The effect of baclofen on walking performance was examined in nine spastic multiple sclerosis patients. In addition, nine healthy subjects were tested as controls. The modulation of the short latency soleus stretch reflex was closer to normal with baclofen compared to the recordings without baclofen, the modulation index being 74% (range: 60 - 100) with baclofen and 62% (range: 20 - 100) without baclofen, P=0.03. In healthy subjects the modulation index was 100% (range: 52 - 100). In the early swing phase the threshold of the soleus stretch reflex was significantly higher during baclofen medication being 139 deg/s (range: 63 - 302) compared with 93 deg/s (range: 37 - 187) without baclofen, P=0.004. The relation between the stretch velocity (input) and the amplitude of the stretch reflex (output) in early swing phase was unchanged being 0. 27 microVs/deg (range: 0.1 - 1.51) in patients with baclofen and 0. 24 microVs/deg (range: 0.08 - 0.79) without baclofen, P=0.25. Baclofen induced no change in input - output properties of the stretch reflex during walking compared with findings in a sitting position at matched EMG activity. There was a significant correlation between clinical spasticity score and stretch reflex threshold in the early swing phase (rho=-0.61, P=0.04) and between clinical spasticity score and the slope of the best linear fit in the early swing phase (rho=0.72, P=0.009).