ALTOGETHER OBVIOUS: HINDSIGHT BIAS AND THE EVALUATION OF INNOVATIONS David Tan Foster School of Business University of Washington [email protected]January 20, 2017 Please do not circulate without permission * I am grateful for helpful comments on earlier versions from Chris Rider, Ming Leung, Deepak Hegde, Ezra Zuckerman, Giacomo Negro, Matt Marx, and seminar participants at the University of Washington, Washington University in St. Louis, and the Strategy Research Forum.
43
Embed
ALTOGETHER OBVIOUS: HINDSIGHT BIAS AND THE ......ALTOGETHER OBVIOUS: HINDSIGHT BIAS AND THE EVALUATION OF INNOVATIONS David Tan Foster School of Business University of Washington [email protected]
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
experience provides guidance about next steps in closely related areas. When individuals remain close to
the locus of prior experience, there are clearer expectations about which next steps are and are not likely
to turn out successfully.
8
From this perspective, it is less indicative of systematic hindsight bias if examiners are more
likely to view inventions in familiar areas of technology as obvious. This is consistent with examiners
judging obviousness based on their own ex ante knowledge. When examiners are more familiar with the
underlying technologies, they have clearer guidance from prior knowledge about which next steps are
likely to turn out successfully. Examiners should perceive less uncertainty and therefore be reasonably
expected to view inventions as more obvious. If this result holds, then there might not be reason to be
concerned about systematic hindsight bias. In this case, examiners are making obviousness rejections in
areas that prior theory suggests they would have perceived more ex ante uncertainty if they had attempted
these inventions without the benefit of hindsight (Nelson & Winter, 1982; March, 1991; Stuart &
Podolny, 1996; Fleming, 2001).
On the other hand, it is more indicative of hindsight bias if examiners are more likely to view
inventions in unfamiliar areas of technology as obvious. For inventions involving unfamiliar
technologies, examiners should have less guidance from their own prior knowledge about which steps are
and are not likely to turn out successfully. But hindsight bias suggests, counterintuitively, that the less ex
ante knowledge individuals have about a domain, the more likely they are to overestimate the obviousness
of outcomes in that domain (Fischhoff, 1975; Hawkins & Hastie, 1990). The logic is that seeing an
outcome ex post only provides information about the successful possibility, whereas ex ante knowledge
about a domain provides more balanced information about the broader space of possibilities that includes
successful as well as unsuccessful outcomes. The less ex ante knowledge individuals have about a
domain, the more their perceptions are weighted towards ex post knowledge about the successful
outcome. As a result, they overestimate the certainty of the successful outcome.
This can be seen in cases when examiners justify the obviousness of an invention based on prior
art that actually points away from the invention (In re Dembiczak, 175 F.3d 994, 1999; Ruiz v. A.B.
Chance Co., 357 F.3d 1270, 2004). Lack of ex ante knowledge about an invention’s areas of technology
means that examiners may be learning about the prior art in these areas for the first time. As a result, they
9
may be inadvertently learning about this prior art through the hindsight lens of the successful invention.
They become vulnerable to seeing only those aspects of prior art that point towards the steps that they
know in hindsight to be successful and vulnerable to overlooking aspects of prior art that point away.
Courts have observed this in cases when examiners justify the obviousness of an invention based on prior
art that actually “teaches away” from the invention (W.L. Gore & Associates, Inc., v. Garlock, Inc., 721
F.2d 1540, 1983; In re Grasselli, 713 F.2d 731, 1983). These cases are notable because they stem from
prior art citations that examiners themselves picked as the best ones to support their conclusions of
obviousness. In these cases, an unbiased reading of the prior art would have revealed that while it may
have contained related elements, the underlying logic advised against the kinds of next steps needed to
arrive at the focal invention. To take these steps in a direction contrary to the guidance of prior art, with
low prior expectations that a workable solution could be found, would have required non-obvious insight
from the inventor. So to conclude that that these steps were obvious would require a biased reading of the
prior art, reflecting hindsight knowledge of the successful outcome.
Qualitative examples
Application 10/071,840 provides an example of how this looks in practice. The invention in this
case is an implantable medical device that inhibits microbial growth. It does so through a combination of
two mechanisms: i) using an antimicrobial agent to kill microbes that come in contact with the device and
ii) preventing these microbes from permanently attaching to and forming a film on the device. In rejecting
several of the claims as obvious, the examiner reasoned that there were already prior inventions that
described: i) using antimicrobial agents on medical devices and ii) adding non-stick coatings to medical
devices. The applicants responded by pointing out that, while literally matching elements of the invention
in question, bringing these elements together is contrary to underlying logic of the prior art:
“… Farber teaches against the incorporation of antimicrobial agents for the following reasons: (a)
even broad-spectrum antibiotics eventually lead to the selection of resistant organisms…
Consequently, Farber resorts to an alternative approach based on interference with the adherence
10
of bacteria to polymeric surfaces of medical devices, thereby eliminating the need for
antimicrobials in his devices.” (emphasis added)
As noted by the applicants, the prior art suggested against using the two mechanisms together. The prior
wisdom for using a non-stick coating was to avoid having to use antimicrobial agents. In contrast, the
applicants’ invention uses the two mechanisms together and in a synergistic way. When microbes attach
to a device, they form a protective film that prevents the antimicrobial agent from killing any further
microbes. The non-stick mechanism is used to maintain a clear surface so that the antimicrobial agent can
continue to act for as long as the device is implanted. Given the contrary prior wisdom about how these
elements should be used, using them together in the focal invention could not be viewed as an obvious
extension of the prior art. The examiner subsequently reversed these rejections and allowed the claims.
Application 10/043,849 provides a similar example. In this case, the invention is an organic
optoelectronic device. In rejecting several of the claims as obvious, the examiner reasoned that the
various mechanisms and components of the invention could be found in the prior art. As in the preceding
example, the applicants pointed out that while the elements could be found, the underlying wisdom
attached to each element would have suggested against combining and adapting them in the manner
necessary to produce the invention in question:
“In this case, however, the invention only appears obvious after studying the here claimed
invention and combining the disclosures of the references with the application of undue
hindsight… For example, without undue hindsight, there is no reason why one of ordinary skill in
the art would be motivated take the layer taught by Hora, which is an electrical isolation layer that
is disposed between the electrodes (i.e., it is a key component of the active region) of an
electroluminescent lamp, and insert it into the OLED structure of Affinito… Indeed, it is
respectfully submitted that one of ordinary skill in the art would actually have been dissuaded
from providing particles within one of the polymer smoothing layers of Affinito… because
adding solid particles to a layer whose function is to cover and smooth underlying features is
antithetical to achieving that function.” (emphasis added)
11
The examiner subsequently reversed these rejections, citing the applicants’ clarification.
While examiners will oftentimes simply include a reference to the applicant’s response when
reversing a rejection (e.g. “Please refer to the Applicant's arguments filed on…”), there are cases when
examiners will explicitly acknowledge their own hindsight bias. Application 10/071,596 provides an
example. In this case, the invention is a method for cutting a hole in a piece of ceramic-matrix composite.
