Most of what we know about parafoveal preprocessing during reading is based on the boundary paradigm in combination with parafoveal masks as a presumably neutral baseline condition. Recent evidence questions the neutrality of the baseline condition by showing that parafoveal masks inflict preview costs. Using a novel, incremental boundary paradigm we studied the effect of parafoveal masks. Manipulating the salience of parafoveal previews, we found that increasing salience of the masks resulted in increasingly longer fixation times on target words, but also on pretarget wordssuggesting preview costs. We conclude that the hidden preview costs of parafoveal masks in the classical boundary paradigm inflate the processing times for the baseline condition and hence lead to an overestimation of the preview benefit. Thus, the present study questions the validity of some of the conclusions drawn on the basis of the classical boundary paradigm.