The brain–machine interface (BMI) has a number of electroencephalogram (EEG) input support tools, including the P300 speller, which uses the event-related potential (ERP), P300. This potential is known to be affected by psychological variables such as the user’s attention and discrimination accuracy. In our study, we investigated the effect of fatigue caused by mental load on discrimination accuracy. P300 can be divided into two types: P3a, which occurs in the frontal center, and P3b, which occurs in the parietal center. As there are few studies using P3a as a discrimination method for P300, we included P3a in the discrimination index to check discrimination accuracy. The results showed that the discrimination index using P3a was generally lower than that using P3b in terms of percentage of correct answers, but some participants showed a high percentage of correct answers, which may lead to an improvement in discrimination rate depending on individual characteristics.