One Way or Another
August 8, 2015 2:24 PM   Subscribe

"In a patent dispute between two pharmaceutical giants arguing over who owns the royalty rights to a lucrative wound-dressing solution, their lordships sat in judgment over an issue that would have tested the mettle of the finest mathematical logicians; and in the process coined a new legal definition of “one”."

"A previous judgment in 2013 ruled in Smith & Nephew’s favour because of a mathematical quirk known among chemists as the “significant figures rule”, which is a way of taking into account errors of measurement.

...So in this earlier judgment “one” meant anything greater than or equal to 0.95 and less than 1.5, which produced an uncomfortable asymmetry that did not go unnoticed in last week’s judgment. How could “one” include something that is 0.05 less but also include something that is nearly ten times this figure more – for instance 0.49 more than 1?"
posted by not_the_water (43 comments total) 9 users marked this as a favorite
 
Saline solution is now patented?
posted by grumpybear69 at 2:47 PM on August 8, 2015


See also: the Indiana Pi Bill, but that didn't pass.
posted by filthy light thief at 3:03 PM on August 8, 2015


This means that any two claims of "less than x%" and "more than x%" (where x is an integer) are now in immediate conflict if in opposing patents or redundant in the same patent, because "less than x" is now interpreted to mean "less than x-0.5" and "more than x" means "more than x+0.5".
posted by Mapes at 3:04 PM on August 8, 2015 [2 favorites]


This article is a parade of face-palming. Searching around I found somebody's blog post that dispenses with philosophical noodling about whether one is God's number and digs into the gory details. Without knowing a thing about patent law, I agree with the Kat's take: both the "integer rounding" interpretation and the "significant figures" interpretations are inappropriate here, and the "average chemist's precision assumptions" ought to be the interpretation they used. (I also agree that Smith & Nephew's claim of a strict-limit range was weird—if a car maker said "this car gets 30–35 mpg" would we call them fraudulent if we measured a car getting 29.98 mpg?)

Oh, and grumpybear69, I'm pretty sure the solution is much more complicated than saline, here's from that blog's quote of the patent text:
(c) subjecting said polymer, during or after step (b) to one or more agents selected from the group consisting of ammonium salts, thiosulphates, chlorides and peroxides which facilitate the binding of said silver on said polymer, the agent being present in a concentration between 1% and 25% of the total volume of treatment, which material is substantially photostable upon drying, but which will dissociate to release said silver upon rehydration of said material.
posted by traveler_ at 3:05 PM on August 8, 2015 [2 favorites]


This is a messy decision. Does a number mean anything +/- 50%? Or half way between it and the "next" number?

1 % is .01. So that covers .005 to .015.

What about .001?
Is it covered by .0005 and .0015 or
.000 and .01 or
.000 and .002 ?

There should be a cut off. If the solution is patented between 1 and 25, everything outside of that is fair game. Same if it is .01 and .25 or 1000 and 25000.
posted by Monday at 3:07 PM on August 8, 2015 [2 favorites]


if a car maker said "this car gets 30–35 mpg" would we call them fraudulent if we measured a car getting 29.98 mpg?

Yes, kinda? I mean, in the world we live in this is the kind of fudging-in-their-favor that advertisers regularly take advantage of, but in a more perfect world companies wouldn't claim to get over 30 mpg unless they actually got over 30 mpg.
posted by Pyry at 3:10 PM on August 8, 2015 [4 favorites]


1 % is .01. So that covers .005 to .015.

If you're measuring to two significant figures, then yes. Or rather, .005 to .014. That's what rounding off is all about.
posted by happyroach at 3:18 PM on August 8, 2015


But in that example mileage is dependent on factors like how and where you drove the car. You could argue that the actual mileage was different if after a bunch of trials the distribution you observed was significantly different from the distribution reported by the manufacturer, but you can't show that the manufacturer was fudging anything from just a single trial.
posted by en forme de poire at 3:18 PM on August 8, 2015 [1 favorite]


It's impossible to get 30 mpg because we don't live in a computer program. Any measurement has uncertainty.
posted by dilaudid at 3:26 PM on August 8, 2015


Mouse on Mars had something to say about this.
posted by symbioid at 3:27 PM on August 8, 2015 [1 favorite]


"Yes, kinda?"

