By: Robert R. Sachs
Having now discussed the underlying clarity issues that the USPTO is seeking to address with its proposal to include glossaries in patent applications and the fundamental linguistic elements that would impede the success of such a requirement, I’ll now discuss some of the problems that would arise in the implementation of patent glossaries.
Identification of Glossary Terms
The very first problem faced by the patent drafter is which words to include in a glossary. Most claims are drafted using terms that are familiar to the drafter and the inventor, and that are by assumption familiar to the examiner as someone expected to be familiar with the relevant technology. Words like computer, memory, module, circuit, logic, database, etc. are commonly used, and patent attorneys expect that examiners understand these terms and that no specific definitions need be provided.
However, it is terms like these that are often the subject of dispute during claim construction, because they are precisely the kind of terms that can have highly varied scope, precisely as technology changes over time. Does a database include a lookup table in memory, or is it limited to a relational database based on Codd’s relational model? Is a circuit limited to an integrated circuit, or does it cover a DSP executing an algorithm? Since it is unlikely that patent counsel would include definitions of such common terms in glossaries, the proposal would have limited, if any, impact on the ambiguities—and arguments—that arise in litigation.
That leaves terms that are specifically constructed for the application, typically “functional” terms that take the form of <function word><structural noun>, such as “determination module” or “identification logic,” or <noun modifier><function word>, such as “user identifier” or “threshold tester.” These are arguably the terms that most need definition and thus the ones that the patent drafter would be most likely to attempt to define as broadly as possible. One problem arises however, the precise contours of these terms are not known at the time the application is drafted—but are worked out between the patent examiner and the patent attorney in view of prior art—paralleling Pinker’s explanation above of how scientists refine the meanings of their lifted metaphorical terms as they better understand the contours of the phenomenon of interest. Further, in some cases, the terms ultimately used in the allowed claims are not employed at all in the initial claims but only arise later on as the patent drafter amends the claim to avoid prior art. If the glossary is fixed when the application is filed, little benefit is gained.
Restrictions on Glossary Definitions
The Office has also proposed that “Alternative definitions for the same claim term that are inconsistent with each other are not permissible.” What does it mean to be “inconsistent with each other?” What if the examiner and the applicant disagree on whether definitions are inconsistent? For example, an applicant may define circuit as both hardware and software (logically consistent, but physically different) while an examiner may consider these inconsistent. Again, this restriction is contrary to the natural use of glossaries and definitions, which routinely list alternative definitions, precisely because words have different meanings, sometimes slight, sometimes significant, depending on the literal and experiential context of their use. To prohibit alternative definitions is to assume that there can be only one possible interpretation of a word regardless of the context of its use. This assumption is simply unreasonable and inconsistent with our use of language.
Amendments to Glossary Definitions
During last month’s Software Partnership Meeting in Berkeley, the question of whether a glossary could be amended during prosecution was discussed. The Office raised the concern that this would introduce the possibility of new matter. However, this should not be a concern. Patent counsel routinely amend their specifications, and they understand the requirements against new matter. Examiners likewise are trained to identify and reject new matter, and can be presumed to do so effectively if such is added to a glossary, either in the form of a new definition that is unsupported, or the revision of an existing definition. Accordingly, if glossaries are to be required, then amendments to the glossaries should be allowed, and treated in the same manner as amendments to the specification.
That said, allowing amendments to a glossary is quite a different thing from requiring amendments. If a patent attorney is required to amend a glossary when a new functional term is added to the claim, this will hamper prosecution. Counsel would be more inclined to either argue claim rejections or amend claims in more obtuse long-winded ways that avoid the introduction of new terms requiring a definition. This is an example of strategic behavior, which I believe is the single most significant impediment to the use of glossaries.
“Rejections” of Glossary Definitions
Related to the question of amendment, was the question as to whether examiners should evaluate glossary definitions and reject those that are insufficient. Let us assume that is the case, and an examiner can reject a definition under § 112 for indefiniteness, in which case the applicant would turn to the rest of the specification to show the meaning of the term. This simply pushes back the problem: if an examiner can reject a glossary definition as indefinite, unsupported, etc. then that same rejection could have been made with respect to the claim. If ultimately the applicant relies on the specification to support a rejected glossary definition, then there is no need for the glossary to begin with. Thus, any issues regarding the meaning of claim terms can be settled in the “old fashioned” way, without relying on glossaries.
I think this point is fundamental and brings into sharp focus why the entire program for glossaries is unnecessary. Any claim term that the examiner believes requires definition can be defined precisely where it belongs: in the claims. And any support that is necessary for a claim term will be found in the specification itself. There is nothing to be gained by having an examiner reject a claim term for indefiniteness or reject its accompanying glossary definition, as a well-founded rejection of the claim achieves all that is necessary: an amendment to clarify the claim and identification of support in the specification. The accompanying glossary definition is simply extra baggage.
