[Sabrina Safrin (guest-blogging), March 5, 2007 at 6:56am] Trackbacks
Can Property Rights Beget Additional Property Rights?

Thank you Eugene for inviting me to guest blog about Chain Reaction: How Property Begets Property. There I argue that intellectual and other property rights can arise in reaction to each other. These reactive property rights may have little to do with the value of the resource in question or efficiency concerns.

The chain reaction evolution of property helps to explain the emergence of more restrictive intellectual and other property regimes. It does not, however, explain the transformation of all property regimes or serve as the exclusive explanation for the process through which all property rights evolve. The theory cautions that once property rights are created, they take on a life of their own and can have adverse consequences.

Over the next several days, my posts will explore examples of and reasons for reactive propertization. My final post will discuss some ramifications of the chain reaction evolution of property rights.

Let us consider one of the most perplexing phenomena of today's patent activity -- the so-called patent paradox. Patent activity has risen dramatically in the last two decades even though paradoxically the expected value of individual patents has diminished. Patent filings generally rose by about 40% between 1993 and 2003. In addition, patent intensity - the measure of patents obtained per research and development dollar - nearly doubled from the mid-1980's to the late 1990s. Meanwhile, empirical research indicates the low average expected value of the overwhelming majority of patents. Studies set the average value of patents at considerably below their average acquisition costs. A paper by Mark Lemley estimates that only 1.5% of patents are litigated, of which courts deem almost half invalid, and only a small additional number are licensed for royalty (as opposed to cross-licensed). A 2005 study by Judge Kimberly Moore shows that most patentees view their patents to hold so little value that they let them lapse before their term rather than pay the periodic maintenance fees.

Why do so many seek patents despite their apparent low expected individual value?

Scholars have offered several excellent theories. Clarisa Long suggests that patents serve a signaling function. They provide information about the invention or firm to, for example, potential investors. Some argue that patents act as lottery tickets. Individuals and corporations obtain patents hoping that one will turn into a winning ticket. Because they do not know in advance which of their patents will ultimately prove the winner, they patent everything. Richard Levin and others have suggested that patents provide a means to measure employee performance. University consideration of the number of patents awarded to individual professors in promotion and tenure decisions exemplifies such use. Meanwhile, Gideon Parchomovsky and Polk Wagner advance a patent portfolio theory. They argue that patents of little individual worth become valuable when bundled together.

Each of these theories helps explain the patent paradox. I believe, however, that something else is going on too. Patent activity itself begets additional patent activity. If we apply insights from those who study group behavior to the patent paradox puzzle, we see that people and corporations may be seeking patents because others have done so.

Group behavior theorists John Miller and Scott Page recently tackled the standing ovation problem. They summarize the problem as follows: A theater performance ends. The audience begins to applaud tentatively. A few audience members stand. "Does a standing ovation ensue or does the enthusiasm fizzle?" Using computational models, They found that the system often converged on the wrong equilibrium. Most people stood even though most did not like the performance. They also discovered that greater pressure to conform led to less efficient aggregation of information. In addition, they found that people sitting in the front had a large impact as almost everyone patterned their behavior off them.

The 1990's frenzy to obtain patent rights over genetic fragments illustrates this copycat behavior. In June of 1991, Dr. Craig Venter, on behalf of the National Institutes of Health (NIH), applied for patents on some 2,700 gene fragments of unknown function that he had sequenced using automated sequencing methods. NIH's attempt to patent and hence control a large quantity of genetic material whose function it had not identified was unprecedented. Academics and industry groups immediately and harshly denounced NIH's action, and uncertainty existed as to whether the PTO would even issue patents on such gene fragments.

Despite these criticisms, legal uncertainty and the enormous expense of preparing and filing patent applications, once word of NIH's applications got out, the patent lemmings began their march. Applications covering hundreds of thousands of genetic fragments began to pour into the PTO -- confronting the office with a 90-year backlog. As a leader in the scientific community, NIH served the societal function of a front row theater-goer standing to applaud. Other researchers and institutions followed its lead.

The increasingly widespread use of defensive patenting also evidences a chain reaction dynamic. Corporations and individuals obtain patents not because of the patents' potential positive value, such as their ability to generate license revenue or to provide a manufacturer with a competitive edge, but rather for maintenance in a patent arsenal. If threatened with a lawsuit, the corporation threatens to counter-sue with one of its warehoused patents. The suits go away, and each corporation returns to business as usual.

In this defensive patenting environment, people patent because others in their field are doing so. Commentators consistently liken the situation to an arms race, the quintessential example of a wasteful tit-for-tat, rather than to an enterprise designed to promote innovation by capturing the actual or the potential value of technological advances.

Have others in the blogsphere come across examples of copycat or reactive propertization?

