Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!seismo!lll-crg!rutgers!sri-unix!hplabs!sdcrdcf!burdvax!psuvax1!berman
From: berman@psuvax1.UUCP (Piotr Berman)
Newsgroups: sci.math,sci.research
Subject: Re: How many people read an average research paper?
Message-ID: <2326@psuvax1.UUCP>
Date: Tue, 11-Nov-86 17:14:05 EST
Article-I.D.: psuvax1.2326
Posted: Tue Nov 11 17:14:05 1986
Date-Received: Wed, 12-Nov-86 04:57:31 EST
References: <7966@watdaisy.UUCP> <2483@phri.UUCP> <236@cartan.Berkeley.EDU>
Reply-To: berman@psuvax1.UUCP (Piotr Berman)
Distribution: sci
Organization: Pennsylvania State Univ.
Lines: 63
Keywords: Science Citation Index
Summary: Citation index is not so reliable
Xref: mnetor sci.math:181 sci.research:23

In article <236@cartan.Berkeley.EDU> larsen@brahms (Michael Larsen) writes:
>In article <2483@phri.UUCP> roy@phri.UUCP (Roy Smith) writes:
>
>>Science
>>Citation Index (put out by ISI Press, I believe; the same people who bring
>>you Current Contents)...
>>lists each year the papers which get cited the most often.  That's
>>really the only way to say "this was an important piece of work".  If more
>>people have cited your paper than any other paper, it's probably the most
>>important.
>
>	This is an interesting theory.  Let's see how it stands up against a
>short trip through the math citations index (CMCI).  The following observations
>can be duplicated by anyone with access to the 1976-80 edition.
>
>1.  Most people would consider Newton's _Principia_ to be a work of some
>importance.  CMCI gives 2 references.
>
>2.  O.K., that example was unfair because the paper in question is fairly old.
>It is hard to choose a contemporary mathematical work whose title most
>people will recognize.  Nevertheless, we can use Math Reviews as an indication
>of stature.  This usually staid index fairly gushed over Deligne's 1972
>paper "La Conjecture de Weil pour les surfaces K3."  I quote:
>
>	This paper is awe-inspiring.  Its powerful technique and arithmetic
>	insight should recommend it to a larger audience, although its
>	sophistication will cause trouble for almost all readers.
>
>How many references did this masterpiece garner from 1976 to 1980? Four.
>And that includes one by the author himself.
>
>3.  How about Durbin-Watson, "Testing for Serial Correlation in Least Squares
>Regression?"  This paper, in which a theorem of Von Neumann is rederived, has
>a reputation for being often cited.  It lives up to it in CMCI: 52 citations.
>
>4.  A random search through the columns of CMCI turned up a 1964 paper by one
>J. B. Kruskal which has 116 citations.  It is quite possible that I am 
>merely exposing my ignorance, but I confess to having heard of neither
>the mathematician in question nor the work.
>
I do not know the paper either, but the paper

  J.B. Kruskal, On the shortest spanning subtree of a graph and
		the travelling salesman problem

is cited by any textbook on data structure and algorithms.  
In general, if someone has a very deep and difficult theorem which 'closes'
certain topic, it will not be cited very much.  On the other hand, even
a weak paper which 'opens' an area of research which becomes very popular,
will be cited very often (often without reading, I guess, many times people
cite citations of others).

>The idea that a reference count is an accurate indication of the quality of
>a scholar must have a strong appeal to the bureaucratic mind.  Unfortunately,
>the real world doesn't seem to work that way.
>
>-larsen @ berkeley.edu.brahms

Here is the catch: there is not such a thing as a precise indication of
quality or importance.  But imprecise indicators have their value, if used
with care.

Piotr Berman