It is known that the preferential accessibility of motor units of the baboon's hand to cortical stimulation depends on the existence of colonies of pyramidal neurons which project monosynaptically to the appropriate spinal motoneurons. A complete description of the architecture of such a colony would include the size, shape and localization of its cortical territory; the maximum quantity of monosynaptic depolarization it can evoke in its target motoneuron; local variations of texture (density of output cells) within the territory; and extent of overlap, if any, between its territory and the territories occupied by other colonies. Systematic intracortical exploration with a stimulating microcathode excites spheres of tissue whose radii, at different current strengths, can be estimated within acceptable limits of accuracy. Such spheres are small in relation to the size of a colony, and can contain no more than a fraction of its pyramidal cells. Piecing these spheres together reveals the size, shape and texture of colonies and of the aggregations of colonies projecting to the motoneuron pools of particular hand muscles (1st dorsal interosseus, flexor and adductor pollicis brevis, extensor digitorum communis). They are localized in that part of area 4 that lies buried in the rostral wall of the central sulcus. Such aggregations occupy cortical territories measuring as much as 6.0 mm $\times $ 5.5 mm. Microstimulation provides critical evidence that they overlap. The words 'discrete' and 'mosaic' therefore have no place in the description of the fine-grained structure of cortico-motoneuronal output in the baboon. To discharge every cell in a colony simultaneously it is necessary to employ surface stimulation. Maximum output can be measured only by intracellular recording from the target motoneuron.