We studied 12 patients with Parkinson's disease (PD): 6 with postural instability (Hoehn and Yahr Stage 3) and 6 without (Stage 2 or 2.5), using a quantitative test based on the clinical pull test. Their findings were compared with those for 12 healthy controls. The patients on their usual medications were pulled either forwards or backwards at the level of the shoulders and asked not to take a step in a series of five trials. Acceleration was monitored for the upper trunk, sacrum, and both tibias. EMG was measured in soleus and tibialis anterior (TA) muscles in all and for thigh and truncal muscles in a subgroup. A target of 0.2 g trunk acceleration was used, but smaller perturbations were used in very unstable patients. All the Stage 3 patients lost balance in at least one trial for the posterior perturbations but none for the anterior ones. None of the Stage 2 patients lost balance. There was increased tonic EMG and agonist activity but no difference in EMG onset or initial force production compared to healthy controls. For posterior perturbations, there were two related disorders that separated the PD patients from controls. There was a significantly higher ratio of sacral-to-applied acceleration and both PD groups showed reduced knee acceleration and shortened latency, more so for the Stage 3 group. The increased sacral-to-C7 acceleration ratio was correlated with the tonic level of activation of the hamstrings (HS), quadriceps, and lumbar paraspinal muscles (PS), while the tibial acceleration latency was also correlated with the level of tonic PS activation. We also found that the size of balance responses, 0-200 ms post-perturbation, correlated significantly with the level of tonic activation in nearly all the muscles studied. We confirmed that PD patients show greater instability posteriorly than anteriorly to applied perturbations. Our findings support increasing axial and limb rigidity as the cause of the impaired pull test rather than postural bradykinesia and suggest that tonic truncal and thigh muscle activation may be an important underlying cause.