This anecdotal evidence is supported by recent more-thorough examinations, some described by Matt Levy of Patent Progress. Michael Risch of Villanova recently published a study showing that of 105 patents whose validity was examined in litigation, 44 were completely invalidated, and only 32 were found to have all valid claims (A Generation of Patent Litigation..", page 38) . Dennis Crouch of the University of Missouri has estimated the allowance rate for applications (when "Requests for Continuing Examination" are included) as about 70% (USPTO Breaks New Ground...) - that is, pretty much everybody who applies gets a patent, though sometimes it takes a while.
Are we to conclude, as Mr. Aharonian of the Internet Patent News email service often seems to advocate, that patent examiners are idiots and the USPTO is run by slimeballs? While the unfortunate limitations of human nature may play a role in the failings of the patent system, in this post we will suggest that the nature of patent prosecution inevitably leads to the often-silly and sometimes harmful results we observe. This brief discussion will then serve as motivation for the changes in the system we will examine in subsequent missives.
A patent consists of an abstract, a specification, drawings, and claims. But when a patent is examined, it is often the case that only the claims play a significant role. (An attorney colleague of mine believes that most examiners don't read anything but the claims, and the specification might as well be skipped.) This is curious, because, as noted by Professor Feldman of UC Berkeley in Rethinking Patent Law*, no one knows what the claims of a patent actually mean until it is litigated.
The claims are supposed to clearly describe what the applicant "claims" to have invented and therefore owns. Claims are usually fairly brief - a quarter of a page or less, often much less - and have a few "elements" which are supposed to describe an invention. An examiner responds to an application by citing one or more pieces of "prior art" (stuff that was published before the application was filed) that he or she alleges contain the purported inventive elements. It is often the case that only one bit of prior art is used; it is rare for more than four or five references to be cited. The applicant then argues that the prior art is not, in fact, relevant, and may also amend the language used in the claims to narrow their scope - although, as noted above, it's never clear what the practical effect is until you're in court.
The cited prior art essentially consists only of patents and patent applications, unless the applicant has provided publications from other sources. Sometimes the cited prior art just flat out has what the applicant thought of. One would think this should result in immediate abandonment of the application - but in fact, many applicants simply find a way to narrow the claims until something gets granted. Sometimes the cited prior art is so bizarrely irrelevant (at least to a person knowledgeable in the field) that one can only assume it is the product of a keyword search with no intelligence applied. But in general we can treat the cited prior art as roughly being taken at random from generally related earlier patents, in the sense that most topics have hundreds or thousands of applications and granted patents, and no one would have the ability or time to read and absorb a substantial fraction of them. And that leaves out all the technical literature and commercial product information that ought to be included but just isn't.
We can thus consider what happens if you randomly sample a complex object. If you actually design something, your code, or chip, or medical device, is unique. It always has a slightly different set of tradeoffs, guesses, wants, and compromises from other examples of similar stuff. So in the sense of not being identical to any single publication or object, every application is inventive, and that's what we see: most applications become granted patents. The applicant can always find a way to claim their stuff that's not quite the same as whatever art was cited against them. This might be ok if the system we used was primarily oriented to preventing exact copies of the patented object, but that's not how our existing system works. (In a future post we will examine some thoughts on how to usefully move towards such an approach to defining what is excluded by a patent right.)
What does not happen is what a practitioner in the field does with a technical publication: read the specification to figure out what the application thinks it has that is new, and then reflect upon the art to figure out if it is, independently of what the applicant thinks they'd like to own. The way the world should be is: If there is no invention in the specification, there can't be an invention in the claims. But that's not how the world actually works.
In summary, patent examination is mainly an exercise in the wordsmithing of claims, not a reflection on the novelty, clarity, and utility (or lack thereof) of a purported invention. It is this fact that explains why the results are often silly and sometimes absurd.
People are people and don't readily modify their behavior, at least once they get old enough to hold a job. What can we do to modify the system to make it work better with the people we have? In the next few posts I'll propose a number of approaches, some old, some perhaps new, to accomplish this end.
*Full disclosure: Professor Feldman's book, as I've noted in a review on Amazon, isn't really very good, at least for anyone who is not a law professor. But she does seem to know the legal side of the system quite well, so I am accepting her view on the issue of claims interpretation.