Is Patent Law Technology Specific?

Although patent law purports to cover all manner of technologies, we have noticed recent divergence in the standards applied to biotechnology and to software patents: the Federal Circuit has applied a very permissive standard of obviousness in biotechnology, but a highly restrictive disclosure requirement. The opposite holds true for software patents, which seems to us exactly contrary to sound policy for either industry. These patent standards are grounded in the legal fiction of the"person having ordinary skill in the art"or PHOSITA. We discuss the appropriateness of the PHOSITA standard, concluding that it properly lends flexibility to the patent system. We then discuss the difficulty of applying this standard in different industries, offering suggestions as to how it might be modified to avoid the problems seen in biotechnology and software patents.

See State Street, 149 F.3d at 1375. Indeed, on remand in that case the district court held the patent invalid under section 102. AT&T Corp. v. Excel Communications, 52 U.S.P.Q.2d 1865 (D. Del. 1999). 23 35 U.S.C. § 112 ¶1 (1994). 24 One classic justification for having a patent system is to encourage inventors to disclose their ideas to the public, who will benefit from this new knowledge once the patent expires. Kewanee Oil Corp. v. Bicron Corp., 416 U.S. 470, 489 (1974) (referring to the "federal interest in disclosure" embodied in the patent laws); see also Edith Tilton Penrose, The Economics of the International Patent System 31-34 (1951). 25 Without the disclosure obligation, patentees could conceivably keep the workings of their inventions secret, relying on that secrecy to extend protection even after the patent has expired. Cf. Pitney-Bowes v. Mestre, 701 F.2d 1365, 1372 n.12 (11 th Cir. 1983) (discussing the policy concerns here). 26 For a detailed discussion of how the law allocates rights between initial inventors and improvers, see, e.g., Mark A. Lemley, The Economics of Improvement in Intellectual Property Law, 75 Tex. L. Rev. 989 For software patents, however, a series of recent Federal Circuit decisions has all but eliminated the enablement and best mode requirements. In recent years, the Federal Circuit has held that software patentees need not disclose source or object code, flowcharts, or detailed descriptions of the patented program. Rather, the court has found high-level functional description sufficient to satisfy both the enablement and best mode doctrines. 27 For example, in Northern Telecom, Inc. v. Datapoint Corp., 28 the patent claimed an improved method of entering, verifying, and storing (or batching") data with a special data entry terminal. The district court invalidated certain claims of the patent on the grounds that they were inadequately disclosed under section 112. The Federal Circuit reversed. It held that when claims pertain to a computer program that implements a claimed device or method, the enablement requirement varies according to the nature of the claimed invention as well as the role and complexity of the computer program needed to implement it. Under the facts in this case, the core of the claimed invention was the combination of components or steps, rather than the details of the program the applicant actually used. The court noted expert testimony that various programs could be used to implement the invention, and that it would be "relatively straightforward [in light of the specification] for a skilled computer programmer to design a program to carry out the claimed invention." 29 The court continued: The computer language is not a conjuration of some black art, it is simply a highly structured language * * * * [T]he conversion of a complete thought (as expressed in English and mathematics, i.e. the known input, the desired output, the mathematical expressions needed and the methods of using those expressions) into a language a machine understands is necessarily a mere clerical function to a skilled programmer. 30 And in Fonar v. General Electric, 31 involving the best mode requirement, the Court explained: As a general rule, where software constitutes part of a best mode of carrying out an invention, description of such a best mode is satisfied by a disclosure of the functions of the software. This is because, normally, writing code for such software is within the skill of the art, not requiring undue experimentation, once its functions have been disclosed. It is well established that what is within the skill of the art need not be disclosed to satisfy the best mode requirement as long as that mode is described. Stating the functions of the best mode software satisfies that description test. We have so held previously and we so hold today. Thus, flow charts or source code listings are not a requirement for adequately disclosing the functions of software. 32 Indeed, the Federal Circuit has gone so far as to hold that patentees can satisfy the written description and best mode requirements for inventions implemented in software even though they do not use the terms "computer" or "software" anywhere in the specification! 33 To be sure, in these latter cases it would probably be obvious to one 29 Id. at 941-42. 30 Id. 31 107 F.3d 1543 (Fed. Cir. 1997). 32 Id. at 1549 (citations omitted). 33 Robotic Vision Sys., Inc. v. View Eng'g, Inc., 112 F.3d 1163 (Fed. Cir. 1997) (best mode); In re Dossel, 115 F.3d 942 (Fed. Cir. 1997) (written description).
In White Consolidated Industries, Inc. v. Vega Servo-Control, Inc., 713 F.2d 788 (Fed. Cir. 1983), by contrast, the Federal Circuit had invalidated a patent for a machine tool control system which was run by a computer program. Part of the invention was a programming language translator designed to convert an skilled in the art that the particular feature in question should be implemented in software. Still, it is remarkable that the Federal Circuit is willing to find the enablement requirement satisfied by a patent specification that provides no guidance whatsoever on how the software should be written. 34 It is simply unrealistic to think that one of ordinary skill in the programming field can necessarily reconstruct a computer program given no more than the purpose the program is to perform. Programming is a highly technical and difficult art. But the Federal Circuit's peculiar direction in the software enablement cases has effectively nullified the disclosure obligation in those cases. 35 And since source code is normally kept secret, software patentees generally do not disclose much if any detail about their input program into machine language, which the system could then execute. The patent specification identified an example of a translator program, the so-called "SPLIT" program, which was a trade secret of the plaintiff. The court held that the program translator was an integral part of the invention, and that mere identification of it was not sufficient to discharge the applicant's duty under section 112. The court seemed concerned that maintaining the translator program as a trade secret would allow White to extend the patent beyond the 17-year term then specified in the patent code.
While White suggests that it is not sufficient merely to identify the program or its functions, more recent Federal Circuit authority is overwhelmingly to the contrary. See, e.g., Dossel, supra, at 946 ("While the written description does not disclose exactly what mathematical algorithm will be used to compute the end result, it does state that "known algorithms" can be used to solve standard equations which are known in the art."; finding this sufficient to describe the invention). 34 One recent decision even found that a specification that provided inconsistent guidance as to how the invention worked was not indefinite. See S3 Corp. v. Nvidia Corp., __ F.3d __ (Fed. Cir. Aug. 3, 2001); compare id. at __ (__, J., dissenting). 35 A recent development in Federal Circuit jurisprudence may suggest another source for a robust disclosure obligation, however. The court has recently reinvigorated the written description requirement in section 112 ¶1, not only in biotechnology cases, e.g., Regents Cir. 1996). We argue below that a broad reading of the written description requirement is largely unique to biotechnology cases. If we are wrong, however, and cases like Lilly represent a general rule, it could mean that most software patents will be held invalid for failure to describe the invention in any detail. But see In re Dossel, 115 F.3d 942, 946 (Fed. Cir. 1997) (rejecting written description argument in a software case, albeit before the Federal Circuit's more recent cases on the issue). programs in the patent. Software patentees during the 1980s and early 1990s tended to write their patents in means-plus-function format 36 in order to satisfy the changing dictates of the Federal Circuit's patentable subject matter rules. 37 Lawyers writing patents in such a format have an incentive to describe their invention in the specification in as general terms as possible, since means-plus-function claim elements will be limited to the actual structure disclosed in the specification and equivalents thereof. 38 635 (1992); D.C. Toedt III, Patents for Inventions Utilizing Computer Software: Some Practical Pointers, 9 Computer Law. 12 (1992) (suggesting disclosure of "pseudocode," i.e., generalized code not written in a particular programming language, to satisfy section 112; and discussing pros and cons of disclosing actual source code). For a policy argument in favor of greater disclosure, see Thomas P. Burke, Note, Software Patent Protection: Debugging the Current System, 69 Notre Dame L. Rev. 1115Rev. , 1158Rev. -1160Rev. (1994: A software patent without source code is like a law review piece filled with case names but missing citations to case reporters. A person of ordinary skill in legal research might be able to track down the full-text of all the opinions. Marbury v. Madison would be found quicker than a state trial court opinion. But, would anyone think that such a practice was enabling or the best mode? As it is now, the disclosure requirements can be met using such devices as specifications, flowcharts, and pseudo-code. Professor Randall Davis of MIT summed it up at the National Research Counsel in 1990: There is almost no way to visualize software. Sure, we have flow charts, we have data-flow diagrams, we have control flow diagrams, and everybody knows how basically useless those are. Flow charts are documentation you write afterward --because management requires them, not because they are a useful tool. A patent is most similar to a real property deed specifying the metes and bounds for a parcel of land. Both documents are not easily understood but succeed if they secure the owners' interests in the specified claims. If the goal is to inform the world of an invention, software professionals have avenues more timely and less expensive than pursuing a patent application. In fostering the trade-off between the interests of inventors and the public, the source code is the best way to explain an algorithm.
