Fig 1. Illustration of the predictions made by the bottom-up, non-word-based, CoG hypothesis.

Under this assumption, saccade amplitudes should be overall modulated by word-processing difficulties regardless of word boundaries, but to greater extents for shorter and less eccentric words, that more greatly benefit from peripheral preview. Are represented the hypothetical frequency distributions of saccades’ landing positions on the line of text for easy (plain lines) vs. difficult (dotted lines) peripheral words (N+1), separately for short and long words (left and right panels), and close, intermediate, and far saccadic launch-site distances to the beginning of the words (from upper to lower panels). Light-grey rectangle areas represent the horizontal extent of the words. Landing positions falling within those areas correspond to within-word landing positions, while landing positions to the right of these areas result in word skipping.