Different methods for ultrasonic velocity determination using broad-band pulse transmission have been investigated in 70 human calcanae in vitro. The work took place within the context of the EC BIOMED1 concerted action Assessment of Quality of Bone in Osteoporosis. Ultrasonic velocities were determined using three different transit time definitions: first arrival (TTV1), thresholding (TTV2), and first zero crossing (TTV3). Phase velocity (PV) was determined over a range of frequencies from 200 to 800 kHz using a new phase spectral analysis technique. The different velocity measurements were compared in terms of their magnitudes and their inter-correlations. There were significant differences of up to 260 m s-1 between different transit time velocities (p < 0.0001), indicating the sensitivity of the measurement to the arrival criteria used. Phase velocities were lower than all of the transit time velocities (p < 0.0001) and decreased with increasing frequency (p < 0.005). A strong correlation (r2 = 0.968) was observed between PV at 400 kHz (PV400) and TTV3, with much weaker correlations between PV and the other transit time velocities. Reproducibility for transit time velocity measurement was optimal for TTV3 (coefficient of variation, cv = 0.41%), and for PV it was optimal at 600 kHz (cv = 0.34%). These data indicate that transit time measurements may be subject to errors due to the modification of the pulse shape during propagation through bone by attenuation and dispersion. Velocity measurement by phase spectral analysis appears to offer advantages over the transit time approach, and should be the method of choice for velocity measurement in trabecular bone. Where transit time velocity measurements are made, the first-zero-crossing criterion appears to be have some advantages over other arrival criteria. We also note that PV measurements provide new information on dispersion which could prove to be relevant to the structural and mechanical characterization of trabecular bone.