The court's reasoning in the enablement and best mode cases has another implication as well. Because the court views actually writing and debugging a program as a "mere clerical function" "within the skill of the art," it follows that the court is unlikely to consider the work of programming itself to be sufficiently innovative to meet the nonobviousness threshold of section 103. After all, the same tests for adequacy of disclosure -would one of ordinary skill in the art be able to make the patented invention without undue experimentation -are also central to the obviousness inquiry. 40 While only a limited number of appellate decisions discuss obviousness in the context of software patents, there is some reason to believe the court is imposing a rather strict standard. The first case involving the obviousness of a software-implemented invention is, perhaps surprisingly, a Supreme Court case from the 1970s. In Dann v.
Johnston, 41 the Court held a patent on a "machine system for automatic record-keeping of bank checks and deposits" invalid for obviousness. The Court took a rather broad view of obviousness in the computer industry, focusing on whether analogous systems to the patentee's had been implemented in computer before, rather than analyzing the precise differences between the patentee's program and the prior art programs. The clear implication of the opinion is that if a reasonably skilled programmer could produce a Under this proposal, a computer system's complete source code would not have to be appendixed to the patent. The applicant would only have to include the source code directly relevant to enabling the claim language. In cases where claims are broadly written (as in a means-plus-function apparatus claim that covers the automation of an entire industry), a nearly complete program listing would be required. 40  program analogous to the patented one, and if there was motivation in the prior art to do so, the patented program is obvious.
The Federal Circuit has found software patents invalid for obviousness in two recent cases, Lockwood v. American Airlines 42 and Amazon.com v. Barnes & Noble. 43 Neither case opined directly on the ease with which computer programs could be produced, but both read obviousness as a rather substantial hurdle to patentability of software. 44 In Lockwood, the question was whether the defendant's own system made the patented claims obvious. The system had been in public use, but American Airlines had kept the workings of the system secret. Nonetheless, because Lockwood's patent was claimed in broad functional terms, the court found that similarly broad functional disclosures in the prior art were sufficient to render the patent obvious. While Lockwood argued that the information provided wasn't sufficient for one skilled in the art to make and use the system, the court pointed out that it was as detailed as the information Lockwood's own patent provided. 45 Thus, the patent's meager disclosure of technical details indirectly contributed to the court's finding of obviousness. In Amazon.com, the court found Amazon's "one-click" shopping feature to be obvious in view of certain references describing the desirability or feasibility of such a system in general terms, and one prior system that delivered data in response to a mouse click. The court rejected 42 44 In In re Zurko, 111 F.3d 887 (Fed. Cir. 1997), the Federal Circuit held that a patented software invention was nonobvious even though each of the elements of the invention could be found in the prior art, where the prior art did not identify the problem to be solved. While Zurko certainly demonstrates that some software patents will be held nonobvious, it is a specific holding of rather limited utility to most software patentees.
arguments that the one-click feature was technically difficult to implement, relying on the fact that the prior art generally described such a system as both desirable and feasible.
The court also gave surprisingly short shrift to Amazon's evidence of secondary considerations of nonobviousness. 46 The likely result of the Federal Circuit's focus on high-level functionality is that improvements in programming techniques will be found non-obvious in view of prior art that solved the same basic problem in a somewhat different way. This was arguably the result in both Dann and Lockwood, and it seems to follow from the court's view in the section 112 cases that programmers are an extremely skilled bunch. While disclosure is a minimal hurdle for software patents, then, obviousness can be a rather tougher one. 47 46 Amazon.com v. Barnesandnoble.com, 239 F.3d 1343, 1366 (Fed. Cir. 2001). To be sure, the court may have treated Amazon's patent more harshly because the case arose on appeal from a preliminary injunction. The court suggested that preliminary injunctions should not be granted if there were any serious question as to the validity of the patent. Id. at 1350-51. Whether it would apply as strict a test of obviousness after trial is not clear. 47 This assumes that courts have access to the prior art necessary to make a realistic obviousness determination. A number of commentators have expressed concern about the difficulty Examiners have in finding software prior art. As Julie Cohen explains: [I]n the field of computers and computer programs, much that qualifies as prior art lies outside the areas in which the PTO has traditionally 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. 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. 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, 47 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. However, it would be inefficient, and probably impracticable, to reexamine every computer program-related patent, and the PTO is unlikely to do so. Julie E. Cohen Most of this criticism has been directed at the failure of the PTO to find (and patent applicants to cite) the relevant prior art. But parties in litigation have more time and money to spend, and are much more Patent scope is necessarily interrelated with obviousness and enablement. 48 The breadth of patent protection is in part a function of how different the invention is from the prior art. Further, patent claims are invalid if they are not fully described and enabled by the patent specification, so the permissible breadth of a patent will be determined by how much information the court determines must be disclosed to enable one of ordinary skill in the art to make and use the patented invention. The scope of the doctrine of equivalents is also a function of obviousness and enablement, since a patentee is not permitted to capture ground under the doctrine of equivalents that it would not have been permitted to claim in the first place. 49 The Federal Circuit's treatment of software validity issues suggests that while the court will find relatively few software patents nonobvious, those that it does approve will be entitled to broad protection. The evidence on software patent claim scope so far is mixed, though there is some evidence tending to support this hypothesis. Most notably, in Interactive Gift Express v. Compuserve, 50 the patentee had designed a kiosk system for printing copyrighted works on demand. The Federal Circuit held that the claims of the patent should be read broadly, to cover any form of online downloading in response to a remote request. 51 In doing so, it reversed the district court's construction of five separate claim elements. As construed by the Federal Circuit, the patent is breathtaking in its likely to find the best prior art than the PTO. The likely result is that while numerous software patents will issue, a large number of those actually litigated will be found obvious. 48 See Chisum, supra note __. 49  scope, and most electronic commerce sites that permit downloading of digital information are likely within its ambit.
The court's treatment of software patent scope under the doctrine of equivalents has been more mixed. Many of these decisions have rejected application of the doctrine of equivalents to read claim language written for one product generation at such a high level of abstraction that it covers accused products from a different generation. Thus, in Alpex Computer Corp. v. Nintendo Co., 52 the Federal Circuit held that a patent claim to a video game output display system was not infringed by a next-generation system that worked in a different way. Alpex's claimed system included a display RAM that stored information corresponding to each pixel of a television screen in a discrete location.
Nintendo's accused device, by contrast, used shift registers to store one "slice" of the video display at any given time. The Federal Circuit rejected a jury finding that the two systems were equivalent. 53 In Digital Biometrics, Inc. v. Identix, Inc., 54 the court construed narrowly a patent claim to "image arrays" storing a two-dimensional slice of video data, and which were merged into a "composite array" storing a fingerprint image.