Cutting holes in ceramic composites is difficult using conventional drilling techniques because the
material is not internally uniform and prone to fraying, splintering, and cracking. The invention proposes
a solution that uses a cutting head smaller than the size of the desired hole and creating the hole by
removing successive layers in a manner akin to milling. In rejecting the claims as obvious, the examiner
reasoned that because i) numerous prior inventions describe methods for cutting holes in composite
materials and ii) techniques for milling metal are well-known, it would have been obvious to adapt
milling techniques to the particular material properties of ceramic-matrix composites for the purpose of
cutting holes. As the examiner subsequently acknowledged, however, the act of combining and adapting
these disparate insights to this unique problem required significant inventive insight and only seemed
obvious in hindsight:
"…the milling operation claimed by Applicant is something of a hybrid operation of drilling and
milling, which Examiner concedes would likely have a different set of parameters than the
parameters utilized in a pure milling operation… to combine the teachings of Handbook with the
teachings of Eriksson to arrive at the presently-claimed invention would require the use of
impermissible hindsight to first modify the teachings of Handbook (before applying them to
Eriksson), since Handbook is providing feed guidelines for a different type of machining
operation (face milling as shown in Figure 10-78) than for the hybrid drilling/milling operation of
the present claims, and since Handbook does not provide any feed guidelines for the selection of
a feed for a composite material.” (emphasis added)
As these examples illustrate, seeing an invention in hindsight primarily provides information about a
successful path of inventive steps. Faithfully judging obviousness from the ex ante perspective of the
12
inventor requires an appreciation of not just the ex ante reasons to expect these steps would work but also
ex ante reasons to expect they would not. When examiners are less familiar with the field of prior art
behind an invention, they are more susceptible to using the invention as a roadmap for learning about this
prior art and inadvertently cherry-picking only those aspects of prior art that point towards the successful
path.
Constraints on examiner familiarity
Ideally, every invention can be matched to an examiner whose prior experience is perfectly
aligned with the various areas of technology in the invention. In practice, however, there are constraints
on the ways in which examiners gain familiarity with the prior art. Each examiner belongs to one of
several technology centers, corresponding to major fields of technology. In the interest of administrative
efficiency, these divisions are necessarily more static and coarse-grained than the evolving universe of
inventive activity. Nested within these technology centers are hundreds of art units, each corresponding to
a narrower subfield and headed by its own supervisory patent examiner. When applications are received
by the USPTO, the Office of Initial Patent Examination makes an initial classification of the content and
dispatches it to the most appropriate art unit. Applications received by an art unit are then divided
randomly amongst examiners to balance workload.
This formal division of labor has important consequences for the evaluation of obviousness
because it constrains how examiners accumulate familiarity with domains of technology. Examiners gain
exposure to technologies based on the applications that are assigned to them in accordance with the
administrative boundaries between technology centers. Within a domain of specialization, examiners are
more likely be aware of not just what technologies exist but beliefs are about the directions in which
technologies are and are not likely to progress (Nelson & Winter, 1982; March, 1991; Stuart & Podolny,
1996). This helps examiners to simulate an inventor’s prior knowledge about the field of search behind an
invention. This includes beliefs not only about which paths of innovation are likely to be successful but,
just as importantly, which paths are likely to be unsuccessful. For inventions that do not align closely with
these domains, examiners are less likely to have prior knowledge about all of the underlying technologies
13
and less likely to be aware of the multitude of unsuccessful steps that inventors could plausibly have
taken. Their views about obviousness are therefore more likely to be overestimated as a result of seeing
only successful outcomes in hindsight. In the language of the obviousness statute, when an invention
aligns with an examiner’s prior experience, the examiner more closely approximates a “person having
ordinary skill in the art”, whereas when it does not align, the examiner must judge obviousness “with
reference to a hypothetical construct that virtually by definition could not exist” (Durie & Lemley, 2008).
Hindsight bias predicts that in these cases, when examiners lack an ex ante reference point, they are more
likely to overestimate obviousness in hindsight.
METHOD
To test this hypothesis, I use data on claim rejections and acceptances for a sample of 12,100
patent applications filed with the USPTO between 2001 and 2002. This sample includes applications that
ultimately resulted in a granted patent (77 percent) and applications that did not (23 percent). Claim-level
data are coded from the text of rejection and allowance notices mailed by examiners as part of the
application process. These can be found in the image file wrapper for each application, available through
the Patent Information Retrieval System (PAIR). Claim-level data permit analysis at a finer-grained level
than the acceptance or rejection of an application as a whole. The data reveal not only which claims are
ultimately accepted but which claims are accepted without rejection and which are accepted after being
initially rejected. Additionally, the text from rejection notices indicates the statutory basis for each
rejection, allowing me to distinguish rejections based on 35 U.S.C. §103 obviousness from rejections for
other reasons.
The sample of 12,100 consists of those applications numbering from 10/000,001 to 10/099,999
for which the image file wrapper was available and there was at least one applicant patent citation2. The
sample excludes cases in which inventions in the same family underwent prior examinations before the
2 The sample size turned out to be a round number purely by chance.
14
focal one, including continuations and non-U.S. applications entering the USPTO via the Patent
Cooperation Treaty (PCT).
Obviousness rejections. The primary dependent variable in my analysis is the percentage of
claims in a patent application that were rejected for being obvious. This is calculated using total number
of pending claims as the denominator and number of claims rejected due to obviousness as the numerator.
In this sample, 25 percent of applications were granted with all claims, and 20 percent resulted in no
granted claims. The remaining applications were granted with some claims allowed and others rejected
and abandoned. Figure 2 shows the distribution of claim allowance rates for applications that did not have
all or no claims allowed.
Final allowance rates understate the rate at which claims were rejected by examiners during the
application process. In particular, they mask rejections that examiners subsequently reversed in light of
applicants’ objections. To test the hypothesized mechanisms in this study, it is important to consider rates
of rejections. On average, 73 percent of the claims in an application are rejected. This is significantly
higher than what can be seen in the rate of ultimately allowed claims. The majority of these rejections are
due to obviousness. This can be seen in Figure 3. On average, 41 percent of an application’s claims are
rejected for being obvious under §103; 32 percent are rejected for not being novel under §102; 15 percent
are rejected for providing insufficient disclosure under §112; and only 2 percent are rejected for being
non-patentable subject matter under §101.
Invention’s technological overlap with examiner experience. The primary independent variable in
my analysis is the degree to which an examiner’s prior experience aligns with the areas of technology in a
focal invention. I measure this based on the technology classes of prior art cited in the applicant’s
information disclosure and the technology classes of the patents that the examiner previously examined3. I
3 I use the technology classes of applicant cited prior art because these are available for applications that were
ultimately granted as well as applications that were not. Unless it results in a granted patent, an application is only
classified under one technology class by the Office of Initial Patent Examination for the purposes of assigning it to
15
construct two vectors: 𝑇𝑎, whose elements are the percentages of an applicant’s prior art in each
technology class; and 𝑇𝑒, whose elements are the percentages of an examiner’s past patents in each
technology class4. The uncentered correlation between these two vectors, 𝑇𝑎 and 𝑇𝑒, provides a measure
of overlap between the technologies in an application and the technologies in which an examiner has prior
experience:
𝑇𝑎′𝑇𝑒
√𝑇𝑎′𝑇𝑎√𝑇𝑒
′𝑇𝑒
This measure ranges from 0, when an examiner has no prior experience with any of the technology
classes cited in an application, to 1, when an examiner has prior experience in all of the technology
classes cited in an application. Figure 4 shows the distribution of this overlap measure in my sample.
Controls. In all multivariate models, I include controls for attributes of the examiner, the
application, and the law firm that prepared the application. To account for differences between examiners,
all multivariate models include fixed-effects and clustered errors at the level of the examiner5.
