RE: arXiv (RE: Why Cornell's Institutional Repository Is Near-Empty)

["Alma Swan" <a.p.swan@talk21.com>]

Rick Anderson wrote:

> Yes, but these are preprints, correct?  If arXiv is hosting the 
> complete postprint contents of any physics journal (without 
> embargo), please let me know so I can cancel my library 
> subscription to that title.

Presumably physics subject librarians around the world know the 
answer to your second point.

I can answer your first point, though, and provide some other 
pertinent figures regarding arXiv:

1. First, (the same figures that I posted to this list last 
November) on the content of arXiv:

When we checked this in September 2005 the numbers of articles in 
arXiv were:

Total records: 336829
Those with a complete journal reference: 173901
Those with an annotation saying "Accepted": 15772

i.e. around 56% of articles have postprint status.

A further 14539 have an annotation saying "Submitted" and may be 
pre- or post- peer review.

The rest have no discernible status and must in most cases be 
preprints.

2. Nature Physics

I only did this exercise once, for the very first issue of this 
journal, and it would be worth repeating on subsequent issues. I 
carried it out because Nature had recently introduced a 6-month 
embargo on self-archiving and I wanted to see how that would be 
handled by the physics community. On the day Issue 1 was 
published in October 2005, I searched three sources for articles 
that appeared in the issue - arXiv, Google (Google Scholar wasn't 
around at that time) and author websites. The results were:

Letters to Nature Physics (6 in total):
2 with postprints in arXiv
2 with preprints in arXiv (one of these also had Nature's own PDF on the
author's home page)
2 not in arXiv (but one of these had Nature's own PDF on the author's home
page)

Articles (2 in total):
2 postprints in arXiv

I defined an article in arXiv as a POSTPRINT if it gave the 
Nature Physics page numbers, or said 'to appear in Nature 
Physics', i.e. it had been accepted for publication.

So only one paper out of 8 was not freely available on the web 
somewhere immediately upon publication. In fact, they were there 
before publication but of course I had to wait for the Issue 1 
ToC before I could carry out the exercise.

Regarding how far before publciation they were available in 
arXiv, the dates of depositing the four postprints (remember that 
the journal publication date was the beginning of October) are 
shown below. The figures in brackets after them show the usage 
activity of those articles from arXiv - as measured by Citebase - 
by 5th October, i.e. nominally two or three days after official 
publication, in the format [number of citations, number of 
full-text downloads]: 11 March [1,10] 27 June [5,11] 26 July 
[0,3] 26 September [3,20]

The 2 preprints had also been read and cited by the time of publication,
too:
i) [3,24]
ii) [0,2]

Alma Swan
Key Perspectives Ltd
Truro, UK