This appeal involves a computer program for the operation of a dental laboratory, and calls upon us to apply the principles underlying our venerable copyright laws to the relatively new field of computer technology to determine the scope of copyright protection of a computer program. More particularly, in this case of first impression in the courts of appeals, we must determine whether the structure (or sequence and organization)1 of a computer program is protectible by copyright, or whether the protection of the copyright law extends only as far as the literal computer code. The district court found that the copyright law covered these non-literal elements of the program, and we agree. This conclusion in turn requires us to consider whether there was sufficient evidence of substantial similarity between the structures of the two programs at issue in this case to uphold the district court's finding of copyright infringement. Because we find that there was enough evidence, we affirm.

Appellant Jaslow Dental Laboratory, Inc. ("Jaslow Lab") is a Pennsylvania corporation in the business of manufacturing dental prosthetics and devices. Appellant Dentcom, Inc. ("Dentcom") is a Pennsylvania corporation in the business of developing and marketing computer programs for use by dental laboratories. Dentcom was formed out of the events that gave rise to this suit, and its history will be recounted below. Individual appellants Edward Jaslow and his son Rand Jaslow are officers and shareholders in both Jaslow Lab and Dentcom. Appellants were defendants in the district court. Plaintiff-appellee Whelan Associates, Inc. ("Whelan Associates") is also a Pennsylvania corporation, engaged in the business of developing and marketing custom computer programs.

Jaslow Lab, like any other small- or medium-sized business of moderate complexity, has significant bookkeeping and administrative tasks. Each order for equipment must be registered and processed; inventory must be maintained; customer lists must be continually updated; invoicing, billing, and accounts receivable, must be dealt with. While many of these functions are common to all businesses, the nature of the dental prosthetics business apparently requires some variations on the basic theme.

Although Rand Jaslow had not had extensive experience with computers, he believed that the business operations of Jaslow Lab could be made more efficient if they were computerized. In early 1978, he therefore bought a small personal computer and tried to teach himself how to program it so that it would be of use to Jaslow Lab. Although he wrote a program for the computer, he was ultimately not successful, limited by both his lack of expertise and the relatively small capacity of his particular computer.

A few months later, stymied by his own lack of success but still confident that Jaslow Lab would profit from computerization, Rand Jaslow hired the Strohl Systems Group, Inc. ("Strohl"), a small corporation that developed custom-made software to develop a program that would run on Jaslow Lab's new IBM Series One computer and take care of the Lab's business needs. Jaslow Lab and Strohl entered into an agreement providing that Strohl would design a system for Jaslow Lab's needs and that after Strohl had installed the system Strohl could market it to other dental laboratories. Jaslow Lab would receive a 10% royalty on all such sales.2 The person at Strohl responsible for the Jaslow Lab account was Elaine Whelan, an experienced programmer who was an officer and half-owner of Strohl.

Ms. Whelan's first step was to visit Jaslow Lab and interview Rand Jaslow and others to learn how the laboratory worked and what its needs were. She also visited other dental laboratories and interviewed people there, so that she would better understand the layout, workflow, and administration of dental laboratories generally. After this education into the ways of dental laboratories, and Jaslow Lab, in particular, Ms. Whelan wrote a program called Dentalab for Jaslow Lab. Dentalab was written in a computer language known as EDL (Event Driven Language), so that it would work with IBM Series One machines. The program was completed and was operative at Jaslow Lab around March 1979.

Presumably with an eye towards exploiting the economic potential of the Dentalab program, Ms. Whelan left Strohl in November, 1979, to form her own business, Whelan Associates, Inc., which acquired Strohl's interest in the Dentalab program. Shortly thereafter, Whelan Associates entered into negotiations with Jaslow Lab for Jaslow Lab to be Whelan Associates' sales representative for the Dentalab program. Whelan Associates and Jaslow Lab entered into an agreement on July 30, 1980, according to which Jaslow Lab agreed to use its "best efforts and to act diligently in the marketing of the Dentalab package," and Whelan Associates agreed to "use its best efforts and to act diligently to improve and augment the previously successfully designed Dentalab package." App. at 1779. The agreement stated that Jaslow Lab would receive 35% of the gross price of any programs sold and 5% of the price of any modifications to the programs. The agreement was for one year and was then terminable by either party on thirty days' notice.

The parties' business relationship worked successfully for two years.3 During this time, as Rand Jaslow became more familiar with computer programming, he realized that because Dentalab was written in EDL it could not be used on computers that many of the smaller dental prosthetics firms were using, for which EDL had not been implemented. Sensing that there might be a market for a program that served essentially the same function as Dentalab but that could be used more widely, Rand Jaslow began in May or June of 1982 to develop in his spare time a program in the BASIC language for such computers. That program, when completed, became the alleged copyright infringer in this suit; it was called the Dentcom PC program ("Dentcom program").4

It appears that Rand Jaslow was sanguine about the prospects of his program for smaller computers. After approximately a year of work, on May 31, 1983, his attorney sent a letter to Whelan Associates giving one month notice of termination of the agreement between Whelan Associates and Jaslow Lab.5 The letter stated that Jaslow Lab considered itself to be the exclusive marketer of the Dentalab program which, the letter stated, "contains valuable trade secrets of Jaslow Dental Laboratory." The letter concluded with a thinly veiled threat to Whelan Associates: "I ... look for your immediate response confirming that you will respect the rights of Jaslow and not use or disclose to others the trade secrets of Jaslow." App. at 1221.

Approximately two months later, on about August 1, Edward and Rand Jaslow, Paul Mohr, and Joseph Cerra formed defendant-appellant Dentcom to sell the Dentcom program.6 At about the same time, Rand Jaslow and Jaslow Lab employed a professional computer programmer, Jonathan Novak, to complete the Dentcom program. The program was soon finished, and Dentcom proceeded to sell it to dental prosthetics companies that had personal computers. Dentcom sold both the Dentalab and Dentcom programs, and advertised the Dentcom program as "a new version of the Dentalab computer system." App. at 178; 1567-73; 1766-69.

Despite Jaslow Lab's May 31 letter warning Whelan Associates not to sell the Dentalab program, Whelan Associates continued to market Dentalab. This precipitated the present litigation.

On June 30, 1983, Jaslow Lab filed suit in the Court of Common Pleas of Montgomery County (Pennsylvania), alleging that Whelan Associates had misappropriated its trade secrets. Whelan Associates responded by filing the instant suit in the United States District Court for the Eastern District of Pennsylvania on September 21, 1983. As set forth in its amended complaint, Whelan Associates alleged that Dentcom's licensing of the Dentalab and Dentcom programs infringed Whelan Associates' copyright in Dentalab, App. at 452-53; that Dentcom's use of the terms "Dentlab" or "Dentalab," violated Pennsylvania common law and 15 U.S.C. Sec. 1125(a) (Lanham Trademark Act of 1946) (false designation of origin), App. at 454-57; and that Dentcom's activities violated various other federal and state laws pertaining to unfair competition and tortious interference with contractual relations, App. at 457-63. Whelan Associates sought injunctive relief, as well as compensatory and punitive damages. App. at 464-69.

Jaslow Lab and its co-defendants answered, denying all liability. They claimed that Whelan Associates' copyright was invalid for two reasons. First, they said that although he had not been listed in the copyright registration, Rand Jaslow had been a co-author (with Elaine Whelan) of the Dentalab program. The omission of Rand Jaslow from the registration form, defendants averred, rendered the copyright defective. App. at 506-09. Second, the defendants maintained that even if Rand Jaslow had not co-authored the program, he owned the copyright because the program had been written by someone employed by him. App. at 507. Defendants also averred that Rand Jaslow had developed the Dentcom system independently, and therefore could not have violated Whelan Associate's copyright, even if the copyright were valid.7 Finally, defendants claimed that their use of "Dentalab" or "Dentlab" violated neither federal nor state law, for, inter alia, those terms are merely general descriptions of goods and services, not names of particular products. App. at 504. Defendants counterclaimed that Whelan Associates had usurped defendants' copyright and that by continuing to sell Dentalab, Whelan Associates was engaging in unfair competition. App. at 505-11. By agreement of the parties, the trade secret action was removed from the Court of Common Pleas to the district court and became a counterclaim. App. at 509-12.