At the application level, I include controls for total number of pending claims, number of prior art
references in the applicant’s information disclosure, the average year of the applicant’s prior art, and total
number of technology classes cited by the applicant’s prior art.
the relevant examiner. The full range of technology classes seen on published patents is determined at the end of
examination process for applications that have been granted. 4 This is consistent with prior measures of technological similarity and measures of examiner specialization, which
express technological experience as vectors of percentages across technology classes (Jaffe, 1986; Cockburn,
Kortum, & Lerner, 2003). 5 All results remain consistent when fixed-effects are at the level of the art unit. Prior studies suggest that there are
persistent differences between examiners and that a significant portion of variation in patenting outcomes can be
attributed to examiner fixed-effects (Cockburn et al., 2003; Lemley & Sampat, 2012). Consistent with these prior
findings, examiner dummies explain a significant portion of variation in obviousness rejections in my sample. An
OLS model with only art unit dummies has an adjusted R-squared of 0.12, while a model with only examiner
dummies has an adjusted R-squared of 0.31, and a model with both examiner and art unit dummies has an R-squared
of 0.32. This suggests that examiners vary substantially within art units, and examiner fixed-effects explain most of
the variation among art units. Since the focus of the analysis is not to estimate individual examiner effects, the main
analyses use within-examiner FE estimators to net out examiner differences, rather than including the full set of
more than 2,000 examiner dummies.
16
To control for variation across law firms that prepare these applications, I construct a measure of
average client size. Law firms representing larger clients that patent more heavily may accumulate
experience differently or have different application strategies compared to firms representing smaller
client that patent less heavily. For each law firm, I identify the assignees whose patents it was responsible
for filing. I count the number of patents granted to each assignee in the prior twenty years and take the
average across assignees for each law firm. This measures the extent to which a law firm works with large
In the case of application 10027417 (Ex. 3), the examiner had argued that the applicant’s
invention simply took an obvious step of modifying an existing catheter invention (Miki et al.) with the
use a heat curing adhesive. As the applicants pointed out, however, the prior art cited by the examiner
explicitly advised against this modification: “the use of a heat curing adhesive agent is not advised”.
Taking this step—and making it work—was therefore contrary to the ex ante expectations in this prior art.
Similarly, in the case of application 10046568 (Ex. 7), the examiner had argued that the applicant’s
invention took an obvious step of modifying an existing nail polish invention (Graves et al.) with the
addition of a polymer. As the applicants point out, however, “a skilled artisan would have had no
reasonable expectation of success in incorporating such a polymer into the compositions of Graves et al.”.
22
Explicit statements in this prior art indicated that the outcome would not be “commercially viable or
useful” due to adverse effects from such a modification. In the case of application 10053085 (Ex. 10), the
applicants noted that the prior art cited by the examiner “expressly criticizes” the inventive step described
in the application. In the case of application 10097257 (Ex. 20), the applicants noted that the prior art
cited by the examiner points away from inks with small particle sizes: “The applicants have, surprisingly,
found that the use of very low particle size polymeric nanoparticles actually reduces the viscosity of the
dispersion, which is counter-intuitive from the known art because smaller particle size usually results in
higher viscosity”.
As these examples illustrate, examiners are often able to find prior inventions containing some
but not all elements of a focal invention. They must then make a judgment about whether the
modifications needed to arrive at the focal invention would have been obvious next steps given the prior
art. When examiners decide that these inventive steps were obvious but support their judgments using
prior art that explicitly advises against these steps, this suggests that examiners’ judgments may reflect
biased readings of the prior art, due to their knowledge about how the focal invention was able to make
these steps work in hindsight.
As an empirical window on hindsight bias, these “teach away” objections are especially
informative because they are less subject to gaming and manipulation on the part of applicants than other
kinds of objections that applicants may raise. First, a “teach away” objection requires that a prior art
citation contain suggestions against a particular inventive step. As the examples illustrate, applicants
cannot freely invoke this objection as a response to every rejection if no such suggestion is actually
present in the prior art cited by the examiner. Second, these are objections to prior art chosen by the
examiner. It is certainly plausible that applicants can always dig hard enough and find some prior art that
teaches away from a given invention. But these are citations chosen by examiners as the best examples to
demonstrate that prior art pointed towards the focal invention as the only plausible next step. To invoke
this objection, it is not enough for applicants to argue that the examiner’s examples are not clear or strong
23
enough. They must show that the examiner’s citations support the opposite conclusion. Applicants
therefore cannot freely invoke this objection in response to every rejection.
Figure 8 compares the number of examiner citations with “teach away” objections across
different levels of technological overlap between an invention and examiner experience. As the figure
shows, when inventions have low levels of overlap with examiner experience, examiners cite significantly
more prior art that teaches away from the focal invention when making obviousness rejections.
Model (1) in Table 7 confirms that this result holds in a multivariate model. The dependent
variable in this model is the number of examiner citations used for obviousness rejections that teach away
from the focal invention. The model is estimated using Poisson regression with fixed-effects and clustered
errors at the examiner level. The model controls for percentage of claims rejected for obviousness and
number of examiner citations used for obviousness rejections. In this model, technological overlap
between an invention and examiner experience has a significant negative effect. This implies that when an
examiner has less familiarity with the areas of technology in invention, the examiner is more likely to
make obviousness rejections based on prior art that teaches away from the invention.
These results are consistent with the hypothesized mechanism behind hindsight bias. Examiners
with less prior familiarity with a field of technology are less likely to have ex ante familiarity with
wisdom in the field, some of which may point towards and some of which may point against the inventive
steps taken by the focal invention. As a result, when they read the prior art in these areas for the first time,
their readings of the prior art are more heavily biased by knowing in hindsight which inventive steps
turned out successfully. By highlighting the successful paths, this hindsight knowledge paints a less
balanced picture than ex ante knowledge. It makes examiners more vulnerable to overlooking aspects of
prior art that pointed away from the invention and therefore vulnerable to overestimating obviousness.
24
Reversals of obviousness rejections
As a final step to explore this mechanism, I examine the likelihood of obviousness rejections
being subsequently reversed. As described above, it is impermissible to reject a claim as obvious based on
a prior art citation that teaches away from the invention. When examiners’ obviousness rejections are
subject to this objection, there should be a higher likelihood that these rejections are subsequently
reversed and the rejected claims ultimately allowed.
To explore this, I matched each individual claim to the examiner citation that was used to reject it
for obviousness. I was able to do this for 156,991 claims. These cover 93 percent of the applications in
my sample and 77 percent of the claims in these applications that were rejected due to obviousness. After
matching rejected claims to examiner citations, I then determined how many of these claims were
ultimately allowed. For 34 percent of the examiner citations used for obviousness rejections, all of the
claims rejected based on these citations were later allowed. For 37 percent of examiner citations used for
obviousness rejections, none of the claims rejected based on these citations was later allowed. Figure 9
shows the distribution of percentage of claims allowed for the remaining 29 percent of examiner citations
for which some but not all of the claims rejected were later allowed.
After matching claims to examiner citations, I then determined whether the examiner citation
used to reject a claim for obviousness was objected to by applicants as teaching away from the focal
invention. Figure 10 compares examiner citations that teach away from the focal invention and examiner
citations that do not. As expected, when examiners reject claims based on prior art citations that teach
away from the focal invention, a significantly higher percentage of these claims are ultimately allowed.
Model (1) in Table 8 confirms that this result holds in a multivariate model. The dependent
variable in this model is the percentage of claims rejected for obviousness that were later allowed. The
unit of observation is an examiner prior art citation. Percentages are calculated specifically for claims
25
rejected based on that examiner citation8. The denominator is the number of claims rejected for
obviousness by a given examiner citation, and the numerator is the number of these rejected claims that
were ultimately allowed. The key variable of interest is an indicator variable that takes a value of one
when the focal examiner citation teaches away from the invention in question (and zero otherwise). In this
model, this variable has a significant positive effect on percentage of claims allowed9. This implies that
when claims are rejected for obviousness based on prior art that teaches away from the focal invention,
these claims are more likely to be later allowed. This provides further support for the mechanism
suggested by prior results: rejections based on prior art that teaches away from the focal invention reflect
hindsight bias by examiners and are more likely to be later reserved.
