Sunlit and shaded slopes have a variety of temperatures based on their orientation with respect to the Sun. Generally, greater slope angles lead to higher anisothermality within the field of view. This anisothermality is detected by measuring changing emitted radiance as a function of viewing angle or by measuring the difference in brightness temperatures with respect to observation wavelength. Thermal infrared measurements from the Lunar Reconnaissance Orbiter Diviner Radiometer were used to derive lunar surface roughness via two observation types: (1) nadir multispectral observations with full diurnal coverage and (2) multiple emission angle targeted observations. Measurements were compared to simulated radiance from a radiative equilibrium thermal model and Gaussian slope distribution model. Nadir observations most closely match a 20° RMS slope distribution, and multiple emission angle observations can be modeled using 20-35° RMS slope distributions. Limited sampling of the lunar surface did not show any clear variation in roughness among surface units. Two-dimensional modeling shows that surfaces separated by distances greater than 0.5-5 mm can remain thermally isolated in the lunar environment, indicating the length scale of the roughness features. Non-equilibrium conditions are prevalent at night and near sunrise and sunset, preventing the use of the equilibrium thermal model for roughness derivations using data acquired at these local times. Multiple emission angle observations also show a significant decrease in radiance at high emission angles in both daytime and nighttime observations, and hemispherical emissivity is lower than is apparent from nadir observations. These observations and models serve as a basis for comparison with similar measurements of other airless bodies and as an initial template for the interpretation of TIR measurements acquired under a variety of geometric conditions.