The first procedural blow was struck by the defendants, who moved almost immediately for a preliminary injunction to enjoin Whelan Associates from using Jaslow Labs' trade secrets. After a three-day hearing, the district court denied the motion in a bench opinion delivered on November 2, 1983. The court held that the defendants had not shown a likelihood of success on the merits because they had failed to prove that any of Jaslow Lab's trade secrets were in the Dentalab program. Moreover, the court held that the defendants' had not proven irreparable harm and that they had approached the court with "unclean hands" on account of their use of the term "Dentlab," which could be and was confused for Dentalab. App. 443-44; 446-47.

A quick victory proving beyond the defendants' reach, the parties prepared for a protracted battle. Discovery proceeded on each side, and a three-day bench trial began on July 9, 1984. At trial, Whelan Associates continued to press all of its claims--copyright violations, unfair competition, and tortious interference with contractual relations. The defendants abandoned the trade secret claim that had failed them in the preliminary injunction battle, but maintained their position that Rand Jaslow owned the copyright to Dentalab. Defendants continued to deny Whelan Associates' allegations.

The principal witnesses were Elaine Whelan, Rand Jaslow, and two expert witnesses, Dr. Thomas Moore for Whelan Associates and Stephen Ness on behalf of the defendants. Whelan and Jaslow testified about the dealings and negotiations between the parties. Dr. Moore and Mr. Ness examined the programs and testified about the programs' similarities and differences. Dr. Moore testified that although the Dentcom program was not a translation of the Dentalab system, the programs were similar in three significant respects. He testified that most of the file structures, and the screen outputs, of the programs were virtually identical. App. at 682-704 (file structures); id. at 681-82 (screen outputs). He also testified that five particularly important "subroutines" within both programs--order entry, invoicing, accounts receivable, end of day procedure, and end of month procedure--performed almost identically in both programs. App. at 704-18.8 Mr. Ness compared the source and object codes9 of the Dentalab, Dentcom, and Datamaster programs, see supra n. 4, and testified at length about the many ways that the programs differed from one another. He concluded that "substantive differences in programming style, in programming structure, in algorithms and data structures, all indicate that the Dentcom system is not directly derived from either of the other systems." App. at 823-24. In his written report, however, which was entered into evidence, Mr. Ness conceded that the Dentalab and Dentcom programs had "overall structural similarities." App. at 1798.

The district court ruled for Whelan Associates on all grounds. Whelan Associates v. Jaslow Dental Laboratory, 609 F. Supp. 1307 (E.D.Pa.1985). It found that Elaine Whelan was the sole author of the Dentalab system (and, hence, that Rand Jaslow was not a co-author) and that the contract between Strohl and Rand Jaslow, see supra n. 2, made clear that Strohl would retain full ownership over the software. Whelan Associates v. Jaslow Dental Laboratory, 609 F. Supp. at 1318-19. The court thus concluded that Whelan Associates' copyright in the Dentalab System was valid, and that Dentcom's sales of the Dentalab program were violations of that copyright.10 Id. at 1320.

The court also found that Rand Jaslow had not created the Dentcom system independently, and that the Dentcom system, although written in a different computer language from the Dentalab, and although not a direct transliteration of Dentalab, was substantially similar to Dentalab because its structure and overall organization were substantially similar. Id. at 1321-22. (The district court's opinion on this point is described infra at 1238-39). This substantial similarity, in conjunction with Rand Jaslow's acknowledged access to the Dentalab system, led the district court to conclude that each sale of the Dentcom program by Dentcom11 violated Whelan Associates' copyright on the Dentalab system. The court therefore awarded Whelan Associates damages for these copyright infringements, and enjoined Dentcom from selling any more copies of the Dentalab or Dentcom programs. Whelan Associates v. Jaslow Dental Laboratory, 609 F. Supp. at 1322-23. The court also held that plaintiffs had exclusive use of the term "Dentalab," and enjoined defendants from using either "Dentalab" or "Dentlab" in their business. Id. at 1324-25.

The parties filed a series of post-trial motions, primarily concerned with damage calculations and attorneys fees. Upon the district court's disposition of these motions, Whelan Associates v. Jaslow Dental Laboratory, 609 F. Supp. 1325 (E.D.Pa.1985), the defendants filed their notices of appeal.12 On appeal, they raise a single issue: whether the district court erred in its finding that the Dentcom program infringes the copyright of plaintiffs' Dentalab system.

We begin with a brief description of computer programs and an explanation of how they are written. This introduction is necessary to our analysis of the issue in this case.

A computer program is a set of instructions to the computer.13 Most programs accept and process user-supplied data. The fundamental processes utilized by a program are called algorithms (mechanical computational procedures) and are at the heart of the program. See Keplinger, Computer Software--Its Nature and its Protection, 30 Emory L.J. 483, 484-85 (1984). These algorithms must be developed by the human creativity of the programmer, and the program therefore cannot contain any algorithms not already considered by humans. Although a computer cannot think or develop algorithms, it can execute them faster and more accurately than any human possibly could. See R. Saltman, Copyright in Computer-Readable Works 59 (1977).

The creation of a program often takes place in several steps, moving from the general to the specific.14 Because programs are intended to accomplish particular tasks, the first step in creating the program is identifying the problem that the computer programmer is trying to solve. In this case, Rand Jaslow went to Strohl and stated that his problem was recordkeeping for his business. Although this was an accurate statement of the problem, it was not specific enough to guide Elaine Whelan. Before she could write the Dentalab program, she needed to know more about Jaslow Lab's business--how orders were processed, what special billing problems might arise, how inventory might be correlated to orders, and other characteristics of the dental prosthetics trade.

As the programmer learns more about the problem, she or he may begin to outline a solution. The outline can take the form of a flowchart, which will break down the solution into a series of smaller units called "subroutines" or "modules,"15 each of which deals with elements of the larger problem. See Note, Defining the Scope of Copyright Protection for Computer Software, 38 Stan.L.Rev. 497, 500-01 (1986). A program's efficiency depends in large part on the arrangements of its modules and subroutines; although two programs could produce the same result, one might be more efficient because of different internal arrangements of modules and subroutines. Because efficiency is a prime concern in computer programs (an efficient program being obviously more valuable than a comparatively inefficient one), the arrangement of modules and subroutines is a critical factor for any programmer. In the present case, the Dentalab program had numerous modules pertaining to inventory, accounts receivable, various dentist-patient matters, and payroll, among others. See App. at 1588-1698 (showing flowcharts of subroutines). Some of the modules were simple; others were quite complex and involved elaborate logical development.

As the program structure is refined, the programmer must make decisions about what data are needed, where along the program's operations the data should be introduced, how the data should be inputted, and how it should be combined with other data. The arrangement of the data is accomplished by means of data files, discussed infra at 1242-45, and is affected by the details of the program's subroutines and modules, for different arrangements of subroutines and modules may require data in different forms. Once again, there are numerous ways the programmer can solve the data-organization problems she or he faces. Each solution may have particular characteristics--efficiencies or inefficiencies, conveniences or quirks--that differentiate it from other solutions and make the overall program more or less desirable. Because the Dentalab program was intended to handle all of the business-related aspects of a dental laboratory, it had to accommodate and interrelate many different pieces and types of data including patients' names, dentists' names, inventory, accounts receivable, accounts payable, and payroll.16

Once the detailed design of the program is completed, the coding begins.17 Each of the steps identified in the design must be turned into a language that the computer can understand. This translation process in itself requires two steps. The programmer first writes in a "source code," which may be in one of several languages, such as COBOL, BASIC, FORTRAN, or EDL.18 The choice of language depends upon which computers the programmer intends the program to be used by, for some computers can read only certain languages.19 Once the program is written in source code, it is translated into "object code," which is a binary code, simply a concatenation of "0 " § and "1 " s. In every program, it is the object code, not the source code, that directs the computer to perform functions. The object code is therefore the final instruction to the computer.^20,21

As this brief summary demonstrates, the coding process is a comparatively small part of programming. By far the larger portion of the expense and difficulty in creating computer programs is attributable to the development of the structure and logic of the program, and to debugging, documentation and maintenance, rather than to the coding. See Frank, Critical Issues in Software 22 (1983) (only 20% of the cost of program development goes into coding); Zelkowitz, Perspective on Software Engineering, 10 Computing Surveys 197-216 (June, 1978). See also InfoWorld, Nov. 11, 1985 at 13 ("the 'look and feel' of a computer software product often involves much more creativity and often is of greater commercial value than the program code which implements the product ..."). The evidence in this case shows that Ms. Whelan spent a tremendous amount of time studying Jaslow Labs, organizing the modules and subroutines for the Dentalab program, and working out the data arrangements, and a comparatively small amount of time actually coding the Dentalab program.

