The fundamental question in determining radiometric accuracy in a Forward Looking Infrared (FLIR) system is "What does the detector see?" The answer is that it "sees" with thermal radiation originating from emitting surfaces in a completely enclosed surround by virtue of the various transfer mechanisms -reflection, refraction, and scattering. To achieve high radiometric accuracy, the detector ideally would "see" only the object. However, due to effects such as lens emission, lens surface reflection, mirror surface emission, calibration source reflection (emissivity less than unity), and vignetting, the detector "sees" many other points in the field of view which are interpreted as part of the target itself. Other factors could contribute, e.g., cosn falloff, distortion, and pupil aberrations. From the viewpoint of the detector, all of these are closely related to vignetting, where the detector is partially "seeing" walls and lens mounts together with reduced intensities of radiation from the object scene. The approach here is to determine radiometric accuracy across the field of view on the basis of contributions from geometrically-defined sources outside of the object. Such a spatial spread of detector view must be combined with the temperature distribution to determine the radiometric accuracy of the system as well as image effects such as shading and narcissus. What the detector "sees" is determined by a computerized backward trace of many rays from the detector. To determine the variation across the field of view, this computation is made for a number of rotations of the scan mirrors. In the case of a FLIR designed for imaging and display, a high relative radiometric accuracy is required. For a FLIR used as an absolute radiometer, absolute radiometric accuracy close to the noise equivalent temperature difference of the system is achievable.