Humans and other primates demonstrate an exquisite ability to precisely shape their hand when reaching out to grasp an object. Here we used a recently developed transcranial magnetic stimulation (TMS) protocol to examine how information about the geometric properties of an object is transformed into specific motor programs controlling hand shape. Pairs of TMS pulses (suprathreshold followed by subthreshold) were delivered at precise intervals to detect changes in the excitability of cortico-cortical inputs to motor cortex when subjects prepared to grasp different objects. We found that at least 600 ms prior to movement, there is an enhancement in the excitability of these inputs to the corticospinal neurons projecting from motor cortex to the specific hand muscles that will be used for the grasp. This enhancement was seen only with pulses delivered with an inter-stimulus interval of 2.5 ms, which is most likely to reflect the facilitation of the second indirect wave of corticospinal activity (the I2 wave). These changes were object- and muscle-specific and the degree of modulation in the inputs was correlated with the pattern of muscular activity used later by individual subjects to grasp the objects. In a number of control experiments we demonstrated that no change in excitability was observed during object presentation alone, under conditions in which subjects imagined grasping the object, or prior to movements involving the same muscles but without an object. This demonstrates a specific cortico-cortical mechanism subserving the transformation from the geometrical properties of an object to the outputs from motor cortex prior to grasp. This mechanism is not involved in movements that do not involve grasp of a physical object. We have now extended these findings to show that excitability of cortico-cortical inputs to the primary motor cortex during visuomotor grasp is transient and occurs in strict temporal relation to the upcoming movement. When subjects see an object but are instructed not to grasp until they are cued to do so, muscle-specific facilitation is observed just at the moment when the movement is cued, long before any overt muscle activity is observed. Paired-pulse modulation of M1 excitability occurred only prior to the movement cue, but not after it. If TMS delivery was used as the cue to grasp a very short time after object presentation (50 or 100 ms), facilitation was absent. It first appeared for TMS cueing movement > 150 ms after presentation. The effect was present when predictable TMS was delivered at the cue to grasp, but was abolished when TMS was given at random in relation to that cue. Our results indicate that the motor cortex does not maintain a state of readiness from object presentation until cue to grasp, but rather excitation occurs only when execution of grasp is required. We suggest that commands related to the selection of appropriate hand shape are stored upstream from M1 and are ‘released’ to M1 immediately prior to grasp execution. Parallel experiments in both anaesthetised and awake behaving monkeys show that single pulse stimulation of the hand representation in the ventral premotor cortex (area F5) can activate neurones in the hand region of primary motor cortex (M1) and that single-pulse stimulation of area F5, while evoking few or no motor effects alone, is able to modulate powerfully outputs from M1 to hand motoneurones and hand muscles. This modulation is seen mainly in terms of the later I wave discharges from M1 corticospinal neurones, suggesting that cortico-cortical inputs from premotor areas have particularly strong inputs to these neurones via late-I wave circuits. We have speculated that it is these inputs, no doubt among others, that are selectively enhanced during object grasp and which are particularly sensitive to the paired-pulse TMS protocol described above.