The use of cone-beam computed tomography (CBCT) has been proposed for guiding the delivery of radiation therapy, and investigators have examined the use of both kilovoltage (kV) and megavoltage (MV) x-ray beams in the development of such CBCT systems. In this paper, the inherent contrast and signal-to-noise ratio (SNR) performance for a variety of existing and hypothetical detectors for CBCT are investigated analytically as a function of imaging dose and object size. Theoretical predictions are compared to the results of experimental investigations employing large-area flat-panel imagers (FPIs) at kV and MV energies. Measurements were performed on two different FPI-based CBCT systems: a bench-top prototype incorporating an FPI and kV x-ray source (100 kVp x rays), and a system incorporating an FPI mounted on the gantry of a medical linear accelerator (6 MV x rays). The SNR in volume reconstructions was measured as a function of dose and found to agree reasonably with theoretical predictions. These results confirm the theoretically predicted advantages of employing kV energy x rays in imaging soft-tissue structures found in the human body. While MV CBCT may provide a valuable means of correcting 3D setup errors and may offer an advantage in terms of simplicity of mechanical integration with a linear accelerator (e.g., implementation in place of a portal imager), kV CBCT offers significant performance advantages in terms of image contrast and SNR per unit dose for visualization of soft-tissue structures. The relatively poor SNR performance at MV energies is primarily a result of the low x-ray quantum efficiencies (∼a few percent or less) that are currently achieved with FPIs at high energies. Furthermore, kV CBCT with an FPI offers the potential of combined volumetric and radiographic/fluoroscopic imaging using the same device.