The journal Immunity says on this webpage
"As a matter of publishing ethics, we cannot consider any paper that contains data that have been published or submitted for publication elsewhere."
Why not? What if the two papers analyze rich sources of data that overlap only partially? What if they have completely different ambitions? Is the only meritorious activity in science the act of gathering data, and to hell with the analysis? And what does this have to do with ethics?
I think you're reading a little bit more into the policy than is actually there - many papers in this type of journal essentially contains data - gels, lab results, etc. That's what they're talking about. What they're not talking about is the corpus of data generated from a whole study.
Why not?
Because the goal is not to allow researchers to publish repeated analyses of the same dataset.
What if the two papers analyze rich sources of data that overlap only partially?
It's likely not what they're talking about - a rich source of data likely means that the data hasn't been published recently. What they don't want is "The same data as last time, now with a marginally different story".
What if they have completely different ambitions?
I'd be impressed if you managed to get two completely distinct analyses out of the same data - and indeed, the stance of the journal would likely be that they'd be willing to accept a decrease in salami slicing of papers at the cost of the occasional genuinely novel analysis that's built entirely off previously published data.
Is the only meritorious activity in science the act of gathering data, and to hell with the analysis?
That's reading way too far into the policy - the paper should also contain that analysis. What they don't want you doing is writing four papers on the same single assay or whatever.
And what does this have to do with ethics?
The idea that a paper should represent a single, stand-alone contribution to science, and that ever more esoteric spins on the same data is a net negative is an ethics in science stance. One you may disagree with, but it is one.
This is a guess but:
The sentence in question comes at the end of the paragraph entitled 'Related Manuscripts', which says they need to see any similar papers you have. This suggests to me that the primary point is that you can't publish the 'same' paper twice, even if you change the wording.
Another potentially relevant point, which I haven't seen mentioned yet, is that you have to be very careful when reusing a dataset - the more questions you ask about some data, the more likely you are to get a positive answer by chance, so that has to be taken into consideration in the analysis. If you present a single result from the data, it's therefore important to know what other questions have been asked.
I think the keyword here is "contains". The paper should not present any data, that has already been presented elsewhere, because that would be considered double publication of the data.
I'm pretty sure that this is not meant as a restriction on using previously published dataset in your papers. For example, I don't think Immunity would want to restrict papers from using data from the Human Genome Project or public census data.
(Am not sure how this is supposed to work with significant extensions of previous datasets though.)
You must only write one paper in which you explain completely how you acquired the data and why exactly you did it the way you did.
The full story. A later paper which reveals that there were some additional motives, not to say parameters, turns the original one into a piece of scientific malpractice, eligible for retraction.
