The complex microstructure of two-phase titanium alloys can produce considerable ultrasonic backscattering noise. The noise introduces problems in detecting small flaws, such as hard-alpha inclusions, by forming a background which can mask the flaw signals. Therefore better understanding of grain noise is required to quantify and increase the detectability of the small flaws. As an aid to understanding the grain noise, an independent scattering model was constructed and studied during last two years by Margetan and Thompson. In that model for the backscattered noise generated by a tone burst, the grain noise is described by following equation (1) N(t)=FOM×M(t) where N(t) is the rms grain noise, FOM is a material characteristic parameter and M is a factor that depends on the detailed description of the experimental configuration as well as the ultrasonic attenuation. The argument, t, is the time delay at which the noise is observed and can be related to a spatial position within the material. Since the model gives an explicit functional form for M, it is possible to use Eq. (1) to infer the FOM from a measurement of N(t).1 Figure 1 presents the results of such a measurement in which the noise was observed, through each of three orthogonal sides of a set of four Ti-6246 specimens, whose history of heat treatment is summarized in Table 1.2 The FOM’s of each of specimens A1, A2 and B2 varied by an order of magnitude, depending on the side of the measurement. However, on specimen C1, which was annealed above the beta transus of 1775 °F, the noise was nearly isotropic. The purpose of this paper is to understand the origin of this anisotropy.