Carolina3 experiments tested the relations between monocular or binocular viewing, visual-field placement of stimuli, and interletter spacing in the distribution of errors over letter positions in the letter span. Exposure time was 100 msec. The most important variable in accuracy within the letter span was found to be the relative location of a letter within the 8-letter line. Errors were fewest for the leftmost letter in the line regardless of variations in absolute retinal locus. These data imply that experiments on the letter span might better be conceptualized as "memory" studies than as "perception" studies.