Julie E. Cohen*
68 S. Cal. L. Rev. 1091 (1995)
Copyright 1995 - HTML Version Prepared by the LII with Permission
Properly understood, copyright does not bar reverse engineering of lock-out routines and the programs they protect, nor does it bar the development of functional analogues to those lock-out routines and other interoperability-related features. As Atari demonstrates, however, in the real world that may be irrelevant. Under the doctrine of equivalents, a reverse engineered product that substantially duplicates the functions performed by a patented lock-out program infringes the patent. n310 On its face, Atari stands for the proposition that a patented lock-out program provides ironclad protection for the computer system in which it is incorporated. What Atari did not consider is whether such a patent can - or should - be valid. n311
Answering those questions in any meaningful way requires consideration of both the theoretical and institutional underpinnings of [p*1153] the patent system. Doctrinally, the federal courts have failed to develop a workable rule for distinguishing patentable from unpatentable computer programs. Institutionally, the Patent and Trademark Office ("PTO"), which functions as the initial gatekeeper and presumptive expert on questions of patentability, has proved ineffective at discharging its statutory mandate where computer programs are concerned. Together, these systemic failures undermine the presumption of validity that attaches to a duly approved patent, and on which the Atari court relied. n312 Judged against a more rigorous standard of patentability by a more demanding gatekeeper, the 10NES might well have been found undeserving of patent protection. Such a standard has been proposed, but so far has received little scholarly or judicial attention. The example of the 10NES patent makes a persuasive case for its adoption, and for changes in the organization and operation of the PTO to respond to the unique challenges posed by computer program-related inventions.
While copyright scholars and computer programmers were debating whether computer programs constituted copyrightable subject matter and whether CONTU had struck the right balance for protecting them, an equally vigorous debate focused on the question whether computer programs satisfied the requirements for patentable subject matter. n313 Over the past two decades, resistance to software patents by the PTO and the courts has all but vanished. Thousands of software patents have been issued and are routinely upheld. n314 Within the scientific, business, and academic communities, however, the debate over whether computer programs are or should be statutory subject matter continues. n315 Regrettably, there is no indication that this debate informed the parties or the court in Atari.
[p*1154]
As provided in section 101 of the Patent Act, any new and useful process or machine is potentially patentable subject matter. n316 On its face, this provision encompasses computer programs. However, longstanding judicially developed doctrines prohibit patent protection for mathematical formulae and mental processes, on the ground that "processes" that simply describe existing natural laws (whether as basic as 2 + 2 = 4 or as complex as e = mc2) or recite steps performable by the human mind do not fall within the category of "useful arts" and are not statutory subject matter. n317 Some commentators have argued that many computer programs are simply mathematical formulae or mental processes made concrete. n318 Others have argued that the "machine-like" nature of computer programs brings them squarely within the class of potentially patentable inventions. n319 A principled basis for evaluating whether computer program-related claims recite statutory subject matter has proved elusive.
1. From Freeman-Walter-Abele to Alappat: The Corruption of the "Otherwise Statutory Process or Apparatus" Standard
Although the Court of Customs and Patent Appeals ("CCPA") decided a number of cases dealing with computer program-related applications in the late 1960s and early 1970s, the history of the current approach begins with Gottschalk v. Benson. n320 In that case, the Supreme Court rejected under section 101 an attempt to patent a [p*1155] computerized process for converting binary-coded numerals to pure binary form, on the grounds that the patent covered no more than a mathematical formula, and if granted would preempt its use. n321 Six years later, in Parker v. Flook, n322 the Court rejected for similar reasons an attempt to claim patentability for a computerized method for continuously recalculating the "alarm limit" during a chemical conversion process. n323 Nonetheless, in both cases the Court stopped conspicuously short of precluding patentability for all computer programs. n324 The Benson Court indicated that it was concerned less with patentability per se than with the potentially preemptive reach of claims based on "mathematical algorithms." n325 Ultimately, the Court made good on its implicit guarantee that a sufficiently narrow patent would be upheld. In Diamond v. Diehr, n326 it held that a process for continuously monitoring the temperature inside a synthetic rubber mold, using a computer and the well-known Arrhenius equation for measuring cure time as a function of temperature and other variables, was patentable subject matter. Central to the Court's decision was that the inventor did not claim all rights to future uses of the Arrhenius equation, but only to the particular application that he had invented in the context of an "otherwise statutory" industrial process. n327
In the wake of the Supreme Court's decisions in Benson, Flook, and Diehr, the CCPA and its successor, the Federal Circuit, elaborated a two-part test for assessing the patentability of computer program-related inventions under section 101. Known as the Freeman-Walter-Abele test, after the three cases from which it emerged, n328 the test first asked whether the invention directly or indirectly recited a [p*1156] mathematical algorithm. n329 If so, it directed the court to inquire whether the claimed invention was no more than the algorithm itself, or whether instead the algorithm was applied as part of an otherwise statutory process or apparatus claim. n330 If the former, patent protection should be denied. n331 In theory, this test appeared to provide a means for limiting patent protection to advances within the "useful arts." n332 In practice, the Federal Circuit's decisions under Freeman-Walter-Abele strongly suggested that that court had little interest in recognizing any meaningful limitation on the patentability of computer program-related inventions. A trio of decisions issued in 1994 has confirmed just that. These decisions establish that the "otherwise statutory process or apparatus" requirement may be satisfied by the simple expedient of drafting claims to include a general purpose computer or standard hardware or memory element that would be necessary for any useful application of the algorithm. n333 As explained below, that course of action ignores constitutional limitations that separate patentable inventions from public domain science, and so threatens the continuing vitality of the patent system. n334
As developed by the PTO and the Federal Circuit prior to 1994, the "otherwise statutory process or apparatus" limitation of Freeman-Walter-Abele was not much of a limit at all. Nearly any physical element or step would suffice to render statutory a claim that recited a mathematical or "mental process" algorithm. That held true even if the physical element or step was well-known or an industry standard, and the mathematical algorithm was the heart of the invention. As a result, although the rule was designed to satisfy the Benson Court's concerns about algorithm preemption, it was incapable of doing so.
