A system and method for voice browsing IVR services using a mobile terminal. A voice application is accessible via a server connected to a network. A call connection is established between the mobile terminal and the server using a dual-mode connection. The call connection includes a voice mode and a data mode for alternately transmitting voice and data via the network. The voice application sends a state-dependent grammar that defines the speech recognition results that the voice application is ready to accept as input or commands at its present state of execution. The voice applications also sends to the mobile terminal state-dependent voice output such as audio prompts and instructions using the voice mode. The user responds orally to the voice output. The mobile terminal processes this voice input using speech recognition facilities. Valid input is extracted from the voice input based on the state-dependent grammar. The mobile terminal sends the valid input to the voice application using the data mode. The voice application updates its state of execution based on the valid input.