The court held that the defendant's systems, which constructed the composite array directly rather than by using two-dimensional slices, did not create "image arrays" within the meaning of the claims. Most 1996). In that case, the Federal Circuit upheld the district court's finding of noninfringement under the doctrine of equivalents, because there were substantial differences between the patent's requirement that a computer program "call on" columns of data one byte at a time and the defendant's product, in which the columns alleged to be equivalent were not in the data matrix, and therefore were not called upon to read data. The court rejected the "conclusory" declaration of plaintiff's expert that the two processes were identical.
Online, 55 the court affirmed a district court decision granting summary judgment of noninfringement under the doctrine of equivalents. The patent claims in that case covered "frames," defined in the specification as pages encoded in character-based protocols. The court rejected Wang's attempt to extend the patent to cover bit-mapped pages, crediting evidence that there were "huge, huge differences" between the two approaches. 56 Other cases have applied the doctrine of equivalents more broadly. In some of those cases, the Federal Circuit has found equivalence between two different types of software programs written in different product generations. More troubling, some cases suggest that software implementations of certain ideas are equivalent to older mechanical implementations. An example is Overhead Door Corp. v. Chamberlain Group, Inc. 57 The patented system claimed a (mechanical) switch connected to a microprocessor, which could store the codes of multiple garage doors. The Federal Circuit held that the claim was not literally infringed by an electronic switch implemented in software. However, the court reversed a grant of summary judgment to the defendants under the doctrine of equivalents, concluding that a reasonable jury could find that the difference between mechanical and software implementations was a mere "design choice." WMS Gaming, Inc. v. International Game Technology 58 is also instructive. In that case, the court held that a claim written in means-plus-function language that relied for its corresponding 55 197 F.3d 1377 (Fed. Cir. 1999). 56 Id. at 1386. In a related context (interpreting equivalent structure in a means-plus-function claim), the court held that Nintendo's video game systems did not infringe GE's television switch patents because the patents, written in means-plus-function format, did not disclose a function for the switches identical to Nintendo's function. structure on a computer programmed with a particular algorithm was limited in literal scope to the particular algorithm chosen and equivalents thereof. However, the court found the defendant's algorithm infringing under the doctrine of equivalents. This latter approach has the potential to expand the scope of patents in the software industry dramatically. 59 Software patents, then, are likely to face serious obviousness hurdles. The few patents that overcome those hurdles need disclose virtually nothing about the detailed workings of their invention, and will likely be broadly interpreted to cover a variety of mechanisms for implementing the basic software invention.

C. Biotechnology Patent Cases
In contrast to the Federal Circuit decisions regarding software, recent decisions involving genetic material have imposed a stringent disclosure standard for patenting macromolecules. The Court has laid particular emphasis on the "written description" requirement of section 112, which requires the patentee to specifically describe the claimed invention as part of the disclosure. The justification for such a detailed description is to demonstrate to others of ordinary skill that the inventor in fact has the invention in her possession; the assumption being that a sufficiently detailed description would not be possible if the inventor were speculating or guessing about its features. 60 58 184 F.3d 1339 (Fed. Cir. 1999). 59 For an argument that a variety of structural tendencies are likely to drive the courts to read software patent claims broadly under the doctrine of equivalents, see Cohen & Lemley, supra note __, at 39-50. This requirement is separate from, and potentially more stringent than that of enablement.
Although the two are closely connected, satisfying one requirement does not necessarily satisfy the other. The classic example offered by one court is the situation in which the description of a particular chemical compound enables one of ordinary skill to make other, related, compounds, yet those other compounds are not described in the patent disclosure. The first compound is both enabled and described; the others are only enabled.
This venerable chemical patenting hypothetical has been brought to life by the Federal Circuit's biotechnology opinions. For example, in Fiers v. Ravel, the court considered the decision of the Patent Office in a three-way interference over patent applications drawn to DNA coding for human fibroblast beta-interferon (β-IF). One of the applicants, Ravel, relied for priority upon his Isreali patent application, which disclosed methods for isolating a fragment of the DNA sequence coding for β-IF and for isolating messenger RNA coding for β-IF. The court considered, inter alia, whether the disclosure in Ravel's Isreali application satisfied the U.S. written description requirement so as to form the basis for a U.S. application. The Federal Circuit upheld a determination by the Board of Patent Appeals and Interferences that Revel's disclosure was not an adequate description, largely because it failed to disclose the nucleotide sequence for the DNA molecule at issue. According to the court's reasoning, disclosing a method for obtaining the molecule was not the same as disclosing the molecule itself: An adequate written description of a DNA requires more than a mere statement that it is part of the invention and a reference to a potential method for isolating it; what is required is a description of the DNA itself. . . . A bare reference to a DNA with a statement that it can be obtained by inventor "has" the invention mentally, and so can give a sufficiently detailed description of that inventive conception -physically creating the invention is mere reverse transcription is not a description; it does not indicate that Revel was in possession of the DNA. 61 Since the Ravel application did not disclose the sequence for the molecule claimed, the court characterized it as disclosing merely "a wish, or arguably a plan, for obtaining the DNA." 62 A similar conclusion was reached in a subsequent Eli Lilly v.
Genentech, where the patent at issue was drawn to a microorganism carrying the DNA sequence coding for human insulin, and supported this claim by disclosing a method for obtaining the human cDNA, as well as the amino acid sequences for the insulin protein.
But relying on the Fiers opinion, the court concluded that the written description requirement again was not met: "Describing a method of preparing a cDNA or even describing the protein the cDNA encodes, as the example does, does not necessarily describe the DNA iteslf." 63 In reaching these holdings, the Federal Circuit has been adamant that the degree of specificity required for an adequate description of nucleic acids requires "structure, formula, chemical name, or physical properties." 64 In Eli Lilly, because "[n]o sequence information indicating which nucleotides constitute human cDNA appears in the patent . . result. Many such genes may achieve that result. The description requirement of the patent statute requires a description of an invention, not an indication of a result that one might achieve if one made that invention. Accordingly, naming a type of material generally known to exist, in the absence of knowledge as to what that material consists of, is not a description of that material. 66 Such failure to describe more than one or two nucleotides is a particular problem where the patent claims are drawn to a broad class of nucleotides. For example, Revel's claim covered all DNA molecules that code for β-IF, but: "[c]laiming all DNAs that achieve a result without defining what means will do so is not in compliance with the description requirement; it is an attempt to preempt the future before it has arrived." 67 The construction of the written description requirement as requiring precise sequence data gains particular significance whenever claims are drawn to an entire genus, or family, of molecules. The patent discussed in the Lilly written description analysis claimed a broad family of DNA molecules coding for insulin in different mammalian species, but it disclosed only one species of DNA, that coding for rat insulin. The court held this to be insufficient to describe the broad class of cDNAs coding for mammalian or vertebrate insulin. Although declining to specify exactly what would be needed to support a broad claim, the court cited previous chemical cases dealing with related groups of small molecules. Based on these cases, the court declared that macromolecules should be treated in the same fashion; the patentee need not not show every member of a claimed genus, but is required to show a "representative" number of cDNAs, illustrating or defining the common structural features of a "substantial" portion of the genus. 66 Id. 67 at 1171 A similarly broad claim was rejected in the Amgen case as failing the standard for enablement. 68 There, the patentee claimed nucleic acid sequences coding for the protein erythropoetin, or for other proteins with the same biological function. The trial judge concluded that because Amgen was unable to specify which of these analogs might have the biological properties claimed, the claims were not enabled. The Federal Circuit panel, however, that the district court had reached the right conclusion for the wrong reason: Whereas the district court focused on the thousands of EPO analogs that could be created by substituting amino acid residues in the polypeptide chain, the appellate court focused on the patentee's failure to disclose the DNA molecules that would code for those analogs. Since the claims were directed to DNA sequences, the issue was not the enablement of the EPO analogs, but rather the enablement of the myriad DNA sequences, which the court held could not be made and used on the basis of a few examples.