The case that best exemplified this problem was In re Iwahashi, n335 which concerned a system for voice pattern recognition. Broadly speaking, the claim in Iwahashi recited a novel pattern recognition algorithm and "a read only memory" ("ROM") device as elements of a claimed apparatus. n336 For the Federal Circuit, the presence of the [p*1157] ROM was sufficient to qualify the claimed invention as a statutory "machine" or "manufacture." n337 As has been elegantly demonstrated, however, the ROM limitation is meaningless because a ROM is a basic component found in all general purpose computers and is necessary whenever complex calculations involving lookup tables, such as the calculation required by the Iwahashi algorithm, are performed. n338 The real innovation claimed in Iwahashi was the new pattern recognition algorithm. By upholding the patent, the court effectively preempted its use - exactly the result it maintained was avoided by the ROM limitation. n339
Iwahashi was not an aberration. Subsequently, in Arrhythmia Research Technology v. Corazonix Corp., n340 the Federal Circuit upheld a patent for a computerized method of analyzing electrocardiograph ("ECG") patterns to detect persons at risk for certain heart diseases. The Arrhythmia patent also clearly preempts the algorithm it contains. The Federal Circuit's opinion briefly raised its application of the "otherwise statutory process or apparatus" limitation to the level of the absurd.
Although the invention claimed in Arrhythmia had a number of steps, those can be grouped into three stages. First, it converted the analog signals obtained using the ECG into digital signals. It then [p*1158] applied a well-known digital filtration technique to analyze the wave characteristics of the digitized ECG data. Finally, it performed additional mathematical analysis of the filtered ECG data to identify particular patterns and values, using a new mathematical model. n341 In holding the invention statutory, the court reasoned: "These claimed steps of "converting,' "applying,' "determining,' and "comparing' are physical process steps that transform one physical, electrical signal into another." n342 But the conversion of analog signals to digital ones is a well-known process, as is the operation of a digital computer to transform one set of digital signals into another. n343 And if digital-to-digital transformation of electrical signals constitutes a physical limitation sufficient to establish otherwise statutory subject matter, then any computer-driven algorithm is statutory subject matter anyway, and no additional "otherwise statutory process or apparatus" need be shown. More recently, the Federal Circuit has clarified that it did not mean to go that far, and that the signals must correspond to some underlying physical steps. n344 According to the court, the manipulation of ECG output counts as such a physical process. n345 Nevertheless, in Arrhythmia, as in Iwahashi, the real innovation was not any physical process or structure, but the mathematical model developed by the inventor for predicting heart failures based on ECG patterns. The discovery that certain naturally occurring ECG patterns are correlated with a likelihood of later heart failure is not something that the patent system can protect.
Finally, in In re Alappat, n346 decided by the en banc court in July 1994 after more than a year of deliberation, the Federal Circuit unambiguously held that a claim for the use of general purpose computing [p*1159] equipment to perform a mathematical operation recites a patentable apparatus - namely, "a combination of known electronic circuitry elements to produce a specific new result." n347 Accordingly, the court ruled, it need not even perform the two-part inquiry required by the Freeman-Walter-Abele line of cases. n348 The PTO had rejected Alappat's claims, which covered a process known as a "rasterizer," developed for controlling the illumination of pixels on a computer screen to minimize discontinuity and jaggedness. n349 The PTO concluded that the applicant's claims were broad "process" claims that stated a mathematical algorithm for pixel control, and would preempt it. n350 In an opinion by Judge Rich, the author of Iwahashi and the court's leading advocate of an expansive approach to computer program patentability, a majority of the Federal Circuit reversed. Citing Iwahashi, the court held that even though the claims were in process form, the application "unquestionably" claimed a machine because it referred to circuitry elements. n351 Regarding the objection that those elements - used for performing mathematical calculations - would be present in any general purpose computing system, the court observed that "a general purpose computer in effect becomes a special purpose computer once it is programmed to perform particular functions pursuant to instructions from program software." n352 In effect, then, Alappat establishes that a mathematical algorithm becomes patentable subject matter merely by virtue of its being programmed into a general purpose computer.