AppSocRes (mail):
It might be worthwhile to mention in passing the impact of Jerome Lemelson's successes with "submarine patents" on current patent law legislation and litigation.
3.5.2007 8:30am
Two words: software patents. The rise of patents on software as business methods (which require no implementation details (like source code)) has been staggering. Large computer corporations apply for thousands (or tens of thousands, IBM I'm looking at you) of these a year. They mostly use them defensively to hold off software patent sutis from other companies. However, when a single, broadly used patent gets into the hands of a small company with no assests or profits, suits tend to proliferate (see Forgent and JPEG for example).
3.5.2007 8:47am
Dennis Crouch (mail) (www):
Sabrina, Great post and article. Can we simply call this "lemming theory"?
3.5.2007 8:58am
Bruce Hayden (mail) (www):
I don't think that too much should be read into the fact that a distinct majority, if not most, patents are not renewed to their full term. Maintenance fees are on a fairly steep curve upwards, for a large entity: $900 @ 3.5 years; $2,300 @ 7.5 years; and $3,800 @ 11.5 years (half that for small entities). On the other hand, the expected utility of most patents is in the other direction: most often by the 11.5 year mark it is obvious that the patent will not pay off, and often even at the 7.5 year mark. So, no surprise that in a corporate setting, where large portfolios are being managed, the first renewal is almost automatic, while the last one is often reserved for those patents that have actually made money.

As to the effect of patenting the genome - that was only one industry. It really didn't have a notable effect on other art areas, esp. those far away from that. In the computer, som electronics, and software art areas, the 900 pound gorilla was IBM. Every year it would file even more applications to keep bragging rights to the most patents issued a year, and every year its take from its patent portfolio would increase. My memory is that by the end of the 1990s, it was making billions each year. Some other U.S. companies were playing the same game, and making increasing returns from their portfolios too.

Meanwhile, the Japanese (and later Koreans, Taiwanese, etc.) were paying for a lot of this through cross-licensing. So, by the early to mid 90s, they were patenting anything that they could think of to build their portfolios. When I was dealing with them in the 1990s, their patents were extremely thin. So, at least in much of the 1990s, a lot of the arms race at the top was between U.S. firms that were making a lot of cross licensing money from their foreign competitors, and their Asian competitors trying to build portfolios to reduce this.

Not that U.S. firms were always on the dominent end of this - up through much of the 1990s, Microsoft was one of the biggest patent targets, and paid huge amounts of money to cross license patents simply because it hadn't bothered to file for patents early on. Even by the mid 1990s, MSFT only had a handful of patent attorneys working for the company, as compared to over a hundred at Motorola and hundreds at IBM.

Back to IBM, what I remember well were all the IBM based metrics I had to deal with: how many patents per milllion $$ sales, how many patents per million $$ of R&D, all based on what IBM was doing. The idea seemed to be that if you weren't patenting at IBM's rate, you would be vulnerable.
3.5.2007 9:03am
mrshl (www):
Does the copycat theory include firms that aggressively patent in order to benefit from or attract cross-licensing agreements? Or is this semi-rational lottery behavior?
3.5.2007 9:04am
Well, from the standpoint of this sometime 'garage inventor', patent law became moot when the patenting process was priced out of reach for average people. Basically, a patent application is a ten-thousand dollar bet that the government will someday issue you a license to try to outspend General Motors on lawyers. Why bother?
3.5.2007 9:09am
Bruce Hayden (mail) (www):

I guess since I make a lot of my income off of software patents, I have to defend them a bit. Even into the early 1990s, the USPTO was still maintaining the illusion that software, per se, was somehow nonstatutory subject matter as being more akin to mathematical algorithms than to engineering. Never mind the tens of billions going into software development even then. So, in the early 1990s, I still found myself filing wiring diagrams for software algorithms. Ultimately, the Federal Circuit pointed out the falacy in this line of thinking by the USPTO (though the agency seems to be trying to reimpose this nonsensical definition again).

Nevertheless, 35 U.S.C. 112 paragraph 1 calls for three things: that the inventor's best mode be disclosed; that the disclosure be enabling; and that the written description match the claims (the later, while always there, became much better developed thanks to Mr. Lemelson, who had claims based on bar code reading based on two 1950s era disclosures). The important one here though is the "enablement" requirement. In order to pass this section of the Patent Act, the disclosure has to enable an experienced practitioner in the relevant art to practice the invention w/o undue experimentation.

So, yes, there are a lot of patents being issued that do not have enabling disclosures, since patent examiners rarely read the specifications very thoroughly, and sometimes not thoroughly at all. But all that means is that if the patents w/o enabling disclosures are ever litigated, they will be most likely be invalidated under 35 U.S.C. 112 p 1.

But that brings us to the other side of this - the in terrorem effect of issued patents. Jerome Lemelson made hundreds of millions of dollars asserting and often starting (but never completing) litigation with what many, if not most, considered blatently invalid patents. Not only were there the "Written Description" problems noted above, but there was also a lot of allegations of hanky-panky early on in their prosecution. Indeed, the entire way the patents were granted was suspicious, as Mr. Lemelson seemed to keep beating on one examiner after another over the first two decades until the cases were finally transferred to a weaker examiner, who finally gave in and let them issue, opening the flood gates.