The same concerns that characterize the Federal Circuit's jurisprudence of biotechnology disclosure -the inadequacy of methodological disclosure, the requirement to specify sequence or structure, and uncertainty of selection within large classes of homologous molecules --have shaped the Federal Circuit's jurisprudence of biotechnology. However, in the case of obviousness, the issue has been the presence of such factors in the prior art, rather than in inventor's disclosure. Thus, the Federal Circuit held in In re Bell that a claim to DNA coding for human insulin-like growth factor (hIGF) was not obvious although the prior art disclosed the amino acid sequence for the hIGF proteins and a method for using that information to obtain the corresponding DNA molecule. 69  coding for heparin binding growth factors (HBGFs) were not obvious in light of prior art disclosure of a partial amino acid sequences and a method for using that information to obtain the corresponding DNA molecule. 70 Each decision rested largely upon the court's perception that the actual sequence of the claimed DNA molecules was uncertain or unpredictable from the prior art. In both cases the court dismissed as irrelevant the biological relationship between the molecules disclosed in the prior art and those claimed by the patent. The amino acid sequences of the proteins disclosed in the prior art are ultimately determined by the sequence of RNA nucleotides coding for the protein, which is in turn determinative of the cDNA claimed in the patent. 71 The correspondence of nucleotide sequences to amino acid sequences is well known as key to the "central dogma" of molecular biology: the transfer of genetic information from DNA to RNA to protein chains. However, certain amino acids can correspond to more than one nucleotide sequence, introducing uncertainty into the inverse relationship: that of amino acid sequence to nucleotide sequence. Because of this redundancy or "degeneracy" in the genetic code, the court noted in Bell that a vast number of possible sequences -about 10 36 -might code for the protein sequences disclosed in the prior art. The plaintiff claimed only one of these, in essence having searched among a large number of possibilities to select the particular sequence coding for hIGF.
Numerous commentators have pointed out that such a search is relatively routine.
But prior art disclosure of method, even admittedly obvious method, was held insufficient 70 51 F.2d 1552 (Fed. Cir. 1995). 71 Neither In re Bell nor In re Deuel dealt with genomic DNA (gDNA) sequences, which are transcribed by cellular proteins to produce a messenger RNA molecule. Both cases considered non-naturally occurring cDNA sequences, which are reverse transcribed from messenger RNAs. The correspondence between to cure such uncertainty of structure. In rejecting the DNA claims in Bell and Deuel, "the PTO's focus on known methods for potentially isolating the claimed DNA molecules is also misplaced because the claims at issue define compounds, not methods." 72 Prior to Bell, the opinion in Amgen had stressed the uncertainty of the methods for gene location available at the time of invention: "[I]t might have been feasible, perhaps obvious to try, to successfully probe a human gDNA library with a monkey cDNA probe, it does not not indicated that the gene could have been identified and isolated with any success. . . . there was no reasonable expectation of success in obtaining the EPO gene by the method that Lin eventually used." 73 By the time of the research at issue in Bell, such methods for searching a large universe of molecules were perhaps painstaking and time-consuming, but had an established likelihood of success.
Yet the court defined the issue in Bell and Deuel not as a matter of the uncertainty of obtaining a particular sequence, but of the uncertainty of predicting or visualizing from the prior art what sequence would be found. Even in the Amgen opinion, the court hinted that the key to obviousness lay in the prediction of an exact sequence, as "[n]either the DNA nucleotide sequence nor its exact degree of homology with the monkey EPO gene was known at the time." 74 And in Deuel, the court explicitly held that "until the claimed molecules were actually isolated and purified, it would have been highly unlikely for one of ordinary skill in the art to contemplate what was ultimately obtained. What cannot be contemplated or conceived cannot be obvious." 75 Thus a likelihood, or even a certainty gDNA and RNA may be very different than than that of cDNA to RNA, especially in eukaryotic organisms where the processing of RNA transcripts may be extensive. 72  of finding a DNA molecule with particular properties was deemed essentially irrelevant to whether structural claims to that molecule are obvious.
The corollary to this holding is that a molecule will be obvious if the sequence, and not particular function, are discernible in the prior art. Prior art description of the "general idea of the claimed molecules, their function, and their general chemical nature" 76 is insufficient to render a molecule obvious. Some commentators have suggested that this requirement for obviousness stands some danger of collapsing into the standard for anticipation: under section 102 of the Patent Act, an invention lacks patentable novelty if is elements are fully described in a prior art reference, and the Federal Circuit's obviousness requirement could be read to require such a prior art anticipation as the effective standard for obviousness. 77 But unlike the requirements for anticipation, the Federal Circuit's biotechnology obviousness standard appears to require that the sequence of the DNA be predictable from the prior art, and not necessarily explicitly described. For example, the court in Deuel suggests that "a protein of sufficiently small size and simplicity, so that lacking redundancy, each possible DNA would be obvious over the protein." 78 Although the Federal Circuit has not explicitly held so, one would also suspect that disclosure in the prior art of a substantial number of homologous sequences would render a new homologue predictable, and so render it obvious -just as the court has held that disclosure of a substantial number of homologues enough to satisfy the written description requirement for that genus. 76 51 F.3d at 1558. 77 Indeed, the Federal Circuit has several times suggested that the two patent standards are closely linked, characterizing obviousness as a sort of continuum with anticipation as the "epitome" or "ultimate" endpoint of obviousness. See Note that these holdings are all of a piece with the court's earlier holdings, such as the rejection, on disclosure grounds, of Revel's claim to all DNA sequences coding for β-IF. Due to degeneracy in the genetic code, Revel could not adequately describe the claimed invention as DNA coding for β-IF; an astronomically large number of possible sequences might do so. And if a functional or narrative description in a patent is insufficient to properly describe a DNA molecule coding for β-IF, so the presence of a functional or narrative description of β-IF protein in the prior art would be insufficient to render the molecule obvious: according to the court, one cannot describe what one has not conceived, and what cannot be contemplated or conceived cannot be obvious. Just as disclosure in a patent of a method for obtaining a particular cDNA is inadequate to properly describe the invention, so disclosure in the prior art of a method for obtaining a particular cDNA cannot render the claimed invention obvious.
The conceptual linkage of obviousness and enablement to the depiction of macromolecular sequences in, respectively, the prior art or the patent disclosure, dictates a particular and predictable result for the availabilty and scope of such biotechnology patents. The expected outcome is that DNA patents will be extremetly narrow but numerous. Under the Federal Circuit's precedent, a researcher will be able to claim only sequences disclosed under the stringent written description --the actual sequence in hand, so to speak. At the same time, the inventor is shielded from obviousness by the lack of such explicit and detailed disclosure in the prior art. This lack of effective prior art seems to dictate that anyone who has isolated and characterized a DNA molecule is certain to receive a patent on it. But the inventor is certain to receive a patent only on that molecule, as the Federal Circuit appears to regard other related molecules as inadequately described until their is sequence is disclosed.
The set of axioms underlying this set of results forms a logical framework that can be predictably extended to other biotechnology inventions besides DNA. For example, one would conclude from the Federal Circuit's analysis that a cDNA should be obvious in light of its corresponding mRNA, since the former is reverse transcribed from the latter, and there is no redundancy or degeneracy in the correspondence between the nucleotides in the two molecules. 79 However, an mRNA or corresponding cDNA needd not render obvious the genomic DNA (gDNA) from which is is derived, since in many organisms, the gDNA will include intervening sequences, or introns, that are not predictable from the mRNA sequence. where highly detailed disclosure is required to satisfy the enablement and written description standards, similarly detailed disclosure is required to render the invention obvious. In software, where little specific detail is needed to satisfy the requirements of disclosure, similarly little detail is needed to render the invention obvious. In each case, the court takes the patentability requirements of non-obviousness and disclosure as firmly tied to a common standard.
