Four eye-tracking experiments were conducted to understand how sentential context and lexical factors affect word recognition during reading. Experiment 1 examined whether readers use preceding sentential context to pre-activate a specific word and whether any processing cost is found when the prediction is wrong. The results showed that readers obtain a processing benefit when the target word was expected based on a strongly constraining context, whereas they experienced a processing cost when the target word was not the expected one even though it was semantically plausible into the context. Experiment 2 investigated how word recognition is influenced by prior activation of lexical information due to word repetition within a sentence. The result showed that gaze duration on a target word with many neighbor words was shorter relative to a target word with few neighbors but only when the target word was repeated; when the target word was not repeated gaze duration did not differ as a function of neighborhood size. This interaction indicates that word recognition at the orthographic level can be influenced by repetition-induced lexical activation. The null effect of the orthographic neighborhood size in the unrepeated condition was unexpected. Previous studies using the lexical-decision task have consistently shown a facilitative effect of orthographic neighborhood size. Therefore, Experiment 3 and 4 studied the role of neighborhood size during sentence reading with better controlled stimuli. The results showed an opposite pattern of results between gaze duration and word skipping such that gaze duration was longer when a word had many neighbors than when one had few neighbors, whereas skipping rates were higher in the many neighbor condition than in the few condition. The results indicate that having many neighbor words inhibits processes responsible for precise recognition of a word, but that it facilitates word skipping by increasing global lexical activity.