Sentence comprehension depends on continuous prediction of upcoming words. However, when and how contextual information affects the bottom-up streams of visual word recognition is unknown. This study examined the effects of word frequency and contextual predictability (cloze probability of a target word embedded in the sentence) on N1, P200, and N400 components, which are related to various cognitive operations in early visual processing, perceptual decoding, and semantic processing. The data exhibited a significant interaction between predictability and frequency at the anterior N1 component. The predictability effect, in which the low predictability words elicited a more negative N1 than high predictability words, was only observed when reading a high frequency word. A significant predictability effect occurred during the P200 time window, in which the low predictability words elicited a less positive P200 than high predictability words. These data suggest that the contextual information facilitate visual-feature detection and orthographic pre-activation in the early stage of visual word processing. There is also a significant predictability effect on the N400 component; low predictability words elicited a greater N400 than high predictability words, although this effect did not interact with frequency. The temporal dynamics of the manner in which contextual information affects the visual word recognition is discussed. These findings support the interactive account, suggesting that contextual information facilitates visual-feature and orthographic processing in the early stage, and semantic integration in the later stage.