And that is really the problem even to this day, at least in the electronics and software areas where I practice. Take a company with $10 billion in sales, and a weak patent asserted against that revenue stream. If a reasonable royalty is 1%, that is $100 million in potential damages. And if the patentee is likely to win 10% of the time, that is $10 million - a year. So, companies are likely to pay off if the offer is $5 million a year, or even better, a $10 million one time payment. Compounding this is the nonlinear shape of the risk curve - that $5 million a year is budgetable, but a $100 million dollar hit (a year, retroactive to being given notice) is the type of item that has to be disclosed on SEC filings.
3.5.2007 9:30am
Bruce Hayden (mail) (www):

Maybe benefit from cross licensing agreements, but not attract. Inevitably, it is the party with the stronger patent portfolio that is the agressor in cross-licensing.

But it isn't just about quantity, but also about quality. In my experience in the electronics and software industry, the way it usually works is that both sides will initially agree to assert a given number of patents, often based on the size of their respective patent portfolios. But just because IBM has 10 times the patents doesn't mean that it gets to assert 10 times the patents, because it may have 8 times the revenue stream too.

Then, each side may assert, for example, 20 patents our of the thousands they have in that area. The other side goes through each patent asserted against them and tries to knock it out, usually under 35 U.S.C. 102 (lack of novelty), since the goal is clarity and nonobviousness (35 U.S.C. 103) can be argued forever. And for those that they can't knock out, they look at how many products are infringed by it. And that is the usual basis for negotiations about who owes whom how much.

If this sounds like a big endeaver, it is. Man years are invested on both sides for the bigger cross-licensing deals - by that makes since since we are often talking tens of millions of dollars.

The dynamic is quite different though if it does not appear that the two companies can cross-block each other. Thus, the small company or sole inventor who may be able to block Microsoft doesn't have to worry about its thousands of patents because the revenues from the patent being asserted would be much greater than any loss that MSFT might be able to cause through blocking of its own.
3.5.2007 9:47am
Rattan (mail):
Bruce has made many interesting points. It should be noted that even if patents are not maintained, to the economy the very act of publishing patents and applications develops a repository of information that is quite unique. I recall a foreign technology expert quite taken aback by the public access to what he considered to be trade secrets-the crucial ingredient for Western dominance. It took some explaining before he realized that translating a patent into actual technology calls upon the skills of one of ordinary skill in the art. And you need many of them. Once you have many of them, you already have a developed economy and you are back to applying for your patents to protect the output of your trained workforce.

One point that Bruce has not commented upon has to do with the conflict between the Patent Office and the controlling case law.

Although the Federal Circuit, which hears most patent appeals, has reinforced patent rights, for instance, by requiring that clear and convincing evidence be provided for invalidating a patent, the Patent Office has not devoted similar resources or attention to justify this presumption in most cases. This inconsistency is seen in the high rate at which patents are invalidated by courts when litigated.

The courts have yet to find a way to control the Patent Office. For cost reasons probablly, the patent office does not search for art beyond other patents in most instances. This makes it relatively easy to patent stuff that is not patented, but otherwise known. The reliance on textual analysis makes it possible to patent stuff that is merely described differently- an outcome that is predictable. The methodology adopted by the Patent Office is at variance with the manner in which claims are interpreted in litigation when all the scales due to different definitions fall away to reveal the true scope.

Courts treat claim interpretation by the Patent Office with considerable deference. The Patent Office rarely uses this to provide express claim interpretations, although its rules allow it to do so. Further, it does not explore the full scope of the claims. Thus, although possible in practice, it is rare to find good estopple based on the file history.

This inattention results in surprisingly broad patent claims issuing with attendent bad publicity. The Lemelson patents are a good example of a situation in which the Courts had two bad choices- invalidate the patents and get rid of the clear and convincing standard, or- due to procedural presumption based constraints imposed upon themselves- allow meritless claims morph into damages (that added upto more than a billion dollars). They chose the latter course with no remonstrations on why were such patents issued in the first place. This kind of a result is common in the presumption-based highly textual legal analysis model favored currently in other areas as well. Some examples are felony cases and public defenders, cases involving classified information, and the like. With style winning over substance due to shift in presumptions the adopted procedure is often too clever by half for real world problems.

Many of these issues are finally being addressed, but with a strange reserve. Most patents are issued without tough negotiations with reliance primarily placed on the presumption of validity and the cost of resisting/defining the vague scope of the patent claims.

The second point I think should be made has to do with empirical measures of patent quality. Unfortunately, few clients and researchers have been grading prosecution with an eye on effectiveness of patents in assertion to allow patents in general to become more valuable as a commodity. Cost of prosecution is a major focus, but it is rarely correlated effectiveness of the perfected rights. Thus cost of prosecution has been falling, the number of issued patents increasing, and with patents difficult to rate as a property, they are paradoxically good as collateral and for defensive use if obtained in large numbers.