Model (2) adds a variable that controls for the percentage of claims rejected for obviousness that
were also rejected for other reasons beside obviousness. As before, percentages are calculated at the level
of a given examiner citation, so the denominator is the number of claims rejected for obviousness based
on that examiner citation, and the numerator is the number of these claims that were also rejected for
other reasons besides obviousness. Model (3) splits this into separate variables corresponding to the three
other rejection criteria: lack of novelty (§102), insufficient disclosure (§112), and non-patentable subject
matter (§101). Each variable is a percentage, in which the denominator is the number of claims rejected
for obviousness based on a given examiner citation, and the numerator is the number of those claims that
were also rejected for one of the other three reasons besides obviousness. In models (2) and (3), with
these controls included, the variable indicating that an examiner citation teaches away from the invention
in question continues to have a significant positive effect on percentage of claims allowed.
8 Results are robust when aggregated at the application level. The models here are at the examiner citation level
because the key variable of interest—whether an examiner citation does or does not teach away from the invention
in question—is at the examiner citation level. The analysis could also be done at the level of the individual claim.
Since the variable of interest is at the examiner citation level, it will not vary across claims rejected by the same
examiner citation. Clustering errors across claim-level observations to account for these duplicate values would
produce the same result as analysis at the examiner citation level. 9 Predicted values are roughly normally distributed around 0.53, with no values below 0.35 or above 0.68, well
within the 0 to 1 range of the dependent variable.
26
Finally, model (4) shows that the effect is robust in a count model of claim allowances. In this
model, the dependent variable is the number of claims that were rejected for obviousness based on a given
examiner citation but were ultimately allowed. The model controls for total number of claims rejected for
obviousness based on that examiner citation. As before, the variable indicating that an examiner citation
teaches away from the invention in question has a significant positive effect on number of claims allowed.
The effect of the “teach away” variable provides a useful window on hindsight bias because it
reflects a relatively concrete reason for overturning obviousness rejections. When studying the allowance
of previously rejected claims, a potential concern is that allowances might reflect examiners being worn
down by especially stubborn applicants. However, as described above, an especially determined applicant
could not freely invoke the “teach away” objection if no such suggestion is actually present in the prior art
cited by the examiner. It is not enough for applicants to argue that the examiner’s examples are not clear
or strong enough. They must show that the examiner’s citations support the opposite conclusion.
Moreover, when applicants can show this, an especially stubborn examiner could not freely deny the
objection10
. The effect of the “teach away” variable indicates cases when rejections are reversed because
examiners had overlooked aspects of the prior art that pointed away from the invention in question. Taken
together, these and the prior results provide evidence consistent with hindsight bias in the evaluation of
obviousness.
DISCUSSION
Searching for new ways to use existing ideas is an important aspect of innovation and can involve
a great deal of ex ante uncertainty (Fleming, 2001). The paradox is that the outcomes of this search often
seem obvious when viewed in hindsight. In this study, I examine the degree to which hindsight bias
affects the evaluation of obviousness in the patent examination process. If examiners are relying only on
ex ante knowledge to assess obviousness, then inventions should seem less obvious when examiners are
10
Note that examiner fixed-effects ensure that the results are not being driven by between-examiner differences in
stubbornness.
27
less familiar with the underlying areas of technology. If, on the other hand, examiners rely more on ex
post knowledge of the outcome when they have less ex ante knowledge, i.e. hindsight bias, then
inventions will seem more obvious when examiners are less familiar with the underlying areas of
technology. I test this hypothesis using a novel data set on claim rejection rates for granted and rejected
patent applications. I find that claim rejection rates due to obviousness are higher when examiners are less
familiar with the underlying technologies in an application. Additional analysis of the kinds of citations
that examiners use to make obviousness rejections, the kinds of objections that applicants raise in
response to rejections, and the rates at which examiners reverse rejections in light of applicants’
objections provide further evidence consistent with hindsight bias.
These results have important implications for understanding the evaluation of innovations. Prior
research has recognized that the value of innovations is not always straightforward to demonstrate. For
instance, innovations with unfamiliar features may be difficult to understand and evaluate and therefore
diffuse slowly (e.g. Rogers, 1962). There may be reluctance to adopt innovations with unfamiliar
characteristics because of doubts about economic value, practicality, or desirability. But these problems
center on doubts about the value of innovations—not about whether an idea is in fact a non-obvious
inventive insight or discovery in the first place. The results in this study suggest, paradoxically, that when
evaluators are less familiar with the underlying areas of technology, an invention may suffer a discount in
terms of perceived innovativeness.
This finding seems counterintuitive because unfamiliarity would seem to imply that an innovation
is contrary to prior expectations. For instance, an insight from prior studies of category-spanning products
is that unfamiliar combinations fly in the face of audience expectations (Zuckerman, 1999; Negro &
Leung, 2013; Bowers, 2015; Leung & Sharkey, 2014). Unfamiliar combinations run counter to audience
beliefs about what does and does not make sense together. This can be a liability when evaluators must
accurately assess a product’s economic value and determine how much there is to gain from investing in
it (Zuckerman, 1999; Bowers, 2015; Leung & Sharkey, 2014). But in a context such as the patent system,
28
where innovations are rewarded precisely for being non-obvious,—i.e. inconsistent with prior
expectations—unfamiliarity would seem on face to be an advantage. Unlike innovation adopters or
investors, patent examiners are not evaluating the economic value of an invention. The evaluation of
obviousness is strictly about the degree to which an invention was an ex ante predictable outcome. If
unfamiliar sets of attributes run counter to audience expectations about what typically goes together, then
this would seem to be an asset in the context of demonstrating non-obviousness.
The theory behind hindsight bias helps to explain why unfamiliar sets of attributes may,
counterintuitively, appear more obvious to evaluators. Prior knowledge about a domain helps to provide
individuals with a clear baseline, which includes both ex ante reasons to expect an outcome will be
successful as well as ex ante reasons to expect it will not. When individuals lack this ex ante baseline,
their perceptions are more likely to be biased in hindsight by their ex post knowledge that the outcome
turned out to be successful (Fischhoff, 1975; Hawkins & Hastie, 1990). In the context of evaluating
innovativeness, an invention may not appear in hindsight to a be an extraordinary achievement if
evaluators have no ex ante reference point for what ‘ordinary’ looks like in that field of technology. Prior
familiarity with a field of technology helps to mitigate this potential bias by making evaluators more
aware of reasons why ‘ordinary’ inventive insight might have suggested many other paths besides the
ultimately successful one. However, when inventions involve unfamiliar areas of technology, evaluators
may not have prior knowledge to provide a relevant benchmark. In these cases, the relevant benchmark
for “ordinary skill in the art” may be “a hypothetical construct that virtually by definition could not exist”
(Durie & Lemley, 2008). In the context of patent examination, there has been a great deal of attention on
the possibility that examiners might be too lax in granting patents as a result of being unfamiliar with new
and changing fields of technology (e.g. Eisenberg, 2004). Indeed, an intuitive expectation is that
innovativeness should be easier to demonstrate to an audience that is ignorant of what else has been done
in a field. This study suggests, however, that there is at least a countervailing tendency. Ignorant
audiences are more likely to discount an innovation as facile: merely and un-extraordinarily novel.
29
Outside of the context of patent examination, these insights shed light on the role of the relative
comparison mechanism in evaluation processes. Familiarity is often thought to affect evaluation outcomes
via evaluator beliefs about whether the underlying elements—e.g. product features, lines of business,
genres, etc.—tend to or ought to go together. But familiarity with particular sets of attributes may also
affect evaluations because evaluators have a better understanding of variation among products with those
attributes, including both successful and unsuccessful variants. Greater experience with variation within a
category helps evaluators to form clearer benchmarks for what an ordinary level of success look like. This
allows evaluators to make relative comparisons among category members (Zuckerman, 1999; 2004).
