We evaluate the predictions of surprisal and cue-based theory of sentence processing using an eye-tracking corpus, the Potsdam Sentence Corpus. Surprisal is a measure of processing complexity based on a probabilistic grammar and is computed in terms of the total probability of structural options that have been disconfirmed at each input word. The cue-based theory characterizes processing difficulty in terms of working memory costs that derive from decay and interference arising during content-based retrieval requests of previously processed material (e.g., to incrementally build the sentence structure). We show that both surprisal and cue-based parsing independently explain difficulty in sentences processing and interestingly, they have an over-additive effect on processing when combined together.