Is this correct and why is that so? Are scheme's call/cc and haskell's callCC semantically different?
Yes, Haskell's `callCC` is a user defined function and Scheme's is a magical primitive. Under the hood Scheme should behave as if *every* expression lived inside the continuation monad. The Haskell equivalent would be something morally equivalent to foo :: Cont r Int foo = callCC $ \cont -> (+) <$> cont 1 <*> pure 1 Also, to do a lot of the fun games we can play with Scheme's call/cc, we need ContT r IO vs plain old Cont. For example, this program works "as expected" and returns 1 foo :: ContT r IO Int foo = do ref <- lift (newIORef undefined) result <- callCC $ \c -> lift (writeIORef ref c) >> pure False if not result then lift (readIORef ref) >>= ($ True) >> return 0 else return 1