Ok. Analogies are messy, so go back to the patent: if they had patented a discrete number, saying "the agent being present in a concentration of 1% of the total volume of treatment", would we reasonably expect that only an atom-precise 1% concentration was patented?

Chemistry, and all natural sciences, don't work that way. Numbers representing measurable properties of reality are always fuzzballs: if it matters you express the mean value (1%) with some information about its spread, like a standard deviation or confidence interval. If it really matters you might express if, say, the value is more sensitive to deviation on one side than the other and give the "skew", or asymmetric confidence intervals (e.g. "the volume of pop in this can will be 350mL +10mL -35mL (CI 95%)"—the sort of thing you allude to where the company wants to have the fudge factor lean in their favor.)

So I agree with the ruling, in part, that there's an intrinsic difference between ranges of numbers that express intentional variation—"this tool has three settings: 1 inch, 10 inches, and 100 inches" and ranges that express inherent variation—"the 1 inch setting is ±1%, 10 inch is ±2%, 100 inch is ±5%". So going back to something analogy-like, if that manufacturer now makes a continuously-variable version of their tool that can take any setting between 1 inch and 100 inches, is that also locking their variation to no longer having any ± on the endpoints? Of course not.
posted by traveler_ at 3:30 PM on August 8, 2015 [1 favorite]


In my opinion the desired precision should have been explicitly stated in the parent. In some processes, +/- 0.5 is tolerable. In other cases +/- 0.0005 is necessary.

I suspect the range was made intentionally vague to include any standard saline but not trace salt, but that's just me and the whole system seems broken.
posted by CBrachyrhynchos at 3:34 PM on August 8, 2015 [3 favorites]


So... every time I brine a turkey, I'm infringing on a patent?
posted by jefflowrey at 3:36 PM on August 8, 2015 [3 favorites]


And along those lines just because your error bars overlap does not necessarily mean that infringement has happened.
posted by CBrachyrhynchos at 3:41 PM on August 8, 2015


Ok. Analogies are messy, so go back to the patent: if they had patented a discrete number, saying "the agent being present in a concentration of 1% of the total volume of treatment", would we reasonably expect that only an atom-precise 1% concentration was patented?

I'm fine with interpreting a single number ("1%") as implicitly defining a range [0.5%, 1.5%]. But "1% to 25%" is already a range and so I'm much less comfortable with implicitly expanding this range to 0.5% to 25.5%. Because if they meant 0.5% to 25.5% why didn't they just say that in the first place? If a car manufacturer says they get 30mpg then it's fine to say that they actually mean 29.5 to 30.5, but if they say they get a range of 30-35mpg but actually get a mean of 29.9 +/- 0.01 in a large study then yeah that's kinda fraudulent.
posted by Pyry at 3:55 PM on August 8, 2015 [3 favorites]


If a car manufacturer says they get 30mpg then it's fine to say that they actually mean 29.5 to 30.5, but if they say they get a range of 30-35mpg but actually get a mean of 29.9 +/- 0.01 in a large study then yeah that's kinda fraudulent.

The point is, that's only fraudulent if they claimed 30.0, not 30.
posted by 445supermag at 4:23 PM on August 8, 2015 [3 favorites]


"And along those lines just because your error bars overlap does not necessarily mean that infringement has happened."

Yeah, that's another good point. I still think the true test ought to have been whether a 0.77% solution and a 1% solution are chemically different, not numerically different.