The Alappat court paid lip service to the concerns about preemption of the "laws of nature" and "abstract ideas" stated by the Supreme Court in Benson and Diehr, n353 but went on to reach a result that ignored those concerns entirely. The Federal Circuit's first significant post-Alappat decisions, In re Warmerdam n354 and In re Lowry, n355 [p*1160] illustrate just how facile the court's conception of the "otherwise statutory apparatus" requirement has become. The court rejected the claims in the Warmerdam patent that were drafted as process claims, on the ground that they recited no more than a mathematical algorithm, n356 but allowed the claims that were drafted to recite "a machine having a memory which contains data ... generated by" the same algorithm described in the process claims. n357 Together with Alappat, Warmerdam teaches that if otherwise unpatentable computer program-related claims are drafted in specific apparatus form, the mathematical algorithm limitation will simply disappear. n358
Lowry involved claims for a method of storing and managing data in a computer memory. The PTO had rejected the claims under the "printed matter" doctrine, on the ground that the claims merely recited the arrangement of data and did not define a new functional relationship between the data and the computer memory that served as the substrate. n359 The Federal Circuit held that "the printed matter cases have no factual relevance where "the invention as defined by the claims requires that the information be processed not by the mind but by a machine, the computer.' " n360 The court reasoned that electronically specified data hierarchies "impart a physical organization on the information stored in the memory." n361 Technically, the "printed matter" rejection was based on section 103 of the Patent Act, which requires that an invention be nonobvious, and not on section 101. n362 (As to section 101, the PTO found, and the court agreed without discussion, that a computer memory was a statutory "article of manufacture." n363) However, a section 103 "printed matter" rejection bears close affinity to a "mental steps" rejection under section 101. In both cases, the objection to patentability is that the claimed invention does not contribute to the "useful arts," but simply restates human thought [p*1161] processes. n364 In the context of computer program-related inventions, the question in both cases is whether the program steps merely mimic those processes. The Federal Circuit's decision categorically rejects any such implication. After Lowry, an algorithm for data arrangement, expressed digitally, is patentable as long as (per Warmerdam) a memory device is specified.
The Atari court did not subject the issued 10NES patent to the then-applicable Freeman-Walter-Abele test. Had it done so, the outcome is a foregone conclusion, and affords yet another illustration of just how nominal the physical limitation required under Freeman-Walter-Abele (and Alappat) need be to satisfy current standards for statutory subject matter. Each of the 10NES claims, in essence, recites a mathematical algorithm for the synchronous generation of initial numbers, followed by mathematical comparison of the results of arithmetic operations performed on those numbers. The first claim recites the following physical limitations: "a main data processor unit for executing a videographics software program" (a video game console); "a main data processor unit for storing the videographics software program" (a video game cartridge); "a first authenticating processor device"; "a second authenticating processor device"; and a reset switch. n365 The description of the inventor's preferred embodiment makes clear that the first and second "authenticating processor devices" are microprocessors, or computer chips. n366 Just as it is "difficult to imagine" any complex system for pattern recognition that does not [p*1162] incorporate a ROM, n367 it is difficult to imagine any computerized lock-and-key system for use with a video game cartridge and machine that does not incorporate microprocessors. In any event, the patent provides for that eventuality; the specification discusses adaptations of the lock-and-key system for floppy disks and magnetic cards, and the claims cover use of the invention with both digital and analog processing devices. n368
In sum, there is no question that based on the then-applicable standard for patentability, the 10NES patent reads on statutory subject matter. Under Alappat, of course, the patent claims an apparatus because it incorporates general purpose computing equipment. Alappat, Warmerdam, and Lowry signal a new era for computer program-related patents, in which anything, or virtually anything, goes. n369 It is difficult to imagine a claim that would not pass muster under the [p*1163] "nominal hardware" standard. n370 However, there is something terribly wrong with a system of decisionmaking that routinely produces exactly the result it claims to prohibit - the result that its highest court has stated would vitiate the purpose of the patent regime. Unfortunately, alternative proposed frameworks for assessing whether computer program-related inventions constitute statutory subject matter are equally unpromising.
2. Rethinking the Mathematical Algorithm Bar
One proposed solution, of course, is to abandon the mathematical algorithm bar altogether on the ground that a mathematical algorithm in digital form constitutes a process like any other, and so is statutory subject matter under the literal language of the Patent Act. Along with Judge Rich of the Federal Circuit, Professor Chisum has long been a leading advocate of this approach. n371 In effect, Arrhythmia, Alappat, Warmerdam, and Lowry reach this result, while giving only the barest nod to the Supreme Court's clear intent to preserve some limitations on patentability. Certainly, classifying all computer programs as statutory subject matter eligible for patent protection would greatly simplify the process of evaluating computer program-related patent applications. However, abandoning the rule against preemption of mathematical formulae would violate the fundamental requirement that patents be granted only for the application of scientific principles, not for their development - that is, not for equations, calculations, or formulations of natural laws. n372 The resulting ease of administration would come at too great a social cost; ultimately, it would [p*1164] frustrate the patent system's constitutionally mandated goal of inducing progress in the useful arts.
