A persistent problem with personal digital assistants (PDAs) is the difficulty of entering data into the devices. The best current solutions to the problem are small soft keyboards and constrained handwriting recognizers. Another solution is use of speech. PDAs do not yet have the power to support full speech dictation, but they do have sufficient power to support voice spelling. Voice-spelling problems include the high acoustic confusability between certain letters of the alphabet and the difficulty of memorizing code words for the letters of the alphabet. This paper describes several User-Centered Design studies conducted to develop a voice-spelling alphabet for PDAs that overcomes these problems, including: (1) the development of a model of user performance to assess the potential of voice spelling as an alternate input method for PDAs, (2) Web-based surveys for determining the words that people tend to associate with the letters of the alphabet, (3) accuracy experiments used to tune the final voice-spelling alphabet, and (4) the development of a graphical user interface for displaying code words as a prompt when voice spelling is used. The results of these studies suggest that it would be worthwhile to develop a working voice-spelling system for PDAs in the future.