Evidence Based Prosecution: Non-Patent Prior Art Leads to Rejections

ScreenShot041Quality patents depend upon our ability to find and examine prior art regardless of its form. Most prior art cited by the USPTO arrives in the form of patent-related references — most notably, issued U.S. patents, published patent applications, and foreign patent documents. We know, for instance, from Columbia Professor Bhaven Sampat that about 90% of cited non-patent prior-art references are provided by the applicant, and that only 10% are examiner references. Link.

In some technologies where patenting has a long and extensive tradition, the lack of non-patent prior art is not a serious problem.  In other areas such as software and biotechnology, non-patent references harbor most of the important prior art. 

Results: My new study of 100,000 recently issued patents reveals that patent that cited non-patent literature are significantly more likely to receive rejections from the PTO.  Specifically, patents citing at least one non-patent reference have a 39% greater chance of receiving a final rejection when compared with patents that issued without citing any non-patent references. Regarding non-final rejections, citing non-patent prior art increased the likelihood of receiving a rejection by 9%.

It turns out that technology area of the invention correlates with both the likelihood of receiving a rejection and the amount of non-patent prior art being cited — thus making it a potential confounding factor. Correcting for technology (Class), the effect of non-patent references are reduced somewhat — patents citing at least one non-patent reference have a 20% greater chance of receiving a final rejection when the data is corrected for technology.

Questions: Why do we see this effect? How should a practitioner react?

Saturation Point: At the other end of the spectrum, it appears that citing more than twenty references does nothing to the likelihood of rejection. As the figure below shows, as the number of cited references increases, the likelihood of receiving a rejection steadily increases until about twenty references are considered.  After that, there is no statistically significant change in final rejections based on a change in the number of cited references.


18 thoughts on “Evidence Based Prosecution: Non-Patent Prior Art Leads to Rejections

  1. 18


    For an interesting discussion of another issue that arises as a result of an examiner’s inability to locate and consider pertinent prior art, see my note in the upcoming November edition of the Federal Circuit Bar Journal entitled, “Throwing a Wrench in the System: Size Dependant Properties, Inherency, and Nanotech Patent Applications.”

  2. 17

    Re: patent haters comment above:

    The current duty of disclosure, as you may know, only requires applicants to disclose information that is material to the patentability of the claims. Thus, your suggestion to cite “relavent” prior art is inconsistent with current law. Further, one thing you may not know is that the more prior art that is cited to an examiner, the LESS likely the examiner is to consider it. Yes, this is paradoxical, but when you consider the time constraints examiners work under it is not surprising that examiners will choose to do their own search rather than waste their time sifting through a hundred potentially irrelavent sources cited by the applicant. Yes, an examiner certifies that they considered the prior art cited by the applicant by initialing the pertionent SB/08. However, if the applicant cites a ton of sources, it is likely that the examiner merely considered the title and the abstract. I used to examine a lot of cases from 3m. They got burned for inequitable conduct one time because it was determined that they did not cite all sources material to the patentability of the claims in one application. Their solution, cite THOUSANDS of documents in each application, 99.9% of which were irrelavent to the case at hand. Obviously Examiners cannot consider all of these references, perform an effective search, and draft an office action in the current time allotted (most examiners have ~18 hours to read the spec, construe the claims, draft an office action, respond to an applicants amendment/remarks, conduct interviews etc.). Currently their is no law preventing companies like 3m from doing this, and the presumption of validity over references cited on the face of a patent gives applicants an incentive to engage in this practice. The proposed rule will limit an applicants ability to do this. Of course, the rule will also give lawyers a lot of heartburn, because it puts them in the drivers seat when it comes to determining which references are relavent and which are not (which opens them up to a lot of potential liability, and opens the applicant/client to potential claims of inequitable conduct during litigation)

  3. 16

    Interesting study. As a former patent examiner (I am now with a patent firm in DC), I have a few thoughts and comments:

    In all likelihood the reason applications that cite NPL are more likely to receive a final rejection is because: a) the examiner is familiar enough with the technology to know where in the NPL to look for relavent prior art; or b) if the NPL is cited by the applicant, it calls the examiner’s attention to the existence of a source of prior art outside the patent literature.

    Points a and b are particularly poignant when one considers the search tools at the Examiner’s disposal. The Examiner Automated Search Ttool (EAST), which most examiners use these days, is fantastic for searching patent literature, but is completely useless when it comes to searching for NPL (aside from possibly Derwent abstracts). While the PTO gives examiners access to literally thousands upon thousands of non-patent literature sources, there is no effective way for an examiner to search all of these journals quickly and effectively. If the examiner chooses to search the NPL themselves, they have to search each source individually (i.e., by individual journal) or in small chunks (i.e., through elsevier or ieee, which allow users to search relatively small collections of journals)

    To my knowledge, the only way an examiner can perform a comprehensive NPL search these days is to request a search from the STIC (scientific techinical informnation center) at the PTO. Because it takes several days for the STIC to return the results of the search, examiners are not inclined to utilize this service because this delays when they can draft an office action and receive a count for the case. This is yet another drawback of the PTO’s quota based system.

    Thus, absent: a) an examiner’s personal knowledge of relavent NPL sources that are likely to contain prior art; or b) an indication from the applicant where the examiner should search (i.e., through the citation of a NPL document), it is unlikely that an examiner will know where in the NPL to search, and thus will choose either not to search the NPL at all, or merely do a cursory search.

    For obvious reasons, the inability of examiners to effectively search the NPL reduces the pool of prior art considered during prosecution, thereby increasing the probability of allowance. However, if an Examiner has personal knowledge of NPL literature that is likely to contain prior art, or the applicant provides some indication of where to look in the NPL literature, the pool of prior art considered by the Examiner increases (at least with respect to some baseline level), resulting in a corresponding decrease in probability of allowance.

