The data queuing performance is analyzed for a fixed-frame-size, integrated voice-data multiplexer (IVDM), where voice has preemptive priority over data, a maximum capacity is available for voice (above which voice freeze-outs occur), and data is queued when there is insufficient capacity. A discrete-state-continuous-time Markov model is used to model a discrete-state-discrete-time IVDM when the average source burst duration is significantly larger than the multiplexer frame time. The correlation of the voice and data source activity is shown to have a significant effect on IVDM data queuing performance. An approximate, simple analytical model for the mean data queue length is derived. The approximate analytical model for mean data queue length is shown to provide a lower bound for exact calculations and simulation results. The concept of maximizing queuing power (throughput/delay) is applied to the data performance.>