Hello Thanks for the quick help with this. I thought about the idea that lookAhead might be the cause of the positioning bug but then discarded that idea because I thought lookAhead should never lead to an error past wherever the input position is now considering it doesn't consume any input. I am aware of the issue with the error message position and the output, I was still working on improving that when I was puzzled by the fact that the error message I specified wasn't even returned to me. As for try, I believe I need it to make sure the input I consume one character at the time, before I know if I will reach another valid match (or any at all) does not stay consumed when my parser fails. I am still very much in the experimental phase as far as writing Parsec combinators beyond very simple stuff is concerned so I am open for suggestions on how to improve it in a way that doesn't need "try". Thanks again for all the help and especially for the patch. After applying it I do get the error message I specified. I noticed there are still some other problems in the code. In particular it doesn't work as intended in cases like this one: parseTest (do; r1 <- anyOf ["Hello", "Hallo", "Foo", "HallofFame"]; r2 <- string "fbla"; return (r1, r2)) "Hallofbla" where it should (according to my goal) return no parse error but instead accept "Hallo" and allow the string parser to consume the rejected suffix but I will try to fix that. Matthias Hoermann