Prediction of modifiers during sentence processing

Many studies have recently shown that comprehenders can anticipate upcoming information on the basis of only part of a sentence. However, all experimental evidence to date has only dealt with the prediction of grammatical heads and their complements. Here, we address the question of whether people also predict modifiers, syntactically non-obligatory elements. Two experiments were conducted to examine whether the processing of a post-nominal relative clause is facilitated when it can be predicted. The results showed some evidence for prediction of the modifiers, which is consistent with a surprisal-based view of processing complexity (Hale, 2001), but not with Gibson's (1998) notion of prediction memory cost, which only applies to heads.