The aim of the study was to assess the variability of EMG signal envelope with electrode location during gait. Surface EMG signals were recorded from 10 healthy subjects from the tibialis anterior (TA), peroneus longus (PL), gastrocnemius medialis (GM), gastrocnemius lateralis (GL), and soleus (SO) muscles. From TA, PL, GL and GM, signals were acquired using a two-dimensional grid of 4 x 3 electrodes (10 x 15 mm in size, as used in most gait laboratories) with 20-mm interelectrode distance in both directions. A similar grid of 3 x 3 electrodes was used for SO. EMG envelope was characterized by its peak value, area after normalization by the peak value, and time instant corresponding to the maximum. The maximum relative change in peak value with electrode location, expressed as a percentage of the peak value in the central location, was (mean+/-SD) 31+/-18% for TA, 29+/-13% for PL, 25+/-15% for GL, 14+/-8% for GM, and 26+/-14% for SO. The maximum relative change in area was 29+/-13% for TA, 73+/-40% for PL, 31+/-23% for GL, 35+/-20% for GM, 20+/-13% for SO, and in the position of maximum, computed as distance from the maximum position in the central channel, it was 5+/-10% of the gait cycle for TA, 26+/-16% for PL, 3+/-2% for GL, 3+/-1% for GM, 3+/-3% for SO. A crosstalk index, defined on the basis of the expected intervals of muscle activation for healthy subjects, indicated that estimated crosstalk was present between TA and PL, in an amount which depended on electrode location. It was concluded that the estimate of muscle activation intensity during gait from surface EMG is variable with location of the electrodes while timing of muscle activity is more robust to electrode displacement and can be reliably extracted in those cases in which crosstalk is limited. These results are valid for healthy subjects, where the level of muscular activity during gait is much lower than maximum.