Pyry, my point is that people use number ranges to express different kinds of variation in the physical quantity. It's a mistake to treat every number range as meaning the same thing, or to assume a different meaning on a range than what was meant. Continuing the mpg example (though I'd rather forget it), you've assumed the manufacturer said 30-35 because they believed the "true" value was more like 33 but there's some uncertainty around that number. But what if new "real world" testing standards attempt to simulate typical driving conditions by testing the fuel efficiency of cars under a range of temperatures, payloads, air pressure, humidity, and all whatnot. Then let's say the manufacturer is saying 30-35 because under the most favorable test conditions they get 35, and under least favorable it's 30.

Wait, let's go on to assume that the bolt-on odometer that you used in your study is made of a different material than theirs. It has a different coefficient of expansion, so the systematic error in its distance measurement as a function of temperature is different than theirs. Your mpg measurements will contain a systematic bias that's not at all reflected in the precision of your tests, as reflected in the 29.9 +/- 0.01 result from your large study. That confidence interval is based on your statistical estimate of the amount of variation you believe your tests are subject to.

That's not the same thing as the amount of variation in their tests, which would presumably be part of their mpg claim, and that they're probably also estimating based on statistics from their tests. In which case, the question "does our 29.9 +/- 0.01 fall outside their claimed 30-35 range?" must be answered using appropriate math and not, as CBrachyrhynchos said, a "do the error bars overlap" test.

Oh, that got long, so I'll sum up my points:
  1. Ranges of numbers can be used to express physically and mathematically different sorts of things and it's an error to assume a particular meaning without justification;
  2. It's incorrect to interpret any range of numbers representing a physical quantity as a mathematically sharp, strictly-bounded range;
  3. Whether two quantities that have surrounding ranges of uncertainty are "the same", "different", or "can't tell" is a question loaded with mathematical and philosophical complications, and requires an understanding of the underlying physical system to choose a correct approach toward answering; and
  4. In this case, the necessary system-understanding is chemistry or pharmacology, and the court erred by using simple rounding and a sharp overlap test.
End of rant.
posted by traveler_ at 4:27 PM on August 8, 2015 [1 favorite]


I am aware of this. But in this case I do not think there is any empirical doubt, given the accuracy to which we can measure concentrations, whether the 0.77% value is, considering uncertainty, less than 1%-- it almost certainly is. So the plaintiff shouldn't be able to wave their hands around and shout "but uncertainty!" as a way to expand the range of their patent from the plainly stated range they gave. If they had meant to patent 0.77% solutions it would have required nothing more than typing "0.5%" rather than "1%" in their patent.

My point is that people should, in patents and advertising, give conservative ranges rather optimistic ones which are bordering on fraudulent. If the car manufacturer is not reasonably confident that people will really see strictly greater than 30mpg in their tests then they should claim a lower number. Because, as it stands, in real life if a manufacturer claims 30mpg or 10 hours of battery life then in practice you're going to get like 25mpg or 9 hours and we shouldn't put up with that because of handwaving about uncertainty.
posted by Pyry at 4:44 PM on August 8, 2015


Given the that the oceans have an average salinity of 3.5% and a human being is around .9% I wonder how the patent claim gets around the UK patent principle of being an actual novel contribution? I mean if only someone had thought of it before.
posted by Ignorantsavage at 4:45 PM on August 8, 2015


I'm with CBrachyrynchos - if this is going to be litigable then it should be in the patent (or defined by law, or in some other way). It's completely unreasonable to place the burden on inventors to guess what techniques the courts will apply next time to decide what extra privileges they should grant the patentholder.
posted by No-sword at 4:48 PM on August 8, 2015


Because if they meant 0.5% to 25.5% why didn't they just say that in the first place?