The common standard connecting these requirements is that under each statutory section, the respective requirements must be addressed with regard to the "person having ordinary skill in the art," sometimes known by the acronym of PHOSITA. 81  as a requirement that the claimed invention, taken as a whole not be obvious to one of ordinary skill in the art at the time the invention was made.
The PHOSITA is equally central to calibrating the legal standard for patent disclosure. As the quid pro quo for her period of exclusive rights over an invention, the inventor must fully disclose the invention to the public. The first paragraph of 35 U.S.C.
sec. 112 requires that this disclosure enable "any person skilled in the art" in the art to make and use the claimed invention. The language this section therefore indicates that the inventor's compliance with the requirement of enablement should be measured with reference to a standard similar or identical to that in section 103; indeed, the language appears to tie the enablement requirement to non-obviousness via this shared metric.
This same language set the metric for several related disclosure doctrines as well.
First, the definition of enablement effects the patentablity requirement of specific utility, as the invention must operate as described in the specification if the inventor is to This places the standard for patentablity on a legally objective, rather than subjective, footing. The PHOSITA standard measures the inventor's achievements against a judicially determined external metric, rather than against an expectation based on whatever level of skill the inventor might actually possess. The standard also has the practical effect, since the question is one of law, of avoiding the requirement that judges and other arbiters of patentability be experts in a given field. A second circumstantial variable is the level of skill that would be considered "ordinary." Unlike the inventor, who almost by definition is presumed to be one of extraordinary skill, 88 the PHOSITA standard contemplates some median level of skill. In assessing that median level, courts may take into account a long list of factors, including the approaches found in the prior art, the sophistication of the technology involved, the rapidity of innovation in that field, and the level of education typical of those in the field. 89 The courts have also endowed the PHOSITA with mediocre personality traits; she is conceived of as an entity that adopts conventional approaches to problem solving, and is not inclined to innovate, either via exceptional insight or painstaking labor. 90 Some care must be exercised in characterizing the PHOSITA, as it is tempting to do so on the basis of an unfounded presumption, which is that the PHOSITA remains constant from section to section of the patent statute. On the contrary, some commentators have recognized as quite possible the possibility that the imaginary artisan found in these different statutory sections, though bearing the same denomination, might well display different and even inconsistent characteristics as between the different sections. 91 The PHOSITA for purposes of obviousness may not necessarily be the PHOSITA for purposes of enablement, written description, definiteness, or equivalence.
Because she is a legal construct designated to embody certain legal standards, the PHOSITA could well change depending on the purpose she is serving at the time.  may occasionally be surprised by their properties, but either outcome is based on the molecules' structural depiction.
The rule in these small molecule cases appears closely related to that announced in biotechnology cases discussed here --the Federal Circuit has declared that DNA "is a chemical, albeit a complex one," and has articulated a desire to treat the patenting of macromolecules in the same fashion as the patenting of more traditional organic molecules. In focusing upon structural depiction as the linchpin of both obviousness and disclosure, the biotechnology cases rely upon, and appear to extend, the line of chemical cases summarized above. But just as we question the application of these rules to macromolecules, we are similarly uncertain that these special rules for obviousness in makes logical sense. At the simplest level, this approach patent is intended to benefit the public; people who work in a given technology must understand the patent as it relates the prior art, so it makes sense to take into account what that person knows in order to decide whether a patent is obvious or has been enabled. From a policy standpoint, the practicality of working in different technologies requires a flexible approach to determining disclosure or obviousness, and the PHOSITA approach gives a court that flexibility.
Consequently, in order to identify the source of the anomalies in biotechnology and software, we look first to the Federal Circuit's application of this standard, rather than to the standard itself. One possibility, which has occurred to previous commentators as well as to us, is that the Federal Circuit application of the PHOSITA standard in these technologies is wrong as a matter of science. One reading of these cases is that the Federal Circuit seems to have substituted caricature for a nuanced understanding of the technology. The court has repeatedly suggested that programming itself is a "mere clerical function" that presumably does not warrant the grant of a patent. The court seems to consider only the ideas or functions of a computer program worth of patent protection. In short, the court thinks of programmers as of extraordinary skill, capable of implementing any idea in a computer program as a matter of course. Sometimes this assumption benefits patentees, notably in enablement and best mode determinations. 103 Other times, notably in obviousness cases, the assumption that programmers are extremely skilled works against the patentee. 104 But as a matter of computer science, the 103 See supra notes __-__ and accompanying text (noting the low standards applied to software patents under section 112). court's assumptions are just wrong. Any programmer will tell you that coming up with an idea for a computer program is rather less than half the battle. 105 Programs can take years to write even under the best circumstances. Some will simply not work. Others will require innovative programming techniques. Even once they are written, most programs have bugs that must we worked out in order for the program to be stable. 106 And in many cases, the process of writing the program changes the idea itself in a sort of continuous feedback loop. 107 Not only is it wrong to say that writing a program is a "mere clerical function" to a skilled programmer, but in fact many of the truly innovative improvements in computer software occur at the level of programming, not the idea to have a computer perform a specific function. 108 In the biotechnology cases, problem is the opposite; the court focuses repeatedly on the "uncertainty" inherent in the field, scoffing at claims drawn to molecular function, and demanding precise disclosure of any embodiment. 109 The court seems to believe that biotechnology is as much a black art as a science. In Bell and Deuel the court's belief in uncertainty benefits the patentee, since it means that knowledge of a protein and a method for deriving the cDNA sequence did not render the cDNA sequence obvious. 110 By contrast, the same assumption about uncertainty hurts patentees in cases like Enzo, Lilly and Amgen, because it precludes them from claiming any DNA sequence they have not actually described in the patent specification. 111 But all of these holdings are based on the assumption that one ordinarily skilled in biotechnology cannot get from a protein to a DNA sequence, or from the DNA sequence of one mammal to the corresponding DNA sequence of another mammal.
Arguably this understanding of the science of biotechnology is simply wrong. 112 Robert Hodges has argued that "the key event is the cloning of the first gene in a family of corresponding genes. Once a researcher accomplishes this very difficult task, the researcher can typically obtain other members of the gene family with much less effort." 113 Indeed, the process is largely automated. Such research is properly compared 110   This observation might be accommodated by an alternative explanation: that the court, rather than stumbling in its application of law to changing technology, is as a matter of law simply creating a unique enclave of patent doctrine for biotechnology. The same explanation might be applied to the software cases: that it is not a unique set of facts applied to the PHOSITA construct that is generating a technology-specific body of patent law, but rather a deliberate manipulation of doctrine itself. Yet this alternative explanation seems to us even less satisfactory than the first --it essentially moves the for the Double Helix, 26 AIPLA Q.J. 381, 418 (1998) ("making the inventions of Amgen, Fiers and Lilly today would be routine"). problem up one level of abstraction to argue that the Federal Circuit is not mistaken as a matter of fact about the state of technology in biotechnology and software, but rather mistaken as a matter of policy about the needs of those industries. If the court is taking the trouble of fashioning individual patentability standards for different areas of subject matter, one would expect that the standards fashioned would be suited to the needs of the different areas addressed --yet that seems not to be the case here.
Yet a third alternative explanation, at least for the biotechnology cases, might be found in a sort of judicial economy. Rather than fashion new doctrines, the Federal Circuit biotechnology cases may be an attempt to extend existing chemical patent doctrines to encompass biotechnology. Such an approach has the advantage of legal predictability, relying on established precedent with which innovators and their attorneys are already familiar. But this explanation fails to explain why such doctrine should not also be extended to encompass software technology -why the explicit listing of computer code is not as critical to obviousness or disclosure as the explicit listing of genetic code seems to be. Moreover, all these explanations beg the question of whether different technologies ought to have their own specific patenting standards, whether those standards are dictated by unique facts, or by legal considerations.