  4. 15

    It seems to me that the “right thing to do” is to disclose all relevant prior art (patent or non-patent) rather than trying to game the system.

    If a patent dispute goes into litigation, I’d much rather have all that prior art “on the record” of the original application and as having been reviewed by the PTO already.

    That sure beats having your opponent dredge up some hitherto “unknown” piece of prior art and bandy it about in front of the court claiming it invalidates your patent. (Of course, they’ll do it anyway…)

  5. 14

    The observation that non-patent prior art produces more rejections is interesting. This is pure speculation, but I have (personally) noticed that examiners like to latch onto the speculative discussions in such references and use them for 103 rejections, which they retain as final. This is especially true in the biotech area.

  6. 13

    I think the reason that there’s more cited NPL in biotech combined with an increased likelihood of rejection is that a biotech case is more likely to claim in slightly more unexplored area than for example a typical chemical case. NPL is predominately cited because the technology is at a more “scientific” than technically developed stage, and for that very reason rejection is more likely. After all the average examiner doesn’t like to go where less than 100 examiners have gone before.

  7. 12

    After you cite 20 references, you might as well cite 100 with no higher risk of a rejection since the curb seems to break at this point.

  8. 11

    I suspect certain subjects (e.g., biotech) include non-patent citations almost as a matter of course. Over a broader sample of art areas, however, a non-patent citation is an indicator that some aspect of the application is relatively more sophisticated. An NPL citation from the applicant’s side can be associated (on average) with better representation, more valuable subject matter, or both of these influences. When we push applications into crowded art areas, we do so because commercial value justifies the extra effort and such extra effort may include more diligently prepared IDS’s. An NPL citation from the Examiner’s side is likewise something I associate with greater sophistication, namely, a more careful and/or a better informed examination.

    I would prefer to use the information Dennis presented as support for providing Examiners’ with broader access to reference library resources. Examiner rewards for making use of NPL research is also something for us to consider. Actual books and periodicals (as in a properly maintained reference library) or their electronic equivalents should be a more consistent part of Examinations in many art areas.

  9. 10

    The problem with non-examiner’s suggestion is that if the new IDS rules are adopted then there will be a tendency to tell searchers not to report any more than 20 references, as if they report 21 the cost of the IDS goes up drastically.

    Unfortunately, the same thing would occur if any reference were more than 25 pages, so in verbose arts such as computers searching before filing could stop altogether.

    IME, a search can sometimes report only a handful of references and still be thorough, but more than 20 are usually cumulative, which I think is the real meaning of the graph.

  10. 9

    So, the trick seems to be to have your searcher find at least twenty patents or patent applications during a search, and determine patentability based upon that. Seems like a safe bet. 🙂

  11. 8

    I wonder how many issued patents that cited no non-patent literature were found to be valid following litigation. I would guess very few indeed.

  12. 7

    I disagree with Joe Smith’s contention that the NPL (non-patent literature) is “popular” and thus more likely to reflect an invention that is old. From my own experience, NPL is more likely accurate (less likely to be prophetic and preposterous; I cannot tell you how many things both wax the floor and cure cancer if patents are to believed) and is quite useful for 112 purposes. Furthermore, using NPL in prior art rejections happens very often, particularly when the inventor(s) is from a university and does not understand a) the duty to disclose and/or b) the statutory bar set forth by 102(b).

  13. 6


    1) I did not do any analysis to determine the type of rejection. It could be a 102 or 103 rejection. Likewise, the rejection could be based on 101 or 112 issues.

    2) Correction for technology slightly reduced chance of receiving a non-final rejection, but the difference (as compared with uncorrected data) is not very much.

  14. 5

    Dear Dennis, referring to your original comments and your correction for technology area above: Is the chance of receiving a nonfinal rejection in an application also reduced after correcting for technology area? Also, did your research include an analysis of 102 v. 103 rejections based on non-patent prior art, and/or when the non-patent prior art became available to the Examiner? Thank you for your comments.

  15. 4

    That non-patent prior art is basically popular, open literature. If a topic is already so widely known that the popular literature is relevant then it becomes more likely that the patent is not really new – as compared with an area which is so obscure that only a few practioners are involved and the prior art is all laid out in patent applications.

  16. 3


    Assuming that the patent applicant does a good job of identifying prior art and drafting the claims to exclude the prior art I cannot understand why there would be more rejections. Do you have any thoughts on this?

    Ronald J Riley,
    President – Professional Inventors Alliance – http://www.PIAUSA.org – RJR”at”PIAUSA.org
    Exec. Dir. – InventorEd, Inc. – http://www.InventorEd.org – RJR”at”InvEd.org

    Change “at” to @
    RJR Direct # (202) 318-1595
    Generally available 9 AM to 9 PM EST

  17. 2

    Kevin. Great question. In this one, I did not distinguish between applicant and examiner cited non-patent art. As Sampat noted earlier, however, the vast majority (90%) of non-patent art cited comes from the applicants.

    It turns out that technology area correlates with both the likelihood of obtaining a rejection (as I noted earlier) and the amount of non-patent prior art being cited. After correcting for that potential confounder, the numbers are reduced some — resulting in a finding that patents citing at least one non-patent reference have a 21% greater chance of receiving a final rejection.

  18. 1

    To clarify: was this non-patent art cited by the Examiner (helpfully identified with an asterisk) or any non-patent prior art? Also, did you control for technology, in view of your recognition that non-patent prior art is more important in biotechnology versus plow art? My back-of-the-envelope calculations suggest that getting a final rejection is more likely in just those art units most likely to have relevant non-patent prior art, so it’s important to understand your controls.

Comments are closed.