If mathematical algorithms are statutory subject matter, patents may issue for computer programs that simply execute complex calculations and are not tied to any "industrial" process. It may be argued that a computer program is not a mathematical formula at all, but an improved process for executing one. n373 In the digital age, however, a rule premised on a distinction between mathematics and process is facile. Complex physical and mathematical discoveries, such as Mandelbrot's theory of fractals or Lorenz' theory of chaos, often see their first and only expression in digital form. To perform the required calculations by hand would take decades, even centuries. The distinction between silicon and paper is too slim a reed on which to hang satisfaction of the statutory subject matter requirement. n374
The requirement that mathematical formulae be excluded from patentability, moreover, is of constitutional stature. The Constitution authorizes Congress to grant "inventors" the exclusive right to their "discoveries" in order "to promote the progress of ... the useful arts." n375 Granting exclusive rights to mathematical formulae merely because they have been expressed in digital form would have the opposite effect. That fundamental mathematical, chemical, and physical principles may be termed "discoveries" is of no moment. n376 A crucial premise of the patent system is that such principles remain in the public domain for future inventors to use. n377 Thus, the patent laws may protect novel and useful applications of those principles, but not the [p*1165] principles themselves. If a mathematical discovery expressed digitally is transmuted into a patentable process, more and more formerly unpatentable discoveries will be removed from the public domain for the statutory seventeen-year term. According patent protection to computerized mathematical algorithms in this indiscriminate fashion would vitiate the terms of the constitutional grant.
The world would be a very different place if Einstein had received a patent on his discovery that e = mc2, or if Millikan or Schrodinger had patented their pathbreaking work in electromagnetism and particle physics, or if any of them had been required to license the principles they used along the way. n378 The truly patentable inventions that these pioneering discoveries enabled might never have occurred, and certainly would have occurred more slowly, if those discoveries had not been freely available. n379 Under a regime that makes distinctions between mathematical formulae based on the mode in which they are expressed, the future of research and development in this country may strongly resemble that world. While a robust public sector might still preserve a core of public domain science, ours may not be up to that task. Even nominally public research is increasingly [p*1166] privatized; joint ventures, faculty consultancies, and technology transfer agreements abound. n380 The Balkanization of scientific knowledge into competing accretions of proprietary rights will greatly weaken existing incentives to inventors. n381
But to except only mathematical algorithms and their equivalents, such as the formal expression of a chemical reaction, from section 101's ambit is to return to the original problem: how to distinguish those computer program-related inventions that are "no more than" mathematical algorithms from those that are properly within the useful arts. The rejection of Professor Chisum's solution only underscores that dilemma.
3. An "Information Processing" Exclusion
Professor Samuelson's proposed solution to the problem of when to treat computer program-related inventions as statutory subject matter is the opposite of Professor Chisum's. She argues that both the mathematical algorithm and mental process doctrines are directed at "information processing," rather than at the "industrial arts," and accordingly proposes excluding information processing-related developments - including, it would appear, all computer programs - from the ambit of the Patent Act. n382
An "information processing" limitation on patentability is problematic for two reasons. First, some information processing developments, such as pioneering advances in data compression technology or color reproduction, are true, qualitative departures from the prior art. n383 Denying patent protection to those inventions would weaken incentives for development, and so would err too greatly on the side of underprotection, confirming the fears of Judge Rich, Professor [p*1167] Chisum, and others who maintain that restricting the patentability of computer program-related inventions will bring about the demise of the patent system as we know it. n384 Computer technology is, broadly speaking, a "useful art," n385 and innovations of appropriate stature should be rewarded, and thereby encouraged, to the same degree as innovations in other fields. Whether the balance of incentives that results from the current overinclusive approach to all computer program-related developments is a healthy one is a separate question. n386
Second, and ultimately far more important, to the extent that an "information processing" exclusion would preclude protection for allcomputer programs, it would shortly render the patent system obsolete. As the flood of software patents issued in recent years demonstrates, industrial processes of all sorts are increasingly computerized. n387 Once the commands required to execute such processes are expressed digitally, each such process has a built-in "information processing" component. If inventions are to be considered "as a whole," as Freeman-Walter-Abele and Alappat require, n388 then no patents may be issued for any of those inventions. And if inventions are not to be considered as whole, how are we to determine which computer program-related inventions are not properly considered information processing devices? The question whether a claimed invention reads on an information processing device merely restates the question whether the claims recite, in essence, no more than a mathematical algorithm, and brings us no closer to answering it. In short, Professor Samuelson's answer to Professor Chisum takes us out of the frying pan and into the fire. While under Professor Chisum's solution to the mathematical algorithm dilemma the patent system would consume the public domain, under Professor Samuelson's the public domain (or a sui generis scheme of quasi-patent protection for information processing inventions) would consume the patent system.
[p*1168]
4. Claim Construction and the Particularity Requirement
Since Iwahashi, some commentators have suggested that the debate over when computer programs constitute statutory subject matter under section 101 is better understood, and more precisely resolved, as a debate over the particularity requirement imposed by section 112 of the Patent Act. n389 Among other things, section 112 requires that the patent describe with particularity - and thereby limit - the claimed invention. n390 Adherents of the section 112 approach argue that the particularity requirement serves as a built-in safeguard against preemption of a particular formula, equation, or sequence of digital steps. n391
The security afforded against algorithm preemption by section 112 is illusory, because, as demonstrated above, there is no assurance that the limitation proposed by the inventor and accepted by the PTO will be a meaningful one, and no guarantee for applicants that rejections for lack of particularity will be made in a consistent manner. Iwahashi, Arrhythmia, Alappat, Warmerdam, Lowry, and Atari all involved very particularized, precise claims. In none was overbreadth a function of vagueness; instead, it arose from the attempt to cast items of general purpose computing equipment as meaningful physical limitations. The inescapable conclusion is that while section 112 prevents the patentee from precluding all uses of a formula claimed as part of an invention, it does not prevent it from precluding all meaningful or practicable uses. n392 Thus, the particularity requirement cannot solve the statutory subject matter problem.
