The study investigated parafoveal preprocessing by the means of the classical invisible boundary paradigm and a novel manipulation of the parafoveal previews (i.e., visual degradation). Eye movements were investigated on 5-letter target words with constraining (i.e., highly informative) initial letters or similarly constraining final letters. Visual degradation was administered to all, no, the initial, or the final 2 letters of the parafoveal preview of the target words. Critically, the manipulation of the parafoveal previews did not interfere with foveal processing. Thus, we had a proper baseline to which we could relate our main findings, which were as follows: First, the valid (i.e., nondegraded) preview of the target words' final letters led to shorter fixation times compared to the baseline condition (i.e., the degradation of all letters). Second, this preview benefit for the final letters was comparable to the benefit of previewing the initial letters. Third, the preview of a constraining initial letter sequence, however, yielded a larger preview benefit than the preview of an unconstraining initial letter sequence. The latter finding indicates that preprocessing constraining initial letters is particularly conducive to foveal word recognition.