Gregg Reynolds wrote::
My original question was motivated by the observation that a human reader of an expression of the form "e >>= f" , on seeing that f is constant, may pull the constant value out of f, disregard e and dispense with the application f e.
While a human reader may well do that, but it would be correct or wrong depending on the definition of >>=. The same is of course true for compilers. By the way, there is no "application f e". An example where it would be wrong to ignore e: sum ([1, 2] >>= const [21]) This expression should evaluate to sum [21, 21] = 42, not sum [21] = 21. There is nothing special with IO or >>=, so there is no need to introduce special cases for IO or >>= in a formal or informal semantics of Haskell. Tillmann