If, as we suggest, the concept of the PHOSITA makes sense, why has the Federal Circuit got it wrong in these industries? We think there are several structural barriers that make it difficult for courts to accurately assess the level of skill in a complex technological art. First, it is worth emphasizing that judges are at a rather serious disadvantage in trying to put themselves in the shoes of an ordinarily skilled scientist.
Judges ordinarily don't have any scientific background, and at least at the district court level most law clerks don't either. Further, district court judges have extremely full dockets with many different types of cases. The average judge may hear no more than one patent case every few years. 114 Few of those will be software or biotechnology cases. 115 So a very busy judge must learn not only patent law but also some difficult science in a very short period of time. Expert witnesses can help, but the Federal Circuit has imposed some limits on the extent to which district courts can rely on such evidence. 116 In particular, courts must avoid the temptation to assume that the expert witness is a person ordinarily skilled in the art. 117  . Since there are over 600 district court judges in the United States, it is obvious that most judges get only a few filed patent cases a year, and well less than one patent trial a year. In fact, many judges get even fewer cases than this number would suggest (though others get more), since the concentration of innovation in certain regions and the permissibility of forum shopping in patent cases cause patent cases to be bunched in a few districts.  Federal Circuit has more time to consider each case, has the full record before it, and gets many more patent cases (and therefore more software and biotechnology cases) than any district court judge would. 119 Arti Rai argues that the Federal Circuit should defer to the PTO, because the PTO better understands biotechnology. Rai, supra note __, at __. We agree with her that the Federal Circuit makes mistakes in this area. We are not persuaded that the PTO can do any better, however, particularly given the minimal time examiners can spend on any one invention. See Mark A. Lemley The Federal Circuit has repeatedly recognized the problem of hindsight bias in its obviousness jurisprudence, 124 and has built rules designed to cope with it there, 125 but in fact hindsight bias risks infecting the PHOSITA analysis in enablement and claim scope as well. Hindsight bias will normally lead factfinders to overestimate the level of skill in the art, since subsequent advances will suggest that the invention couldn't have been that Finally, the backward-looking nature of the legal system itself creates a problem that is in some sense the opposite of the hindsight bias. Legal rules are based on stare decisis. The law accumulates nuance over time by respecting and building on the body of existing precedent. Only rarely will courts expressly reject their prior decisions. This system has worked well over time in producing thoughtful legal rules. 127 Judges trained in this process will naturally tend to apply it to factual issues they see repeatedly as well.
Indeed, doing so seems economical as well, since revisiting those factual determinations appears redundant. Thus, once the Federal Circuit has ruled on the level of skill in a particular art, the temptation is strong for both that court and district courts to apply that determination in subsequent cases as well. This tendency is evident in both software and biotechnology cases. In the software cases, the court in Northern Telecom held that the patentee needn't disclose the actual code implementing a program in order to satisfy the enablement or best mode requirements. The court in that case acknowledged that determinations of the level of skill in the computer industry should be made on a case-bycase basis. 128 But subsequent Federal Circuit decisions have not inquired separately into the level of skill in the art, or explored the complexity of the program before them in much detail. Instead, they have tended to rely on prior cases holding that code need not be disclosed. In biotechnology the linkage is even stronger. In re Bell concluded that knowledge of an amino acid sequence produced by a gene, coupled with a plan for identifying the DNA sequence of the gene, did not render the DNA sequence itself  In re Deuel, 51 F.3d 1552, 1559 (Fed. Cir. 1995). In Bell, the prior art disclosed the amino acid sequence for the proteins of interest, and a method for cloning genes. By contrast, the art in Deuel disclosed only a partial sequence. Nontheless, the passage of time between the priority dates of the applications -almost 10 years [check] -was ignored by the court, which did not focus on or even mention when the inventions occurred. the invention was made, while enablement is tested at the time the application was filed.
Clearly the application cannot be filed until after the date of invention, and in some cases several years elapse between the two. 134 The knowledge in the art can change during this period, sometimes dramatically. Second, and more important, the level of skill in the art will normally change between the dates of different inventions. It is hazardous, therefore, to rely on one court's statement of the level of skill in the art as determinative or even evidentiary of the level of skill in the same art at a different time. The level of skill in the art is a factual question that must be determined anew on the particulars of each case. 135

III. Should Patent Law Be Technology-Specific?
If we are right that the nominally unified standards of patent law are in fact technology-specific because they rely so heavily on the level of skill in the art, the question remains whether this heterogeneity is desirable. In the last section, we suggested that the Federal Circuit has, either as a matter of fact or as a conclusion of law, improperly characterized the level of skill in the art in the biotechnology and software industries. But even if these problems are overcome, there seems little question that the effect of reliance on the PHOSITA will be to impose different standards in different industries. In section A, we discuss the policy implications of this approach for software 134 The law permits a one year grace period between any public act and the filing of a patent application. 35 U.S.C. § 102(b). But many inventors wait even longer between invention and the filing of an application. This is permissible, so long as they do not put the invention on sale or in public use in the interim, and do not abandon it. 35 U.S.C. § 102(c). 135 For a detailed discussion, see Dillen, supra note __, at 1039-44. The CCPA recognized this in In re Driscoll, 562 F.2d 1245, 1250 (C.C.P. A. 1977), and the Federal Circuit in Enzo Biochem v. Calgene, Inc., 188 F.3d 1362, 1374 n.10 (Fed. Cir. 1999). But it has proven a hard rule to adhere to. and biotechnology. In section B, we discuss the more general implications of patentspecificity for patent reform.

A. The Proper Scope of Software and Biotechnology Patents
Nonobviousness is a function of uncertainty. 136  projects unless the expected reward is correspondingly greater. This higher level of perceived risk may be to some extent counterbalanced by lowering the standard for patentability of high-cost projects, increasing the likelihood of financial reward.
Application of this innovation-based model to software and biotechnology reveals problems with the current legal standards. As we have seen, the courts have portrayed computer programming as relatively straightforward and biotechnology as more uncertain. As a result, software patents are more likely to be found obvious, but if valid tend to be broad in scope and require only minimal disclosure. 138 Biotechnology patents, by contrast, are relatively easy to prove nonobvious but require detailed disclosure and are extremely narrow. 139 Arguably this result is exactly backwards as a matter of innovation policy. As we shall see, common policy arguments responding to market conditions peculiar to the two industries call for narrow software patents and broad biotech patents. At a minimum, the PHOSITA test --even if correctly applied --will not produce the results that seem optimal for each industry.

Optimal Software Patent Policy
The computer industry is characterized by a large number of rapid, iterative improvements on existing products. 140  ideas, and often on prior code itself. 141 This incremental improvement is desirable for a variety of reasons. First, it responds to the hardware-based architectural constraints of the software industry. Data storage capacity, processing speed, and transmission rates have all increased steadily over time. 142 Programs written during an older period therefore faced capacity constraints that disappear over time. It makes sense to improve those products progressively as the constraints that limit the functionality of the programs disappear. Second, incremental improvement of existing programs and ideas tends to render programs more stable. It is received wisdom that you should avoid version 1.0 of any software product, because its maker is unlikely to have all the bugs worked out.
Iterative programs built on a single base tend to solve these problems over time. This is most obviously true when actual computer code is reused, but it is true even when tested algorithms or structures are replicated in new programs. Third, iterative improvement helps preserve interoperability, both among generations of the same program and across programs. 143 141 On reuse of existing code, both within and across companies, see Lemley & O'Brien, supra note __. 142 Moore's "law" provides that historically the speed of microprocessors has doubled every 18 months. It is well known that data storage capacity and transmission rates have shown similarly exponential increases. 143 For the same reason, reverse engineering has had a respected place as a legitimate means of creating interoperability. Virtually all recent copyright decisions have endorsed reverse engineering in some circumstances.  . 1999) (acknowledging the right to reverse engineer for some purposes, but holding it unjustified in this case).