When a set of attributes does not cohere with any domain of evaluator experience, there is greater
ambiguity about what the relevant benchmark is. Only recently has research begun to isolate the effects of
this relative comparison mechanism. Recent findings show, for instance, that rather than appear
extraordinary for not fitting in, unfamiliar combinations are less likely to be rewarded as extraordinary
(Bowers, 2015).
This study yields new insights about the relative comparison mechanism by showing how
evaluators form benchmarks when they are less familiar with a domain. The results suggest that
evaluators may fail to appreciate achievements in unfamiliar domains because they lack prior familiarity
with what an ordinary level of success looks like. As a result, they lack a clear reference point against
which extraordinary levels of achievement can be contrasted. In the context of patent examination,
obviousness rejections represent an evaluative outcome that directly reflects an examiner’s relative
comparison of an invention to the prior art. The prior art citations that examiners use to make obviousness
rejections shed light on how examiners form benchmarks for “ordinary skill in the art”. The results
suggest that when examiners are less familiar with a domain, reference points tend to be constructed
based on guidance from successful outcomes and therefore biased towards reasons to expect success. By
providing fine-grained evidence of these previously unstudied micro-mechanisms, this study contributes
30
novel insights about how reference points and relative comparison shape evaluation processes in
unfamiliar domains.
In terms of isolating these micro-mechanisms from other related explanations for discounts
against unfamiliar features, this context provides several advantages. A common explanation for
evaluative discounts against unfamiliar features is greater cognitive confusion. The context of
obviousness rejections is useful for isolating the relative comparison mechanism because the very nature
of an obviousness rejection implies the opposite of cognitive confusion. For a claim to be rejected for
obviousness, an examiner must first have determined that the claim does indeed contain a novel inventive
step—indicating that examiners are not failing to recognize novel features. An obviousness rejection then
implies that not only did these inventive steps seem easy to understand but there was so little ambiguity
about these steps that they were the only conceivable ones to take. Indeed, hindsight bias in patent
examination is essentially rooted in the tendency for examiners to experience less cognitive confusion
than they should. After seeing an invention in hindsight, examiners are only able to see how the
underlying features make sense. They are unable to imagine reasons why these would not have made
sense ex ante. When applicants argue that prior art teaches away from a novel feature they add, they are
essentially trying to demonstrate that this would have seemed more confusing ex ante than it may seem ex
post. They are explicitly pointing to ways in which these features clash with prevailing beliefs about what
does and does not make sense, trying to recreate a sense of the cognitive confusion that is no longer
apparent to examiners in hindsight.
A general insight stemming from these findings is that an achievement is more likely to appear
extraordinary to an audience that appreciates the ease of failure than to an audience that knows only of the
focal success. Similar observations have been made in other evaluative contexts in which savvy audiences
seem to be “the most understanding and forgiving audience” (Becker, 1982, p. 54). An example of
success can appear unremarkable if that is the only example audiences have seen. Inexperienced or
outside evaluators may be less capable of appreciating the ex ante difficulty of work in a domain because
31
they are less aware of how easily it could have been done less successfully. The generality of these
insights should be tested in other contexts, but they have the potential to shed new light on how
experience in a domain affects the way individuals conduct evaluations.
32
REFERENCES
Becker, H. S. 1982. Art Worlds. Berkeley: University of California Press.
Bowers, A. 2015. Relative comparison and category membership: the case of equity analysts. Organization Science, 26:
571-583.
Cockburn, I. M., S. Kortum, and S. Stern. 2003. Are all patent examiners equal? The impact of characteristics on patent
statistics and litigation outcomes. In Patents in the Knowledge-based Economy, W. M. Cohen and S. A. Merrill
(eds). National Academies Press: Washington, DC: 19-73.
Durie, D. J. and M. A. Lemley. 2008. “A realistic approach to the obviousness of inventions.” William and Mary Law
Review, 50: 989-1020.
Eisenberg, R. S. 2004. Obvious to whom? Evaluating inventions from the perspective of PHOSITA. Berkeley Technology
Law Journal, 19: 885-906.
Fischhoff, B. 1975. “Hindsight ≠ foresight: the effect of outcome knowledge on judgment under uncertainty.” Journal of
Experimental Psychology: Human Perception and Performance, 1: 288-299.
Fleming, L. 2001. “Recombinant uncertainty in technological search.” Management Science, 47: 117-132.
Hall, B. H., A. B. Jaffe, and M. Trajtenberg. 2001. The NBER patent citations data file: lessons, insights, and
methodological tools. NBER Working Paper 8498.
Hawkins, S. A. and R. Hastie. 1990. “Hindsight: biased judgments of past events after the outcomes are known.”
Psychological Bulletin, 107: 311-327.
Hegde, D. 2012. “Funding and performance at the US Patent and Trademark Office.” Nature Biotechnology, 30: 148-
150.
Lemley, M. A. and B. Sampat. 2012. “Examiner characteristics and patent office outcomes.” Review of Economics and
Statistics, 94: 817-827.
Lemley, M. A. 2001. Rational ignorance at the patent office. Northwestern University Law Review, 95: 1495-1532.
Leung, M. D. and A. J. Sharkey. 2014. “Out of sight, out of mind? Evidence of perceptual factors in the multiple-category
discount.” Organization Science, 25: 171-184.
Mandel, G. N. 2008. “Another missed opportunity: the Supreme Court's failure to define nonobviousness or combat
hindsight bias in KSR v. Teleflex.” Lewis and Clark Law Review, 12: 323-342.
March, J. 1991. Exploration and exploitation in organizational learning. Organization Science, 2: 71-87.
Negro, G. and M. D. Leung. 2013. “’Actual’ and perceptual effects of category spanning.” Organization Science, 24:
684-696.
Nelson, R. R. and S. G. Winter. 1982. An Evolutionary Theory of Economic Change. Cambridge, MA: Harvard
University Press.
Rogers, E. M. 1962. Diffusion of Innovations. New York: Free Press.
Stuart, T. J. and Podolny, J. 1996. Local search and the evolution of technological capabilities. Strategic Management
Journal, 17: 21-38.
Witherspoon, J. F. (Ed.). 1980. Nonobviousness: the Ultimate Condition of Patentability. Bureau of National Affairs.
Zuckerman, E. W. 1999. “The categorical imperative: securities analysts and the illegitimacy discount.” American
Journal of Sociology, 104: 1398-1438.
Zuckerman, E. W. 2004. “Structural incoherence and stock market activity.” American Sociological Review, 69: 405-432.
33
Figure 1. Rounds of rejection for accepted and rejected patent applications
This figure shows the distribution of rejection rounds for 72,252 patent applications filed in 2001 and 2002 that were ultimately
accepted or ultimately rejected and abandoned as of 2014. The sample represents cases with application numbers from 10/000,001 to
10/099,999. A rejection round refers to the mailing of a rejection notice by the examiner. Applicants always have the opportunity to
respond to and challenge a rejection notice. An application is only ultimately rejected if an applicant chooses to abandon the
application instead of responding. In this sample, 76 percent of applications ultimately resulted in a granted patent.
Figure 2. Percentage of claims in an application that were ultimately allowed
This figure shows the percentage of claims that were ultimately allowed for a sample of 12,100 patent applications. The figure does
not include applications that resulted in all claims being granted (25 percent) or applications that resulted in no claims granted (20
percent).