The absence of a principled basis under section 101 of the Patent Act for separating statutory inventions from claimed inventions that are "no more than" nonstatutory mathematical algorithms is disturbing. However, the intense focus on statutory subject matter ignores the existence of other statutory requirements for patentability. In particular, the requirements set forth in sections 102 and 103 of the [p*1169] Patent Act, that a claimed invention be novel and nonobvious to those ordinarily skilled in the field, n393 may be used to accomplish what the statutory subject matter inquiry cannot achieve: a rule that permits analytic dissection of claims into statutory and nonstatutory elements for purposes of identifying which computer program-related inventions are patentable.
Intuitively, the most troubling aspect of many computer program-related patents is that they appear to reward the inventor for recognizing the obvious - that a given function may be performed more efficiently or more accurately if computerized - and using general purpose computer equipment and standard programming techniques to computerize it. Other computer program-related patents simply reward the programmer for developing otherwise unpatentable mathematical formulas. In that sense, the objections to the Iwahashi, Arrhythmia, and Alappat patents, like the objection to the Lowry patent, are really based on obviousness and lack of novelty. In each case, the new and nonobvious element was not the physical apparatus, which was well-known and widely used, but rather the otherwise unpatentable mathematical algorithm developed to analyze the data and compute the desired output. n394
In response to Iwahashi, Richard Stern, former chief of the Department of Justice's Intellectual Property Section, proposed reconceiving the standard for nonobviousness for computer program-related inventions. His solution, which may be termed the "innovative programmer" standard, adds a third step to the Freeman-Walter-Abele test. If a claimed invention recites a mathematical algorithm, but appears to be statutory subject matter when taken as a whole, the examiner must ask whether the claimed invention would have been obvious to "a person of ordinary skill ... who: (a) knew the particular algorithm; (b) desired to accomplish the function or task to be performed; and (c) desired to do so with the aid of a computer ...." n395 If not, it is nonobvious, and so patentable. By taking general purpose computer equipment and the mathematical algorithm as part of the prior art for purposes of assessing nonobviousness, the innovative programmer standard is intended to avoid the pitfalls described in Part IV.A, above.
[p*1170]
The innovative programmer standard is similar to an approach known as the "point of novelty" test, which has been disfavored by the courts. That treatment was ill-considered. The point of novelty test offered a logically and doctrinally viable method for assessing the patentability of computer program-related inventions. The innovative programmer standard improves upon it, by tailoring the test to the unique nature of computer software.
The point of novelty approach first surfaced in the early days of computer programming. As set forth in In re Abrams, n396 it involved analytic dissection of the claimed invention to determine whether its novelty resided in its physical elements or in "one or more of the so-called mental steps." n397 However, the court then confused the section 102 requirement of novelty with the section 101 requirement of statutory subject matter. It held that if the novelty of the claimed invention resided in "mental steps," the claimed invention was not statutory subject matter. n398 In Application of Musgrave, n399 Judge Rich rejected the Abrams approach as based on a logical fallacy. He reasoned:
If [Abrams] were the law, a given process ... could be statutory during the infancy of the field of technology to which it pertained, when the physical steps were new, and non-statutory at some later time after the physical steps became old, acquiring prior art status, which would be an absurd result. Logically, the identical process cannot be first within and later without the categories of statutory subject matter, depending on such extraneous factors. n400
The Musgrave court concluded that because an invention cannot be considered statutory subject matter only at certain points in time, each claimed invention must be assessed as a whole for purposes of the inquiry required by section 101. n401
Eight years after its rejection by the CCPA, the Abrams point of novelty analysis resurfaced in the Supreme Court's opinion in Parker v. Flook. n402 Without citing either Abrams or Musgrave, the Court observed that "the proper analysis for this case" was that "the process itself, not merely the mathematical algorithm, must be new and useful. [p*1171] Indeed the novelty of the mathematical algorithm is not a determining factor at all .... It is treated as though it were a familiar part of the prior art." n403 The Court then went on to duplicate the fallacy of Abrams, however. It ruled that if the novelty of the claimed invention resided in the mathematical algorithm, then the claimed process could not constitute statutory subject matter under section 101. n404 In fact, properly stated, the ill-fated Abrams/Flook test for patentable subject matter turns entirely upon a novelty and nonobviousness analysis. A slight reformulation of the test avoids Judge Rich's objection: A claimed invention that, taken as a whole, is (or may be) statutory subject matter is nonetheless unpatentable if its novelty and nonobviousness inheres in its nonstatutory elements. This reformulated test is similar to the innovative programmer standard proposed by Stern in that both would require the examiner to separate the mathematical algorithm from the other elements of the claimed invention when conducting the separate inquiry into novelty and nonobviousness. n405
In Diamond v. Diehr, n406 the Court recognized its error of logic. Adopting the CCPA rule, it held unambiguously that " "novelty' ... is of no relevance in determining whether the subject matter of a claim falls within the 101 categories of possibly patentable subject matter," and that the claim "must be considered as a whole" for that determination. n407 After Diehr, then, it is undisputed that a claimed invention may not be dissected into its component parts for purposes of the statutory subject matter inquiry. However, the Court also rejected the Flook point of novelty analysis on independent grounds. It opined that a rule requiring that mathematical algorithms be considered part of the prior art "would, if carried to its extreme, make all inventions unpatentable because all inventions can be reduced to underlying principles of nature which, once known, make their implementation [p*1172] obvious." n408 Although this analysis has received some critical support, n409 it is in its own way as ill-considered as Flook's confusion regarding section 101.
