The Federal Circuit has released its long-awaited opinion in McRo v. Bandai, reversing the lower court’s decision that the claims were ineligible subject matter. McRo’s invention in U.S. 6,307,576 was a method used in 3D computer animation, specifically how to synchronize positioning of the lips and facial expression of a 3D character to correspond to the character’s dialogue, thus providing a realistic impression of the character speaking. Here is the claim:
A method for automatically animating lip synchronization and facial expression of three-dimensional characters comprising:
obtaining a first set of rules that define output morph weight set stream as a function of phoneme sequence and time of said phoneme sequence;
obtaining a timed data file of phonemes having a plurality of sub-sequences;
generating an intermediate stream of output morph weight sets and a plurality of transition parameters between two adjacent morph weight sets by evaluating said plurality of sub-sequences against said first set of rules;
generating a final stream of output morph weight sets at a desired frame rate from said intermediate stream of output morph weight sets and said plurality of transition parameters; and
applying said final stream of output morph weight sets to a sequence of animated characters to produce lip synchronization and facial expression control of said animated characters.
The court held that this claim is not directed to an abstract idea, but reached this result in a new way, using preemption. Here's how they did it.
In step 1 of the Mayo test, the court’s focus is to unpack the “directed to” inquiry: what does it mean for a claim to be “directed to” an abstract idea? “First, we determine whether the claim at issue is ‘directed to’ a judicial exception, such as an abstract idea.” Note the use of quotations here on directed to, which signals that the court is attempting to give this phrase specific content. This is consistent with other recent cases, such as Enfish and Rapid Litigation Management, in which the court likewise focused on the nature of “directed to.”
Next, the court turns to a framing of the abstract idea exception, as intended to exclude claims on results themselves, rather than on processes that produce the results: “The abstract idea exception prevents patenting a result where ‘it matters not by what process or machinery the result is accomplished.’”
This articulation of the abstract idea exception is in line with the various statements that patent claims should recite “how” a particular process achieves the desired result. The panel in Electric Power found the claims there ineligible, noting that the “[i]nquiry, therefore, must turn to any requirements for how the desired result is achieved.” (emphasis in original) The Electric Power panel (pun intended) used this rationale to explain the Enfish as well, stating that “the claims at issue [in Enfish] focused not on asserted advances in uses to which existing computer capabilities could be put, but on a specific improvement—a particular database technique—in how computers could carry out one of their basic functions of storage and retrieval of data.” In Rapid Litigation Management, the court again focused the step 1 inquiry on whether the claim recited just the desired result—the ability of liver cells to survive multiple freeze-thaw cycles—or a particular way of obtaining that result. In Vehicle Intelligence, the claims simply recited “screening an equipment operator by one or more expert systems to detect potential impairment of said equipment operator”; the court held that “None of the claims at issue are limited to a particular kind of impairment, explain how to perform either screening or testing for any impairment, specify how to program the ‘expert system’ to perform any screening or testing, or explain the nature of control to be exercised on the vehicle in response to the test results.” Thus, McRo firmly articulates the framing of the abstract idea analysis as whether the claims simply recite a desired result, or a particular way of achieving the desired result.
Admittedly, this focus alone does not make the step 1 inquiry perfectly easy in practice. But it does move the analysis away from what I call “death by analogy.” Currently, the fashion is to look for a real world “pre-computer” analogy for the claimed method, and once that is found, pronounce the claim directed to an abstract idea. This “analogy” approach elevates the creativity of the litigant over the substance of what is actually claimed—it makes patent eligibility depend not on the skill of the draftsperson, but on the skill of the advocate. And we have seen the bizarre results of this approach: courts finding analogies in reference librarians (Cogent Medicine v. Elsevier; Collarity v. Google), chefs making risotto (Kinglite Holdings Inc. v. Micro-Star International), smoke signals (OpenTV v. Apple), and the telephone operators in the television show Lassie (Stanacard v. Rubard). More fundamentally, this approach belies the fact that inventions solve human problems, which are more or less the same over time: food, clothing, shelter, transportation, entertainment, and of course, work. Thus, most problems have pre-computer analogues with pre-invention (and often pre-computer) solutions. That simply is the nature of innovation, not an indictment of human creativity.