A. The elements of a copyright infringement action--To prove that its copyright has been infringed, Whelan Associates must show two things: that it owned the copyright on Dentalab, and that Rand Jaslow copied Dentalab in making the Dentacom program. Sid & Marty Krofft Television Prods., Inc. v. McDonald's Corp., 562 F.2d 1157, 1162 (9th Cir.1977); Reyher v. Children's Television Workshop, 533 F.2d 87, 90 (2d Cir.), cert. denied, 429 U.S. 980, 97 S. Ct. 492, 50 L. Ed. 2d 588 (1976); 3 Nimmer On Copyright Sec. 13.01 (1985) [referred to hereinafter as "Nimmer"]. Although it was disputed below, see supra 1228, the district court determined, and it is not challenged here, that Whelan Associates owned the copyright to the Dentalab program. We are thus concerned only with whether it has been shown that Rand Jaslow copied the Dentalab program.

As it is rarely possible to prove copying through direct evidence, Roth Greeting Cards v. United Card Co., 429 F.2d 1106, 1110 (9th Cir.1970), copying may be proved inferentially by showing that the defendant had access to the allegedly infringed copyrighted work and that the allegedly infringing work is substantially similar to the copyrighted work. Ferguson v. National Broadcasting Co., 584 F.2d 111, 113 (5th Cir.1978); Sid & Marty Krofft Television Prods. Inc., supra; Universal Athletic Sales Co. v. Salkeld, 511 F.2d 904, 907 (3d Cir.), cert. denied, 423 U.S. 863, 96 S. Ct. 122, 46 L. Ed. 2d 92 (1975); Midway Mfg. Co. v. Strohon, 564 F. Supp. 741, 753 (N.D.Ill.1983). The district court found, and here it is uncontested, that Rand Jaslow had access to the Dentalab program, both because Dentalab was the program used in Jaslow Labs and because Rand Jaslow acted as a sales representative for Whelan Associates. See Whelan Associates v. Jaslow Dental Laboratory, 609 F. Supp. at 1314.22 Thus, the sole question is whether there was substantial similarity between the Dentcom and Dentalab programs.23

B. The appropriate test for substantial similarity in computer program cases--The leading case of Arnstein v. Porter, 154 F.2d 464, 468-69 (2d Cir.1946), suggested a bifurcated substantial similarity test whereby a finder of fact makes two findings of substantial similarity to support a copyright violation. First, the fact-finder must decide whether there is sufficient similarity between the two works in question to conclude that the alleged infringer used the copyrighted work in making his own. On this issue, expert testimony may be received to aid the trier of fact. (This has been referred to as the "extrinsic" test of substantial similarity. Sid & Marty Krofft Television Prods., Inc. v. McDonald's Corp., 562 F.2d at 1164-65.) Second, if the answer to the first question is in the affirmative, the fact-finder must decide without the aid of expert testimony, but with the perspective of the "lay observer," whether the copying was "illicit," or "an unlawful appropriation" of the copyrighted work. (This has been termed an "intrinsic" test of substantial similarity. Id.) The Arnstein test has been adopted in this circuit. See Universal Athletic Sales Co., 511 F.2d at 907.

The district court heard expert testimony. See supra 1228; infra at 1246-1248. It did not bifurcate its analysis, however, but made only a single finding of substantial similarity. See Whelan Associates v. Jaslow Dental Laboratory, 609 F. Supp. at 1321-22. It would thus appear to have contravened the law of this circuit. Nevertheless, for the reasons that follow, we believe that the district court applied an appropriate standard.

The ordinary observer test, which was developed in cases involving novels, plays, and paintings, and which does not permit expert testimony, is of doubtful value in cases involving computer programs on account of the programs' complexity and unfamiliarity to most members of the public. See Note, Copyright Infringement of Computer Programs: A Modification of the Substantial Similarity Test, 68 Minn.L.Rev. 1264, 1285-88 (1984). Cf. Note, Copyright Infringement Actions: The Proper Role for Audience Reactions in Determining Substantial Similarity, 54 S.Cal.L.Rev. 385 (1981) (criticizing lay observer standard when objects in question are intended for particular, identifiable audiences). Moreover, the distinction between the two parts of the Arnstein test may be of doubtful value when the finder of fact is the same person for each step: that person has been exposed to expert evidence in the first step, yet she or he is supposed to ignore or "forget" that evidence in analyzing the problem under the second step. Especially in complex cases, we doubt that the "forgetting" can be effective when the expert testimony is essential to even the most fundamental understanding of the objects in question.

On account of these problems with the standard, we believe that the ordinary observer test is not useful and is potentially misleading when the subjects of the copyright are particularly complex, such as computer programs. We therefore join the growing number of courts which do not apply the ordinary observer test in copyright cases involving exceptionally difficult materials, like computer programs, but instead adopt a single substantial similarity inquiry according to which both lay and expert testimony would be admissible. See E.F. Johnson Co. v. Uniden Corp., 623 F. Supp. 1485, 1493 (D.Minn.1985); Hubco Data Products Corp. v. Management Assistance Inc., 2 Copyright L.Rep. (CCH) p 25,529 (D.Idaho Feb. 3, 1983) (enunciating bifurcated test, but relying entirely on expert testimony); Midway Mfg. Co. v. Strohon, 564 F. Supp. 741, 752-53 (N.D.Ill.1983) (relying entirely on expert testimony to find substantial similarity); see also Fed.R.Evid. 702 ("If [expert testimony] will assist the trier of fact to understand the evidence or to determine a fact in issue, a witness ... may testify thereto in the form of an opinion or otherwise."). That was the test applied by the district court in this case.24

C. The arguments on appeal--On appeal, the defendants attack on two grounds the district court's holding that there was sufficient evidence of substantial similarity. First, the defendants argue that because the district court did not find any similarity between the "literal" elements (source and object code) of the programs, but only similarity in their overall structures, its finding of substantial similarity was incorrect, for the copyright covers only the literal elements of computer programs, not their overall structures. Defendants' second argument is that even if the protection of copyright law extends to "non-literal" elements such as the structure of computer programs, there was not sufficient evidence of substantial similarity to sustain the district court's holding in this case. We consider these arguments in turn.25

V. THE SCOPE OF COPYRIGHT PROTECTION OF COMPUTER PROGRAMS

It is well, though recently, established that copyright protection extends to a program's source and object codes. Stern Electronics, Inc. v. Kaufman, 669 F.2d 852, 855 n. 3 (2d Cir.1982) (source code); Apple Computer, Inc. v. Franklin Computer Corp., 714 F.2d 1240, 1246-47 (3d Cir.1983) (source and object code), cert. dismissed, 464 U.S. 1033, 104 S. Ct. 690, 79 L. Ed. 2d 158 (1984); Williams Electronics, Inc. v. Artic International, Inc., 685 F.2d 870 (3d Cir.1982) (object code). In this case, however, the district court did not find any copying of the source or object codes, nor did the plaintiff allege such copying. Rather, the district court held that the Dentalab copyright was infringed because the overall structure of Dentcom was substantially similar to the overall structure of Dentalab. Whelan Associates v. Jaslow Dental Laboratory, 609 F. Supp. at 1321-22. The question therefore arises whether mere similarity in the overall structure of programs can be the basis for a copyright infringement, or, put differently, whether a program's copyright protection covers the structure of the program or only the program's literal elements, i.e., its source and object codes.

