I am trying to find information on one of the many exchanges between Fisher and (I believe, but cannot be sure) Neyman. I believe the exchange took place at one of the Royal Statistical Society meetings, but I cannot be sure anymore. I read about it years ago and have forgotten it.

The discussion was about Fisher's randomization inference (i.e. the famous Lady Tasting Tea experiment). Fisher's interlocutor pointed out that the null hypothesis was what we, today, call a sharp or strong null -- no effects for any units. Fisher replied that this null was the one of most scientific interest, as opposed to "no effect, on average, across all units."

Again, I think this was a discussion with Neyman, but I cannot be sure. Anyone who can describe this discussion in more detail or -- ideally -- point me to a source discussing it would be greatly appreciated.

  • 5
    $\begingroup$ I think this is a perfectly on-topic question for Cross Validated, but it's worth mentioning there is a separate History of Science and Mathematics Stack Exchange. I'm not sure how many statistics history questions they get there, but if you don't get a strong response here it might be worth asking for this question to be migrated there. (I wouldn't recommend cross-posting.) $\endgroup$
    – Silverfish
    Feb 19, 2016 at 15:10
  • 1
    $\begingroup$ HSM is a small Stack Exchange but very interesting - they have some extremely knowledgeable posters about the development of mathematics. There are quite a few people on CV who are interested in the history of statistics and particularly the historical disputes about hypothesis testing, so I reckon you might be best to leave it here for now. (If you do ever want to have it migrated, I believe you can "flag" your own post to get moderator attention.) $\endgroup$
    – Silverfish
    Feb 19, 2016 at 15:27
  • 6
    $\begingroup$ The paper N. read, & F.'s contribution to the ensuing discussion, are here: Neyman et al. (1935), "Statistical Problems in Agricultural Experimentation", Supplement to the Journal of the Royal Statistical Society, 2, 2, pp 107-80. It concerned Latin squares & RCBDs rather than FET. A recent discussion (with many useful references) is Sabbaghi & Rubin (2014), "Comments on the Neyman–Fisher Controversy and Its Consequences", Statistical Science, 2, 20, pp 267-284. $\endgroup$ Feb 19, 2016 at 15:38
  • 4
    $\begingroup$ Unfortunately Salsburg's book is just not reliable on history. The story of the lady tasting tea was first in Fisher's book on design of experiments, first published in 1935. A fuller version is in Joan Fisher Box's 1978 biography of her father. Salsburg's version, which is his lead anecdote, gets both place and people wrong, as I pointed out in my review in Biometrics 57: 1273-1274 (2001). Although many people have found the book entertaining, it's full of similar historical mistakes and confusions. For several other examples (far from a complete list), see my review. $\endgroup$
    – Nick Cox
    Feb 19, 2016 at 16:46
  • 1
    $\begingroup$ Scortchi, if you submit your comment as an answer, I'll accept it. That was the exchange I was looking for, though I remember it being a bit more clear (for some reason I recall Fisher stating explicitly that the strong null was the one of scientific interest) -- but I think I was just reading too much into their words. Thanks, though, this is exactly what I was looking for! $\endgroup$ Feb 20, 2016 at 10:35


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.