RE: Does the arXiv lead to higher citations and reduced publisher downloads?

["David Goodman" <David.Goodman@liu.edu>]

There are other possible interpretations as well.

If we focus on the equivalent downloads for OA and TA articles, 
then, since is pure mathematics where all early use can be 
assumed to be browsing, this might mean that

1. all browsing was done on the publishers site --possibly 
because of its known completeness or

2.  those browsing on arXiv, upon encountering a citation of 
abstract to a TA article

 	a. did not get it, because of the lack of access
 	or
 	b. did not bother getting it, because the OA articles
 	are or would be assumed to be superior,
 	or

2.1 those browsing on arXiv would not have found many TA articles 
represented even by an abstract or link.

These are not intended to be the only alternate hypotheses any of 
us could easily find others.  We do not need more hypotheses, we 
need more data and more analysis. Accurate use data has only 
recently become available, and it is not possible from this and 
similar studies to decide between hypotheses. We should not 
over-interpret what is, after all, preliminary data. (Additional 
information that would be helpful would be continuing publishers' 
series with data for all recent years, and similar data by 
journal from arXiv and other repositories.)

In other subjects than matematics, where much of the 1st and 2nd 
year use would be for known references, the situation is much 
more complex. As SH and all others say here and elsewhere, many 
factors must be taken into account, and we do are not yet able to 
do this completely. But the different factors are not in 
competition--they all contribute to the overall advantages of OA. 
As scientists, we want to know the mechanism, and assume as we do 
everywhere, that it will lead to further understanding and even 
better practical results. As practical librarians or publishers, 
we look to the resolution of important but previously undecidable 
problems.

Citation data (as contrasted with use data) has been available 
for many years, and studies have long been in progress. Many of 
our assumptions, such as the long half-life of journals in 
mathematics, are based on such studies. For Phil's citation data, 
the results are similar to those found elsewhere, and their 
interpretation has similar ambiguities.

Fortunately, the many reasons why OA is good do not depend on the 
interpretation. If these results had been made available only in 
a TA journal, it would be long before these discussions could 
even begin, and not all would be able to participate.

Dr. David Goodman
Associate Professor
Palmer School of Library and Information Science
Long Island University
dgoodman@liu.edu