Apparent string shortening concomitant with letter crowding

In our previous studies of the crowding effect, we have observed that human observers tend to underestimate the length of a letter string (the number of letters in the string) when the letters are close to visual acuity, and the interletter spacings are small. In this study, we asked our observers to identify letters in randomly presented four-letter and five-letter strings. We found that, when a priori knowledge of the lengths of letter strings was not available, the probability of underestimating string length increased with decreasing interletter spacing. The causes of underestimation errors appeared to be the omission of an interior letter and the merging of two neighboring letters. Since our experiments were conducted in the foveal region, neither spatial uncertainty nor split attention can explain the underestimation errors. The effect of the point spread function of the eye on closely packed letter strings is discussed.