In a previous experiment [P. L. Scilley, "Evaluation of a vibrotactile auditory prosthetic device for the profoundly deaf," unpublished Masters thesis, Queen's University, Kingston, Canada (1980)] two normal subjects learned to identify 70 and 150 words, respectively, using the Queen's Tactile Vocoder. In the present experiment, the most advanced subject continued word learning until a tactile vocabulary of 250 words was acquired. At this point randomized tests were given to obtain an indication of final performance level. From these data conditional probabilities of correct response for each stimulus word and significant confusions were obtained, which provides insight into the advantages and present limitations of the tactile vocoder.