There have been some discussions lately in the science blogosphere about paper authorship in the biological/biomedical sciences, and I think it would be useful to tie those discussions together and add a few more thoughts here.
(1) The idea that paper authorship should be resolved explicitly (and in writing) at the beginning of a collaboration sounds very nice, and is a decent idea. But check out this scenario:
Two labs are collaborating on a project with the understanding that one lab's PI is the senior/corresponding author of the first manuscript that will result and that lab's trainee is the first author. All of a sudden, when the project starts to smell really good, the secondary lab's Pi starts suggesting all sorts of additional experiments to be done--surprise surprise--in her lab! Sometimes this is just an attempt to shift the balance of effort from one lab to the other, so as to open the door for "renegotiation" of senior/corresponding and first authorship.
Now what the fuck do you do? Fun, huh!?
(2) If you want scientists who, themselves, are trainees--either grad students or post-docs--and not 9-5 technicians to contribute experimental effort to a manuscript, it is appropriate that the effort be contributed on a collaborative basis, and with the expectation of co-authorship. To expect a scientist for whom authorship matters--i.e., anyone other than a technician--to contribute experimental effort to a manuscript without authorship is totally fucking ridiculous.
(3) What happens when someone contributes substantial experimental effort to a project, but the experimental results don't end up in the paper? Some have argued that such an individual should categorically not be an author. I say it's not so fucking simple. If the outcome of those experiments--positive or negative, or even failed--influenced the overall course of the project, then there is a good argument that the scientist who performed those experiments should be an author.
(4) The idea that lab heads who didn't "do" anything on a project--design or perform experiments or write the manuscript--should not be authors is completely fucking ridiculous. Who created and maintained the conceptual, methodological, and physical environment within which the project was performed? That merits authorship.
(5) Provision of a reagent--such as a novel clone, cell line, chemical compound, genetically altered organism, etc--to another lab prior to first publication of that reagent *always* merits authorship of the trainee(s) who generated the reagent and the PI of the lab in which the reagent was generated. While the generators of the reagent can, of course, decline such authorship, they are *always* entitled to it.