Evidence Based Prosecution II: Independent and Dependent Claims


Part II of my new series on evidence based prosecution focuses simply on a descriptive feature of the patent application.  How many independent claims should you file? Although we don’t answer that question here, we can answer the question of how many are usually filed.  Using data collected from patent applications published thus-far in 2006, I constructed the chart above comparing the average number of independent claims in a patent application to the total number of claims in the patent. 

The results fit very nicely to the equation:

y = 0.0921x + 1.1993, where y is the number of independent claims and x is the total number of claims.

The fit has a very high R(^2) value of 0.9967. The actual predictive value of of the expected number of independent claims loses its power when you consider the variance of the average number of independent claims. In the chart below, high-low bars mark the rang associated with one standard deviation above and below the average number of independent claims.


The equation does predict that an application with 20 total claims will include 3 independent — that is perhaps the most common combination of claims. As the total claims pass 40, the variance begins to climb even further.

Back to Part I: As a follow-on to my earlier post on Sensitivity to Claim Fee Variation, the following figure is an overlay showing the difference between patents the few months before and after December 8, 2004.  On that day, the PTO instituted a new fee schedule that increased the cost of filing a patent and added an additional fee for every claim filed beyond twenty. The data here is slightly different from that reported earlier to remove applications that published outside (before or after) one month from the eighteen–month publication time-line.


Notes: Evidence Based Prosecution:


  • Thanks to Jeff Steck for his suggestion of the new title of the series. 

17 thoughts on “Evidence Based Prosecution II: Independent and Dependent Claims

  1. 17

    Ultram (Tramadol) is used to relieve moderate to moderately severe pain. It also may be used to treat pain caused by surgery and chronic

    conditions such as cancer or joint pain. This medication is sometimes prescribed for other uses; ask your doctor or pharmacist for more


    link to drugs-generic.info

    Tell your doctor of all prescription and nonprescription medications you may use, especially of: carbamazepine, narcotic pain relievers (e.g.,

    codeine), drugs used to aid sleep, antidepressants (e.g., SSRI-types such as fluoxetine or fluvoxamine), MAO inhibitors (e.g., furazolidone,

    linezolid, phenelzine, procarbazine, selegiline, tranylcypromine), psychiatric medicine (e.g., nefazodone), “triptan”-type drugs, anti-anxiety

    drugs (e.g., diazepam), sibutramine.

    link to drugs-generic.info

  2. 16

    Hi, I’m writing my law review note on the proposed USPTO rules. I’d like to get some data on the average number of claims per granted patent, and if possible to have this subdivided by technology area, e.g. biotech/mechanical/etc. I haven’t had any luck finding sucy a data set — does your sampling include these statistics? Can anyone point me in a good direction for obtaining this data? Thanks!

  3. 15

    David French wrote, “I once had an examiner ask me on the phone at the end of an interview why so many applicants file applications with so many claims. He is right: most of those claims are inappropriate.”

    As David Jaglowski noted, claim differentiation is helpful to reduce the risk of unexpectedly narrow claim construction rulings.

    In addition, because prosecution history estoppel limits the application of the doctrine of equivalents, applicants often submit several seemingly redundant claims that are only slightly different from each other. The slight variations may help the applicants avoid prosecution history estoppel for some of those claims. But even if all of the claims are still subject to prosecution history estoppel, they still provide different possibilities for literal infringement. If an applicant is willing to pay for those additional possibilities by paying increased claim fees, I see nothing wrong with it.

  4. 14

    “So don’t bother adding a dependent claim that says that the tire tread comprises a rubber compound.”

    You forget the utility of claim differentiation. Even if the dependent claim isn’t patentably distinguishable, those claims can discourage some instances of “narrow interpretation” that mysteriously fail to qualify as instances of reading-in from the specification.

  5. 13

    With the PTO’s desire to substantively examine fewer claims, encouraging multiple dependent claims could be an effective way to simplify patent examination while allowing applicants to claim the fullest extent of the invention under each independent claim. The USPTO could make multiple dependent claims cost the same as regular dependent claims to encourage their use, but only examine them for formal issues to reduce workload.