While big ticket litigation commonly receives much attention, it is rare to hear the name of the prosecutors or firms who are successful in obtaining particularly useful patents.

Thus, if possible, a small inventor should get patents as they are their best trump card in negotiations. These patents should be good and focussed on enforcement, which means higher costs and attention paid during prosecution. Small firms develop a patent portfolio to stave off pressures by other firms and to be able to counter-sue. They also obtain patents because memorializing their art in the form of claims is far more effective than as internal memoranda, published articles and the like because such descriptions are (presently) easily distinguished in litigation. For large companies, patents are important because absent such appropriate memorialization of their art, they can be leveraged even more as happened to RIM recently. In this form, extensive patents are more than just assertable rights, they are the preferred form of memorializing broad technical contributions suitable for legal proceedings. Equivalent non-patent art is difficult to obtain, often destroyed by companies themselves as part of their document retention policies, and construed narrowly by the courts to boot.
3.5.2007 11:45am
markm (mail):
"Basically, a patent application is a ten-thousand dollar bet that the government will someday issue you a license to try to outspend General Motors on lawyers. Why bother?"

Developing an idea to the point that it can be produced in large quantities and marketed usually costs hundreds of thousands. Spending ten thousand to document this work is cheap. The only problem is, somewhere between the lawyers and the PTO, it has become customary to use a style that makes the patent documentation close to worthless...
3.5.2007 12:28pm
Bruce Hayden (mail) (www):
For a long time, the problem with the USPTO and esp. software patents was that they were rejecting them as non-statutory when the reality was that they didn't want to deal with them. The result was that they didn't build up the database of prior art that they should have had. Most recently though, they seem to be going back to their position that many software patents are nonstatutory for some bogus reason or another, despite fairly strong precedents to the contrary by the Federal Circuit.

But this is just the tip of the iceberg with that agency right now. Their backlog has gotten out of hand, and instead of really addressing the problem, which they may not be able to do due to funding reasons (they are supposed to be self-funding, but that doesn't work if Congress steals a lot of their fees). In any case, we are currently at a place where examiners routinely violate USPTO regulations (i.e. the MPEP) in giving final rejections based on new art or ignoring amendments, leaving the applicants to file for continuing prosecution (RCE), necessitating a new filing fee. But despite that, they are still in trouble, so their latest ploy is a proposal to limit the number of continuations, divisions, and claims. But most of those continuations are a result of examiners ignoring the rules in the first place. Nevertheless, the patent community is almost unanimous in their opposition to this plan, which is why the agency announced that it was proceeding to the rule making to put them into place. Of course, since continuations would now be limited, the typical response will now be to file an appeal with the Board of Patent Appeals &Interferences, which is already overloaded to the extent that the agency now has preappeal procedures in place that short circuit the majority of appeals in favor of the applicants before they get to the Board.

In other words, the agency is a mess. A lot of us hated it when Clinton put a gay copyright attorney in charge of the agency to make diversity points, but looking back, he was a much better director it seems than the current one, despite not knowing anything about patents.
3.5.2007 12:49pm
Bruce Hayden (mail) (www):
FYI, the prior art is out there for software patent applications, it just isn't easy to search and is often not available in the form of an issued patent or published patent application. Nevertheless, the ACM, IEEE, and other groups over the years have published a lot of prior art. It just isn't cheap or easy to research.

But if you really want the prior art for software inventions searched, one of the best resources is Greg Aharonian of He also has a free Patenting Newsletter that comes out every couple of days where he rants about the USPTO, but does keep us up to date on what is happening with the agency.
3.5.2007 1:02pm
The Original TS (mail):

Thanks for posting here. I'm quite looking forward to it, even though I have yet to do the required reading . . .

The theory cautions that once property rights are created, they take on a life of their own and can have adverse consequences.

I think this is very true. It's been particularly obvious in the trademark arena. In a way, it's almost an obvious corrolary to granting property rights. Just as someone with enforceable rights in a piece of real estate will tend to take better care of it, people with enforceable IP rights will tend to try and expand their scope and value. People pushing IP rights in court (or in Congress), such as trademarks, typically have far more to gain than any individual defendant. Inevitably, this pushes both judges and legislators towards granting more expansive rights even though those rights are demonstrably inefficient and completely contradict the orginal basis for recognizing the IP right in the first place. I hope you'll be blogging on this topic this week.
3.5.2007 1:09pm
Ken Arromdee:
Developing an idea to the point that it can be produced in large quantities and marketed usually costs hundreds of thousands. Spending ten thousand to document this work is cheap.

This is not, of course, true when the "idea" is computer software, some of which is given away for free.
3.5.2007 2:01pm

Developing an idea to the point that it can be produced in large quantities and marketed usually costs hundreds of thousands. Spending ten thousand to document this work is cheap.

