Do readers obtain preview benefit from word N + 2? A test of serial attention shift versus distributed lexical processing models of eye movement control in reading.

Two experiments tested predictions derived from serial lexical processing and parallel distributed models of eye movement control in reading. The boundary paradigm (K. Rayner, 1975) was used, and the boundary location was set either at the end of word n - 1 (the word just to the left of the target word) or at the end of word n - 2. Serial lexical processing models predict that there should be preview benefit only when the boundary is set at word n - 1 (when the target word will be the next word fixated) and no preview benefit when the boundary is set at word n - 2. Parallel lexical models, on the other hand, predict that there should be some preview benefit in both situations. Consistent with the predictions of the serial lexical processing models, there was no preview benefit for a target word when the boundary was set at the end of word n - 2. Furthermore, there was no evidence of parafoveal-on-foveal effects.