Speed of sound (SOS) measurements, typically made using 1 MHz broadband pulses, are increasingly used in the clinical diagnosis of bone disorders. Previous in vitro studies indicate that broadband ultrasound pulses are susceptible to distortion in cancellous bone, leading to imprecise arrival time and SOS measurements. We investigated the effect of bandwidth and frequency on SOS by comparing measurements made using 1 MHz broadband with 1 MHz and 300 kHz narrowband toneburst signals in 15 human proximal femur cancellous bone specimens. There was no significant difference in the value of SOS measured from the leading edge of 1 MHz broadband, 1 MHz toneburst and 300 kHz toneburst signals. Values of SOS in later regions of 1 MHz and 300 kHz tonebursts fell significantly (p < 0.001) when compared to earlier regions. This decrease in SOS levelled off by the third complete cycle of 300 kHz toneburst signals, reaching a plateau value of 1961 +/- 239 m s-1. No plateau SOS value was obtained in 1 MHz tonebursts. The reproducibility of SOS, as measured by the coefficient of variation, was higher for later regions of 300 kHz tonebursts than for the leading edge of 300 kHz toneburst and 1 MHz broadband signals (p < 0.005). The correlation between ultrasound measured modulus and compressive Young's modulus improved when 300 kHz tonebursts (r2 = 0.83) rather than 1 MHz broadband (r2 = 0.77) signals were used to calculate SOS. The improved SOS reproducibility of later regions 300 kHz tonebursts suggest that it may be beneficial to use such signals rather than 1 MHz broadband pulses in SOS measurement. Since no reliable SOS measurements could be obtained from any region of 1 MHz tonebursts, the use of high frequency toneburst signals in cancellous bone has little value.