Title 17 U.S.C. Sec. 102(a)(1) extends copyright protection to "literary works," and computer programs are classified as literary works for the purposes of copyright. See H.R.Rep. No. 1476, 94th Cong., 2d Sess. 54, reprinted in 1976 U.S.Code Cong. & Ad.News 5659, 5667. The copyrights of other literary works can be infringed even when there is no substantial similarity between the works' literal elements. One can violate the copyright of a play or book by copying its plot or plot devices. See, e.g., Twentieth Century-Fox Film Corp. v. MCA, Inc., 715 F.2d 1327, 1329 (9th Cir.1983) (13 alleged distinctive plot similarities between Battlestar Galactica and Star Wars may be basis for a finding of copyright violation); Sid & Marty Krofft Television Productions, Inc., 562 F.2d at 1167 (similarities between McDonaldland characters and H.R. Pufnstuf characters can be established by " 'total concept and feel' " of the two productions (quoting Roth Greeting Cards v. United Card Co., 429 F.2d 1106, 1110 (9th Cir.1970)); Sheldon v. Metro-Goldwyn Pictures Corp., 81 F.2d 49, 54-55 (2nd Cir.1936); Nichols v. Universal Pictures Corp., 45 F.2d 119, 121 (2d Cir.1930) (copyright "cannot be limited literally to the text, else a plagiarist would escape by immaterial variations"). By analogy to other literary works, it would thus appear that the copyrights of computer programs can be infringed even absent copying of the literal elements of the program.26 Defendants contend, however, that what is true of other literary works is not true of computer programs. They assert two principal reasons, which we consider in turn.

A. Section 102(b) and the dichotomy between idea and expression--It is axiomatic that copyright does not protect ideas, but only expressions of ideas. This rule, first enunciated in Baker v. Selden, 101 U.S. (11 Otto) 99, 25 L. Ed. 841 (1879), has been repeated in numerous cases. See, e.g., Mazur v. Stein, 347 U.S. 201, 217, 74 S. Ct. 460, 470, 98 L. Ed. 630 (1954) ("Unlike a patent, a copyright gives no exclusive right to the art disclosed; protection is given only to the expression of the idea--not the idea itself." (citation omitted)); Universal Athletic Sales Co., 511 F.2d at 906; Dymow v. Bolton, 11 F.2d 690, 691 (2d Cir.1926); see generally A. Latman, The Copyright Law 31-35 (5th ed. 1979); 1 Nimmer Sec. 2.03[D]. The rule has also been embodied in statute. Title 17 U.S.C. Sec. 102(b) (1982) states:

In no case does copyright protection for an original work of authorship extend to any idea, procedure, process, system, method of operation, concept, principle, or discovery, regardless of the form in which it is described, explained, illustrated, or embodied in such work.

The legislative history of this section, adopted in 1976, makes clear that Sec. 102(b) was intended to express the idea-expression dichotomy. See H.R.Rep. No. 1476 at 57 reprinted in 1976 U.S.Code Cong. & Ad.News at 5670 (Sec. 102(b) is intended to "restate ... that the basic dichotomy between expression and idea remains unchanged.") See also Apple Computer, supra, 714 F.2d at 1252.

Defendants argue that the structure of a computer program is, by definition, the idea and not the expression of the idea, and therefore that the structure cannot be protected by the program copyright. Under the defendants' approach, any other decision would be contrary to Sec. 102(b). We divide our consideration of this argument into two parts. First, we examine the caselaw concerning the distinction between idea and expression, and derive from it a rule for distinguishing idea from expression in the context of computer programs. We then apply that rule to the facts of this case.

1. A rule for distinguishing idea from expression in computer programs--It is frequently difficult to distinguish the idea from the expression thereof. No less an authority than Learned Hand, after a career that included writing some of the leading copyright opinions, concluded that the distinction will "inevitably be ad hoc." Peter Pan Fabrics, Inc. v. Martin Weiner Corp., 274 F.2d 487, 489 (2d Cir.1960). See also Knowles & Palmieri, Dissecting Krofft: An Expression of New Ideas in Copyright?, 8 San.Fern.Val.L.Rev. 109, 126 (1980) (arguing that there can be no meaningful distinction between idea and expression). Although we acknowledge the wisdom of Judge Hand's remark, we feel that a review of relevant copyright precedent will enable us to formulate a rule applicable in this case. In addition, precisely because the line between idea and expression is elusive, we must pay particular attention to the pragmatic considerations that underlie the distinction and copyright law generally. In this regard, we must remember that the purpose of the copyright law is to create the most efficient and productive balance between protection (incentive) and dissemination of information, to promote learning, culture and development. See U.S. Const. Art. I, Sec. 8, cl. 8 (Copyright Clause) (giving Congress the power to "promote the Progress of Science... by securing for limited Times to Authors... the exclusive Right to their respective Writings").27

We begin our analysis with the case of Baker v. Selden, which, in addition to being a seminal case in the law of copyright generally, is particularly relevant here because, like the instant case, it involved a utilitarian work, rather than an artistic or fictional one. In Baker v. Selden, the plaintiff Selden obtained a copyright on his book, "Selden's Condensed Ledger, or Bookkeeping Simplified," which described a new, simplified system of accounting. Included in the book were certain "blank forms," pages with ruled lines and headings, for use in Selden's accounting system. Selden alleged that Baker had infringed Selden's copyright by making and selling accounting books that used substantially the same system as Selden's and that reproduced Selden's blank forms. No one disputed that Baker had the right to use and promulgate Selden's system of accounting, for all parties agreed that the system could not be copyrighted, although the Court opined that it might be patentable. Id., 101 U.S. at 102. Nor did the parties dispute that the text of Baker's book on accounting did not infringe Selden's copyright. The dispute centered on whether Selden's blank forms were part of the method (idea) of Selden's book, and hence non-copyrightable, or part of the copyrightable text (expression). Id. at 101.

In deciding this point, the Court distinguished what was protectible from what was not protectible as follows:

[W]here the art [i.e., the method of accounting] it teaches cannot be used without employing the methods and diagrams used to illustrate the book, or such as are similar to them, such methods and diagrams are to be considered as necessary incidents to the art, and given to the public.

Id. at 103. Applying this test, the Court held that the blank forms were necessary incidents to Selden's method of accounting, and hence were not entitled to any copyright protection. Id. at 104.

The Court's test in Baker v. Selden suggests a way to distinguish idea from expression. Just as Baker v. Selden focused on the end sought to be achieved by Selden's book, the line between idea and expression may be drawn with reference to the end sought to be achieved by the work in question. In other words, the purpose or function of a utilitarian work would be the work's idea, and everything that is not necessary to that purpose or function would be part of the expression of the idea. Cf. Apple Computer, Inc. v. Formula Int'l, Inc., 562 F. Supp. 775, 783 (C.D.Ca.1983) ("Apple seeks here not to protect ideas (i.e. making the machine perform particular functions) but rather to protect their particular expressions ..."), aff'd, 725 F.2d 521 (9th Cir.1984). Where there are various means of achieving the desired purpose, then the particular means chosen is not necessary to the purpose; hence, there is expression, not idea.28

Consideration of copyright doctrines related to scenes a faire and fact-intensive works supports our formulation, for they reflect the same underlying principle. Scenes a faire are "incidents, characters or settings which are as a practical matter indispensable ... in the treatment of a given topic." Atari, Inc. v. North American Philips Consumer Elecs. Corp., 672 F.2d 607, 616 (7th Cir.), cert. denied, 459 U.S. 880, 103 S. Ct. 176, 74 L. Ed. 2d 145 (1982). See also See v. Durang, 711 F.2d 141, 143 (9th Cir.1983). It is well-settled doctrine that scenes a faire are afforded no copyright protection.29

Scenes a faire are afforded no protection because the subject matter represented can be expressed in no other way than through the particular scene a faire. Therefore, granting a copyright "would give the first author a monopoly on the commonplace ideas behind the scenes a faire." Landsberg v. Scrabble Crossword Game Players, Inc., 736 F.2d at 489.30 This is merely a restatement of the hypothesis advanced above, that the purpose or function of a work or literary device is part of that device's "idea" (unprotectible portion). It follows that anything necessary to effecting that function is also, necessarily, part of the idea, too.