As with courts, the overwhelming majority of commentators endorse a right to reverse engineer copyrighted software, at least for certain purposes. Predominantly copyright, but also trade secret and contract law. One factor militating in favor of stronger intellectual property protection in software is the ease of duplication of digital information in the networked world. But copyright protection is much better suited than patent to preventing exact duplication. Copyright law has also been modified to better prevent such copying in the computer context by allowing copyright owners to control access to copy-protected works. See 17 U.S.C. §1201 (the Digital Millenium Copyright Act). 146 Cohen & Lemley, supra note __, at 39-50 (incremental nature of software innovation, lack of good prior art, rapid pace of change, and the difficulty of characterizing code inventions in words all contribute to broad readings of software inventions). They write: The pattern of cumulative, sequential innovation and reuse that prevails in the software industry creates the risk that software patents will cast large shadows in infringement litigation. Specifically, we believe that because innovation is especially likely to proceed by building on existing code in other programs, the temptation for the trier of fact to find equivalence of improvements will be correspondingly greater. Id. at 41. absence of other forms of protection, a large number of narrow software patents may be the best way of protecting these low-level innovations. 148 Patent protection for such incremental software inventions should be relatively narrow, and in particular should not generally extend across several product generations. 149 But broad software patents are precisely what the court's PHOSITA approach has produced. By defining software inventions in broad terms, supported by very little in the way of detailed disclosure, the Federal Circuit has encouraged software patents to be drafted broadly and to be applied to accused devices that are far removed from the original patented invention. By implication, the Federal Circuit's standard also seems to suggest that many software patents on low-level incremental improvements will be invalid for obviousness. In software, then, the Federal Circuit's standard seems to have it backwards.

Optimal Biotechnology Patent Policy
If any technology fits the criteria of high cost and high risk innovation, it is certainly biotechnology. Biotechnology products appear in a wide variety of economic sectors, from pharmaceuticals to foodstuffs to industrial processes. 150 1025 (1990). As noted below, we share this concern, but believe the solution is to narrow the scope of those patents. 149 155 See, e.g., Robert P. Merges, Patent Law and Policy 519 (2d ed. 1997). 156 One way to think of this is to reconceive patents as a financing mechanism: by providing definable rights, patents enable companies to obtain the funding they need to turn an invention into a product. See receive a patent, but one that is trivial to evade. Under this standard, no one is likely to receive a patent broad enough to support the further costs of development. 157 An additional drawback to the proliferation of narrow biotechnology patents may be the development of what has been termed an "anti-commons" in the biotechnology industry. 158 The anticommons is characterized by fragmented property rights, the aggregation of which is necessary to make effective use of the property. 159 Aggregating such fragmented property rights entails high search and negotiation costs to locate and bargain with the many rights owners whose collective permissions are necessary to complete broader development. This type of licensing environment may quickly become dominated by "hold outs" who refuse to license their essential sliver of the pie unless bribed. 160 Because a given project will fail without their cooperation, "hold-outs" may be prompted to demand a bribe close to the value of the entire project. 161 And, of course, every property holder needed for the project is subject to this same incentive; if everyone holds out, the cost of the project will rise substantially, and probably will rise prohibitively Unfortunately, if patent law is to accommodate the uncertainty of biotechnology innovation, this proliferation of narrow biotechnology patents may be nearly impossible to avoid under the reciprocal structure of obviousness and enablement in the PHOSITA patent doctrine. 162 In order for the invention to avoid obviousness, it must be deemed beyond the skill of the PHOSITA to construct given the level of disclosure in the prior art. Yet this means that in disclosing the invention, the inventor must tell those of ordinary skill a good deal more about how to make and use it, effectively raising the standard for enablement and written description. The Federal Circuit's insistence that the results biotechnology research are unforeseeable or unpredictable avoids the problem of obviousness, but this inevitably results in an extremely stringent standard for disclosure and description.

B.
Can Others suggest that a form of sui generis patent-like protection is appropriate. 174 Still others who endorse the general framework argue that the courts should apply patent law to software in somewhat different ways than they do in other industries. 175 Similarly, scholars have suggested that biotechnology patent standards should deviate from the general patent law rules. 176 Some argue that certain types of biotechnological patents should be unpatentable altogether. 177 Others suggest that the disclosure requirements 171 Cite Boucher-Berman; Claritin patent relief. 172 See, e.g., Samuelson, Benson Revisited, supra note __. 173 should be loosened, 178 that the obviousness standard should be lowered, 179 or that the scope of DNA sequence patents should be restricted. 180 They have variously argued that the Federal Circuit should defer to the PTO, 181 or conversely that the PTO should defer to the Federal Circuit. 182 Calls to modify patent law are a natural response to the different effects patent law has in different industries. The economic effects of patents are quite different in different industries. Thus, in a perfect world the patent system might well be tailored to give optimal incentives to each different industry.
We do not live in a perfect world, however. In the real world, a number of factors caution against tailoring the patent law to the needs of particular industries. The most obvious barrier is legal -the TRIPs agreement prohibits member states from discriminating in the grant of patents based on the type of technology at issue. 183 As noted above, however, the United States has not faithfully followed this treaty mandate.
Neither has the EU, which has industry-specific rules for compulsory licensing of pharmaceuticals and for the patentability of software and business methods. 184 Even if industry-specific patent legislation were legal, we are not fully persuaded that it is a good idea. First, while economics can make useful policy suggestions as to how patents work in different industries, we are skeptical of its ability to dictate in detail the right patent rules for each industry. 185 Economic theory is more useful in making general suggestions about how the patent system can be adapted than it is as the basis for a whole series of new statutes. delivery systems might be thought of as medical devices, pharmaceuticals, or biotechnology; presumably a different law would apply depending on how the invention was characterized. Even technologies that seem radically different, like biotechnology and software, may unexpectedly converge, as recent developments in bioinformatics and proteomics have made clear. 186 Further, a significant percentage of inventions fall into more than one field. 187 And of course new fields arise regularly; imagine trying to fit all modern inventions into categories created 50 or 100 years ago. As a result, it will prove impossible to carve up innovation into static fields.
This point raises a related one. The history of industry-specific statutes suggests that many turn out to be failures because they are drafted with current technology in mind, and are not sufficiently general to accommodate the inevitable change in technology. The best example is the Semiconductor Chip Protection Act. 188 Passed after six years of debate, the SCPA created a detailed set of rules designed to protect semiconductor mask works. But it has virtually never been used. 189 The most likely reason is that the particular focus of the SCPA --duplication of mask works --is obsolete because of changes in the way semiconductor chips are made. Industry-specific patent statutes risk a similar fate. 190 Finally, and of most concern, technology-specific patent legislation will encourage rent seeking. Patent law has some balance today in part because different industries have different interests, making it difficult for one interest group to push through changes to the statute. Industry-specific legislation is much more vulnerable to industry capture. It is no accident that the industry-specific portions of the patent law are among the most complex and confusing sections, 191 and that they have had some pernicious consequences. 192 The copyright model --in which industry-specific rules and exceptions have led to a bloated, impenetrable statute that reads like the tax code 193 --is hardly one patent law should emulate. 194

General Modifications to Accommodate Policy Concerns
An alternative to industry-specific patent legislation is some sort of general change to the patent law that permits the courts to reach more efficient results in particular industries. General changes are more promising than industry-specific legislation for the reasons we suggest above. 195 In this section, we consider two possible changes to the patent law that might solve the biotechnology-software problem we identify in this article. Because the level of knowledge for the enablement PHOSITA is measured at the time a patent is filed, rather than as of the date of invention, a larger pool of prior art will frequently be imputed to the knowledge of the enablement PHOSITA. The accumulation of prior art between invention and filing could in theory allow an invention to enjoy both a low threshold of obviousness and a low threshold of disclosure.
