The Halogen Occultation Experiment (HALOE) onboard UARS measures profiles of limb path solar attenuation in eight infrared bands. These measurements are used to infer profiles of temperature, gas mixing ratios of seven species, and aerosol extinction at five wavelengths. The objective of this paper is to validate profiles of temperature retrieved from atmospheric transmission measurements in the 2.80-μm CO2 band. Temperatures are retrieved for levels above where aerosol affects the signals (35 km) to altitudes where the signal-to-noise decreases to unity (~85 km). At altitudes from 45 to 35 km the profile undergoes a gradual transition from retrieved to National Meteorological Center (NMC) temperatures and below 35 km the profile is strictly from the NMC. This validation covers the uncertainty analysis, internal validations, and comparisons with independent measurements. Monte Carlo calculations using all known random and systematic errors determine typical measurement uncertainties of 5 K for altitudes below 80 km. Comparisons of coincident HALOE sunrise and sunset measurements are an indicator of the upper limit of measurement uncertainty. The sunrise-sunset comparisons have random and systematic differences which are less than 10 K for altitudes below 80 km. Comparisons of HALOE to lidar and rocket measurements typically have random differences of ~5 K for altitudes below 65 km. The mean differences for the correlative comparisons indicate that HALOE temperatures have a cold bias (2 to 5 K) in the upper stratosphere and stratopause. ¿ American Geophysical Union 1996