Last Fall I put together 'an improvement to lawn tractors' which is probably novel and certainly useful. It would make a manufacturer's lawn tractors significantly more attractive to buyers and could be implemented for essentially no cost. It cost me maybe fifty dollars to make a practical working model and, trust me, "spending ten thousand to document this work" is not cheap!
3.5.2007 2:09pm
logicnazi (mail) (www):
Bruce Hayden,

In what way is the claim that software is a mathematical algorithm false?

The best point on issue is that the Curry-Howard isomorphism shows that proofs and computer programs can be naturally identified. In other words any programing language is just another syntax in which you can write proofs. Every program behavior (function from input to output) naturally corresponds to a mathematical statement and the body of function calls and procedure calls constitute a proof of this statement. Since it is easy to view any program as a function (yes, even interactive ones I'll explain if you want) this tells us that programs and proofs really are the same thing.

If this is too abstract and mathematical for you I will offer the (less compelling) but easier to understand argument. Consider a classical example of a mathematical algorithm that is clearly unpatentable, the euclidean algorithm. The easiest way to describe the way this algorithm works is as follows.

To calculate the gcd of two numbers x,y
Check if y is 0 if so then stop the answer is x.
If y is not 0 repeat the process with (y, rem x/y)

Frankly I don't see almost any substantial difference between this description which is pretty much the way you would find it written in any mathematical textbook and describing it as (thanks wikipedia):

function gcd(a, b)
if b = 0 return a
else return gcd(b, a mod b)

There is no way you can claim that one of these is patentable and the other is not. Certainly anyone who knows how to read mathematics can easily translate between a description given in the above form (even if very long and complex) and the same thing written in a real computer language.

Now the gcd algorithm is a bit simple but the area of math I work in (recursion theory) routinely proves by literally describing a program. Sure we don't write out function calls or use other syntax but it is a plain language description of a program. It would be a pain to read but if you took one of these constructions and actually coded it up in a programming language the practitioners in my field would accept it as the hard part of the proof (the easy part usually being showing it works).

However, this is clearly a problem for you. What I do is almost as pure mathematics as you can get with the results all being about infinite objects and totally useless in practice (we are only interested in what the program would produce if you let it run forever). These results are published in journals and treated no differently than some construction in group theory. Yet as most of the proof is just a description of what a program does you can't claim that one is patentable and the other is not.

Your best approach might be to try and argue that proofs and descriptions of programs aren't patentable because they don't 'do' anything they just describe how something can be done. Yet this doesn't give you the result you want as it would say that one could patent the gcd algorithm and that while descriptions of it wouldn't be covered by the patent all implementations of it would be. Avoiding a patent on the gcd algorithm so anyone who wants to compute a gcd on a computer has to pay up is exactly what the no mathematical algorithms rule is designed to prevent.

Besides this distinction doesn't even hold up. Surely rigorous formal proofs still count as proofs and there are some nice computer programs (like Coq) for manipulating formal proofs on computers. These programs read in the formal proof and do specific tasks based on the content of the proof (check it is valid, try to fill in gaps etc..). Yet this is not different in kind than what an interpreter does for a computer language.


In short there is pretty clearly no principled distinction that can be drawn between mathematical algorithms/proofs and computer programs. Now there are some reasonable policy and economic arguments that might (I'm not yet convinced) establish that we should just stipulate some totally arbitrary boundary. Say just decide if it would be interesting to a working mathematician. However, even here I don't think this can work because of areas like computer graphics that blur the line between software patent and algorithm patent.
3.5.2007 2:23pm
Sabrina Safrin (mail):
I just wanted to chime in on a few of the blogs.

- Dennis, thank you for your kind words. I don't think we can call this the "lemming theory" because, in addition to copycat behavior, other factors, such as a breach of cooperative norm and fear of exclusion, contribute to reactive propertization. (I will discuss these over the next few days so stay tuned.)

- I agree with those who say that software patents have played a major role in the "patent paradox." But it is not the only industry implicated. The semiconductor industry, the computer hardware sector as well as the biotechnology sector all appear prone to defensive patenting. A sector here, a sector there - pretty soon we are talking about a sizable percentage of issued patents.

I suspect that we will see an increasing number of business method patents flowing through the system too as copycat behavior works its way through financial firms, and words like, "you should patent that" increasingly are heard around Wall Street water coolers.

- I note that I have not seen Rattan's point about patents serving as a preferred form of memorializing technological contributions in the extensive patent paradox literature. Seems like a fresh and intriguing insight.

- Mrshl queried whether the copycat theory included cross-licensing arrangements and Bruce provided a helpful snapshot of the dynamic around the negotiating table. In terms of the literature, Lemley found that less than 5% of patents generated royalties or were litigated so this data would seem to leave out cross-licensing arrangements.

I think cross-licensing should be divided into 2 catagories:

1) Cross-licensing to obtain necessary technologies. Firms who swap useful patented technologies with each other are obtaining something valuable in exchange for their patents even if we can't measure what they are obtaining in monetary terms or in terms of market advantage through exclusionary power. Thus, patents obtained for bona fide technological cross-licensing present less of a paradox.

