I am not a Twain scholar (my field is CS and Stat) but I think Dagnall's point is that before someone asserts a new-found "truth" there ought to be more going for the author than "maybe", "perhaps", "seems", etc. I know in stat words like that just don't belong in a paper except in describing the situation before your paper or in the section on future directions. Dave Gomberg