Legal Commentary

This is a continuation of my series on patentable subject matter.  The first post introduced my argument that judicial limits on patentable subject matter should be abandoned in favor of adherence to the statutory categories: processes, machines, manufactures, and compositions of matter.  My second post discussed how Supreme Court precedent in the area is based on repeated dicta with little analysis.  This post extends that discussion in a particular area ? computer software.  Analysis of these cases reveals just how difficult it is for courts to apply judicially developed limitations on patentable subject matter.

The Supreme Court has addressed software patents directly on three occasions.  I address each in turn.

Gottschalk v. Benson, 409 U.S. 63, 67 (1972).  In Gottschalk, the Supreme Court considered a patent relating to the mathematical conversion of ?binary coded decimals? into binary pure binary format, a conversion that was known and could be done by pencil and paper.  The opinion?s text implies that the Court was more concerned with the inventor?s failure to describe the process in such a way that made clear that the applicant actually invented the claimed invention.  The real concern appeared to be that the claim fell short of the specification and novelty requirements.  Furthermore, a pure algorithm with no practical purpose was not ?useful? as required by 35 U.S.C. §101.

Parker v. Flook, 437 U.S. 584 (1978).  In Flook, the Supreme Court considered a claim related to automobile catalytic converters.  The claimed method was for determining the level of temperature, pressure, or flow rate necessary to trigger an alarm; it included a mathematical algorithm to determine the proper ?alarm limit.?   The Court ruled that the only allegedly ?new? part of the three step method was the mathematical algorithm.   The Court then held that discovery of a mathematical algorithm cannot be novel even if the algorithm was previously unknown: ?Whether the algorithm was in fact known or unknown at the time of the claimed invention, as one of the ?basic tools of scientific and technological work? . . . it is treated as though it were a familiar part of the prior art.?    In other words, the Court ruled that a scientific principle cannot be novel, because it must have existed in nature.  The Flook Court admits that its rules will bring no clarity: ?The line between a patentable 'process' and an unpatentable 'principle' is not always clear.?

Flook cites Gottschalk v. Benson for the notion that ?pure mathematical algorithms? are unpatentable subject matter.  While Gottschalk does say this, it was in dicta that Gottschalk by its own terms called a nutshell of the actual holding, which is that one may not patent a non-useful algorithm where the particular method for carrying out the process is neither described nor novel.  This is an example of unchallenged dicta later parroted as a bright-line rule.

Diamond v. Diehr, 450 U.S. 175 (1981).  Only three years later, in Diamond v. Diehr, the Court again considered whether a patent should issue where a claim used a mathematical algorithm, this time as part of a process for processing and curing rubber.  The process included a well known algorithm relating to the time required to cure rubber.  The patent applicant argued, and the Court agreed, that the process could be novel and useful because the claimed invention described a process for accurately measuring the temperature that was later used in the mathematical algorithm.   Thus, the Court ruled that the patent could not be rejected on subject matter grounds.  The decision did not turn on the mathematical nature of one of the steps; indeed, the process could have contained a non-mathematical step that was well known.   What was important was that such a known step, when combined with the other elements of the claim, became novel and non-obvious.

The Federal Circuit, which hears all patent appeals, currently applies the Diamond v. Diehr standard while simultaneously giving lip service to the notion that mathematical algorithms are not patentable.

These three cases show the difficulty (even folly) of trying to apply judicially created restrictions to computer software.  Sure, it?s easy to say that a mathematical algorithm is unpatentable, but every software program boils down to an algorithm of one type or other.  How do we know whether the algorithm is part of a Diehr process or whether it is a Flook principle of nature?  Do we follow Flook?s point of novelty analysis (which has been rejected elsewhere) to isolate the algorithm, or do we follow Diehr?s holistic analysis to see how the algorithm fits in to a more complex process?

The directly contradictory outcomes of Flook and Diehr show that courts have great difficulty applying the standard.  Many would say that Flook is simply wrongly decided, but it has never been overruled and is still cited today in favor of barring certain patents. 

Rules about ?natural phenomena? and ?natural products? fare little better.  After all, virtually every invention is based upon or extends some form of natural phenomenon.  Determining when something ceases to be natural can be very difficult, as the recent Metabolite v. Lab Corp. case shows.

There is, however, a better way.  My next post will discuss a way that courts can systematically apply the rules while also not allowing the patenting of algorithms that are too abstract.