2) Cross-licensing to avoid litigation. This kind of cross-licensing is just another form of defensive patenting. Litigation minded patenting does not encourage innovation and contributes to the drag on the system pointed out by some of today's bloggers. I suspect that defensive cross-licensing creates pressure to patent everything under a when in doubt patent model.

3.5.2007 2:37pm
Bruce Hayden (mail) (www):
To be brief, the problem isn't that a program can be expressed algorithmically or as an equation, but rather that because of that, it assummed to be akin to a law of nature. Yes, there are equations that express laws of nature, and there are equations that express certain mathematical relationships or facts. And those are the sort of thing that are not typically and should not be patentable.

The distinction though is whether the algorithm/ equation is developed through a lot of engineering effort, or represents a fundamental truth. For example, Microsoft Windows does not represent a fundamental truth, regardless of what Bill Gates believes, but rather the culmination of hundreds of millions of dollars of R&D. Compare this to the law of nature represented by E=MC^2.

My point is that E=MC^2 is unpatentable because it is a representation of a basic truth. It wasn't developed, but rather was discovered. But the courts analogized that since you can make the sort of transformation that you suggest, that software is equivalent to an equation, and thus is equivalent to that sort of fundamental truth. And as the courts gained more experience in the area of software, it became increasingly evident to them (and in particular, to the Federal Circuit) that this was not an appropriate analogy.

The statute at issue is 35 U.S.C. 101 which states:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Software claims, properly drafted, fall under the first three of those four categories. The "process" category is obvious for "method" claims. A general purpose computer programmed with a novel and nonobvious program is considered a special purpose computer (since it acts differently from a general purpose one due to the added programming), and thus is a "machine". And we typically also file "Beuregard" type floppy disk claims containing the software for an article of "manufacture". Note though, there is no explicit "equation" or "algorithm" exception to this statute. Rather, the legislative history suggests interpreting this statute as broadly as possible.
3.5.2007 3:26pm
Bruce Hayden (mail) (www):
Maybe another way of looking at this is that a decade or so ago, I was dealing with some members of the Russian (formerly Soviet) Academy of Science. They would send in patent disclosures consisting essentially of an equation, X. I would ask them what it was good for, and they would reply that it could be used for Y (and how it could be used so), and that is what we filed the patent for, Y utilizing equation X. X alone had no utility, and was thus non-statutory subject matter. But when X was used to accomplish Y in a certain way, we now had utility, and it was statutory under 35 U.S.C. 101.
3.5.2007 3:37pm
Bruce Hayden (mail) (www):

I am not sure how really important your distinction between the two types of cross-licensing is. Yes, to some extent, you could look at the first as "freedom of action" or "gatekeeper" patents. This is more typical in the biotech industry, where one narrow patent can block everyone else.

My experience is primarily with the second situation, where cross licensors have mutually blocking patents. A has patents A1, A2, and A3 that block B, and B has B1, B2, and B3 that block A. In the semiconductor industry, they typically don't have two or three that block, but rather hundreds, if not thousands.

But you can't necessarily say that the patents don't provide value there. Indeed, in many cases, the patents are a result of hundreds of millions, if not billions of dollars of R&D effort. And typically the companies making the R&D expenditures are the ones inventing the novel and nonobvious stuff incorporated into everyone else's products. So, in many cases, you can look at net cross-licensing revenues as payment for the extra R&D expended by the company receiving monies from the other.

The basic difference between these two types of licensing is that of technology - how much overlap is there? Or, how many patents does any one product infringe? You really can't do the gatekeeper sort of thing in much of the semiconductor industry because any single semiconductor device invariably infringes hundreds, if not thousands, of patents owned by the manufacturing company's competitors. But theirs infringe too.
3.5.2007 3:51pm
BruceM (mail) (www):
My solution, as I've advocated before: Only grant 100 patents per year. And 100 per year is rather generous, half of those will be worthless. I'd be happier with 10 or 15 per year.
3.5.2007 5:08pm
Hunter McDaniel (mail):
Microsoft Windows is not a fundamental truth, but the computing algorithms embedded with it largely are fundamental truths. The appropriate model here is copyright protection, not patents.

Star Wars is copyrighted, but George Lucas didn't get a patent on any of the plot elements.
3.5.2007 5:52pm
logicnazi (mail) (www):
What I said above argues that it is an interprative error of current law to allow software patents (or circuit diagrams implementing them). I think that software IP really requires a new form of IP in addition to copyright and patents.

The big issue for developing software is not coming up with better ways the program can behave but in coming up with better ways to organize the program so humans don't get lost. I tend to think that unless a compelling case can be made for the economic necessity of IP it should be avoided. I think this case has certainly not been made with software patents and they could have a very great cost. However, we want companies to make profits on the large investments they make developing software. Copyright is the current solution but these seems quite inappropriate.

