A 3-D numerical finite-difference model to calculate the thermal condu
ctivity of a composite with Two or more constituents has been develope
d to better understand how the relative quantities and distributions o
f the component materials within a sample affect the whole sample cond
uctivity. It is found that different distributions of similar componen
ts within a sample can yield whole sample conductivity differences gre
ater than a factor of 2. Comparisons are made between the numerical mo
del results and the geometric and real-space renormalization group the
ory (RSRG) models and the effective medium approximation (EMA) for ran
dom distributions. Although good agreement exists between the numerica
l model and the EMA for the ranges of constituent conductivities consi
dered, the geometric and RSRG models give lower values.