Fact intensive works are given similarly limited copyright coverage. See, e.g., Landsberg, 736 F.2d at 488; Miller v. Universal City Studios, Inc., 650 F.2d 1365, 1372 (5th Cir.1981). Once again, the reason appears to be that there are only a limited number of ways to express factual material, and therefore the purpose of the literary work--telling a truthful story--can be accomplished only by employing one of a limited number of devices. Landsberg, 736 F.2d at 488. Those devices therefore belong to the idea, not the expression, of the historical or factual work.

Although the economic implications of this rule are necessarily somewhat speculative, we nevertheless believe that the rule would advance the basic purpose underlying the idea/expression distinction, "the preservation of the balance between competition and protection reflected in the patent and copyright laws." Herbert Rosenthal Jewelry Corp. v. Kalpakian, 446 F.2d 738, 742 (9th Cir.1971); see also Apple Computer, 714 F.2d at 1253 (quoting Kalpakian ); supra n. 27. As we stated above, see supra at 1231, among the more significant costs in computer programming are those attributable to developing the structure and logic of the program. The rule proposed here, which allows copyright protection beyond the literal computer code,31 would provide the proper incentive for programmers by protecting their most valuable efforts, while not giving them a stranglehold over the development of new computer devices that accomplish the same end.

The principal economic argument used against this position--used, that is, in support of the position that programs' literal elements are the only parts of the programs protected by the copyright law--is that computer programs are so intricate, each step so dependent on all of the other steps, that they are almost impossible to copy except literally, and that anyone who attempts to copy the structure of a program without copying its literal elements must expend a tremendous amount of effort and creativity. In the words of one commentator: "One cannot simply 'approximate' the entire copyrighted computer program and create a similar operative program without the expenditure of almost the same amount of time as the original programmer expended." Note, 68 Minn.L.Rev. at 1290 (footnote omitted). According to this argument, such work should not be discouraged or penalized. A further argument against our position is not economic but jurisprudential; another commentator argues that the concept of structure in computer programs is too vague to be useful in copyright cases. Radcliffe, Recent Developments in Copyright Law Related to Computer Software, 4 Computer L.Rep. 189, 194-97 (1985). He too would therefore appear to advocate limiting copyright protection to programs' literal codes.

Neither of the two arguments just described is persuasive. The first argument fails for two reasons. In the first place, it is simply not true that "approximation" of a program short of perfect reproduction is valueless. To the contrary, one can approximate a program and thereby gain a significant advantage over competitors even though additional work is needed to complete the program. Second, the fact that it will take a great deal of effort to copy a copyrighted work does not mean that the copier is not a copyright infringer. The issue in a copyrighted case is simply whether the copyright holder's expression has been copied, not how difficult it was to do the copying. Whether an alleged infringer spent significant time and effort to copy an original work is therefore irrelevant to whether he has pirated the expression of an original work.32

As to the second argument, it is surely true that limiting copyright protection to computers' literal codes would be simpler and would yield more definite answers than does our answer here. Ease of application is not, however, a sufficient counterweight to the considerations we have adduced on behalf of our position.

Finally, one commentator argues that the process of development and progress in the field of computer programming is significantly different from that in other fields, and therefore requires a particularly restricted application of the copyright law. According to this argument, progress in the area of computer technology is achieved by means of "stepping-stones," a process that "requires plagiarizing in some manner the underlying copyrighted work." Note, 68 Minn.L.Rev. at 1292 (footnote omitted). As a consequence, this commentator argues, giving computer programs too much copyright protection will retard progress in the field.

We are not convinced that progress in computer technology or technique is qualitatively different from progress in other areas of science or the arts. In balancing protection and dissemination, see supra at 1235 & n. 27, the copyright law has always recognized and tried to accommodate the fact that all intellectual pioneers build on the work of their predecessors.33 Thus, copyright principles derived from other areas are applicable in the field of computer programs.

2. Application of the general rule to this case--The rule proposed here is certainly not problem-free. The rule has its greatest force in the analysis of utilitarian or "functional" works, for the purpose of such works is easily stated and identified. By contrast, in cases involving works of literature or "non-functional" visual representations, defining the purpose of the work may be difficult. Since it may be impossible to discuss the purpose or function of a novel, poem, sculpture or painting, the rule may have little or no application to cases involving such works. The present case presents no such difficulties, for it is clear that the purpose of the utilitarian Dentalab program was to aid in the business operations of a dental laboratory.34 See supra 1225. It is equally clear that the structure of the program was not essential to that task: there are other programs on the market, competitors of Dentalab and Dentcom, that perform the same functions but have different structures and designs.

This fact seems to have been dispositive for the district court:

The mere idea or concept of a computerized program for operating a dental laboratory would not in and of itself be subject to copyright. Copyright law protects the manner in which the author expresses an idea or concept, but not the idea itself. Albert E. Price v. Metzner, 574 F. Supp. 281 (E.D.Pa.1983). Copyrights do not protect ideas--only expressions of ideas. Universal Athletic Sales Co. v. Salkeld, 511 F.2d 904, 908 (3d Cir.1975). There are many ways that the same data may be organized, assembled, held, retrieved and utilized by a computer. Different computer systems may functionally serve similar purposes without being copies of each other. There is evidence in the record that there are other software programs for the business management of dental laboratories in competition with plaintiff's program. There is no contention that any of them infringe although they may incorporate many of the same ideas and functions. The 'expression of the idea' in a software computer program is the manner in which the program operates, controls and regulates the computer in receiving, assembling, calculating, retaining, correlating, and producing useful information either on a screen, print-out or by audio communication.

Whelan Associates v. Jaslow Laboratory, 609 F. Supp. at 1320 (emphasis added). We agree. The conclusion is thus inescapable that the detailed structure of the Dentalab program is part of the expression, not the idea, of that program.

Our conclusion is supported by SAS Institute, Inc. v. S & H Computer Systems, Inc., 605 F. Supp. 816 (M.D.Tenn.1985), the only other case that has addressed this issue specifically, in which the court found that a program's copyright could extend beyond its literal elements to its structure and organization. In SAS, plaintiffs supported their allegation of copyright infringement of their computer program with evidence of both literal and organizational similarities between its program and the alleged infringer. Id. at 822, 825-26. The court found a copyright infringement, and although it did not discuss in detail its evaluation of the evidence, it is apparent that the organizational similarities of the programs were relevant to its decision. After a brief discussion of the programs' literal similarities, the court said:

In addition, the copying proven at trial does not affect only the specific lines of code cited by Dr. Peterson in his testimony. Rather, to the extent that it represents copying of the organization and structural details of SAS, such copying pervades the entire S & H product.

Id. at 830. Although the SAS court did not analyze the point in great depth, we are encouraged by its conclusion.

The Copyright Act of 1976 provides further support, for it indicates that Congress intended that the structure and organization of a literary work could be part of its expression protectible by copyright. Title 17 U.S.C. Sec. 103 (1982) specifically extends copyright protection to compilations and derivative works. Title 17 U.S.C. Sec. 101, defines "compilation" as "a work formed by the collection and assembling of preexisting materials or of data that are selected, coordinated, or arranged in such a way that the resulting work as a whole constitutes an original work of authorship," and it defines "derivative work," as one "based upon one or more preexisting works, such as ... abridgement, condensation, or any other form in which a work may be recast, transformed, or adapted." (emphasis added). Although the Code does not use the terms "sequence," "order" or "structure," it is clear from the definition of compilations and derivative works, and the protection afforded them, that Congress was aware of the fact that the sequencing and ordering of materials could be copyrighted, i.e., that the sequence and order could be parts of the expression, not the idea, of a work.

Our solution may put us at odds with Judge Patrick Higginbotham's scholarly opinion in Synercom Technology, Inc. v. University Computing Co., 462 F. Supp. 1003 (N.D.Tex.1978), which dealt with the question whether the "input formats" of a computer program--the configurations and collations of the information entered into the program--were idea or expression. The court held that the input formats were ideas, not expressions, and thus not protectible. Synercom did not deal with precisely the materials at issue here--input formats are structurally simple as compared to full programs--and it may therefore be distinguishable. However, insofar as the input formats are devices for the organization of data into forms useful for computers, they are similar to programs; thus, Synercom is relevant and we must come to grips with it.

