Dual-source computed tomography (CT) utilizes two x-ray tubes and two detectors simultaneously for the purpose of obtaining 83 msec temporal resolution, 160 kW of x-ray power reserve, or dual-kV (dual-energy) scan capabilities. One inherent constraint of such a design is cross-scatter radiation, which occurs when x-rays from tube A are scattered by the patient and detected by detector B, or vice versa. In the evaluated dual-source CT scanner, an on-line cross-scatter correction technique is used to address this limitation. The technique, available using the 14x1.2-mm collimation, measures scattered radiation along the z axis using detector rows beyond those corresponding to the 16.8 mm nominal total beam width. These direct measurements of scattered radiation are used to correct the measured projection data (scattered and primary radiation) for cross-scatter. A semi-anthropomorphic thorax phantom was used with increasing thicknesses of tissue-equivalent material to simulate small, medium, large and extra-large patients. Phantoms were scanned using single-source and dual-source protocols at 80, 100, 120 and 140 kV, and the mean and standard deviation of the CT numbers in a water-equivalent cylinder located centrally within the phantom measured. For this comparison, images reconstructed using only tube A data from the dual-source acquisition were compared to the single-source images, also obtained using tube A. The differences in the mean and standard deviation of the measured CT numbers between the dual-source tube A images, which were corrected for cross-scatter, and the single-source images, where no cross-scatter existed, were determined for all tube energies and phantom sizes. The differences in mean CT number ranged from -5.2 to 1.3 HU, and the differences in standard deviations ranged from -4.5 to 3.0 HU. We conclude, therefore, that use of the evaluated on-line cross-scatter correction algorithm results in negligible differences in CT number and image noise between single-source and dual-source image data, independent of phantom size and tube potential.