Two-color soot optical pyrometry is a widely used technique for measuring soot temperature and volume fraction in many practical combustion devices, but line-of-sight soot temperature and volume fraction gradients can introduce significant uncertainties in the measurements. For diesel engines, these uncertainties usually can only be estimated based on assumptions about the soot property gradients along the line of sight, because full three-dimensional transient diesel soot distribution data are not available. Such information is available, however, from multidimensional computer model simulations, which are phenomenologically based, and have been validated against available in-cylinder soot measurements and diesel engine exhaust soot emissions. Using the model-predicted in-cylinder soot distributions, uncertainties in diesel two-color pyrometry data are assessed, both for a conventional high-sooting, high-temperature combustion (HTC) operating condition, and for a low-sooting, low-temperature combustion (LTC) condition. The simulation results confirm that the two-color soot measurements are strongly biased toward the properties of the hot soot. For the HTC condition, line-of-sight gradients in soot temperature span 600 K, causing relatively large errors. The two-color temperature is 200 K higher than the sootmass-averaged value, while the two-color volume fraction is 50% lower. For the LTC condition, the two-color measurement errors are half as large as for the HTC condition, because the model-predicted soot temperature gradients along the line of sight are half as large. By contrast, soot temperature and volume fraction gradients across the field of view introduce much smaller errors of less than 50 K in temperature and 20% in volume fraction. Published by Elsevier Inc. on behalf of The Combustion Institute.