0
5
10
15
20
25
30
35
40
45
0 1 2 3 4 5 6 7 8 or more
% o
f a
pp
lica
tion
s
Rounds of rejection
Accepted applications Rejected applications
34
Figure 3. Percentage of claims in an application that were rejected
This figure shows the percentage of pending claims that were rejected in a sample of 12,100 applications. All percentages are
calculated using total number of pending claims as the denominator. In the first column, the numerator is the number of claims
ultimately allowed. In the second column, the numerator is number of claims rejected for any reason. In the next four columns, the
numerators are the numbers of claims rejected based on each of the four patentability standards.
Figure 4. Invention’s technological overlap with examiner experience
This figure shows the distribution of technological overlap between the prior art cited by an applicant’s information disclosure and the
patents that an examiner has previously examined. A value closer to 1 indicates high overlap, and a value closer to 0 indicates low
overlap.
0.0%
10.0%
20.0%
30.0%
40.0%
50.0%
60.0%
70.0%
80.0%
Ultimately allowed Rejected:
All reasons
Rejected:
Obvious§103
Rejected:
Not novel§102
Rejected:
Insufficientdisclosure
§112
Rejected:
Non-patentablematter
§101
% o
f p
en
din
g c
laim
s
35
Table 1. Descriptive statistics
These tables show descriptive statistics and correlations for the key independent variables used in all analyses. The unit of observation
This figure shows the number of times that an assigned examiner requested that an application be transferred elsewhere before it was
ultimately assigned and examined. The unit of observation is the application. Dotted lines show the 95 percent confidence interval.
The figure compares low, medium, and high levels of an invention’s technological overlap with examiner experience, split at the
terciles.
Table 2. Fixed-effects Poisson model of number of transfer requests
This table reports results from a Poisson model of transfer requests. The unit of observation is the application. The model includes
fixed-effects and clustered standard errors at the examiner level.
(1)
ln(total claims) 0.156**
(0.057)
ln(law firm's average client size) 0.003
(0.018)
# technology classes in applicant prior art 0.046***
(0.012)
Average year of applicant prior art 0.001
(0.005)
ln(applicant prior art citations) -0.194***
(0.042)
Invention's technological overlap with examiner experience -0.811***
(0.107)
Observations 7,191
Number of examiners 1,146
Chi-sqr 78.98 df 6
Standard errors clustered by examiner; *** p<0.001, ** p<0.01, * p<0.05, + p<0.10
0.14
0.19
0.24
0.29
0.34
Low Medium High
Nu
mb
er o
f tr
an
sfer r
eq
uest
s
Invention's technological overlap with examiner experience
36
Figure 6. Percentage of claims rejected due to obviousness
The vertical axis in this figure shows the percentage of pending claims in an application that were rejected due to obviousness during
the examination process. Percentages are calculated using total number of pending claims as the denominator and number of claims
rejected due to obviousness as the numerator. The figure compares applications with low, medium, and high levels of an invention’s
technological overlap with examiner experience, split at the terciles. The unit of observation is the application. Dotted lines show the
95 percent confidence interval.
Table 4. Regression models of claims rejected due to obviousness
This table reports results from models of claims rejected due to obviousness. In column 1, the dependent variable is the percentage of
total claims in an application that were rejected due to obviousness. The model is estimated using linear regression. In column 2, the
dependent variable is number of claims rejected due to obviousness with a control for total number of claims. The model is estimated
using Poisson regression. In all models, the unit of observation is the application. All models include fixed-effects and clustered
standard errors at the examiner level.
(1) (2) Linear Poisson
% of total claims that were
rejected for obviousness
# of claims rejected for
obviousness
ln(total claims) -0.000 0.914***
(0.006) (0.017)
ln(law firm's average client size) 0.002 0.007 (0.002) (0.005)
# technology classes in applicant prior art -0.000 -0.001
(0.001) (0.003) Average year of applicant prior art -0.001 -0.001
(0.001) (0.002)
ln(applicant prior art citations) -0.003 0.005 (0.004) (0.012)
Invention's technological overlap with examiner experience -0.033* -0.085*
(0.014) (0.038) Constant 2.176+
(1.144)
Observations 12,100 10,257
Number of examiners 2,707 1,901
F-stat 1.614 . Chi-sqr . 2892
df 5 6
Standard errors clustered by examiner; *** p<0.001, ** p<0.01, * p<0.05, + p<0.10
0.38
0.39
0.4
0.41
0.42
0.43
0.44
Low Medium High
% o
f cla
ims
reje
cte
d f
or o
bvio
usn
ess
Invention's technological overlap with examiner
experience
37
Figure 7. Examiner citations’ technological overlap vs. invention’s technological overlap with examiner experience
This figure shows the technological overlap between examiner experience and the examiner citations used to make obviousness
rejections. The figure compares low, medium, and high levels of an invention’s technological overlap with examiner experience, split
at the terciles. The unit of observation is the application. Dotted lines show the 95 percent confidence interval.
Table 5. Technological overlap between examiner citations and examiner experience
This table reports results of a linear regression model in which the dependent variable is technological overlap between examiner
experience and the citations used by the examiner to make obviousness rejections. The unit of observation is the application. The
model includes fixed-effects and clustered standard errors at the examiner level.
(1)
Technological overlap
between examiner citations
and examiner experience
ln(total claims) -0.005
(0.006)
ln(law firm's average client size) -0.001 (0.002)
# technology classes in applicant prior art -0.007***
(0.001) Average year of applicant prior art 0.000
(0.001)
ln(applicant prior art citations) 0.031*** (0.004)
% claims rejected for obviousness -0.005
(0.013) Number of examiner citations used for obviousness rejections -0.005**
(0.001)
Invention's technological overlap with examiner experience 0.289*** (0.015)
Constant 0.498
(1.278)
Observations 6,582
Number of examiners 2,001
F-stat 57.66
df 7
Standard errors clustered by examiner; *** p<0.001, ** p<0.01, * p<0.05, + p<0.10
0.6
0.65
0.7
0.75
0.8
0.85
Low Medium High
Exa
min
er c
ita
tio
ns'
tech
no
log
ica
l
overla
p w
ith
exa
min
er e
xp
erie
nce
Invention's technological overlap with examiner experience
38
Table 6. Examples of applicant objections to examiner citations that teach away from focal invention
These examples come from the text of applicant responses to examiners’ rejection letters. They point out ways in which the prior art
cited by examiners to make obviousness rejections actually teach away from the inventive steps described by the applicants.
Example
number
Application
number
Text from applicant response to examiner rejection (emphases added)
Ex. 1 10002145 In summary, Vlodavsky explicitly teaches against the sugars of the currently claimed invention by teaching that to retain
high inhibitory activity against metastasis N-substituted oligosaccharides require a molecular size of about 4000 daltons or
more… Moreover, Vlodavsky has only negative things to say about the metastatic inhibitory effect of sugars within the
scope of the claimed invention (i.e., sugars with a molecular weight of 3000 daltons or less), such as the pentasaccharide was
completely devoid of activity against tumor metastasis.
Ex. 2 10024633
Rather, the very fact that Saitoh is an ink jet printing paper requires it to be water absorbing - if this were not the case, you could not ink jet print with the inks as disclosed in that patent. Furthermore, Saitoh does not suggest or motivate one to
employ such a weatherproof coating and, in fact, teaches away from the same since such a coating would render the ink
jet printing paper of Saitoh inoperative for its intended purpose - that is, to absorb the aqueous-based ink applied thereto.
Ex. 3 10027417 Moreover, Miki et al. teach away from using a heat curing adhesive at column 21, lines 33-39: "When a heat curing
adhesive agent is used, the catheter shaft and balloon are inevitably exposed to heat during the heating required for curing. As
a result, it is entirely possible that the balloon diameter will shrink, the balloon bursting pressure will decrease, and the catheter shaft will undergo thermal degradation, and these all lead to diminished balloon catheter performance, so the use of a heat
curing adhesive agent is not advised."
