At least when your typing, you get a word wrong and immediately fix it. With Voice Recognition, you have to parse back through the whole message and pick out th

At least when your typing, you get a word wrong and immediately fix it. With Voice Recognition, you have to parse back through the whole message and pick out the corrections one at a time, many corrections if it has a 20.4% error rate. Humans don't do well with that, they look at something and their brain fills it in correctly for them, it's easy to miss typos in your own work. When I used to do helpdesk and someone had password problems I'd have them retype their usernames, because they would not notice it was wrong, but it was often the problem.