  6. 12

    David, I think that this is a useful conversaiont for an additional post. There are a number important uses for dependent claims — even when those claims don’t add much from a novelty perspective.

  7. 11

    David French writes:

    I once had an examiner ask me on the phone at the end of an interview why so many applicants file applications with so many claims. He is right: most of those claims are inappropriate.

    I explain to my clients that dependent claims are “embellishments” on the main invention represented by the independent claim(s) and are there only as back-ups. Dependent claims are irrelevant if the preceding independent claim is valid. If the preceding independent claim is invalid then the dependent claims become relevant. But they are only really relevant if they add a patentable distinction. So don’t bother adding a dependent claim that says that the tire tread comprises a rubber compound.

    Another analogy that I use is that of the Gatling gun. If the first barrel has a dud cartridge, then you rotate to the second barrel. But by the time you get to a barrel that fires, you may not be aiming at your original target.

    As long as clients are impressed by the number of claims in their patents, I suspect that patent attorneys will continue to deliver what the clients seem to want.

  8. 10

    Kerry, Great comment. There are a couple of reasons why I did the graph with total claims as the independent variable. The graph looks much better on the blog this way — I get 40 points of data along the axis. The data on independent claims would start to run thin around 10 independent claims. Also, we shouldnt let ourselves be confused by the names — independent claim does not mean independent variable.


  9. 9

    It is interesting that you choose to plot independent claims against dependent claims instead of the other way around. You then use regression to explain the variation in independent claims by the number of dependent claims. By definition, a “dependent” claim depends on an “independent” claim. Therefore, I would have done the regression the other way around to ensure statistical meaning.

  10. 7

    “I constructed the chart above comparing the average number of independent claims in a patent application to the number of dependent claims in the patent. ”

    In fact, the chart says it compares the number of independent claims to the number of TOTAL claims, not the number of DEPENDENT claims. The chart as labelled is consisent with the regression equation you give so I expect that the labelling is correct and the description in your text is wrong.

  11. 6

    Interesting analysis that corroborates what most of us intuitively knew. I’ve always thought that U.S. continuation practice is more in tune with the way innovation works than the European system – you don’t stop developing the product after you file a patent application – so I’m concerned about the changes in continuation practice on the horizon, and since the changes in continuation practice were first proposed I’ve felt that using fee increases would be a better alternative.

    One nitpicky point: the fee increase in December 2004 was not a result of a change in PTO rules but a statutory change, and raised some fees by over 100% (e.g. the jump in costs for claims in excess of 20 or independent claims in excess of 3). Rule-made fee increases, at least during the several years preceding December 2004, had tended to be on the order of 1-2%, reflecting the increase in some index (cost of living index?), and didn’t apply to all fees. I suspect that if you examine your data more, you’ll see that the number of claims didn’t change in response to those earlier, rule-made fee increases, but changed only in response to the dramatic fee increase in December 2004.

  12. 4

    > … few months before and after December 8, 2004. On that day, the PTO instituted a new fee schedule that … added an additional fee for every claim filed beyond twenty.

    That isn’t correct – while the extra claims fees have been increasing for years, along with all the other fees, there’s been a fee for addional claims over twenty for a long time, certainly much longer than the 26 years I’ve been practicing. I have a patent book which dates to 1956, and it says that there’s a fee for each claim over 20 (only $1.00, albeit in 1956 dollars), so the practice was in effect fifty years ago, at least. I don’t have any references older than that.

  13. 3

    Interesting stuff. A tasty stat would be how many multi-claim (say >35 claim) applications are subject to restriction. Or what is the correlation between restriction and number of claims, if any.

    I always counsel young lawyers to stop at around 20 and hold any other claims in abeyance until one gets to the point of “knowing” that the pending claims are allowable, then introduce multitudes of claims if necessary and the client is willing to foot the bill. A colleague (RIP Melvin) called this file wrapper expansion.

  14. 1


    What exactly are you using as your data set. I assume that it is not based on ALL published US patent applications since 1/1/2006 but rather a representative sampling. If it is based on all the published US patent applications I would love to know how you are able to process that much data.

Comments are closed.