Copyright lasts far too long for computer programs, which ultimately are designed to produce practical results not (usually) be works of art. Also it doesn't seem to quite capture what we want to protect. We want to protect all that work choosing the best structure and organizing principles of the program which doesn't seem quite patentable either.

I would propose a third class of IP protection for software that would replace both copyright and patent protections for software. In order to get IP rights for your software you would have to make the source code publicly availible just like a patent application makes the way the device works publicly availible. This would discourage useless demands for money (the shareware programer who writes the code for fun then thinks 'might as well demand $5') and solve worries about the source code holder disappearing. While one might worry that this would make unnoticed copying too easy since everyone who wants IP protection needs to show their source it would be easy to catch violations. Also the duration of the software patent would be much shorter than copyright. The length of copyright is causing great inefficency in the computer market as it allows vendors to rest on past laurels via the benefits of lock-in.

On the other hand I would do a bit to extend the protection for overall program structure. As I understand it now (could be really wrong) if one were to look over a competitor's code and copy the broad structure (ohh we break this part down by attaching the functions to the objects and this part should use a linked list) but wrote different code with little syntactic similarity it wouldn't violate copyright. We don't want to go too far with this but since this is the big investment in software it should be protected.


As far as the original post I thought it was widely accepted that it was the need to deter others from sueing with overbroad patents that was encouraging the patent rush. The legal uncertainty with what is patentable certainly encourages the desire to cover your ass and look for the lotto ticket.

One problem with the patent office is that there is a diffuse interest to deny patents but a very concentrated interest involved in the process that wants the patent granted. The idea of asking for public input is a good first step (I think they started) but what really needs to be done is to spend more money on researching the patent applications and either massively increase turnover of patent reviewers or start outsourcing the stuff to keep up with quick new fields.
3.5.2007 6:45pm
Dan Simon (mail) (www):
Cross-licensing to avoid litigation. This kind of cross-licensing is just another form of defensive patenting.

Sabrina, it sounds as though you've identified (correctly, in my view) "defensive patenting" as the underlying source of the patent explosion, and are now trying to develop a(n intellectual) property-based theory to explain why people file defensive patents. But I would argue that there's really nothing property-specific about this behavior. Rather, wherever the law is capricious, people tend to spend money in unproductive ways, trying to insure themselves against legal risks. Competent, careful doctors buy expensive malpractice insurance. Fair, well-managed companies hire expensive consultants to lead anti-harassment and diversity workshops. Product manufacturers spend fortunes anticipating and warding off liability lawsuits. And technology companies file reams of expensive "defensive patents". If there's something specific to property in general--let alone intellectual property--about this kind of risk minimization, I really can't see it.
3.5.2007 11:55pm
Andrew Morton:

The increasingly widespread use of defensive patenting also evidences a chain reaction dynamic. Corporations and individuals obtain patents not because of the patents' potential positive value, such as their ability to generate license revenue or to provide a manufacturer with a competitive edge, but rather for maintenance in a patent arsenal. If threatened with a lawsuit, the corporation threatens to counter-sue with one of its warehoused patents. The suits go away, and each corporation returns to business as usual.

This is a salient observation, and an important one in understanding why the so called Lemming behavior takes place, in my opinion. In many instances, these are just two sides of the same coin. As you pointed out, the semiconductor industry is on of several industries that is a prime example here. There is great utility in the high tech industry in general in obtaining patents not so much for using them per se, but rather because its simply easier to patent everything that might get approved and cross license if there's a lawsuit later. Therefore, everyone rushes to patent everything they can. As you point out, this can't really be regarded as a good thing, at least in so far as anyone deriving real benefit from the patents in question.

I think there's a more than a one sided equation though. That is, some companies have incentives to take up what might be called "aggressive patenting." In terms of the value created or the behavioral motives, this could be something of a distinction without a difference with respect to defensive patenting. Thanks to the speed with which things change and develop in areas such as semiconductors and software (at least to date), it's not surprising to find that just about any patent that can be obtained might find use as a litigation tool during its active life.

A company like Intergraph, which exited the hardware business in 1991 and was unable to productize their ideas, can still get use out of their patent portfolio to sue Intel in 1997. I'm not sure if Intel knew anything of Intergraph's patents or their possible applicability to Intel's work at that time. Similarly, companies like SCO can take the patents they have (or at least, the patents they claim to have) and sue vendors of UNIX related software, even though SCO appears to have no intentions to make a viable product (and indeed, can't even claim to have created the IP in question). For companies in those positions, they can hold onto a patent portfolio that they perhaps intended to use at one point, or that they came into through some business dealing, and use it later to extract value from other firms.

To use Intergraph as an example, they held several legitimate patents on various semiconductor implementations, but clearly weren't using them or getting any value from them after their exit from hardware. For a company in their shoes, there's no incentive to obtain cross-licenses in that same area, so seeking royalties or filing suit is the only way to get value out of a now worthless (to them) patent. In a similar way, a company might have incentives to patent ideas tangential to their business that they have no intention of using, but which could be used to obtain fees from other industry players who might find useful application for similar ideas later on.

