Previous research has demonstrated that familiarity with the orthographic structure within a letter string can facilitate the processing of the component letters. The current research was directed at discovering the psychologically relevant properties of this structure. Two fundamental descriptions were independently varied in the construction of six-letter nonword strings. A probabilistic description based on the frequency of occurrence of letters in each position was factorially combined with a rule-governed description defined in terms of graphemic and phonological constraints. College sophomores and sixth-grade readers were asked to indicate whether or not a predesignated target letter was present in these strings. For both groups of readers, orthographic regularity and summed positional frequency were found to have only a small facilitative effect on reaction time (RT). In contrast, RTs to say "no" increased dramatically with increases in the number of letters in the catch string that were physically similar to the target letter. In another experiment, the letter string was presented for a short duration, followed immediately by masking stimulus and then the target letter. College students indicated whether or not the target was present in the test string. Accuracy of performance was critically dependent on the orthographic regularity and summed positional frequencies of the letters in the test string. No effect of letter similarity was observed. The large differences that were observed between these two tasks were accounted for in terms of the stages of processing that are critical for performance in the tasks.