To answer whether the claims are directed to a result per se or a series of specific operations to achieve the result, the court returns the lens to the proper subject: what is actually claimed. Claim drafters have control over what is claimed, whether they claim results or operations—but they do not have control over what analogies may be imagined by a litigant or a court. A meaningful rule in any domain of law is one that allows a person to adapt their behavior to conform to the rule—if there is nothing one can do to change one’s behavior to satisfy the rule, then the rule can only be applied arbitrarily. If this sounds familiar, it should be: it is why many, including judges, have said the Mayo test is still one of “I know it when I see it.”
How then determine whether a claim is “directed to” a result or a specific operation? By focusing on specific claim limitations. As if to signal the importance of this requirement, the court set out the discussion under a distinct heading:
1. Specific Limitations
The district court determined that claim 1 of the ’567 patent is “drawn to the [abstract] idea of automated rules-based use of morph targets and delta sets for lip-synchronized three-dimensional animation.” We disagree. We have previously cautioned that courts “must be careful to avoid oversimplifying the claims” by looking at them generally and failing to account for the specific requirements of the claims. TLI Commc’ns, 823 F.3d at 611; see also Diehr, 450 U.S. at 189 n.12. Here, the claims are limited to rules with specific characteristics. As the district court recognized during claim construction, “the claims themselves set out meaningful requirements for the first set of rules: they ‘define a morph weight set stream as a function of phoneme sequence and times associated with said phoneme sequence.’” They further require “applying said first set of rules to each sub-sequence . . . of timed phonemes.” Id. Whether at step one or step two of the Alice test, in determining the patentability of a method, a court must look to the claims as an ordered combination, without ignoring the requirements of the individual steps. The specific, claimed features of these rules allow for the improvement realized by the invention.
Notice that the court immediately follows the caution against oversimplification with an analysis of the claim limitations themselves, and then seals that analysis with a clear directive that courts must not ignore claim requirements and specific, claimed features in both step one and step two. This unmistakable and intentional arrangement in the argument is a clear rejection of the approach of many district courts to treat step 1 as a “quick look” or search for the “purpose,” “gist,” or “heart” or “basic character” of the invention.
This approach is also consistent with how the court analyzed the claims in Enfish, rejecting the district court’s characterization of the claims as “at too high a level of abstraction”: “Here, the claims are not simply directed to any form of storing tabular data, but instead are specifically directed to a self-referential table for a computer database.” Claiming “any form of storing tabular data” would have been claiming the result. And the Enfish court immediately followed the above by an analysis of the actual claim limitations:
For claim 17, this is reflected in step three of the “means for configuring” algorithm described above. For both pairs of claims 31 and 32, this is reflected in other claim language, discussed infra at 20.
The court determined that the patents’ self-referential concept could be satisfied by creating a table with a simple header row. But that is simply not the case. For example, step three of the algorithm described above explains that the table stores information related to each column in rows of that very same table, such that new columns can be added by creating new rows in the table. See J.A. 325 (describing four-step algorithm); see also ’604 patent, col. 2 ll. 53–65 (describing “the present invention,” including a description where “columns are entered as rows in the table and the record corresponding to a column contains various information about the column,” thereby “render[ing] the table self-referential”). It is beyond debate that this is more than simply a header row.
Another detail here is that McRo confirms Enfish’s holding that the improvement provided by the specific claim limitations can be considered in step 1, and need not be relegated to step 2. Note the last sentence above: “The specific, claimed features of these rules allow for the improvement realized by the invention.” And as in Enfish, the McRo court looks to the specification for a discussion of this improvement: “As the specification confirms, the claimed improvement here is allowing computers to produce “accurate and realistic lip synchronization and facial expressions in animated characters.” Notice that the phrasing of the improvement is not in terms of technical details but in terms of an improved result.
