Re-evaluating split-fovea processing in word recognition: Effects of word length