Ex. 4 10029584
Because Houser describes a graft that achieves an effective seal and fixation within a vessel without auxiliary means, and McDermott describes using the inflatable channels as fixation means without the use of a stent, the skilled artisan reading the
references would not look to the teachings of McDermott to modify the graft of Houser. Further, since McDermott is
explicitly desirous of minimizing the profile of the graft and states that the graft lacks a stent for maximal compression,
McDermott teaches away from the inclusion of a stent to the graft.
Ex. 5 10035420
Applicants submit that, contrary to the Examiner's assertion, the cited passage from Dorin et al., when taken in context, does
not provide motivation to adjust the pH of a formulation of IFN-p3 to about 3.0 to about 5.0. … to the contrary, this cited
passage specifically teaches away from a pH range of about 3.0 to about 5.0… Upon understanding the teachings of Dorin
et al. that during storage of an IFN-p3 polypeptide, maintenance of pH is "critical," and that the pH of the formulation is most
preferably "adjusted to 6.5," one of skill in the art (who would understand that each pH unit represents a tenfold difference in hydrogen ion concentration) would not be motivated to adjust the pH of a formulation of IFN-p to about 3.0 to about 5.0,
which is fifty to five thousand times more acidic than the preferred pH of 6.5 taught by Dorin et al.
Ex. 6 10043392 As noted above, injection molding and press-fitting processes operate on much different principles, which are not compatible with one another… If the socket housing of Borgen et al. was cooler than the socket liner, then it would be inoperable for its
intended purpose. Similarly, if the socket liner of Borgen et al. was hotter than the socket housing, then it would be inoperable
for its intended purpose. Accordingly, one of ordinary skill in the art would be discouraged from combining the injection
molding process of Jackson with the press-fitting process of Borgen et al.
Ex. 7 10046568 Moreover, Graves et al. actually teaches away from upsetting its careful chemical balance because of the adverse bubbling effects of certain additives, like film formers, polymers, and solvents. See col. 1, lines 39-67, discussing how the incorporation
of a particular copolymer into the nail formulations therein caused such adverse effects that "the product was not commercially
viable or useful." … In light of this teaching, one of ordinary skill would not have been motivated to incorporate at least one first polymer, as presently claimed, into the nail enamel compositions of Graves et al., for fear that it may render them
unsatisfactory for their intended purpose. See MPEP § 2143.01. And certainly a skilled artisan would have had no
reasonable expectation of success in incorporating such a polymer into the compositions of Graves et al. in an effort
achieve the claimed invention, in light of this teaching.
Ex. 8 10050432 Thus, Orthoefer et al. '122 teaches that sulfurous ion, such as salts of sulfite or bisulfite, should be added during the water
extraction step of the soy starting material to decrease the viscosity of the extract which, in turn, decreases the viscosity of the final soy isolate product. On the other hand, Sailer '554 explicitly teaches against the use of sulfurous ion in the water
extraction step of the soy starting material or any subsequent steps in the Sailer '554 process. For the foregoing reasons, Sailer
'544 and Orthoefer et al. '122 clearly teach away from one another regarding how to obtain a soy product having a desired viscosity. Therefore, one of ordinary skill in the art would not combine the disclosures of Sailer '554 and Orthoefer et al. '122
in the manner relied upon by the Examiner.
Ex. 9 10053013 Furthermore, Goertzel et al. teaches way from the claimed invention. Goertzel et al. assigns access based on virtual locations,
where "access is assigned based upon the least trusted location (i.e., the "weakest link") through which a user's packets are
being routed." Col. 6, lines 14-17. Goertzel et al. recites "[i]ndeed, a RAS 681, 682 dial-up user may be closer in physical
distance than [a] user at a remote office 641, connecting via a T1 line, even though the dial-up user will ordinarily be considered less secure." Col. 5, lines 20-23. Goertzel et al. thus teaches away from reliance on physical locations to
determine and select security levels, providing strong indicia of nonobvious of the present invention…
Ex. 10 10053085 A review of Cable and Isenberg makes clear that the cited references teach away from combination with one another. The Background section of Cable '903 expressly criticizes the electrodes described by Isenberg in US Patent 4,582,766, noting that
in devices made according to Isenberg, the electrolyte is bound to the electrode, which bonding results in undesirable
mechanical and structural complications (see Cable '903 at col. 2, lines 38-41)… The cited references likewise teach away from the claimed invention. While the claimed invention recites that the anode and electrolyte are secured to one
another, the cited Cable '903 reference expressly criticizes such bonding and instead advocates the disposition of so-called
"microslip zones" between the electrolyte and the electrode components (Cable '903 at col. 5, lines 53-68).
39
Ex. 11 10057198 Thus, Shanbrom conforms to the standard approach within the pharmaceutical industry of not heating and therefore not
vaporizing the medicament subject to thermal decomposition. Thus, Shanbrom teaches away from the present invention in
which degradable physiologically active compound is heated to form a vapor. As Shanbrom teaches to avoid heating of the
medicament, Shanbrom cannot be said to teach a method for decreasing the decomposition of the medicament upon heating…
and thus teaches away from the instant invention.
Ex. 12 10058495
Indeed, the very art on which the Examiner relies teaches the commonly known need to anneal UHMWPE to remove free radicals therefrom subsequent to irradiation crosslinking… No one skilled in the art would be motivated to radiation
crosslink the finished acetabular cup of Devanathan, as proposed by the Examiner, since to do so would require post-
irradiation annealing at a temperature which closely coincides with the melting temperature of PMMA. Unlike UHMWPE which remains workable above its melting temperature, PMMA turns to a liquid (with the approximate viscosity of
honey) above its melting temperature. As such, post-irradiation quenching would completely distort, and effectively destroy,
Devanathan's acetabular cup.
Ex. 13 10083205 In addition, the Ward et al. and Beaupre references teach away from perforating the aluminum vapor barrier layer of Beaupre.
Ward et al. teaches that any holes in the vapor barrier material are undesirable. Ward et al. states that "vapor can
penetrate [pin holes] to wet the insulation or condense on the underside of the superimposed metal decking. As is well known, wetted insulation has significantly reduced resistance to thermal conductivity than does the same insulation when dry." Ward
et al. col. 1, lines 43-52. Beaupre teaches that it is undesirable to deteriorate the vapor barrier characteristics of a foil vapor
barrier, as would certainly be the case if the foil layer were perforated… Claim 8 is not obvious in view of Beaupre, Ward et al. and Denenberg, because the Beaupre and Ward et al. patents teach away from perforating the aluminum vapor barrier layer
of Beaupre.
Ex. 14 10087662 The skilled reader of this reference would not consider it obvious to replace the silicon nitride taught by Glaser with silicon dioxide… Glaser repeatedly lauds the benefits of using high refractive index material for the lower coating… Further, Glaser
expressly indicates that the refractive index of the described layers should not be substantially modified… Thus, Glaser
teaches against any modification that would involve a substantial change in the refractive index of the described
films… Since replacing silicon nitride (or bismuth oxide or tin oxide) with silicon dioxide would result in a very substantial
decrease in refractive index, Applicant submits that this modification would be far from obvious.
Ex. 15 10090293 No such suggestion or teaching is provided by Reiley or Ferree. Rather, Ferree does not mention facet joint replacement or facet joint ailments, and Reiley teaches against the use of artificial discs where facet joint ailments are present… Indeed, not
only Reiley, but the general state of the prior art teaches away from the combination of facet joint replacement with use
of an artificial disc. As expressed in Applicants' specification, "contraindications for artificial discs include arthritic facet joints, absent facet joints, severe facet joint tropism or otherwise deformed facet joints."