Because saying that means they are actually saying .45% to 25.54%. It's turtles all the way down.
posted by happyroach at 4:51 PM on August 8, 2015 [1 favorite]


Apparently one isn't the loneliest number after all
posted by TedW at 5:13 PM on August 8, 2015 [2 favorites]


"people should, in patents and advertising, give conservative ranges rather optimistic ones "

I remember in my first year of grad school, a very senior scientist taking me aside and pointing out that by being conservative in my estimates of some factors, I was introducing an anti-conservative bias in the final statistics if interpreted politically. I thought back to that a few years later when I overheard an experienced rocket scientist lecturing some new students on how, if they kept being "conservative" in their safety margins like that, we'd end up with a unlaunchable brick. When numbers really matter (as I would hold patents up as a place where they are taken extremely seriously) ranges of variation must be expressed, or estimated, or tracked; in ways appropriate to their subject and not considered "conservative" or "optimistic". One place where I emphatically agree with Lord Kitchen's resoning is here:
First, the scope of any such claim must be exactly the same whether one is considering infringement or validity.
I get that you're sensitive to fraudulent blurring of claims. But in patent law, or chemistry, or patent law about chemistry, uncertainty must not be considered "handwaving". And one claimant's conservative range is another's optimistic one.

Then, realize both sides waved their hands and made absurd, math-based claims as to these numbers. One side saying "you should use integer rounding, not significant figures" and the other side saying "you should use strict boundaries, not significant figures". Meanwhile the significant figures approach was also invalid from a chemistry perspective, making this entire case a hot mess.

"given the accuracy to which we can measure concentrations"

Given good analytical methods, we can measure concentrations way beyond the per-cent level, down to per-trillion even. I think you aren't actually aware of my point, because a big part of it was that "considering uncertainty" is not at all one single thing, and is not at all necessarily a question of the uncertainty in measurement. Perhaps it's the uncertainty in relevance to the chemical process? Perhaps it's the level of precision in their specification?

It's a mistake to think of this as a problem of "expanding the range" because these are different ranges that mean different things. The average temperature where I live is about 45°F. Well that seems odd. Of course it doesn't account for seasons, so let's put a range on it: the temperature where I live ranges from 15°–70°. That accounts for seasons, but not daily cycles. Ok, so it ranges from 7°–85°. But that's not representative of reality! In coldest winter it ranges from 7°–28°, while in hottest summer it's 51°–85°. But that's still not correct, because those are still average temperatures. What about weather variations, making some days hotter or colder than average? Ok, today has an average high of 82° but it was cloudy and drizzly this specific day and just barely touched 70°. The range of record high and low highs for today is 63°–98°. The range of record high and low lows is 36°–62°. Would I be correct in saying that the range of expected August 8th temperatures where I live is 36°–98°?

But wait again—these are all recorded temperatures, not at all the same thing as true temperatures because thermometers have inherent error. Perhaps I could estimate the true temperatures by averaging over several different thermometers, but if that data isn't available I can still use an estimate of the inherent uncertainty in a thermometer's accuracy to put ranges on all these numbers, say ±1%.

Count how many different sources of variation, each with their own sub-range, I found: seasonally the range is 15°–70°. Daily there's a 31° spread centered on the value in the range (this time of year). Weather-variation causes a 35° spread centered on that value (at the daily high) and thermometer uncertainty adds a ±1% spread to that (for the sake of argument). Four different sources of variation, that to be handled correctly must be modeled as separate processes, with their own error bars or ranges or whatever you want to use. The question "what temperature is it where I live" ends up not just being a fuzzball, but a four-dimensional fuzzball with different widths at different slices along various dimensions.

Likewise, the question "what solution concentrations are covered by this patent" is not a question of a single range of numbers. Each source of variation is a separate dimension in the underlying statistics. The Smith & Nephew solution is itself not a 0.77% solution, but 0.77%±x where I don't know their x. Does that infringe on the ConvaTec patent? Do these two hyperdimensional blobs overlap in a chemistry-meaningful way? That's the question.
posted by traveler_ at 5:55 PM on August 8, 2015 [4 favorites]


Why are medicines and treatments patentable? Why is a medical procedure "lucrative"?

