E. Wijsman wijsman at
Fri Dec 5 19:52:43 EST 1997

Actually, the discussion was partly around the issue of citations when
there is a limitation to the total number of cited papers.  These
situations, as a proportion of all reference lists, are rare, and I think
it is important to distinguish what one has to do in this situation from
the usual situation where the number of citations is not limited.  I think
that in the more normal situation, one should try to cite the relevant
papers, even if this does require several citations.  

It is helpful if the authors of the software include in their
documentation a statement of how they want to have the software referenced
(if they don't, they shouldn't complain if only one citation is used).  It
is also fair to expect referencing to reflect the actual program use.  If
I use version 1.0 of some package in my analysis, I shouldn't necessarily
be expected to reference the paper describing version 2.0. If version 2.0
deals with some more difficult issues (e.g., loops) but my data are
comfortably analyzed by the simpler approach, I shouldn't need to
reference the more complicated approach just because it is the newest
version (especially if I didn't use it).  However, if I used version 2.0
because I had the complications, then I might also need to reference the
papers describing both versions since presumably 2.0 would build on 1.0. 
Sandy Hasstedt in her documentation for PAP is really good about stating
which references should be used - it is up to the user to check the
subroutine documentation for the references to the particular papers
describing specific algorithms for the specific subroutines used in a
particular analysis, but at least then one gets an appropriate set of mix
& match references that do describe the specific analyses actually

Alex is right that the appropriate papers should be referenced in most
situations.  It does help, however, for the program developers to provide
appropriate guidelines for referencing their work.

Research Professor			1914 N 34th St., suite 209
Div. of Medical Genetics and		Seattle, WA   98103
Dept. Biostatistics 			(Note:  do not mention the
BOX 357720, University of Washington 	 Univ. of Washington, and
Seattle, WA   98195-7720		 use this address only for	
phone:  (206) 543-8987			 express mail)
fax:    (206) 616-1973			email:  wijsman at

On 5 Dec 1997, Alex Schaffer wrote:

> The recent discussion thread suggested that it is a reasonable
> policy to cite no more than one paper per software package.
> I think with this policy linkage analysts shoot themselevs
> in the foot severely by communicating to software developers
> that they are not interested in improving exisiting packages.
> If I want to recruit a student to improve some exitsing software, the
> possibility that a resulting paper may be cited many times is
> important. I know we all should be doing science for its own sake, but
> let's face the fact that for students having an oft-cited paper makes
> a big difference in career prospects.
> If I have to tell a student: "Look, a lot of people will use your
> improved code, and we can write a good paper, but nobody will cite
> your paper, because I already wrote one," How would you expect the
> student to react?
> The sole reason I ask FASTLINK users to cite two papers is so that
> the (then) student authors who are on one of the two papers can 
> all get proper credit. 
> In most journal settings an extra citation costs the author nothing,
> except a little extra effort. 

More information about the Gen-link mailing list