It may be that this differential knowledge should be emphasized, in order to decouple the tight reciprocity of obviousness and enablement. In the case of biotechnology, this approach might bring policy expectations into line with doctrinal results. The characteristics of the industry would seem to demand fewer and broader 195 Of course, general modifications may have disproportionate impact on specific industries. Thus, if Wells is right that Internet business methods have low costs of development and are easy to bring to patents that are easier to obtain, which is to say, patents with a relatively low obviousness threshold, but also a relatively low enablement threshold. These might be provided by the temporal disparity in the two PHOSITAs. Since the body of prior art grows during the period between invention and the filing of a patent, the corpus of knowledge imputed to the enablement PHOSITA will be larger than that imputed to the obviousness PHOSITA. Thus, the level of disclosure required to enable one of ordinary skill at the time of filing could well be lower than that required to enable one of ordinary skill at the time the invention was made, because the enablement PHOSITA is expected to know more. Conversely, an invention may well be non-obvious at the time it is made, although it would not be at the time a patent is filed. This will depend to some extent upon when the invention is considered to be "made" for non-obviousness purposes. If the inventor must rely upon her filing date as the date the invention is made, then the knowledge imputed to the enablement and obviousness PHOSITAs may be coterminous; however, if the date of invention can be related back to an early time of conception, the disparity between the two bodies of prior art may be substantial.
This same effect may hold true for the PHOSITA of the written description requirement, albeit to a lesser extent. The written description requirement substantially overlaps with the degree of disclosure required for enablement, but is likely to require something more. Because the unique purpose of the written description requirement is to demonstrate what the inventor had in his possession at the time the patent was filed, courts have been understandably reluctant to assume that details missing from the disclosure could be supplied by the prior art knowledge imputed to the PHOSITA. Thus, market, see Wells, supra note __, at 770-72, they are more likely to be found obvious under the standard we propose. the inventor is less able to rely upon the level of knowledge in the prior art to establish a less stringent requirement for written description. Nonetheless, some courts have suggested that there is some flexibility in the written description requirement, although that has not been the trend in the Federal Circuit biotechnology cases. If the less stringent holdings of those older, cases could be revitalized, the outcome for biotechnology might be brought into line with the policy expectations for that technology.
However, this consonance between policy outcomes and the emphasis on disparities in prior art may be peculiar to biotechnology, and not capable of generalization to other technologies. In particular, reliance upon the differential in prior art between obviousness and enablement may not yield the optimal result in the case of computer software. As we have described, the profile of that industry militates in favor of narrower and more sharply defined patents; or in other words, toward a higher threshold of patentability for both obviousness and disclosure. But, in this case, the differential in prior art between the time of invention and the time of filing pushes in the wrong direction, away from a stringent enablement standard. No matter how high the threshold for obviousness may be set, the passage of time between invention and filing will place more knowledge at the disposal of the PHOSITA at the latter event, favoring less disclosure rather than more. This means that even a decoupled PHOSITA standard won't achieve ideal results in both biotechnology -where the standards may need to be relaxed somewhat -and software -where they need to be tightened.
Consequently, it may be inadequate to rely upon the knowledge differential already found in the PHOSITA standard in order to correct the mismatch of policy outcomes and doctrinal analysis. What may be requited is to decouple the section 103 and section 112 PHOSITA standards altogether, recognizing that the PHOSITA contemplated for purposes of obviousness is simply not the PHOSITA contemplated for purposes of disclosure. Although tight reciprocity of the two standards, mediated by a common PHOSITA construct, makes for an appealing and intellectually elegant doctrinal framework, theoretical esthetics might be required to give way to technological pragmatics. Again, there are precursors latent in the case law that could be developed into such a doctrinal shift; recall that the section 103 PHOSITA has been portrayed by some courts as a bit of an innovator, while the section 112 PHOSITA has not. Certainly the two constructs are conceived as being engaged in very different inquiries, the first seeking some motivation to compile prior art knowledge into an invention, and the second drawing upon prior art knowledge to supplement an invention disclosure.
Divorcing the two inquiries could allow each standard the freedom to independently accommodate the incentive needed by a given industry.

b. Considering Innovation, Not Just Invention
Second, courts or Congress might modify the standard of obviousness to account for the cost and uncertainty of production and development, not just invention. We think that the divergence between legal and policy outcomes is attributable to the fact that software inventions are much easier to bring to market than biotechnology inventions.
The court repeatedly intones the maxim that biotechnology is an "uncertain art." 196 We 196 See, e.g., In re Vaeck, 947 F.2d 488, 496 (Fed. Cir. 1991) (biotechnology less "predictable" than mechanics or electronics). think, however, that it is not so much invention as product development, production and regulatory approval that is uncertain in biotechnology.
From a policy perspective, the result is the same: biotechnological inventions need more incentive than other types of inventions if they are actually to make it to market. This suggests that the lowered obviousness standard found in the Federal Circuit's biotechnology decisions is the correct outcome, but that the court has reached the correct result for the wrong reasons. The justification in biotechnology for a lower standard of obviousness is not, as the Federal Circuit has suggested, the cost or difficulty of invention -numerous commentators have rightly pointed out that the costs of isolating valuable macromolecules are now trivial, and the expectation of success in such an undertaking is high. The justification in biotechnology for a lower obviousness standard is rather that the costs of bringing such products to market are prohibitive. Product development can be encouraged by offering patentees a greater certainty of obtaining the financial rewards from an exclusive patent right.
A number of commentators have suggested that the court should take account of the higher costs and greater uncertainty of innovation in biotechnology by modifying the nonobviousness standard to account for what might be called "economic nonobviousness." 197 Karen Boyd argues that permitting evidence of economic nonobviousness would improve biotechnology policy, and that the court could start with a rough assumption that all biotechnological inventions are high-cost inventions for which a finding of economic nonobviousness would be appropriate. 198 Rob Merges makes the more general argument that the courts should reduce the obviousness standard either in cases in which the invention itself is uncertain, or in which the commercialization process is long and costly. 199 We think Merges and Boyd are on the right track in pointing to the cost and uncertainty of innovation as a whole -not just invention -as relevant factors in the nonobviousness determination. 200 But we also think that courts in the biotechnology relevancy," 204 the Federal Circuit regularly speaks of them as a required part of any obviousness inquiry. 205 And the secondary considerations that have been used tend to be economic indicia of a patent's value: the commercial success of a patent, long-felt need for the invention and the failure of others to produced it, market response to the invention by copying or acquiescence, and simultaneous invention. 206 One may reasonably question the economic value of some of this evidence, 207 but it is clear that the courts are already willing to consider economic indicia of nonobviousness of their own volition in a variety of patent cases. As a result, introducing evidence of development cost and uncertainty will not substantially disrupt the existing processes for determining patent validity.
A case-specific focus on development cost as a plus-factor also avoids many of the problems with industry-specific patent legislation discussed above. 208 Because courts could easily adopt this test themselves, no legislation is necessary. This not only avoids administrative costs, but it also avoids the very real problem of industry rent-seeking.
Further, because it is a general policy, it can be applied to all types of patent cases.
While the results will be tailored to the characteristics of particular industries, the legal rules will not be. Indeed, by creating a framework in which courts can take account of legitimate policy concerns, the Federal Circuit may actually be able to reduce the judicial "drift" towards industry-specific legal rules that we observed in Part I.

Conclusion
Patent law is becoming technology-specific. The legal rules applied to biotechnology cases bear less and less resemblance to those applied in software cases.
While there are good policy reasons to treat the two industries differently, the current legal rules are not expressly informed by the economics of the industries, but by an ad hoc combination of judicial policymaking and stare decisis. Not surprisingly, they don't reflect optimal patent policy in either biotechnology or software. We have offered some explanations for this phenomenon, along with two specific suggestions -decoupling the obviousness and enablement standards and adding development cost to the secondary considerations of nonobviousness. These suggestions will help optimize patent policy in general and biotechnology and software law in particular.