With recent advances in cognitive biometrics, user authentication using brain-computer interfaces (BCIs), namely a pass-thought system, has received much attention in the cryptographic community. However, as the performance of BCIs hinges upon human factors, the feasibility of a pass-thought system needs to be examined. In this paper, we show that classification accuracy can be increased by increasing the number of test trials. More importantly, we propose a new information-theoretic measure (termed communication rate) based on a binary asymmetric channel model, for measuring the performance of pass-thought systems. We show that the maximum spelling rate of a pass-thought authentication system lies within the acceptable speed range of user comfort, indicating the practicality of this system. The relationships among classification accuracy, number of symbols, number of trials, and communication rate are investigated. The communication rate of a pass-thought authentication system is found to be directly proportional to accuracy and number of symbols, but it is inversely proportional to the number of trials.