The answers are not "because otherwise there would be no incentive to make them".
posted by maxwelton at 6:08 PM on August 8, 2015 [2 favorites]


Ok, so previously I was flogging this idea that the patented range, being a physical quantity not a mathematical abstraction, must be interpreted according to the nature of its underlying process. I had chemistry in mind, but disclaimed knowing anything about patent law. But from reading some of the blog comments, I did learn this much:
A sub-range selected from a broader numerical range of the prior art is considered novel, if each of the following three criteria is satisfied (see T 198/84 and T 279/89):
(a) the selected sub-range is narrow compared to the known range;
(b) the selected sub-range is sufficiently far removed from any specific examples disclosed in the prior art and from the end-points of the known range;
(c) the selected range is not an arbitrary specimen of the prior art, i.e. not a mere embodiment of the prior art, but another invention (purposive selection, new technical teaching).
European Patent Office, guidelines for examination, part G chapter VI section 8

So taking "patent law precedent" as the underlying process that part of the question is moot: you have to be "sufficiently far removed" from the end-points of a known range to not infringe. Choosing a value with no novelty, just to avoid a patented range, is considered infringing. So there it is: 1%–25% is legally a fuzzy range, not a sharp one. And 'The meaning of "narrow" and "sufficiently far removed" has to be decided on a case-by-case basis.'
posted by traveler_ at 6:12 PM on August 8, 2015 [1 favorite]


Apparently one isn't the loneliest number after all

The problem is that "0.77 is the loneliest number does" not scan as well.
posted by ricochet biscuit at 6:36 PM on August 8, 2015


Why are medicines and treatments patentable? Why is a medical procedure "lucrative"?

The answers are not "because otherwise there would be no incentive to make them".


No, not all of the answers.
posted by Revvy at 6:53 PM on August 8, 2015


It's incorrect to interpret any range of numbers representing a physical quantity as a mathematically sharp, strictly-bounded range

But the court is interpreting it as a mathematically sharp, strictly-bounded range, right? Just not the one actually asserted in the patent.
posted by escabeche at 7:07 PM on August 8, 2015 [3 favorites]


I mean is there any argument that it's OK to slap someone with a moving violation for driving 58 when the speed limit is 60? After all, on the significant digit account, they're within the correct interpretation of the range (60,infinity) of illegal speeds.
posted by escabeche at 7:08 PM on August 8, 2015 [1 favorite]


But in this case I do not think there is any empirical doubt, given the accuracy to which we can measure concentrations, whether the 0.77% value is, considering uncertainty, less than 1%

Yes, and really, the salient point has to be whether Smith &Nephew can reliably produce a 0.77% solution (which, of course, they can). If they can measure and produce at 0.77% (with an error less than 0.23%), then it seems clear they aren't infringing. Otherwise, every writer of patent applications would be incentivized to write numbers in the least precise way possible, so as to make things as broad as possible (e.g. writing 1000 as 1x103 so as to claim the range from 500 to 1500).
posted by ssg at 7:24 PM on August 8, 2015


"But the court is interpreting it as a mathematically sharp, strictly-bounded range, right? Just not the one actually asserted in the patent."

Well this gets into the more philosophical side of the math. Certainly when testing something against a fuzzy range you're going to end up, eventually, with some value falling on one side of a judgment and some value falling on the other side of the judgment. So you might call that a sharp boundary. But the underlying process is generating that judgment based on the underlying fuzzy values -- a process with the wonderful name of "defuzzification" -- so the fuzzy nature of the range is exposed depending on the process. Say, if there was a prior art formula with a concentration of 0.8% +/- some x, different widths of x could change whether 0.8% was inside or outside that sharp-seeming boundary.

"I mean is there any argument that it's OK to slap someone with a moving violation for driving 58 when the speed limit is 60? "

I don't know what the law is, but I can tell you under this formulation what it ought to be: drivers' speeds should be classified into "definitely under the limit", "definitely over the limit", and "too close to tell". Legally it would be in keeping with innocent-until-proven-guilty that only the "definitely over" class gets tickets. This is also in keeping with the two-sided symmetry of the patent law principle I quoted before, "First, the scope of any such claim must be exactly the same whether one is considering infringement or validity."