As Professor Burk has noted, the Diehr Court's reasoning has its roots in Justice Frankfurter's concurrence in Funk Bros. Seed Co. v. Kalo Inoculant Co. n410 The Funk Bros. Court disallowed a patent for a mixed culture of naturally occurring bacteria on the ground that it claimed a preexisting phenomenon of nature n411; Justice Frankfurter objected that a " "laws of nature' " rationale "could fairly be employed to challenge almost every patent." n412 The Diehr Court was also influenced by its recent decision in Diamond v. Chakrabarty, n413 which established the patentability of laboratory-grown bacteria. Together, Chakrabarty and Diehr might be read to establish that the touchstone for patentability must be the simple test of whether the claimed invention is "manmade," and that the invention must be evaluated as a whole for all purposes. In fact, there is no logical reason that analytic dissection may not be employed during the inquiry into novelty and nonobviousness required under sections 102 and 103 of the Patent Act. To the contrary, such an analytic approach is a necessary part of the process of evaluating any claimed invention against existing prior art. The real question, dismissed in Diehr with a reference to "extremes," is whether the patent laws forbid examiners and courts from dissecting out newly discovered scientific and mathematical principles. Because, as I have discussed, the patent laws cannot reward new and nonobvious advances in mathematics, the answer must be yes. n414
Given the difficulties that attend any effort to separate unpatentable principles from their patentable applications, the Court's "slippery slope" argument is unpersuasive. As the "otherwise statutory process or apparatus" rule developed in Freeman-Walter-Abele illustrates, n415 a rule requiring isolation and exclusion of unpatentable elements is not the only approach to patentability that poses a danger of [p*1173] overbreadth. When carried to its extreme, as it has been by the Federal Circuit, the "otherwise statutory" rule can be applied to render any process that contains computer components patentable. Arguably, the danger of reduction to "phenomena of nature" is particularly acute in the case of biological inventions, which may too frequently be viewed as black boxes that produce results based on their (naturally determined) biochemical properties, even when those results did not formerly exist in nature. n416 Even so, however, generalizing that danger to all inventions is unwarranted. Although it is certainly possible to analyze computer program-related inventions in a way that reduces them to bedrock principles of conductivity and electromagnetism, computer programs as a class, unlike bacteria or proteins, do not exist in nature in any form. Rather, they are a wholly human-made class of articles that employ mathematical principles to accomplish results. Thus, although in the abstract the boundary between mathematical principle and application defies precise articulation, n417 it should be possible to separate the two in particular cases.
Ironically, Diehr is a case in point. The mathematical equation used in the rubber-curing process at issue in Diehr was the well-known Arrhenius equation. n418 The fact that the mathematical formula was "a familiar part of the prior art" n419 did not preclude the Court from finding the claimed application of the formula to be new and nonobvious, even as it rejected that test. Of course, isolating the mathematical formula component of a claim is not always so easy; if it were, the controversy over computer program-related patents would not exist. The advantage of a point of novelty approach over the "otherwise statutory process or apparatus" formulation developed in Freeman-Walter-Abele is that, unfettered by section 101's requirement that the claimed invention be taken as a whole, it allows courts and examiners to dissect principles and mathematical proofs, which are not patentable, from their applications, which may be, and determine where the inventive act lies and whether it is the kind of inventive act the patent system can reward.
[p*1174]
The innovative programmer standard differs from the reformulated Abrams/Flook test, however, in that it adds general purpose computing equipment to the list of givens. n420 As the analysis in Part IV.A demonstrates, that recommendation is sound and its adoption long overdue. A mathematical principle expressed digitally, though not patentable, may be new; general purpose computing equipment is not even that. We are fast approaching an era in which any industrial function can be directed by a general purpose computer with the appropriate software. As with any other useful art, the patent laws should reward only genuinely new and nonobvious advances in the application of computer technology, not the comparatively mundane, though complex, process of adapting a general purpose computer to a particular use with existing programming techniques.
In light of Diehr and the Federal Circuit's aggressive stance on software patentability, it is perhaps not surprising that the innovative programmer standard has drawn little response. For those who believe that any mathematical algorithm expressed digitally is a patentable process - or, with a "nominal hardware" or memory limitation, a patentable apparatus or article of manufacture - a standard designed to narrow the rule and close the floodgates will not constitute much of an improvement. However, the innovative programmer standard should satisfy those who concede that computerized mathematical formulae are unpatentable, but worry that there is no reliable means of identifying which claimed inventions fall into that category. The standard offers a doctrinally sound basis for dissecting out those elements of a claimed invention that alone cannot properly be considered patentable, either because they are public domain building blocks available to all, or because the idea of using them would be obvious to anyone conversant with computers. Decisions about whether the invention is sufficiently novel and nonobvious to qualify for patentability can then be made based on what the dissection reveals.