The next move by the court is the most interesting intellectually, as it brings into step 1 the problem of preemption—and then solves it. The claim recites the use of “a first set of rules that define output morph weight set stream as a function of phoneme sequence and time of said phoneme sequence.” The structure of the claim is prototypical. First, the rules are “obtained”; these rules link morph weights to a “function” (any function!) of two factors: phoneme sequence and time of sequence. Next, the data that the rules will be applied to are retrieved by “obtaining a timed data file of phonemes having a plurality of sub-sequences.” Then there are two generating steps, one which generates intermediate morph weight sets “by evaluating said plurality of sub-sequences against said first set of rules,” and a second which generates “final weight sets.” Last, the final morph weights are applied (in any manner!) to a sequence of characters “to produce lip synchronization and facial expression control of said animated characters.” This is a fairly well-crafted claim: all of the nouns and verbs “do work,” there is nothing extraneous in the claim, and the “logic” is very tight.
But notice that there is no specific limitation of how the morph weights are “functions” of the phoneme sequence. This was the rationale of the district court for why the claims were abstract. To get over that hurdle, that court does something entirely new. First, the court notes “The claimed rules here, however, are limited to rules with certain common characteristics, i.e., a genus.” The court reminds us that claiming a genus can be acceptable depending on the scope of the claim. Then, in what I believe is a first in patent eligibility law, the court expressly links this concept to eligibility:
Claims to the genus of an invention, rather than a particular species, have long been acknowledged as patentable. E.g., Diamond v. Chakrabarty, 447 U.S. 303, 305 (1980) (patentable claim to “a bacterium from the genus Pseudomonas containing therein at least two stable energy-generating plasmids, each of said plasmids providing a separate hydrocarbon degradative pathway”). Patent law has evolved to place additional requirements on patentees seeking to claim a genus; however, these limits have not been in relation to the abstract idea exception to § 101. Rather they have principally been in terms of whether the patentee has satisfied the tradeoff of broad disclosure for broad claim scope implicit in 35 U.S.C. § 112. E.g., Carnegie Mellon Univ. v. Hoffmann-La Roche Inc., 541 F.3d 1115, 1122 (Fed. Cir. 2008). It is self-evident that genus claims create a greater risk of preemption, thus implicating the primary concern driving § 101 jurisprudence, but this does not mean they are unpatentable.
The narrower concern here is whether the claimed genus of rules preempts all techniques for automating 3-D animation that rely on rules.
I believe that this issue of whether genus claims would be inherently “abstract” was the core reason the McRo decision was delayed. As the court notes, the problem that genus claims risk preemption is self-evident. Once it identified the problem, the court had to solve it in order to find the claims eligible.
In a recent blog, I suggested that the court did not have a deep understanding of the nature of technology, in particular, computer technology. Perhaps I spoke too harshly—or perhaps the court simply got lucky here—but either way, the court’s recognition of and ultimate approval of using “rules” at the genus level to create automated processes as patent eligible is extremely important. Let me say that again:
The court’s recognition of and ultimate approval of using “rules” at the genus level to create automated processes as patent eligible is extremely important.
This is because the vast majority of software inventions are based on a framework of domain-specific rules that are applied to domain-specific data to solve a particular problem. First, using programmatic rules is a design approach found universally in software. At a low level, every programming language supports some form of conditional logic, such as “IF…THEN” statements, CASE/SWITCH statements, branching, etc.. These statements are used to control the program operations, as well as to define rules for processing given inputs to produce desired outputs (as well as for other functions). At a higher level, entire areas of software development are based on rule frameworks, such as rule-based systems, including decision support systems, business intelligence tools, CAD design tools, etc. (Yes, there are expressly “non-rule” based programs, like neural networks; but that is only at the high level; the which may be characterized as rules.)
Second, it is the use of programmatic rules that fundamentally enables computerized automation. Without rules that define the operation of a program over a range of relevant inputs, no system would be able to operate “automatically.” Without rules defining the relationships between inputs and outputs, your navigation application would not have given you the best route to work this morning, your hybrid or electric car's ECU (engine control unit) would not have been able to optimize the power and energy delivery, nothing would have controlled the timing of the traffic lights along that route, or scheduled the arrival of your train; your news application would not have provided you with stories of interest to you to read along the way; the elevators in the lobby would not have coordinated to stop at your floor; and you would have been unable to securely log into your computer, Shepardize your cases and spell-checked your brief. In short, everything you take for granted would not have been there for you.
