A speech recognition system having a user interface that provides both visual and auditory feedback to a user. The user interface includes an audio sound or speech generator that produces three distinct sounds: an "on" sound signifying that the speech recognition system is on and actively awaiting vocal input; an "off" sound indicating that the speech recognition system is off and in a sleep mode; and a "confirm" sound noting that an utterance has been recognized. The "on" sound is triggered by a key "wake up" command or by depression of button. Once awake, the speech recognition engine expects to receive an utterance within a predetermined response time. The "confirm" sound signals the start of the response time. If the response time lapses before a recognizable utterance is entered, the "off" sound is played. The user interface further includes a visual component in the form of a graphic that changes with the tolling of the response period. In one implementation, the count graphic is a progress bar that counts down or shortens in proportion to the passage of the response period. When the response time runs out, the progress bar disappears entirely. On the other hand, if the speech engine recognizes an utterance within the response period, the user interface plays the "confirm" sound and restarts the countdown graphic. The user interface may also change the color of the graphic elements briefly to reflect a correct voice entry.