Previous research shows letter-fragment masks and non-letter-fragment fields have different effects on performance with briefly presented alphabetic targets. However, popular accounts of these differences ignore mask configuration. Over a series of experiments, configurational effects of letter-fragment (LF) and non-letter-fragment (NLF) masks were compared. When the configuration of LF masks matched word boundaries, performance with exterior letter pairs from words improved, whereas performance with illegal exterior-letter pairs and single letters was unaffected. When the same changes were made to NLF masks, only an overall drop in performance occurred, with no selective effect on target type. Thus, although LF mask configuration selectively affected lexical processing, NLF mask configuration produced substantially different effects, indicating problems with contemporary accounts of masking differences that ignore influences of mask configuration.