More subtly, rule frameworks are typically assembled in classes or types (i.e., genus), and it is the identification and articulation of the types themselves that are often the key inventive acts. Once the various genera of rules for the domain are invented, the specific rule instances are then created—and many of these may also be inventive in themselves. But what is frequently of most competitive value is the overall architecture of the rule framework (the genus), not the specific rules (the species). Limiting eligibility to specific rules here would have utterly devalued McRo’s inventive contribution. And so too would it have crippled protection for future software.
So how then did the court solve the preemption problem? It circles back to its starting point, the difference between claim results and claiming specific processes, even repeating its quotation from O’Reilly v. Morse, and tying in Enfish (as I suggested above) and Rapid Litigation Management:
The abstract idea exception has been applied to prevent patenting of claims that abstractly cover results where “it matters not by what process or machinery the result is accomplished.” Morse, 56 U.S. at 113; see also Mayo, 132 S. Ct. at 1301. “A patent is not good for an effect, or the result of a certain process” because such patents “would prohibit all other persons from making the same thing by any means whatsoever.” Le Roy v. Tatham, 55 U.S. 156, 175 (1853). A patent may issue “for the means or method of producing a certain result, or effect, and not for the result or effect produced.” Diehr, 450 U.S. 175, 182 n.7. We therefore look to whether the claims in these patents focus on a specific means or method that improves the relevant technology or are instead directed to a result or effect that itself is the abstract idea and merely invoke generic processes and machinery. Enfish, LLC v. Microsoft Corp., 822 F.3d 1327, 1336 (Fed. Cir. 2016) (“Enfish”); see also Rapid Litig. Mgmt. Ltd. v.CellzDirect, Inc., No. 2015-1570, 2016 WL 3606624, at *4 (Fed. Cir. July 5, 2016).
Now the court couples the solution to preemption-by-automation-using-rules with the “directed to” analysis, again putting this analysis under a specific heading to indicate the linkage of the argument:
2. Claims Directed To
Claim 1 of the ’576 patent is focused on a specific asserted improvement in computer animation, i.e., the automatic use of rules of a particular type. We disagree with Defendants’ arguments that the claims simply use a computer as a tool to automate conventional activity. While the rules are embodied in computer software that is processed by general-purpose computers, Defendants provided no evidence that the process previously used by animators is the same as the process required by the claims. See Defs.’ Br. 10–15, 39–40. In support, Defendants point to the background section of the patents, but that information makes no suggestion that animators were previously employing the type of rules required by claim 1. The prior art “animator would decide what the animated face should look like at key points in time between the start and end times, and then ‘draw’ the face at those times.” Defs.’ Br. 10. The computer here is employed to perform a distinct process to automate a task previously performed by humans. McRO states that animators would initially set keyframes at the point a phoneme was pronounced to represent the corresponding morph target as a starting point for further fine tuning. J.A. 3573 at 8:53 (McRO’s Claim Construction Presentation). This activity, even if automated by rules, would not be within the scope of the claims because it does not evaluate sub-sequences, generate transition parameters or apply transition parameters to create a final morph weight set. It is the incorporation of the claimed rules, not the use of the computer, that “improved [the] existing technological process” by allowing the automation of further tasks. Alice, 134 S. Ct. at 2358. This is unlike Flook, Bilski, and Alice, where the claimed computer-automated process and the prior method were carried out in the same way. Flook, 437 U.S. at 585–86; Bilski, 561 U.S. at 611; Alice, 134 S. Ct. at 2356.