This example also demonstrates the importance of handling multiple sources of uncertainty: what is the +/- we can reasonably expect on cars' built-in speedometers? What is the +/- on a speed radar's measurement precision? How frequently can a driver be expected to check the speedometer, and how much will an average-skill driver drift in speed between checks? A good law would apply all its fancy "reasonable expectation" and "due diligence" and "semper ubi sub ubi" tests to all these ranges of variability in order to generate a good fuzzy-width for the "too close to tell" band. But even a lazy law (like the one I suspect we have) allows for calibration variation as a way to get out of speeding tickets from what I hear.
posted by traveler_ at 7:50 PM on August 8, 2015 [1 favorite]


I think the court is putting the uncertainty in the wrong place. The uncertainty should not be in the range specified in the patent. The uncertainty should be in the measurement process -- in our ability to determine whether the completely certain range specified in the patent was infringed.

It seems to me that being "sufficiently far removed ... from the end-points of the known range" is a nice way of making the standard of evidence vague enough for practical purposes. But it is still (I think) intended to come back to some kind of measurement issue. Suppose Company A patents the range [x, y] and Company B produces something with value z. If we have good reason to think that z is outside [x, y], then Company B has not infringed. If not, then Company B has infringed. The uncertainty here is not in the specified range [x, y] but in whether z falls in the specified range.

With respect to traveler_'s question: if they had patented a discrete number, saying "the agent being present in a concentration of 1% of the total volume of treatment", would we reasonably expect that only an atom-precise 1% concentration was patented? I'm not sure what I would expect. But it seems to me that the problem here is a poorly-written patent. It allows, I think, anyone to avoid infringing so long as they have good evidence that the concentration in their thing is different from 1%. If we can reliably distinguish concentrations of "atom-precise 1%" from "atom-precise 1% plus one atom," then I suppose the answer should be that only atom-precise 1% was patented.

My view here makes what is enforceable depend on technological innovation, but that strikes me as entirely reasonable.
posted by Jonathan Livengood at 8:11 PM on August 8, 2015


A sub-range selected from a broader numerical range of the prior art is considered novel, if each of the following three criteria is satisfied (see T 198/84 and T 279/89):
(a) the selected sub-range is narrow compared to the known range;
(b) the selected sub-range is sufficiently far removed from any specific examples disclosed in the prior art and from the end-points of the known range;
(c) the selected range is not an arbitrary specimen of the prior art, i.e. not a mere embodiment of the prior art, but another invention (purposive selection, new technical teaching).
[my emphasis]
I don't think this is relevant to our case. The defendant's case was that a concentration of .77% lay outside the patented range. This section addresses the case where a new application lies within a patented range, and it says that such an invention may still be patentable.
posted by Joe in Australia at 9:49 PM on August 8, 2015


My point is that people should, in patents and advertising, give conservative ranges rather optimistic ones which are bordering on fraudulent. If the car manufacturer is not reasonably confident that people will really see strictly greater than 30mpg in their tests then they should claim a lower number. Because, as it stands, in real life if a manufacturer claims 30mpg or 10 hours of battery life then in practice you're going to get like 25mpg or 9 hours and we shouldn't put up with that because of handwaving about uncertainty.

I think maybe gas mileage here is not a helpful analogy here because mileage is such a variable and context-dependent phenomenon compared to salinity. I mean, you can drive a car such that it gets arbitrarily bad mileage -- if I idle in my driveway for an hour and then stop the test I'll have gotten 0mpg. (Conversely, you can also get way above the advertised mpg in many cases by using weird hypermiling tricks.) And again, if the advertised mpg is 32.5 +/- 2.5 mpg (95% CI) and one time you measure 29.95 mpg, taking that as evidence of fraud because the lower bound given should be a "strict cutoff" seems bizarre and unreasonable to me -- particularly when you even say here that the car manufacturer should be "reasonably confident," which seems to suggest that the manufacturer cannot actually be 100% confident (which I would agree with). I'm obviously fine with holding people accountable for what they advertise, but that's a really different claim from saying that if you ever get a value outside of the lowest advertised range then the advertiser must have been fudging things or acting fraudulently.
posted by en forme de poire at 12:17 AM on August 9, 2015 [1 favorite]


I mean is there any argument that it's OK to slap someone with a moving violation for driving 58 when the speed limit is 60? After all, on the significant digit account, they're within the correct interpretation of the range (60,infinity) of illegal speeds.