Under the innovative programmer standard, the patents at issue in Iwahashi, Arrhythmia, Alappat, Warmerdam, Lowry, and Atari might never have been granted. Atari is illustrative. Take away the two microprocessors and the reset switch that constitute the "nominal hardware" and all that remains is a specification for a series of mathematical functions. Take those functions as part of the prior art and all that remains is the relatively mundane task of assembling the physical [p*1175] building blocks of the lock-and-key system. The question the examiner should have asked was whether the combination of those physical elements would have been obvious to any ordinarily skilled programmer who wished to devise an effective lock-out device. n421 Because I am not an ordinarily skilled programmer, I do not attempt to answer that question. Instead, the next section examines the remaining factors that would bear on whether the examiner reached the correct answer.
Developing a workable and doctrinally sound approach to assessing the patentability of computer program-related inventions is only half the battle. Because decisions on patentability are inherently fact-specific, the technical competence of the decisionmaker - the PTO - is critical in determining whether the rules governing patentability will be applied correctly. Technical competence alone cannot cure the lack of a coherent standard, but it can significantly narrow the margin of error and reduce the number of erroneously granted patents. In the years since Diehr opened the floodgates to computer program-related claims, the PTO has failed in that respect because it lacked both the personnel and the resources to evaluate such claims for technical merit and to make informed judgments regarding nonobviousness. In response to the howls of criticism that followed the issuance of a patent covering a standard industry method of multimedia data retrieval, n422 the PTO has set in motion several organizational and procedural changes designed to address its shortcomings. Those changes, though important, do not go far enough.
[p*1176]
1. The Decisionmakers
In theory, the patent examination process consists of a rigorous analysis of the claimed invention by an examiner trained in the same scientific field. Until 1994, however, the PTO's list of accepted technical backgrounds for patent examiners did not include computer science, and the PTO did not hire computer scientists as examiners or technical staff members. n423 It is impossible to overstate the consequences of that policy, or lack of it, for the overprotection of computer program-related developments. A duly examined and approved patent is presumed valid. n424 To overcome this presumption in subsequent infringement litigation, a challenger must present clear and convincing evidence of some error by the patent examiner. n425 This standard is a high one, often prohibitively so. Because patent examination requires technical competence that most judges do not possess, judges are reluctant to second-guess an examiner's decision. As a matter of logic, however, the presumption of validity is justified only if the examiner's qualifications warrant it.
Computer program-related applications are particularly difficult to evaluate because they may require knowledge of two different fields - computer science and some other field, such as chemistry or medicine, in which the invention is intended for use. For example, an application for an improved method of processing and interpreting seismic waves to detect the presence of fossil fuel deposits would be assigned to Class 367, for "Communications, Electrical: Acoustic Wave Systems and Devices," while the application for an improved method of analyzing electrocardiograph readings to predict and treat heart disease, at issue in Arrhythmia, was assigned to Class 128, for "Surgery." n426 What both inventions have in common, however, is that they are based in part on computer applications - applications that neither the examiner trained as a geologist nor the examiner trained in the biological sciences is equipped to evaluate. Moreover, the prospecting application might rely in part on programming techniques originally developed in a different industrial context altogether, such [p*1177] as chemistry or metallurgy. Yet the examiner trained in geology assigned to examine the prospecting patent would have no particular knowledge of patents recently granted in those fields.
In 1994 the PTO revised its hiring criteria to include computer science among the accepted technical backgrounds and, in mid-1994, hired nine individuals with computer science degrees to assist the examiner corps. n427 Although this is an important step in the right direction, the PTO will not be fully equipped to handle computer program-related claims until computer scientists have been accorded examiner status and can veto applications that are insufficiently novel or nonobvious to merit patent protection. Where appropriate, each such application should be assigned two examiners, one to evaluate the claimed invention with respect to the intended field of use and one to evaluate it with respect to the state of the computer science-related prior art; the approval of both should be required before a patent may issue. As to all computer program-related patents issued before the addition of the new computer science personnel to the PTO's technical staff, courts deciding issues of patent validity should discount the presumption of validity, and give greater weight to expert testimony offered by the parties.
2. Prior Art
The determination whether a claimed invention is novel and nonobvious is based on examination of relevant prior art in the applicable technical field. The applicant must cite prior art that bears on the claimed invention as part of the patent application, and the examiner also must look beyond what is cited before granting the patent. n428 The presumption of validity that applies to an issued patent is based not only on the examiner's technical qualifications, but also on the examiner's search of the prior art. Once again, that presumption is only as good as the resources that support it.
