The objective of the work was to develop an ultrasonic inversion procedure which (1) discriminates, (2) sizes, and (3) determines the orientation of two-dimensional (crack-like) and· three-dimensional (void-like) defects in materials. Adaptive learning networks (ALN's) were used to estimate directly the defect size and orientation parameters from the spectrum of the echo transient. A 19-element hexagonal synthetic array measured the scattered field within a 60-degree solid angle aperture. The ALN' s were trained on theoretically generated spectral data where the crack forward scattering model was based on the Geometrical Diffraction Theory and the void model was based on the exact Scattering Matrix Theory. The theoretically trained models were evaluated on both theoretical and experimental data. Excellent results were obtained, and the errors for size and odentation estimates were, in general, less than 10%. The significance of this work is that: (1) the ALN approach to defect characteristics provides a systematic procedure for discovering relationships in the data which could otherwise be overlooked, and (2) significant economic benefits can be gained by simulating difficult-to-produce defect reflector scenarios. Furthermore, a result of this work has been the development of an algorithm which can ultimately be applied in field and industrial use.