Unless I'm really misunderstanding you, this is only true if you assume both speedometers and police radar are capable of calculating MPH to only a single significant digit.
posted by en forme de poire at 12:30 AM on August 9, 2015 [1 favorite]


Ignorantsavage, I'm a bit confused by your comment - the patent isn't for saline solution (which as you say is nothing new). As paragraph 2 of the judgment says,

ConvaTec is the owner of European Patent (UK) No 1,343,510 ("the Patent") which is concerned with a process for the silverisation of gel-forming fibres used in wound dressings. Silver is a known antimicrobial agent but a problem with silver-containing materials is that they are often sensitive to light and discolour on exposure to it for any length of time. The patented invention is said to solve this problem by providing a method of making light stabilised silverised antimicrobial materials. In broad terms, the method comprises three steps. The first involves preparing a solution comprising an organic solvent and a source of silver. The second involves subjecting gel-forming fibres containing particular polymers to the silver solution. The third (which may be carried out during or after the second) involves subjecting the gel-forming fibres to an agent, such as a chloride salt, which facilitates the binding of the silver to the polymers. Importantly, the relevant patent claim says that the agent must be present in a concentration of "between 1% and 25% of the total volume of treatment". It is the meaning of this phrase which has given rise to these proceedings.
posted by Major Clanger at 1:13 AM on August 9, 2015


Only on rereading this did I notice that the FPP doesn't actually link to the court's judgment, which discusses in detail many of the issues raised in the comments here.

In fact, the original article doesn't either. I'm not so surprised, because few news stories do link to original source material: it's seen as inviting a reader to leave the site, and if the story has an overt or even subtle editorial slant then the writer may not want readers to look at the source.

Yes, someone fairly quickly linked to the IPKat article that does link to the judgment and discusses it in detail, but I don't get the feeling that many of the people who commented here went and read that piece.

Is it worth extending the guidance on FPPs to suggest that where the subject is a written document (or a decision set down in a document) then the FPP should try to link to that document if available rather than just to a journalistic report of it?
posted by Major Clanger at 2:00 AM on August 9, 2015 [3 favorites]


With respect to the learned judge, whose summary of the law is self-evidently correct, the rule is even more peculiar than he supposes. The degree of rounding depends not only on the number of significant figures, but also on the units that are used.

Suppose a patent is expressed in the old Imperial units of pints, quarts, and gallons; a pint being half a quart and a gallon being four quarts. The applicant wants to patent a compound that requires a certain compound, and can express the lower bound as being:
One quart (which will be taken as including everything from .95 to 1.5 quarts)
Two pints (which will be taken as including everything from 1.5 to 2.5 pints, or .75 to 1.25 quarts)
A quarter of a gallon - which will presumably be converted to .25 and rounded from .245 to .255 gallons, or .98 to 1.02 quarts.

It's very clear that the applicant should measure the lower bound in pints.

Suppose that the applicant can express the upper bound as:
Four quarts (= 3.5 to 4.5 quarts);
Eight pints (= 7.5 to 8.5 pints, or 3.75 to 4.25 quarts); or
One gallon (= .9 to 1.5 gallons, or 3.6 to 6 quarts)

Once again the answer is clear: express the upper bound in gallons! So by clever selection of units, our applicant can express a range that is nominally 1-4 quarts but will actually be taken as .75 to 6 quarts, a range nearly twice as great as the nominal one.
posted by Joe in Australia at 4:04 AM on August 9, 2015


… and don't forget, Joe, that there are numerous definitions of pints and gallons, so you could pretty much patent (I exaggerate) all the things. But when I am made Benevolent-but-Slightly-Testy-Technocrat-God-Emperor-Engineer, Fortran EN representation for all numbers shall be the law …
posted by scruss at 8:27 AM on August 9, 2015


Joe in Australia: I don't think this is relevant to our case. The defendant's case was that a concentration of .77% lay outside the patented range.