And like airline fees, extra baggage costs money: the cost of prosecution will certainly increase if patent counsel has to deal with two separate rejections—of claims and of a glossary—with two sets of amendments, additional arguments, showings of specification support and so forth. Nothing is free in patent prosecution.
Glossaries and definitions typically are not used by inventors in their explanations of their inventions. Except for those whose work demands precise definitions such as mathematicians, inventors do not routinely provide rigorous definitions of the terms describing their inventions, but instead rely on the creative use of language in view of the common experiential context of their community. As a result, the task of manufacturing a glossary will be principally left to patent counsel, and we can expect then definitions that do not fully capture the understanding of the inventor or the relevant knowledge of the community of those of skill in the art. We can also expect predictable responses by patent attorneys to potentially limiting rules, as discussed next.
Strategic Behavior Arising from the Requirement of Glossaries
It may here be replied that all of these objections are beside the point because the burden rests upon the patent drafter to take all of these issues into consideration when formulating the glossary definitions, and to carefully construct the definition that best reflects the scope of the claim that they are entitled to. If the drafter chooses poorly, they must accept the consequences of too narrow a definition, such as non-infringement, or too broad of one, invalidity.
This is trivially true but fails to consider what happens in the real world of patent prosecution—how patent attorneys will likely behave in the face of any set of rules. As a general principle, patent attorneys will seek to maximize flexibility in the meaning and use of particular words, and minimize the likelihood of a narrowing definition. And the more rules or the more restrictive the rules governing the implementation of the glossaries—limits on length of definitions, number of alternatives, etc.—the more strategic the behavior by patent counsel to avoid the impact of such rules. This behavior may be intentional, but need not be: patterns of strategic behavior can arise entirely without a party intentionally trying to manipulate the rules.
For example, if glossaries are optional, then some attorneys will simply avoid them entirely. Consider the practice of many attorneys to simply dispense with a summary of invention. This arose in response to a trend in Federal Circuit cases to construe claims based primarily on how the invention was described in the summary. Rather than adapt to these cases and draft a summary in broad and non-limiting terms, these practitioners simply eliminated what they perceived as the source of risk. Other drafters took a different route: they simply repeated the independent claims in prose form in the summary (a practice that is all too common today—and whose practitioners often have no idea why they do it, they were simply taught this style). In either case, zero useful information is provided to the examiner or any other reader.
If the only use of a glossary is by the examiner and the court to narrow the scope of the claim—which must be assumed—then there is no reason a patent attorney would include one absent compulsion. Obviously, no benefit is obtained by either the examiner or the public for understanding the claims if no glossary is provided. “Incentives” such as faster examination, or other procedural benefits are simply the patent equivalent of the adage “a moment on the lips [the glossary], forever on the hips [the claim scope].” Or to put it more directly, glossaries are perfect subjects for a Miranda warning: “Everything you say can and will be used against you in a court of law.”
Let us assume then that the Office requires glossaries in future applications. We can predict a number of different approaches by patent drafters to minimize their impact. First, it is likely that any glossary will have a minimal number of entries. One way this will be achieved is by avoiding the use of specially defined terms, relying perhaps on more long-winded and complex expressions in place of shorter more succinct terms or expressions. For example, instead of using the technical term encoder in a claim and having to provide a glossary definition thereof, counsel may instead use the more cumbersome digital information format conversion module, relying on the customary meanings of the individual words or providing general definitions for each of these words. No additional clarity is gained by this approach, but the risk of a narrowing definition is arguably avoided.
Another approach would be to follow the claims-as-summary-of-invention model: entries in the glossary will simply be taken directly from the patent claims themselves, preserving as carefully as possible the exact phrasing and limitations of any particular elements. For example, consider a method claim for encoding an audio bitstream, with the following step:
receiving the audio bitstream, the audio bitstream containing encoded data that represents audio content of an audio input signal, wherein the audio bitstream also contains encoding parameters generated by an audio encoder when encoding the audio input signal to produce the audio bitstream…
The glossary entry for “audio bitstream” could be simply:
An audio bitstream is a bitstream that contains encoded data that represents audio content of an audio input signal, and also contains encoding parameters generated by an audio encoder when encoding the audio input signal to produce the audio bitstream.
The net result is that no useful information is provided to the examiner, but again the drafter has minimized the risk of a narrowing claim interpretation. In short, rules mandating glossaries with particular features will only further turn patent attorneys into “clever draftsmen,” precisely the opposite of the intended goal.
USPTO Pilot Program
During the Berkeley Roundtable, the Office discussed an initial study it did of some 72 patents having glossaries to determine whether there were any identifiable metrics to support the argument that glossaries were beneficial. The metrics included a reduction in the number of false positive rejections (rejections that should not have been made) and false negative rejections (rejections that should have been made but were not), as well as a reduction in the number of office actions. The results were inconclusive, leading the Office to propose a pilot program whereby applicants can choose to participate and include a glossary. Apparently, the Office intends to again evaluate these and similar prosecution metrics to determine whether glossaries are beneficial.
