There are several advantages of performing a scattering rather than a transmission measurement for NDE. The signal depends upon the composition of the volume element, defined by the overlap of the incident and scattered collimated beams rather than the transmission ray sum integrated along one direction. The scattering method requires access from one side and this can be a major advantage if the object is part of a large structure. In this technique the contrast due to localized defects will be an order of magnitude larger than that obtained through transmission when the beam widths are less than the defect size and this has obvious consequences for discerning small defect structures. In the present work a small scale research instrument based on the Compton scattering of 59.54 keV gamma radiation from Am-241 isotope has been developed to investigate the sensitivity of the technique for the characterization of defects incorporated in copper and stainless steel phantoms. The basic components of the scanner consist of a radioisotope source enclosed in a labyrinth of lead containing a small cylindrical aperture defining the primary beam and a collimated HPGe detector providing high resolution energy dispersive analysis of the scattered energy spectrum. The sample to be scanned is positioned centrally on a target mount with the primary and scattered beam directions inclined at an angle of 45 degree to the major surface. A scattering angle of 90 degree was chosen to minimize the scattering voxel and to increase the positional sensitivity. Data are accumulated in a 4096 channel personal computer analyzer (PCA) covering the energy range 10 to 70 keV. The variations in the integrated Compton peak intensities of the cylindrical holes of sizes varying from 2 to 14 mm are recorded on to a floppy disc for later analysis. The results presented are taken directly from the PCAand have not been corrected for photoelectric absorption, detector efficiency or indeed multiple scattering effects showing that the voids can be detected directly from the raw data. The beam absorption is low and all the defects are clearly visible as minima in the intensity vs. position curve. Indeed each void can be accurately located to within 5% of beam width and moreover the area of each minimum is found to be directly proportional to the void volume. The voided scattering volume is then proportional to the cylinder's cross sectional area. The beam absorption is reflected in the poorer contrast in the recorded signal. This dependence on absorption can however be removed by performing a second measurement on a plate of identical dimensions but which is free from defects. A direct subtraction of these two recovers the contrast levels. Further experiments are in progress to characterize the voids embedded at different depth levels of the bulk material.