Ex. 16 10090358 Substituting the invention of Scherson into the invention of Henley would defeat this purpose: creation of a sustained, oxygen-
enriched environment through oxygen generation from ambient air under hyperbaric pressure. Indeed, the negative pressure supplied by Henley to drain wound surfaces would entirely teach away from the hyperbaric pressure teaching of Scherson for
maintenance of an oxygen-rich environment. Accordingly, one of skill in the art would not think to modify Henley by
combining it with the device of Scherson, because, at the least, the two inventions teach away from each other.
Ex. 17 10093199 In fact, Yamaguro teaches to the contrary. As explained above, the invention of Yamaguro is designed for low temperature
engine environment. If a heater were installed at the hydrogen generating unit, the heat would dramatically affect the catalyst operation. Specifically, as shown in Figure 7 of Yamaguro, even a slight amount of additional heat at the catalyst inlet is
detrimental to NOx conversion efficiency (col. 6, lines 20-24). … He teaches away from the introduction of any additional
heat in the exhaust or near the exhaust inlet (Figure 7; col. 6, lines 20-24).
Ex. 18 10094253 Additionally, the primary teaching of Hossom is that components of the test device should be removable to increase the
flexibility of the device and to optimize the device for the particular assay to be performed. This runs directly counter to the
claimed apparatus, i.e., a guide tray having a one-piece construction. Thus, Hossom teaches away from providing a guide
tray having a one-piece construction.
Ex. 19 10094265 As stated above, Melis et al, clearly states that antagonists (or blockers) of the dopamine receptors reduce or eliminate the
sexual stimulation seen with agonists in mammals (see at least col. 2 on page 20, col. 1 on page 22, and col 2, under fluman Studies, page 32). Therefore, Kulagowski et al, in US 5,792,768 not only does not present a motivation to use the claimed
antagonists in treating sexual dysfunction, but also teaches away from the present invention. Accordingly, one skilled in the
art would not have a reasonable expectation of success in using the compounds of Kulagowski et al, in US 5,792,768 in
treating sexual dysfunction, especially in view of the teachings of Melis et at.
Ex. 20 10097257 However, those skilled in the art know that there are a multitude of reasons why the 1-15 nm particle size regime is generally
not accessible to traditional emulsion polymerization or dispersion polymerization techniques… For the invention described in Chandler, the practitioner would want the maximum effective concentration range to be as large as possible, and thus the
particle size to be as large as possible or practical. The applicants have, surprisingly, found that the use of very low particle
size polymeric nanoparticles actually reduces the viscosity of the dispersion (paragraph 0015), which is counter-intuitive
from the known art because smaller particle size usually results in higher viscosity… Given that the fields of invention
are so diverse, Chandler teaches away from our invention, and that one would anticipate increased viscosity by moving to
lower particle size, anyone of ordinary skill in the art would not consider Chandler to be prior art for the inventive step of very low particle size (1-15 nm) cross-linked polymeric nanoparticles with attached chromophores.
40
Figure 8. Examiner citations that teach away from focal invention
This figure shows the number of citations used by examiners to make obviousness rejections that actually teach away from the focal
inventions. Data come from applicants’ responses to examiners’ rejection letters. The figure compares high, medium, and low levels
of technological overlap between the focal invention and the examiner’s prior experience. The unit of observation is the application.
Dotted lines show the 95 percent confidence interval.
Table 7. Examiner citations that teach away from focal invention
This table reports results of a Poisson regression, in which the dependent variable is the number of examiner citations used for
obviousness rejections that teach away from the focal invention. The unit of observation is the application. The model includes fixed-
effects and clustered standard errors at the examiner level.
(1)
# examiner citations that teach away from
focal invention
ln(total claims) 0.270+ (0.154)
ln(law firm's average client size) -0.024
(0.061) # technology classes in applicant prior art -0.062*
(0.025)
Average year of applicant prior art -0.007 (0.015)
ln(applicant prior art citations) 0.486***
(0.100) % claims rejected for obviousness 0.819*
(0.334)
Number of examiner citations used for obviousness rejections 0.215*** (0.031)
Invention's technological overlap with examiner experience -0.949**
(0.321)
Observations 1,331
Number of examiners 242 Chi-sqr 149.7
df 8
Standard errors clustered by examiner; *** p<0.001, ** p<0.01, * p<0.05, + p<0.10
0.025
0.035
0.045
0.055
0.065
0.075
0.085
Low Medium High
# e
xa
min
er c
ita
tio
ns
tha
t te
ach
aw
ay
fro
m f
oca
l
inven
tio
n
Invention's technological overlap with examiner experience
41
Figure 9. Claims rejected due to obviousness that were later allowed
This figure shows the percentage of claims rejected due to obviousness that were later allowed. The unit of observation is the
examiner prior art citation, and percentages are calculated specifically for claims rejected based a given examiner citation. The
denominator is the number of claims rejected for obviousness by a given examiner citation, and the numerator is the number of these
rejected claims that were ultimately allowed. For 34 percent of the examiner citations used for obviousness rejections, all of the claims
rejected based on these citations were later allowed. For 37 percent of examiner citations used for obviousness rejections, none of the
claims rejected based on these citations were later allowed. This figure shows the distribution of percentage of claims allowed for the
remaining 29 percent of examiner citations for which some but not all of the claims rejected were later allowed.
Figure 10. Claims rejected due to obviousness that were later allowed
The vertical axis in this figure shows the percentage of claims rejected due to obviousness that were later allowed by the examiner.
The unit of observation is the examiner prior art citation, and percentages are calculated specifically for claims rejected based a given
examiner citation. The denominator is the number of claims rejected for obviousness by a given examiner citation, and the numerator
is the number of these rejected claims that were ultimately allowed. The figure compares allowances for examiner citations that do and
do not teach away from the focal invention. Dotted lines show the 95 percent confidence interval.
49.0%
51.0%
53.0%
55.0%
57.0%
59.0%
61.0%
63.0%
65.0%
67.0%
69.0%
Does not teach away Teaches away from focal invention
% o
f cla
ims
reje
cte
d f
or o
bvio
usn
ess
th
at
were l
ate
r a
llo
wed
Examiner citation used to reject claims for obviousness
42
Table 8. Regression models of claims allowed after being rejected
This table reports results from regression models of claims allowed after being rejected due to obviousness. The unit of observation is
an examiner prior art citation. All variables at the examiner citation level are computed using claims that were rejected for
obviousness using that examiner citation. In columns 1 through 3, the dependent variable is a percentage, in which the denominator is
the number of claims rejected for obviousness by a given examiner citation, and the numerator is the number of these rejected claims
that were ultimately allowed. These models are estimated using linear regression. In column 4, the dependent variable is the number of
claims that were rejected for obviousness using a given examiner citation but were ultimately allowed. Fixed-effects and clustered
errors are at the level of the examiner.
(1) (2) (3) (4)
Linear Poisson
% of claims rejected for obviousness
that were ultimately allowed
# of claims rejected due to
obviousness that were
ultimately allowed
Application level variables
ln(law firm's average client size) 0.006 0.005 0.005 0.003
(0.004) (0.004) (0.004) (0.010) # technology classes in applicant prior art -0.002 -0.002 -0.002 -0.004
(0.002) (0.002) (0.002) (0.006)
Average year of applicant prior art 0.002 0.002 0.002 0.003 (0.001) (0.001) (0.001) (0.003)
ln(applicant prior art citations) 0.013 0.014 0.014 0.032
(0.009) (0.009) (0.009) (0.020)
Examiner citation level variables
ln(# claims rejected due to obviousness) 0.006 0.006 0.006 0.995*** (0.005) (0.005) (0.005) (0.016)