We combine a robust multiple-scattering radiative transfer algorithm with the Thermal Emission Spectrometer (TES) spectral data set in order to characterize the properties of Martian aerosol particles. Because of the importance of accurate model input when performing such retrievals, we include self-consistent and physically plausible treatments of surface emissivity and atmospheric aerosol dielectric functions, as well as gaseous absorption effects. Considerable effort is expended in the identification and discussion of potential sources of error and uncertainty. Significant results stemming from this analysis are a new dust aerosol dielectric function that appears to well represent the IR spectral behavior sampled by TES for a wide range of dust loading conditions, two distinct populations of water ice particles with reff of ~1--2 ¿m and ~3--4 ¿m; and distinct departures in dust particle sizes during the 2001A global dust storm from the canonical 1.6--1.7 ¿m values. Very consistent aerosol size distributions are obtained when 9 ¿m dust and 12 ¿m ice optical depths retrieved from this analysis are compared to visible optical depths retrieved from TES solar band emission phase function sequences <Clancy et al., 2003>. Direct comparison of our optical depths to those available from the Planetary Data System (PDS) (as provided by the TES science team) reveals a systematic bias toward τ values which are 20--30% (or more) too small. Much of this offset stems from the fact that TES PDS aerosol optical depths are actually an approximation to τabsorption, which is ~30% lower than τextinction for Mars dust aerosols. Additional biases in TES optical depths arise from assumptions of fixed surface emissivity and temperature.