Adult Schistocerca were tethered and allowed to walk on a styrofoam ball. Animals with both neck connectives cut could still walk. Therefore no descending fibres from the head ganglia are necessary to initiate walking. The roles of the brain and suboesophageal ganglion (so.g.) in walking were examined with the use of extracellular microstimulation to evoke behaviour. Pulses lasting 0.1 ms were delivered to the neck or circum-oesophageal (co.) connectives through a glass microelectrode fine enough to be used to record from single fibres. The calculated first approximation maximum area of current spread was a sphere 1.5 $\mu $m in diameter. Double-recording experiments indicated that usually one to three and maximally seven fibres were stimulated. The evoked behaviour was videotaped. Walking, jumping, struggling or grooming was evoked from different electrode positions. The effects were reproducible at each position while some behaviours were reproducible from animal to animal. These latter were grouped together as being evoked by the same `locus'. At seven loci, termed `subroutine' loci, only specific movements were evoked in a resting animal. When the animal was active these movements were incorporated into the ongoing behaviour, e.g. a lifting of the prothoracic leg could be used as part of a swing in walking or as a grooming movement. Seventy-three loci evoking walking were derived from 163 stimulating positions. Ten loci were common to both connectives, 37 were found only in the co. connective and 26 only in the neck connective. The optimal stimulus parameters for most positions in the co. connective were single 200-400 ms trains at 20-50 Hz. At most neck connective positions several such trains to continual stimulation were required to evoke walking. Increasing the stimulus frequency above the optimum produced abnormal movements or struggling. There was no evidence that the behaviour type evoked depended on stimulus frequency. The 73 loci evoked different types of walking or turning. They differed in direction, speed, step size, positioning of the legs relative to the body and individual movements. Some loci evoked walking starting with stance or with swing of specific legs; other loci evoked walking combined with extra movements of the antennae, palps or individual legs. Thus, parameters of leg movement, of coordination and of coupling between the sides (as in turning), even the phase of the local pattern generators, can be determined by influence from the head ganglia. The effectiveness of stimulation at each locus depended on the behavioural context at the time of stimulation, e.g. most loci were ineffective in evoking their type of walking if the animal was already walking or otherwise active. Therefore, I describe the function of each locus as `recommending' its specific type of walking. These `recommendations' may also suppress other behaviours: stimulation at some walking positions immediately stopped ongoing flight. Experiments where lesions were combined with microstimulation suggested that both the brain and the so.g. contribute recommendations initiating and determining the type of walking. The brain may be more concerned with initiating walking while the so.g. is involved with both initiation and maintenance. Furthermore, the so.g. plays a unique role in mixing information from both brain and so.g. halves and distributing it to both sides of the body. I suggest that walking is normally initiated by many fibres acting in consensus, sending their recommendations for specific walking types to the local pattern generators. The type of walking output (speed, direction, form) is determined by the `across-fibre pattern' of all the inputs to the pattern generators. The many fibres acting in consensus give the system the ability to match walking output to a large variety of inputs, therefore providing the flexibility required of any model purporting to explain this enormously adaptable behaviour.