The authors have developed a computer program to calculate the response of a 20-cm-diameter phoswich detector [3-mm thick NaI(Tl) primary detector] to a source of low-energy photons distributed in the lungs of a heterogeneous (MIRD) phantom, approximating ICRP Reference Man. Monte Carlo techniques are employed to generate photons and trace their fates in the thorax of the phantom. The acceptable points of photon interactions (photoelectric and Compton) in skeletal, lung and ordinary tissue are determined by the Coleman technique. The calculations yield the exit photon energy spectrum, which is then smeared with an experimentally determined Gaussian resolution function to convert into the pulse-height spectrum observable with the detector. The computer program has provisions for incorporating the effects of the iodine K X-ray escape as well as the variable intrinsic efficiency of the detector. Computed calibration factors (cpm/microCi integrated over the full spectrum) are given for several low-energy photon sources, viz. 238Pu, 239Pu, 241Am, 244Cm, 246Cm, 250Cf and 103Pd, distributed uniformly or as points in the lungs of the phantom, with the phoswich located centrally over and in contact with the chest. Examples of generated exit photon and the corresponding pulse-height spectra are included. The spectral changes observed in these generated spectra, which are also discerned in experimental pulse-height spectra, are discussed in detail. Thus we observed photopeak energies of 18.4 and 55.5 keV for UL X-rays and 241Am gamma-rays, respectively. It is shown that consideration of the total flux of escaping photons (i.e. both uncollided photons and those escaping after collision) improves the calibration factors by about 50% for 239Pu, 70% for 103Pd and as much as 340% for 241Am gamma-rays. In addition, calibration factors are calculated for point 239Pu and 103Pd sources located at different sites in the phantom lungs and data on the escape efficiencies of 16-, 53-, 90- and 185-keV photons from different parts of the phantom are presented. The suitability of the MIRD phantom as a calibration device for low-energy photon in vivo spectrometry is examined in the light of these results.