Within the last decade, many glaring defects relating to the organization and accessibility of the PTO's collection of prior art have been remedied. The PTO's database of issued patents has been automated, and examiners have access to commercially available on-line [p*1178] databases as well. n429 However, in the field of computers and computer programs, much that qualifies as prior art lies outside the areas in which the PTO traditionally has looked - previously issued patents and previous scholarly publications. Many new developments in computer programming are not documented in scholarly publications at all. Some are simply incorporated into products and placed on the market; others are discussed only in textbooks or user manuals that are not available to examiners on line. n430 In an area that relies so heavily on published, "official" prior art, a rejection based on "common industry knowledge" that does not appear in the scholarly literature is unlikely. n431 Particularly where the examiner lacks a computer science background, highly relevant prior art may simply be missed. In the case of the multimedia data retrieval patent granted to Compton's New Media, industry criticism prompted the PTO to reexamine the patent and ultimately to reject it because it did not represent a novel and nonobvious advance over existing technology. n432 However, it would be inefficient, and probably impracticable, to reexamine every computer program-related patent, and the PTO is unlikely to do so.
Even when an examiner overlooks relevant prior art, the patent remains clothed in the presumption of validity. An accused infringer may offer the missed prior art to show invalidity, but the court must find the evidence of invalidity clear and convincing. n433 In Atari, that hurdle proved too high. A missed piece of prior art relating to a computer reset switch was a central element of Atari's defense against Nintendo's patent infringement claims. n434 Judge Smith declined to grant summary judgment in favor of Atari on the invalidity issue. After examining the reset reference, the district judge ruled that she could not conclude as a matter of law that Nintendo's application of [p*1179] the reset feature would have been obvious to a programmer of ordinary skill. n435 Subsequently, the jury also rejected Atari's invalidity argument. Absent the presumption of validity, the outcome could well have been different at either procedural stage.
Before the PTO can evaluate computer program-related applications competently, its database of relevant prior art must be made complete. Thus far, however, the PTO has taken no steps to do this. In response to the PTO's inaction, the Software Patent Institute in Ann Arbor, Michigan has begun assembling "a database of computer science folklore - techniques that are in use, but not widely published." n436 As conceived, the database also will include computer science textbooks and computer program-related magazine items. n437 If realized, such a database might constitute a significant step toward eliminating the prior art problem, but only if the PTO uses it on a regular basis. A more reliable alternative would be to place a database of computer program-related prior art within the PTO's control and require that it be used. Much of the raw material for such a database already is present in the archives of the Copyright Office and the Library of Congress. Every computer science textbook and computer system manual in which copyright is claimed must be deposited with the Copyright Office as part of the copyright registration process, and textbooks published with notice of copyright are deposited with the Library of Congress. n438 Under regulations adopted pursuant to the 1976 amendments to the Copyright Act, every computer program that is the subject of a copyright registration also is deposited, at least in partial form, with the Copyright Office, and every computer program "published" with notice of copyright is deposited with the Library of Congress. n439 It would be a relatively simple matter to amend the copyright and patent statutes to require that a separate copy of every computer program, every piece of supporting documentation, and every computer-related textbook or magazine deposited with the Copyright Office or the Library of Congress also be filed with the PTO. Accessing the knowledge contained in this prior art then would be a matter of hiring individuals with the requisite specialized knowledge to index and maintain the collection.
[p*1180]
3. Shifting the Burden of Production
In 1994, after the Compton's New Media debacle, the PTO supported an additional change in its examination procedures that reflected a newfound lack of confidence in its ability to examine computer program-related patent applications. The Patent Term and Publication Reform Act, submitted to Congress but not enacted in 1994, would have required that patent applications be disclosed to the public eighteen months after the patent application is filed. n440 Interested parties could then bring relevant prior art not contained in the PTO's database to the examiner's attention. A bill to require publication after eighteen months has been reintroduced in the 104th Congress, n441 as well as another bill that would afford third parties increased opportunities for participation in patent reexamination proceedings. n442 Both amendments would, in effect, shift the burden back to the computer industry, the PTO's most vocal critic, to police itself.
Ironically, the strongest opposition to publication at eighteen months may come from inventors themselves, who risk losing trade secret protection for their products if their patent applications are denied. n443 It might be argued that the disclosure provisions would deter inventors from seeking patent protection, with the result that many valuable, patentable inventions would be kept from the public. This is a valid argument against the amendment only if the applicants turned away are worthy, however, and there is no empirical evidence to suggest that this would be the case. Absent such evidence, it is at least as [p*1181] likely that the current system of nondisclosure encourages applications of dubious merit, submitted on the off chance that a patent will issue. If a patent is denied, the innovation can still be held as a trade secret. Under a system that subjects each application to public scrutiny, the applicants who weed themselves out may well be the undeserving ones. Weakening the incentives to seek patent protection for computer programs thus may yield results that are entirely consistent with the purposes of the Patent Act and the criteria for patentability. The proposed amendment regarding reexamination, a process that cannot be triggered until a patent is issued, will not produce a like effect.
However, neither publication of patent applications nor increased opportunities for reexamination will reduce the need for a qualified, well-informed examiner corps. Even under a public comment regime, the examiner assigned to each computer program-related application must be able to digest and evaluate the significance of materials submitted by interested members of the public, and must make the final decision as to patentability. The PTO cannot and should not cede that critical function to the public, particularly where, as a practical matter, the "public" will be made up largely of opponents of software patents and the prospective patentee's competitors. Accordingly, enactment of a provision for publication of patent applications should not foreclose continued self-examination by the PTO.