And several paragraphs later:
The limitations in claim 1 prevent preemption of all processes for achieving automated lip-synchronization of 3-D characters. McRO has demonstrated that motion capture animation provides an alternative process for automatically animating lip synchronization and facial expressions. Even so, we have recognized that “the absence of complete preemption does not demonstrate patent eligibility.” Ariosa Diagnostics, Inc. v. Sequenom, Inc., 788 F.3d 1371, 1379 (Fed. Cir. 2015). The narrower concern here is whether the claimed genus of rules preempts all techniques for automating 3-D animation that rely on rules. Claim 1 requires that the rules be rendered in a specific way: as a relationship between subsequences of phonemes, timing, and the weight to which each phoneme is expressed visually at a particular timing (as represented by the morph weight set). The specific structure of the claimed rules would prevent broad preemption of all rules-based means of automating lip synchronization, unless the limits of the rules themselves are broad enough to cover all possible approaches. There has been no showing that any rules-based lip-synchronization process must use rules with the specifically claimed characteristics.
There are several distinct arguments here that must be teased out. The court starts with the narrower argument that the claims do not merely automate—and thus preempt—what animators have traditionally done: “Defendants provided no evidence that the process previously used by animators is the same as the process required by the claims….[ the prior art process] even if automated by rules, would not be within the scope of the claims because it does not evaluate sub-sequences, generate transition parameters or apply transition parameters to create a final morph weight set.” Here the court is evaluating the actual claim limitations and demonstrating that they do not perform the same steps as traditionally employed. Because of that, the claims are not merely “tools to automate conventional activity.” This formulation finally puts some useful meat on the otherwise naked concept of using computers as “tools,” an otherwise dangerous formulation since by design that’s exactly what computers are.
From this narrower preemption focus, the court moves to the more general one: “preemption of all processes for achieving automated lip-synchronization.” Here, the court again does something very important: it acknowledges and reframes Ariosa. Ariosa seemed to put the kibosh on non-preemption arguments, stating “Where a patent’s claims are deemed only to disclose patent ineligible subject matter under the Mayo framework, as they are in this case, preemption concerns are fully addressed and made moot.” Patent examiners routinely rely on this statement, and ignore any non-preemption argument by saying that preemption is not a separate test (a position reiterated by the U.S. Patent and Trademark Office (USPTO) in the May memo to examiners). The panel here makes clear that a demonstration of meaningful non-preemption is sufficient to establish that a claim is not “directed to” an abstract idea, and thus eligible at step 1, without proceeding on to step 2 of Mayo test. In other words, non-preemption offers another way of doing the “streamlined eligibility analysis” set forth in the USPTO's Interim Guidance.
And this principle is not just stated as an aside, as the court repeats it in various ways:
- “There has been no showing that any [i.e. all] rules-based lip-synchronization process must use rules with the specifically claimed characteristics.”
- “Defendants again rely only on the patents’ description of one type of rules, but the description of one set of rules does not mean that there exists only one set of rules, and does not support the view that other possible types of rules with different characteristics do not exist.”
- “The only information cited to this court about the relationship between speech and face shape points to the conclusion that there are many other possible approaches to automating lip synchronization using rules.”
- “Honda shows, however, that the interaction between vocalization and facial expression is very complex, and there are relationships present other than those required by the claimed rules.”
- “This complex interaction permits development of alternative rules-based methods of animating lip synchronization and facial expressions of three-dimensional characters, such as simulating the muscle action underlying characters’ facial expressions.”
- “By incorporating the specific features of the rules as claim limitations, claim 1 is limited to a specific process for automatically animating characters using particular information and techniques and does not preempt approaches that use rules of a different structure or different techniques.”
- “Under these circumstances, therefore, we need not assume that future alternative discoveries are foreclosed.”
This repeated, extended analysis was certainly not necessary to make the court’s point, but it nonetheless made these (and other statements). There is simply no plausible way to say that preemption does not matter. Given that Judge Reyna wrote both McRo and Ariosa, I think that Ariosa is being narrowed or limited because otherwise the decisions are in conflict. And since McRo provides a substantive analysis of why and how preemption can be used, it deserves priority of Ariosa, which offers no explanation at all.
The McRo court did not just raise preemption as a mode of analysis, they went further: they bound the analysis to evidence and facts.
