Magnetic resonance (MR) was used to measure SF6 gas velocities in beds filled with particles of 1.1 mm and 0.5 mm in diameter. Four pulse sequences were tested: a traditional spin echo pulse sequence, the 9-interval and 13-interval pulse sequence of Cotts et al. (1989) and a newly developed 11-interval pulse sequence. All pulse sequences measured gas velocity accurately in the region above the particles at the highest velocities that could be achieved (up to 0.1 ms(-1)). The spin echo pulse sequence was unable to measure gas velocity accurately in the bed of particles, due to effects of background gradients, diffusivity and acceleration in flow around particles. The 9- and 13-interval pulse sequence measured gas velocity accurately at low flow rates through the particles (expected velocity <0.06 ms(-1)), but could not measure velocity accurately at higher flow rates. The newly developed 11-interval pulse sequence was more accurate than the 9- and 13-interval pulse sequences at higher flow rates, but for velocities in excess of 0.1 ms(-1) the measured velocity was lower than the expected velocity. The increased accuracy arose from the smaller echo time that the new pulse sequence enabled, reducing selective attenuation of signal from faster moving nuclei.