Central to Judge Higginbotham's analysis was his conviction that the organization and structure of the input formats was inseparable from the idea underlying the formats. Although the court acknowledged that in some cases structure and sequence might be part of expression, not idea, see id. at 1014, it stated that in the case of input formats, structure and organization were inherently part of the idea. The court put its position in the form of a powerful rhetorical question: "if sequencing and ordering [are] expression, what separable idea is being expressed?" Id. at 1013.

To the extent that Synercom rested on the premise that there was a difference between the copyrightability of sequence and form in the computer context and in any other context, we think that it is incorrect. As just noted, the Copyright Act of 1976 demonstrates that Congress intended sequencing and ordering to be protectible in the appropriate circumstances, see supra p. 1237-1238, and the computer field is not an exception to this general rule. Although Congress was aware that computer programs posed a novel set of issues and problems for the copyright law, Congress did not then make, and has not since made, any special provision for ordering and sequencing in the context of computer programs. There is thus no statutory basis for treating computer programs differently from other literary works in this regard.35

Despite the fact that copyright protection extends to sequence and form in the computer context, unless we are able to answer Judge Higginbotham's powerful rhetorical question--"if sequencing and ordering [are] expression, what separable idea is being expressed?"--in our own case, we would have to hold that the structure of the Dentalab program is part of its idea and is thus not protectible by copyright. Our answer has already been given, however: the idea is the efficient organization of a dental laboratory (presumably, this poses different problems from the efficient organization of some other kinds of laboratories or businesses). Because there are a variety of program structures through which that idea can be expressed, the structure is not a necessary incident to that idea.36 See supra 1239.

B. The CONTU Report--Defendants' second argument against copyright protection for non-literal elements of computer programs relies not on venerable principles of copyright law, but on the report of a special congressional commission. In 1974, concerned that the rapidly developing field of computer technology was outpacing the extant copyright laws, Congress passed Pub.L. 93-573, Sec. 201, 88 Stat.1873 (1974) creating the National Commission on New Technological Uses of Copyrighted Works (CONTU) to study and report on the problems and issues of new technology and copyright.

In 1976, before CONTU reported its findings to Congress, Congress passed a new copyright law to replace the one that had prevailed since 1909. Pub.L. No. 94-553, 90 Stat. 2541 (1976) (codified at 17 U.S.C. Sec. 101 et seq.). The new law's only explicit accommodation to computer technology was 17 U.S.C. Sec. 117, which dealt with copyright issues arising upon certain uses of computer programs not in issue here.

The Commission's Final Report [hereinafter "CONTU Report"] was delivered to the President on July 31, 1978. The CONTU Report recommended legislation:

The new copyright law should be amended: (1) to make it explicit that computer programs, to the extent that they embody an author's original creation, are proper subject matter for copyright; (2) to apply to all computer uses of copyrighted programs by the deletion of the present section 117; and (3) to ensure that rightful possessors of copies of computer programs may use or adapt these copies for their use.

CONTU Report at 1. Congress responded to these suggestions in the 1980 Computer Software Copyright Act by adding a provision to 17 U.S.C. Sec. 101 defining computer programs, see Pub.L. No. 96-517, Sec. 10(a), 94 Stat. 3028 (1980) (codified at 17 U.S.C. Sec. 101) (quoted supra n. 13). It also replaced the old Sec. 117 with a new Sec. 117 outlining users' rights of adaptation.

As neither the definition of "computer program" nor 17 U.S.C. Sec. 117 is relevant here, it would appear that the CONTU Report is of marginal relevance, at best. Nevertheless, defendants argue that it is not only relevant, but that it compels a decision in their favor. Specifically, defendants argue that the CONTU Report recommends that copyright protection be limited to protection of the literal elements of computer programs, and that the CONTU Report is to be taken as authoritative legislative history, which this court is bound to follow. Appellants' Br. at 15-21. This argument is unpersuasive, however.

In the first place, the CONTU Report never suggests that copyright protection should be limited to the literal code. To the contrary, in discussing the limits of copyright protection, the Report referred to the dichotomy between idea and expression discussed above, and said:

[T]he separation of idea from form of expression ... is better realized through the courts exercising their judgment in particular cases [than by a per se rule].... Flow charts, source codes, and object codes are works of authorship in which copyright subsists....

CONTU Report at 21 (emphasis added). While the first part of this passage suggests that the Commission may have reserved judgment as to the copyrightability of non-literal elements of computer programs, the second part--particularly the reference to the copyrightability of flowcharts--demonstrates that the Commission intended copyright protection to extend beyond the literal code. Thus, the CONTU Report does not support defendants' position in this case.37

Even if the CONTU Report had advocated a strict limitation of the copyrightability of computer programs, defendants' argument would still fail, for the CONTU Report is not binding on us in this case. Defendants correctly note that some courts have treated the CONTU Report as a surrogate legislative history. See, e.g., Micro-Sparc, Inc. v. Amtype Corp., 592 F. Supp. 33, 35 n.7 (D.Mass.1984) ("The CONTU Report ... comprises the entire legislative history of Sec. 117."); Midway Mfg. Co. v. Strohon, 564 F. Supp. 741, 750 n. 6. The courts have said this because Congress adopted the CONTU Report's recommendations without alterations and without any committee reports. Defendants argue that this court, too, is bound by the CONTU Report for the same reason.

The argument fails, however, for the only statutory provision relevant to this case is Sec. 102(b), see supra 1234-35, in which no changes were made as a result of the CONTU Report. Therefore, it follows that the CONTU Report cannot be a substitute for legislative history in this case. In effect, defendants ask us to apply the precepts of the CONTU Report regardless of what legislative provision we are considering. We cannot do this, however, for the CONTU Report has force only insofar as it can be said to represent the will of Congress. There is no sense in which it represents the will of Congress with respect to provisions not amended in response to the Report.38

Defendants' second argument is that even if copyright protection is not limited to computer programs' literal elements as a matter of law, there is insufficient evidence of substantial similarity presented in this case to support a finding of copyright infringement. The defendants claim that all three parts of Dr. Moore's expert testimony as to the similarity of the programs, see supra at 1228, were flawed, and also that the district court erred in evaluating the relative weight of Dr. Moore's and Mr. Ness' testimony. We consider these arguments in turn.

A. File structures--Defendants claim that Dr. Moore's examination and conclusions with respect to file structures are irrelevant to the question whether there was a copyright violation. Defendants analogize files to blank forms, which contain no information but merely collect and organize information that is entered from another source. They argue, relying on Baker v. Selden, that, as a matter of law, blank forms cannot be copyrighted. Thus, they conclude, neither can file structures be part of the copyright of a program. See Appellants' Br. at 34-38; Reply Br. at 7-8.39

Defendants' description of the file structures is indeed correct. Dr. Moore himself described a computer's file as "a storage place for data, and it's really no different in a computer than it is in a file drawer, it's like a manila folder that contains all the data on a particular subject category in a computer." App. at 682. (Another analogy, particularly accessible to lawyers, is to a very complex cataloguing structure like the structure of Lexis or Westlaw without any entries yet made.) Defendants' legal conclusion is not correct, however. Although some courts have stated that the meaning of Baker v. Selden is that blank forms cannot be copyrighted,40 this circuit, like the majority of courts that have considered the issue,41 has rejected this position and instead have held that blank forms may be copyrighted if they are sufficiently innovative that their arrangement of information is itself informative. Apple Computer, Inc., 714 F.2d at 1250. See also Manpower, Inc. v. Temporary Help of Harrisburg, Inc., 246 F. Supp. 788 (E.D.Pa.1965) (upholding copyrightability of form for vacation schedules).42

This is not to say that all blank forms or computer files are copyrightable. Only those that by their arrangement and organization convey some information can be copyrighted. Cf. 1 Nimmer at 2-201: "Thus books intended to record the events of baby's first year, or a record of a European trip, or any one of a number of other subjects, may evince considerable originality in suggestions of specific items of information which are to be recorded, and in the arrangement of such items." (footnote omitted). Defendants do not contend, however, that the file structures convey no information, and it appears to us that the structures are sufficiently complex and detailed that such an argument would not succeed. As we have noted, supra at 1239, there are many ways in which the same goal--the organization of the business aspects of a dental laboratory--might be accomplished, and several of these approaches might use significantly different file structures.43 The file structures in the Dentalab and Dentcom systems require certain information and order that information in a particular fashion. Other programs might require different information or might use the same information differently. When we compare the comprehensiveness and complexity of the file structures at issue here with the "blank forms" at issue in the cases mentioned above, we have no doubt that these file structures are sufficiently informative to deserve copyright protection.