- “Defendants’ attorney’s argument that any rules-based lip-synchronization process must use the claimed type of rules has appeal, but no record evidence supports this conclusion.”
- “There has been no showing that any rules-based lip-synchronization process must use rules with the specifically claimed characteristics.”
- “Defendants provided no evidence that the process previously used by animators is the same as the process required by the claims. ”
Further, the court made clear that simply relying on the specification itself to show preemption will not always be sufficient:
- “In support, Defendants point to the background section of the patents, but that information makes no suggestion that animators were previously employing the type of rules required by claim 1.”
- “Defendants again rely only on the patents’ description of one type of rules, but the description of one set of rules does not mean that there exists only one set of rules, and does not support the view that other possible types of rules with different characteristics do not exist.”
The focus on evidence outside of the specification to support a preemption argument is a fundamental shift in patent eligibility for two reasons. First, on a Rule 12 motion, just as the patentee is limited to the intrinsic record, so too is the defendant. To date, defendants often make preemption arguments that at best rely on bits and pieces of the specification, and some district courts have adopted such arguments uncritically. The court’s repeated emphasis on the need for evidence of preemption suggests that defendants may find it more difficult to raise this argument on such a motion.
Second, the evidentiary framing further shows that the burden is on the defendant to establish ineligibility. The current zeitgeist is that once a defendant makes the argument of ineligibility, the psychological (though not legal) burden is on the patentee not merely to rebut the argument, but to prove the claim is eligible. This is a result of how the Mayo test is framed: deciding whether the claim is “directed to” an exception—thus orienting the analysis toward the “failure condition”—and then deciding if there is “something more.” Since the first step often disembowels the claim, there’s nothing of the claim left over that could constitute something more, except "patent-ese"-type convoluted language.” (In re TLI Commc’ns). In other words, courts find what they are looking for: tell them to look for ineligible subject matter and they will find it.
Consider where we would be now if the Mayo test had been reversed:
- Step 1: Does the claim recite a useful, practical application? If so, stop, the claim is eligible.
- Step 2: If the answer to Step 1 is no, is there any limitation that is not a law of nature, abstract idea or natural phenomenon? If so, then the claim is eligible. If not, then the claim is ineligible.
If this were the Mayo test, most patents would be found eligible precisely because instead of hunting for abstractions and looking for the elusive transformations, the courts (and the USPTO) would be looking for eligible subject matter—and they would find it.
But we have the Mayo test we have. And this is why I think evaluating preemption in Step 1 is such an important development. It would be expected that preemption be considered in Step 2, to determine if limitations are “meaningful” or “significant.” This is precisely how Judge Lourie, writing the plurality opinion in CLS v. Alice, implemented preemption:
The § 101 inquiry next proceeds to the requisite preemption analysis. With the pertinent abstract idea identified, the balance of the claim can be evaluated to determine whether it contains additional substantive limitations that narrow, confine, or otherwise tie down the claim so that, in practical terms, it does not cover the full abstract idea itself.
While this is analytically appealing, in practice arguing a lack of preemption at step 2 has typically failed to save the claims.
In sum, this opinion makes two major contributions to eligibility. First, the court has established that preemption is not simply a mere conclusory label that can be slapped on a patent claim after the Mayo test is applied. Rather, it is an actual analytical tool that calls for specific types of reasoning and evidence. Second, by coupling the recognition of rules with concept of the genus/species distinction, the decision approves a general pattern by which many software inventions can be claimed: a set of rules that “define” a type of output Y (e.g., morph weights) that are a function of one or more types of inputs X (e.g., phoneme sequence and time). So long as the scope of the claimed types does not preempt all other types to achieve the same result, the claim should be eligible.
But I’ve still not exhausted the richness of McRo. I'll discuss additional insights and lessons in an upcoming blog.
Overall, McRo is a fundamental and important decision, one that I hope the district courts and USPTO will follow closely. I applaud the panel for getting McRo right on every major point of analysis and putting preemption, the core concern of Section 101, squarely back in the analysis.