That also came up in the comments. Apparently the patent describes a number of different ranges as prior art, or something, but only claims the one particular range? Anyway since that legal section covers subranges selected from prior art apparently it's relevant here. And also there's something complicated about the other ranges that affects this decision. I haven't found access to the European Patent but supposedly its US equivalent is identically-worded:
Preferably, the agent is present in a concentration between 0.01 and 50% of the total volume of treatment. In some embodiments, the concentration of agent is between 0.01-25%, 0.01-10%, 0.01-5%, 0.1-5%, 0.1-25%, 0.1-10%, 1-25%, 1-10%, 1-5%, 5-25%, 10-25% or 25-50% of the total volume of treatment.
posted by traveler_ at 10:19 AM on August 9, 2015


traveler_, that is what is quoted in the court's judgement as paragraph [0028]. The court's judgement (thanks for linking to it, Major Clanger) refers to that here:

61. How then did the judge arrive at the opposite conclusion? In my judgment he fell into error in the following three important respects. First, he considered that the skilled person would reject the whole numbers approach because he would see that many of the ranges in paragraph [0028] of the specification are defined by limits that are not whole numbers. The judge was plainly correct that some of the ranges in paragraph [0028] are defined by limits expressed in whole numbers and that others are not, and Professor Kennedy did not at any stage suggest otherwise. Some of those limits are expressed to an accuracy of zero decimal places (that is to say, in whole numbers), others are expressed to an accuracy of one decimal place, and yet others are expressed to an accuracy of two decimal places. But all this shows is that the author knew full well how to express numbers with different degrees of precision, and that when it came to the claim, he chose limits expressed to an accuracy of zero decimal places.

Emphasis mine. This is the most ridiculous bit of reasoning in this judgement. The mess of [0028] shows no consistency or precision at all with expressing figures. None of those numbers are shown with any more precision than is necessary to convey the value; that is each of those figures expresses one significant digit except for the 25s, which possess two. To me stating a range of, for example, 0.01-25% shows quite the opposite; the author had either no idea or no desire how to express numbers with precision and clarity.

Relatedly I was surprised by 49, where the judgement says it doesn't matter "that the concentration limits which appear in the claim have in a sense been chosen by the patentee as opposed to calculated by him or determined by experiment." The 1-25% range of interest is entirely made up.

I find the original judgement (that this overturned) that 1-25% effectively means 0.95% to 25.5% to be far more reasonable, but part of me would want to consider both 1 and 25 to be precise values, with no error bars at all. They're not the result of measurements.
posted by mountmccabe at 12:01 PM on August 9, 2015


I wonder how long it will be before marketing people get hold of this and use it like they have the phrase "up to"

"there are 100 times 1 pieces in this packet"
means anywhere between 50 and ........whatever you get my point??
posted by Burn_IT at 2:31 PM on August 9, 2015


Major Clangor,

I was lazy and thought that the science reporter was competent. When the article said, "The ConvaTec patent covered any salt solution “between 1 per cent and 25 per cent of the total volume of treatment”, " it seemed fairly clear. As this was in a paper other than the Daily Mail or similar tabloid I thought that the reporter would be mostly reliable. I shall endeavor to remember to do all of the research a science reporter should have done in the future. Thank you for the corrective.
posted by Ignorantsavage at 4:32 PM on August 9, 2015


« Older BadBadNotGood: the kids are alright (and jazz...   |   We can never keep, never predict our fortunes for... Newer »


This thread has been archived and is closed to new comments