A vehicle-mounted voice recognition apparatus including a voice input unit for acquiring an inputted voice, a voice recognition unit for performing voice recognition on the acquired inputted voice, a guidance information output unit for providing a guidance on the basis of the result of the voice recognition, and a recognized word understanding degree determining unit for determining a user's degree of understanding of recognized words to the guidance according to the number of timeout times that a timeout has occurred in the user's operation which is monitored while the voice recognition unit performs the voice recognition, and the number of correction times that a correction has been made to the user's operation, so as to change the guidance provided by the guidance information output unit.