At http://book.realworldhaskell.org/read/concurrent-and-multicore-programming.h... , just after the code for `sillySort`, it says <quote> "In this mistaken example, we evaluate each sorted sublist to WHNF. Since the outermost constructor in each case is just a single list constructor, we are in fact only forcing the evaluation of the first element of each sorted sublist! Every other element of each list remains unevaluated." </quote> My ruminations on the subject ended up in the following analysis, which is quite contrary to that of the authors': In the general case of the type inside the list being just an instance of Ord, it is *not* the case that the first element is necessarily evaluated; it is evaluated just enough (along with the elements of the *rest* of the lists greater and lesser) to enable < and >= to do their job inside the list comprehensions. Now, if these elements are Int:s or Integer:s then certainly, yes, the first element *is* evaluated, but so is the rest of the elements of the lists greater and lesser. So, greater and lesser are not only reduced to WHNF, but to NF in the case of sillySort working on [Int] or [Integer]. Am I mistaken in my analysis? If so, where? /Fredrik -- haskellBlog: http://www.monoid.se/categories/haskell/