Saturday, February 11, 2012

Against joint theory-experiment papers

In an early post I argued against experimentalists feeling a compulsion to present a theoretical "explanation" for their data. I should have also expressed concern about how in a similar vein many theory papers waffle on about how the calculations presented are relevant to specific experiments.

Reading through More and Different, I found an interesting Physics Today column that Phil Anderson wrote in 1990, Solid State Experimentalists: Theory should be on tap, not on top.
He argues for re-instatement of Cornelius Gorter's rule that theory and experiment should be published in separate papers. It is quite possible that they will both not be correct.
Much more serious is the distortion of priorities, of communication and of the refereeing process that occurs when excessive weight is given to theoretical interpretation. We don't want to lose sight of the fundamental fact that the most important experimental results are precisely those that do not have a theoretical interpretation; the least important are often those that confirm theory to many significant figures.
Although written in the early days of high-Tc superconductivity Anderson's arguments and concerns seems just as relevant today.


  1. Your link to
    Journal Club for Condensed Matter Physics
    is dead...

  2. This sounds like an appealing idea and I certainly agree with the goal of remembering which results are most important, but I think in reality there are some difficulties with this kind of approach.

    I think it makes a lot of sense when experimentalists are studying a new system and find interesting results, as long as they are confident about them.

    However, when experimental results seem to contradict an existing theory, there's a substantial chance that the experiments are wrong, i.e. they are not really measuring what the experimenters think. In the cases where the theory is indeed wrong, the demonstration of this without a replacement theory offers an enormous opportunity to theorists, but it's very difficult to distinguish the two cases.

    Of course, this is never an argument for bad theory, but I would be more willing to believe an experimental result if there is an explanation for why it defies what was believed.

    From the other direction, I think the more abstract theory can survive without reference to experiment, but much theory needs to refer to experiment in some way. For instance, theorists need to be clear about what would really be experimentally measured if their theory is correct. It's too easy for theorists to report what is honestly internal to their theory instead of experimentally relevant variables.