B. Screen outputs--Defendants' second line of argument is slightly confusing. Defendants appear to argue that to the extent that the district court relied upon the similarity of the screen outputs of Dentalab and Dentcom its finding of substantial similarity was erroneous because (1) the screen outputs are covered by a different copyright from the program's, and/or (2) the screen outputs bear no relation to the programs that produce them. Although these arguments are not always clearly distinguished, Appellants' Br. at 38-41; Reply Br. at 8-10, the distinction is important because whereas the first argument is weak, we feel that the second is more persuasive.

It is true that screen outputs are considered audio-visual works under the copyright code, see Williams Electronics, 685 F.2d at 874; Midway Manufacturing Co., 564 F. Supp. at 749 (distinguishing audiovisual copyright in display of videogame from copyright in program that creates the audiovisual display), and are thus covered by a different copyright than are programs, which are literary works, see supra at 1233. It is also true that Whelan Associates asserts no claim of copyright infringement with respect to the screen outputs. But the conclusion to be drawn from this is not, as defendants would have it, that screen outputs are completely irrelevant to the question whether the copyright in the program has been infringed. Rather, the only conclusion to be drawn from the fact of the different copyrights is that the screen output cannot be direct evidence of copyright infringement. There is no reason, however, why material falling under one copyright category could not be indirect, inferential evidence of the nature of material covered by another copyright.

Thus, the question is whether the screen outputs have probative value concerning the nature of the programs that render them sufficient to clear the hurdles of Fed.R.Evid. 401 and 403. The defendants argue that the screen outputs have no probative value with respect to the programs because many different programs can create the same screen output. Defendants rely on Stern Electronics Inc. v. Kaufman, 669 F.2d at 855 ("many different computer programs can produce the same 'results,' whether those results are an analysis of financial records or a sequence of images and sounds."), and Midway Manufacturing Co., 564 F. Supp. at 749 ("it is quite possible to design a game that would infringe Midway's audiovisual copyright but would use an entirely different computer program."). Neither court, however, was presented with the question that faces us today, the evidentiary value of screen outputs in a suit for infringement of the underlying program.44

Insofar as everything that a computer does, including its screen outputs, is related to the program that operates it, there is necessarily a causal relationship between the program and the screen outputs. The screen outputs must bear some relation to the underlying programs, and therefore they have some probative value. The evidence about the screen outputs therefore passes the low admissibility threshold of Fed.R.Evid. 401.45

It is still possible, of course, that the risk of unfair prejudice would outweigh the probative value of the evidence, thus contravening Fed.R.Evid. 403. In support of this position one might argue that, because the screen outputs are vivid and easily understood (at least as compared with the obscure details of computer programs), they might have disproportionate influence on the trier of fact. However, because the portions of programs that relate to the screen outputs are often so small a part of the full program, they might say very little about the underlying program. This combination of ease of perception and slight probative value, the argument concludes, works a Rule 403 violation because the screen outputs appear significantly more probative than they in fact can be.

This argument has force, but we feel that it is ultimately unpersuasive. In the first place, the defendants point to no place in the record where they objected to testimony about the screens' similarities. Our independent review of the record has revealed no such objections. See, e.g., App. at 681-82 (Dr. Moore testifying about screen similarities without objection from defendants). Therefore, because this could not be plain error, the objection was waived. Second, even if the objection had not been waived, we do not believe that admission of the evidence relating to screen similarity suffices for a Rule 403 exclusion. Screen outputs are not so enticing that a trier of fact could not evaluate them rationally and with a cool head. The party against whom the screen outputs are introduced can easily explain their limited probative value. Given the substantial deference with which we review the district court's decision on this matter, see McQueeney v. Wilmington Trust Co., 779 F.2d 916, 922 (3d Cir.1985), and the parties' ample opportunities to advance their arguments before the district court, we do not believe that the district court committed error in considering the evidence relating to screen outputs.

C. The five subroutines--With respect to the final piece of evidence, Dr. Moore's testimony about the five subroutines found in Dentalab and Dentcom, defendants state that they "fail to understand how a substantial similarity in structure can be established by a comparison of only a small fraction of the two works." Appellants' Brief at 43, see also Reply Br. at 12. The premise underlying this declaration is that one cannot prove substantial similarity of two works without comparing the entirety, or at least the greater part, of the works. We take this premise to be the defendants' argument.

The premise does not apply in other areas of copyright infringement. There is no general requirement that most of each of two works be compared before a court can conclude that they are substantially similar. In the cases of literary works--novels, movies, or plays, for example--it is often impossible to speak of "most" of the work. The substantial similarity inquiry cannot be simply quantified in such instances. Instead, the court must make a qualitative, not quantitative, judgment about the character of the work as a whole and the importance of the substantially similar portions of the work. See, e.g., Harper & Row Publishers, Inc. v. Nation Enterprises, 471 U.S. 539, ---- - ---- & ---- n. 8, 105 S. Ct. 2218, 2233-34 & 2233 n. 8, 85 L. Ed. 2d 588 (1985); Atari, Inc. v. North American Phillips Consumer Electronics Corp., 672 F.2d 607, 618 (7th Cir.) ("When analyzing two works to determine whether they are substantially similar, courts should be careful not to lose sight of the forest for the trees."), cert. denied, 459 U.S. 880, 103 S. Ct. 176, 74 L. Ed. 2d 145 (1982); Hoehling v. Universal City Studios, Inc., 618 F.2d 972, 979-80 (2d Cir.) (warning against same danger), cert. denied, 449 U.S. 841, 101 S. Ct. 121, 66 L. Ed. 2d 49 (1980). See also Universal Pictures v. Harold Lloyd Corp., 162 F.2d 354 (9th Cir.1947) (finding copyright violation for copying of 20% of plaintiff's film); In re Personal Computers and Components Thereof, 1983-84 Copyright L.Dec. (CCH) p 25,651 at 18,931 (Int'l Trade Comm'n Mar. 9, 1984) (18%-25% identity is sufficient for substantial similarity). Compare Elsmere Music, Inc. v. National Broadcasting Co., 482 F. Supp. 741, 744 (S.D.N.Y.), aff'd, 623 F.2d 252 (2d Cir.1980) (similarity uncontested by defendants where four notes out of 100 measures and two words out of 45 were identical) with Jewel Music Publishing Co. v. Leo Feist, Inc., 62 F. Supp. 596, 597 (S.D.N.Y.1945) (finding no substantial similarity where three bars of an eight-bar line were identical and the line appeared 24 times in both songs).

Computer programs are no different. Because all steps of a computer program are not of equal importance, the relevant inquiry cannot therefore be the purely mechanical one of whether most of the programs' steps are similar. Rather, because we are concerned with the overall similarities between the programs, we must ask whether the most significant steps of the programs are similar. See Midway Mfg. Co. v. Strohon, supra, 564 F.Supp. at 753. This is precisely what Dr. Moore did. He testified as follows:

What I decided to do was to look at the programs that had the primary, or let's say most important, tasks of the system, and also ones which manipulate files, because there are a lot of programs that simply print lists, or answer a question when you ask him it, but I thought that the programs which actually showed the flow of information, through the system, would be the ones that would illustrate the system back.

App. at 704. Dr. Moore's testimony was thus in accord with general principles of copyright law. As we hold today that these principles apply as well to computer programs, we therefore reject the defendants' argument on this point.

D. Sufficiency of the evidence--Defendants' final argument is that the district court erred in evaluating the testimony of Dr. Moore and Mr. Ness. They contend that, properly evaluated, Mr. Ness' testimony was sufficiently strong and Dr. Moore's sufficiently weak, that there was not sufficient evidence of substantial similarity for plaintiff to prevail.

