A multichannel cochlear prosthesis was evaluated using the method of Connected Discourse Tracking. Data were obtained from five subjects over a 10-week period. Significant learning effects were obtained both with and without the prosthesis. The method of orthogonal polynomials was used to obtain a statistically precise fit for each learning curve. The curves differed both in terms of shape and average rate of learning. The two best subjects showed substantial improvements, reaching tracking rates in excess of 90 words-per-minute. A method for representing prosthesis-based improvements, which takes learning effects into account, is developed and discussed.