This perhaps isn't so bad as purely defensive patenting, at least insofar as it might encourage individuals or organizations that have the intellectual capital to develop useful ideas, but lack the financial capital to actually realize them. Nonetheless, it's an interesting quandary that companies might be incentivized to patent potentially useful ideas that one has no intention of actually using or productizing. It seems that there's a tricky balance to be struck here, as there is obvious utility to this in some instances, but it presents a situation ripe for abuse as well.

One of the issues pointed out in the original post that has been looked at is the value of patents in bundles or groups. This would also seem to be a significant factor in the race for more patents in many of the aforementioned industries. Because so many of the players in these industries rely heavily on each other for widgets that complement and inter-operate, it makes sense for them to patent little pieces that don't necessarily have obvious application on their own. This is one more motivation to take up defensive patenting; these things are really only functional in conjunction with other related ideas (that the holding company may or may not have patents to) and this is why they serve as valuable trading chips in potential disputes.

As a side note, I have to say that at least to some extent, I side with Mr. Hayden, at least to the extent that software implementations should be patentable in some way. It's not obvious to me that the way in which software patents are currently granted is necessarily the greatest, but I do think there is a need to protect and incentivize software engineering. If we're going to grant patents on an ASIC that implements a particular function in hardware, for example, it seems absurd to me that we wouldn't protect the same implementation in software that simply makes use of a general purpose piece of hardware. The value in the ASIC is the function it performs, and the fact that its transistors are laid out this way or that is secondary in my opinion. If the novel idea in something like an ASIC is the manner in which the transistors are laid out, then that should be the patent and it should see application across a broad variety of silicon devices.
3.6.2007 12:35am
theobromophile (www):
Great post! :)

I used to work for an engineering firm that patented its inventions and found that it could make more money off infringement suits than by trying to license the technology. (As a small firm, it lacked the resources to commercialise its inventions and did mostly Phase I and Phase II research.) Patents also served the above-mentioned signalling function: having a patent demonstrated ability to make novel, functional products, which helped it to get government grants.

Despite this, I'm going to disagree with the idea that defensive patenting or over-patenting is unproductive. As pointed out earlier, a single patented invention may be of little use to a company, but several patented inventions, together, may be useful. (Many "inventions," in fact, are covered by several dozen patents.) If a company has developed one possible component of an invention but is waiting to see if enabling technology will develop, it will have to get a patent before it knows whether the technology is viable or not. The costs of getting several useless patents may be outweighed by the value from the few that end up being marketable. Without actual numbers, it's hard to determine if patents are part of a cost of doing business or if IP protections are snowballing.

Patent filings generally rose by about 40% between 1993 and 2003. In addition, patent intensity - the measure of patents obtained per research and development dollar - nearly doubled from the mid-1980's to the late 1990s.

As someone noted above, this is partially due to the fact that people can now patent business processes. What is really misesd, though, is that the United States signed on to the Patent Cooperation Treaty during this time. The US has a first to invent rule, which should discourage (or at least mitigate) the effects of competitive filing. Other countries have a first-to-file rule; however, a US inventor who files a PCT application within 12 months of filing a US application can claim the earlier priority date. The inventor has thirty months from the date of the American filing to file in specific countries. Through this treaty, there is a tremendous incentive for people to file in the US if there is any desire to protect their inventions abroad.
3.6.2007 5:12am
Bill Sommerfeld (www):
As a software engineer I absolutely agree with Sabrina's observations. cynically, I see software patents as a racket invented by lawyers to extract money from engineering organizations.

Does patent cross-licensing save on R&D expenses? Hardly. You need to license actual implementations of the patent, and even there you'll undergo significant expenses integrating someone else's pile of code into your product.
More likely, what happened is that in the ordinary course of development, you reinvented something covered by someone else's defensive patents, they reinvented your wheel, and the rational solution is a nonaggression pact.

Does the patent database form a repository of inventions useful to engineers? Absolutely not! Engineers are not allowed to look at anyone else's patents, lest they subject their employer to treble damages due to knowing infringement. Even if we did look, they're written in stilted and specialized patent-ese bearing no relationship to how engineers describe inventions to each other.

Software patents do not "promote the progress of science and useful arts". They just keep patent lawyers fed and clothed.
3.6.2007 3:12pm
theobromophile (www):
I've always thought that software should be covered exclusively under copyright laws and not patent law. Given that there are a fairly limited number of ways to write code, it just seems more rational to allow for independent invention in the software field.

Under Diamond v. Charkabatry, almost anything is patentable - rather, the PTO will grant patents in almost any field and will let Congress place limitations as it will. Sec. 101 reads: Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.

Software isn't really a machine - that's the computer. They aren't a process (for example, you could patent a process for developing film), manufacture, and they certainly aren't a composition of matter. As I recall, they are considered to be something that, when applied to a computer, makes it perform a certain function and are therefore process patents.
3.6.2007 10:23pm