OBSOLESCENCE OF THE PATENT LITERATURE

DOIhttps://doi.org/10.1108/eb026614
Pages32-52
Published date01 January 1976
Date01 January 1976
AuthorC.V. CLARK
Subject MatterInformation & knowledge management,Library & information science
OBSOLESCENCE OF THE PATENT LITERATURE
C. V. CLARK
Medical Research Council Laboratory Animals
Centre,
Carshalton,
Surrey
Patents deserve bibliometric study both for their own
sake
and because their
formality can be exploited. Here the year by year issue of US patents since
1836 is used to correct for growth their apparent aging, as obtained by a
synchronous study of citations made by (a) US patent examiners and
(b) periodicals. Apparent and corrected aging are treated in terms of condi-
tional probabilities. In (a), whereas the recent apparent aging of chemical
patents is much faster than that of the whole, after correction the rates are
very
close.
Using a very broad (trichotomous) subject classification, no cut-
off dates for novelty searches can be established if total recall
is
the
goal.
The
strictly retrospective method of such searches is invoked to explain why a
linearity found here in one type of corrected aging function has also been
found for a search file truncated at 1920. In (b) the now classic exponential
form of aging applies back to the twenties, but older patents are cited too
frequently to conform. The deviation is graphically even more striking
after correction for growth and is probably due to citations made for
historical purposes.
INTRODUCTION
IT HAS recently been confirmed1 that despite the frequent neglect of the patent
literature as a source of technical information, much is published there which
appears nowhere
else.
A series of reviews and critical bibliographies of the litera-
ture of patent documentation2 and a review of work on literature obsolescence3
reveal that there
is
a corresponding neglect by library and information science of
patent literature in general and of its obsolescence in particular.
It seems likely that the neglect of the patent literature is due to its forbidding
formality, but we argue that this formality is an advantage to bibliometric work
in that it makes for a simpler, more easily quantified system, the study of which
will throw light on the processes of aging in literature in
general.
It
is
the purpose
of this paper to begin such a study,
as
well as to report some quantitative findings
about a scientifically and economically important literature.
OBSOLESCENCE IN GENERAL
We can distinguish as many classes of obsolescence as uses which decline, and
informally classify use on the experience of both documentalists and users. But
we are far from having a quantitative index for each type of
use,
and some of the
indices which are employed, notably citations, have no direct correspondence to
an informal, common-sense classification of
uses,
while
a
classification based only
on available indices is formally incomplete and at odds with common-sense.
However, let
us
assume that we have
a
well defined
use
and
a
measurable index
of it, u. A comprehensive study of the obsolescence of a collection of documents
would yield the values of t2 and of
u
for
all
tl, where t2
is
the publication date and t1
Journal
of
Documentation,
Vol. 32, No. 1, March 1976, pp. 32-52
32
March 1976 OBSOLESCENCE OF PATENT LITERATURE
the time at which u is measured. Since t1 and t2 do not determine u, the results of
such a study could best be represented as the conditional probability distribution
where, in an obviously motivated notation, U, T1 and T2 are discrete random
variables. We also introduce T = T1
T2, the age of a document at the time its
use is measured.
Line and Sandison3 distinguish between the following approaches to studying
obsolescence: (a) synchronous, (b) diachronous, (c) diasynchronous. The most
that can be found by each approach is:
It can thus be seen that, theoretically, synchronous and diachronous studies are
each a restricted type of the full diasynchronous study, even though, experimen-
tally, a diasynchronous study may consist of a sequence of synchronous ones for
different values of t1.
CITATION STUDIES
We now restrict ourselves to studies in which citations are the index of
use.
The
times t1 and t2 are now both publication dates, of respectively the citing and the
cited documents. They do not refer to instants but to finite intervals to which
documents are assigned, say
1
year for t1 and 1 or 5 years for t2.
Usually the citing and cited literatures are kept as similar as possible, though
there are considerable methodological and experimental problems, notably with
regard to subject matter. Formally, each document d1 in
L1,
the citing literature,
will need an identifier x1 and a publication date
t1,
and similarly for the cited
literature L2. Thus
The set of citations, S, is the binary relation
In a synchronous study, t1 is the same for all d1 and therefore could be omitted,
with t2 being unambiguously replaced by t = t1
t2,
which
is
the age of the cita-
tion
(d1,
d2)
as well as of d2
itself.
The sets S and L2 can be partitioned into equivalence classes according to the
age of d2, so that St is the set of citations of age t and L2,t the set of all documents
of the same age. What is usually done in experiments is to find | St | by counting
citations according to their age. This gives access to a conditional probability
distribution of T:
The conditioning event C is that a particular document is cited by another,
(d1,
d2)
єS, and not just by some document or other,
(Ǝd1)
(d1,
d2)
єS.
The result (7)
may be obtained as follows.
33

To continue reading

Request your trial

VLEX uses login cookies to provide you with a better browsing experience. If you click on 'Accept' or continue browsing this site we consider that you accept our cookie policy. ACCEPT