This approach is perhaps interesting, but it is both irrelevant and flawed. This impact of glossaries is not a “data driven” problem, for the reasons I have discussed above; the issues are systemic, resulting from the nature of language and the behavior of individuals in complex, rule bound systems. Similarly, surveys of examiners for their opinions (even with management not present) as to whether they think glossaries would be helpful are simply not relevant: examiners, though critical to the prosecution process, are not the ultimate “consumers” of patents. Of course we would expect examiners to support any change that makes, or appears to make, their work easier and quicker, regardless of the long-term consequences of the change, or whether that change in fact improves the “quality” of patents. The simple fact is that the majority of examiners have no knowledge of, or experience in, the legal and practical issues that arise during patent litigation (e.g., claim construction, prosecution history estoppel, etc.), or how patents are licensed or used in technology transactions—all of which can be critically impacted by the scope of claims, and thus by extension, by the presence of glossaries.
Second, the proposed metrics—or any metrics that derive from the prosecution process—are simply measuring the wrong thing. What matters is not the number of office actions, how many rejections were in error, etc. What matters is what happens to patents after they are granted. Ideally, one would want to know whether patents with glossaries are less frequently litigated (i.e., because the scope of the claims was more readily determined by the parties), resulted in larger royalties or licensing fees, or had other indicia of improved value. Of course, these questions are years away from being capable of determination, and largely outside the reach of the Office, and so naturally the Office has decided to measure attributes more under its immediate control and within its purview.
But the measurement problem remains—just because the Office can measure internal prosecution metrics does not mean that it should or that they are meaningful. Let us assume that the pilot program demonstrates that patents with glossaries have a shorter pendency and fewer rejections errors (false positives and false negatives). Does this tell us anything about whether the patent claims are actually less ambiguous in the “real world”? It does not—precisely for the many reasons I set forth above. If examiners tend to allow patents more quickly because they see a glossary of claim terms and think that the claims will necessarily be limited by the definitions regardless of what they actually cover—and if experience is any guide, this is precisely what some examiners will do—the resulting claims are just as likely to be broad and potentially vague as patents without glossaries.
The response to this argument is that I’m assuming examiners are not always diligent, and that’s not the case in practice and certainly not in the pilot program. Let us assume then that the examiners in the pilot program are extremely conscientious about reviewing the glossary and making sure all terms are well defined, etc., setting aside all of the issues I have discussed. Let us assume that the metrics demonstrate “better” outcomes, however measured. These results still cannot be meaningfully used to extrapolate the impact of glossaries’ rules on patent applicants in general. This is because the participants in the pilot program are self selected—they do not represent applicants operating under required rules, but instead those who choose to be in the pilot program. As a result, they are more likely to pay particular attention to carefully crafting their glossary definitions, to be motivated to provide complete and detailed responses to any objections or rejections, and to respond to all other aspects of how the glossaries impact prosecution in a non-representative fashion. Self selection inherently results in a form of sampling bias called “self-selection bias,” in which the decision of certain individuals (here, patent applicants) to participate in a study is correlated with the behaviors being studied. Short of randomly sampling applicants and requiring them to provide a glossary (which obviously the Office cannot do) the problem of sampling bias cannot be eliminated in this pilot program, though it can be minimized by careful study design.
There is a third major problem with the pilot program, at least as it was described at the Berkeley Roundtable. What is the relevant control group? To decide whether glossaries make a difference in prosecution, one has to ask: compared to what? It is not sufficient for the Office to simply compare its prosecution metrics to known metrics that it currently has on hand—average pendency, error rates, etc.—because the population from which these known metrics were derived is by definition not similar to the population of the pilot program (self-selected as it may be). Instead, the Office has to create a control group of applications that is sample matched to the applications in the pilot program. Sample matching here would be along the lines of selecting one or more applications from the same art unit, classification and examiner for every pilot program application. Ideally, there would be multiple matching samples for each pilot program application, and the metrics for these samples would be compared to the corresponding metrics of the pilot applications to determine whether the pilot applications have statistically significant differences in metrics. In short, the Office should consult an expert statistician to assist in the pilot program design.
The use of glossaries or other forms of definitions implicates fundamental issues in the nature of language and meaning, and I have only touched upon a few of these in my letter to the USPTO. The Office, for all of its experience in the examination of patents, does not have deep expertise in these areas. To further understand the issues I have set forth, and to develop a theoretically sound and practically effective approach, the Office should consult with experts in linguistics and philosophy of language. The Office’s visit to Berkeley to discuss the use of glossaries was unintentionally felicitous, as some of the best linguists, philosophers and other scholars of language are resident there: George Lakoff, Chair of Berkeley’s Department of Linguistics; John Searle, Philosophy; and Geoffery Nunberg, School of Information, to name a few. It is my sincere hope that the USPTO will seek advice from such experts before mandating the use of glossaries in patents. Like Humpty Dumpty, all the king’s horses and all the king’s men won’t be able to fix the problem once it falls apart.