that secondary meaning in itsmark was established prior to the junioruser’s first useTTTT

4 McCarthy § 24:103 (footnote omitted).Stated differently, the defendant’s first di-luting use of a famous mark ‘‘fixes the timeby which famousness is to be measured’’for purposes of the FTDA. Nissan MotorCo., 378 F.3d at 1013.

The district court concluded that ‘‘Roset-ta Stone Marks are famous and have beensince at least 2009, when Rosetta Stone’sbrand awareness reached 75%.’’ RosettaStone, 730 F.Supp.2d at 550. The courtexplained that ‘‘[t]he Marks need not havebeen famous when Google revised itstrademark policy in 2004. Instead, Roset-ta Stone must only show that at any timeafter its Marks became famous, Googlebegan using a mark or trade name incommerce that was likely to cause dilutionof the Rosetta Stone Marks.’’ Id.

According to Google, however, even ifROSETTA STONE had become a famousbrand by 2009, it was not famous whenGoogle began its alleged facilitation of theuse of ROSETTA STONE in 2004. In-deed, Rosetta Stone alleges in its Com-plaint that the use of ROSETTA STONEand other trademarks as keywords in Goo-gle’s AdWords program ‘‘lessen[ed] the ca-pacity of Rosetta Stone’s famous and dis-tinctive TTT Marks to distinguish RosettaStone’s products and services from thoseof others, and has diluted the distinctivequality’’ of the marks. J.A. 56. The useof Rosetta Stone’s mark as a keywordtrigger began at least as early as 2004.Google points to survey evidence reflectingthat, in 2005, two percent of the generalpopulation of Internet users recognizedROSETTA STONE without being prompt-

ed while 13 percent recognized ROSETTASTONE with prompting.

In response, Rosetta Stone argues thatGoogle first began permitting the use ofRosetta Stone’s mark in sponsored ad textin 2009, by which time it had become fa-mous. Thus, Rosetta Stone’s position isthat the phrase ‘‘commences use’’ in§ 1125(c)(1) refers to any diluting use incommerce, not merely the first. This ar-gument, of course, undercuts RosettaStone’s own Complaint, which clearly as-serts that Google diluted Rosetta Stone’smark beginning in 2004 by permitting theuse of trademarks such as ROSETTASTONE as keyword triggers. RosettaStone asks us to ignore this alleged dilut-ing use for purposes of § 1125(c)(1). Thestatute does not permit the owner of afamous mark to pick and choose whichdiluting use counts for purposes of§ 1125(c)(1). See Nissan Motor Co., 378F.3d at 1013 (‘‘If TTT first use for purposesof § 1125(c) turned on whatever use themark’s owner finds particularly objectiona-ble, owners of famous marks would havethe authority to decide when an allegedlydiluting use was objectionable, regardlessof when the party accused of diluting firstbegan to use the mark.’’). The fame ofRosetta Stone’s mark, therefore, should bemeasured from 2004, when Rosetta Stonealleges Google’s diluting use of its markbegan.

Alternatively, Rosetta Stone suggeststhat it produced evidence showing that itsmark was famous in 2004. It is, however,unclear from the voluminous record pre-cisely which evidence reflects ROSETTASTONE’s fame in 2004, and we think thebetter course is for the district court tohandle this fact-intensive question of whenRosetta Stone’s mark became famous inthe first instance, particularly since otherfacets of the dilution claim will be recon-sidered on remand. Thus, on remand, the

173BOUCHIKHI v. HOLDERCite as 676 F.3d 173 (5th Cir. 2012)

district court should reconsider whetherROSETTA STONE was a famous markfor purposes of its dilution claim againstGoogle. That will require the court firstto determine when Google made its firstostensibly diluting use of the mark. Sec-ond, the court must decide whether Roset-ta Stone’s mark was famous at that point.In making the latter determination, thedistrict court should assess fame in light ofthe relevant statutory factors, see 15U.S.C. § 1125(c)(2)(A), as well as thestrong showing required to establish fameunder this statute, see, e.g., I.P. LundTrading ApS v. Kohler Co., 163 F.3d 27, 46(1st Cir.1998) (explaining that to satisfythe famousness requirement, ‘‘a mark hadto be truly prominent and renowned’’ (in-ternal quotation marks omitted)).

VII. Conclusion

For the foregoing reasons, we affirm thedistrict court’s order with respect to thevicarious infringement and unjust enrich-ment claims. We vacate, however, thedistrict court’s order with respect to Ro-setta Stone’s direct infringement, contribu-tory infringement and dilution claims, andwe remand the case for further proceed-ings on those three claims.

AFFIRMED IN PART, VACATED INPART, AND REMANDED

,

Zoubir Al Tilimsani BOUCHIKHI,Petitioner,

v.

Eric H. HOLDER, Jr., U.S. AttorneyGeneral, Respondent.

No. 10–60982.

United States Court of Appeals,Fifth Circuit.

March 8, 2012.

Background: Alien, a native and citizen ofAlgeria, petitioned for review of an orderof the Board of Immigration Appeals(BIA) dismissing his appeal from an immi-gration judge’s (IJ) decision pretermittinghis application for asylum and denying hisapplication for withholding of removal.

Holdings: The Court of Appeals held that:

(1) alien failed to file asylum applicationby one-year deadline following his ar-rival;

(2) IJ did not abuse its discretion in con-cluding that alien’s witness’s attain-ments did not qualify him as an expert;and

(3) sufficient evidence supported IJ’s de-termination that alien was not eligiblefor withholding of removal.

Petition dismissed in part and denied inpart.

1. Aliens, Immigration, and CitizenshipO403(2)

Generally, the Court of Appeals hasjurisdiction to review an immigrationjudge’s (IJ) determinations of fact under asubstantial evidence standard.

2. Aliens, Immigration, and CitizenshipO398

The Court of Appeals has authority toreview only an order of the Board of Im-migration Appeals (BIA), but its task is

Documents

144 676 FEDERAL REPORTER, 3d SERIES - … · provides no guidance for our analysis here. ... The United States District Court ... 100,000 impressions over six years,