We have described the testimony of Dr. Moore and Mr. Ness, see supra 1227; see also infra 1247-48, and it is recounted in the district court opinion, Whelan Associates v. Jaslow Dental Laboratory, 609 F. Supp. at 1316. The district court explained its evaluation of the evidence as follows:

I conclude ... that Dr. Moore, plaintiffs' expert, had greater knowledge as to the particular programs at issue. Dr. Hess [sic], the defendants' expert witness, reviewed only the source and object codes of the IBM-Series 1 [Dentalab program], the IBM Datamaster [see supra n. 4], and the IBM-PC Dentcom system. He never observed the computer in operation nor viewed the various screens or the user's manual. He stated he was not familiar with EDL coding. More basically, however, his comparison as to dissimilarities was between the IBM Datamaster and the IBM-PC Dentcom systems. Plaintiff contends that the IBM-PC Dentcom is a copy of the IBM Series 1 System--not the IBM-Datamaster system. Dr. Hess's conclusions were that although the overall structures of those systems is similar, the code in the IBM-PC Dentcom is not "directly derived" from either plaintiff's IBM Series-1 or its IBM-Datamaster system. To the extent that Dr. Moore's testimony supports plaintiff's contentions of copying, I find his testimony more credible and helpful because of his detailed and thorough analysis of the many similarities.

Id. at 1321-22. Determinations of credibility and the relative weight to be given expert witnesses are, of course, left primarily to the discretion of the district court. Our review of the record convinces us that the district court's analysis of the two experts' opinions was far from being erroneous. As the district court pointed out, Mr. Ness had studied the programs, but he had never observed them at work in computers. Id. at 1321. The district court also pointed out that Mr. Ness was unfamiliar with EDL coding. Id. These factors suffice to support the court's conclusion.

In addition, we believe, on re-reading the trial transcript, that although Mr. Ness' testimony was quite competent, Dr. Moore's was more persuasive on the issues relevant to this appeal. Whereas the greater part of Mr. Ness' testimony was concerned with the dissimilarities between the two programs' source and object codes, see, e.g., App. at 824-32, Dr. Moore discussed the crucial issue in this case, the similarities and differences in the programs' structures. For example, when he discussed the programs' invoicing subroutine, Dr. Moore testified as follows:

In the DentaLab system, the same kind of thing again, same information is up there, description, unit price, extension, items and program reads all those things in from te [sic ] number of files actually, and displays them and then gives the operator a number of options to change the order as it appears on the screen to skip this one, to cancel it, or to accept it.

The same choices are given in Dentalab systems, change, skip, cancel.

Assuming that the order is accepted, both systems then calculate the money, calculate the amount of money that will be billed, and at this point they use the price code to find which of the four prices are to be charged for this particular customer. Both systems do that. They pick that one of the four prices and calculate the total amount, write [sic] then the record of this invoice that has been formed to show the invoice's file, sets the flag in the order's file to show that this order has now been invoiced so that it doesn't get reinvoiced.

Q. What is a flag?

A. Well, a flag would be, in this case, a certain location is marked I for invoices, just an indicator that invoicing has been done on this record.

Q. Both used it?

A. Both used a flag. I don't remember whether Dentcom uses a letter I or some other symbol, but there is a flag there that it's a field number 12, in which it's indicated that this file has been invoiced or this order has been invoiced.

* * *

* * *

Q. Do you have any comment about the invoicing?

A. Well, I think it should be clear, it was clear to me from going through these programs that there is a very marke[d] similarity between the two, that they, item by item, are doing pretty much the same thing with the same fields in the same files, and accomplishing roughly the same results.

So there was quite a match, line by line, between these two, flow in these two.

Q. What do you conclude from that?

A. Well, back together with the file's structure, sort of set up with the--how the programs have to proceed. I would think that the person who designed or constructed the DentaLab system must have been thoroughly familiar with the Dentcom system.

The person who constructed the Dentcom system must have been familiar with the series one system, because the same file structure and same program steps are followed, same overall flow takes place in both systems.

App. at 710-12 (emphasis added). Dr. Moore's testimony about Dentcom's and Dentalab's month-end subroutines also demonstrates the structural comparisons in which he engaged:

Q. What did you find in month-end?

A. Okay. Month-end, the calling program in Dentcom, this obviously is done at the end of each month.

In the Dentcom system there is a program called MOEND, which chains all these other programs, that is, MOEND calls MOPRDL, and after that program runs, goes back to MEEND, calls the print sale and so on.

In the Dentalab system there is a supervisory program also called MOEND, and that system calls or runs a series of programs doing various functions.

Now if we look at the functions done by the programs in order, we find that they are the same except for a flipping of the order in the first two things.

The Dentcom system, it first prints product group report, and then prints the monthly customer sales analysis.

In Dentalab, just reverses, prints sales analysis first, product group report second.

After that, both systems do the same thing in the same order.

They now do accounts receivable aging, since a month has gone by they have to update all the 30 days, 60 days, et cetera, calculate service charges. Then they print the monthly AR reports that had to do with service charges, only those that involve service charges, they both do that. Then they both print the age file balance, balance report, and following that they print the month and accounts receivable report. That's the total accounts receivable rport [sic].

Then they both go through and look for accounts that are not active that month, and print a list of these accounts, accounts not serviced, an account that doesn't have any access.

The final thing that the Dentcom system does is to calculate the new AR total for the entire lab, which I mentioned is contained in the company file.

DentaLab doesn't keep that total, so that's the last item, that is not as far as I can tell, done by DentaLab. I may have said--did I say Dentcom keeps that total? DentaLab does not. That's the only difference.

App. at 716-18 (emphasis added). Dr. Moore testified in similar detail and to similar effect about the other three subroutines that he felt were particularly important, order entry, accounts receivable, and day's end.46 This testimony, in addition to Dr. Moore's exhaustive comparison of the two programs' file structures, see supra 1227, and his testimony about the screen outputs, id., demonstrates the marked similarity between the programs. Defendants' argument as to sufficiency of the evidence therefore fails.47

We hold that (1) copyright protection of computer programs may extend beyond the programs' literal code to their structure, sequence, and organization, and (2) the district court's finding of substantial similarity between the Dentalab and Dentcom programs was not clearly erroneous. The judgment of the district court will therefore be affirmed.

                                  APPENDIX A
                             ORDER ENTRY PROGRAMS
Dentcom PC Systems                          Dentalab System
Primary Menu, choose (1)                    Primary Menu, choose (1)
Production MENU                             Production SCHEDULING
Production menu choose (1)                  Production Menu, choose (1)
ORDER ENTRY                                 ORDER ENTRY
ORDER ENTRY program                         DL1000 program (Order entry)
"ENTER ACCOUNT NAME KEY:                    "ENTER ACCOUNT OR NAMEKEY"
_________________________________           "Check CUSTMAST for valid customer.
Check CUSTMAST for valid customer.  If       Read CUSTMAST file on this
                                              customer.
valid, read CUSTMAST file for this
  customer
If yes, increment order # in ORDERS         Increment sequential order no. in
                                              ORDERS.
                                            Display customer name, address.
Display entry screen, patient shade,        Display entry screen (6.6), patient
  mould,                                      shade,
remake, call Dr.?                           remake, call
Dr.?
case/span #, Drs. request date, final case  Pan #, Dr's. request date.
statut T, F, B, R.
IS THIS SCREEN CORRECT?
If yes ask for first department number.     Ask for first department number.
                                              Display
Display dept. order screen (P10) (list of   dept. order screen (6.8-6.11) (list
                                              of items
item in this dept. from ITEMMAST)           in this dept. from ITEMMAST)
User entry choices                          User entry choices
System adds days in dept. from COMPANY      System adds days in dept.  (DAYVAL)
                                              to
to present date to find due out date.  Time  present date to find due out date.
                                              System
is of AM. "Noon" or PM. System              computes workload for dept/day out
                                              by
accummulates case load by product of item   product of load factor X quantity.
load factor X quantity.
ITEMMAST                                    ITEMMAST
Adds this to load already in DEPTLOAD       Add this to load already in DAYVAL
                                              for
*1248_ for date out.                               date out.