Being able to stop an ongoing motor response when required has been considered a distinctive feature of action control. The cortical substrates of this kind of action inhibition have been extensively investigated. A large number of studies identify both the right caudal inferior frontal gyrus (rcIFG) and the pre-supplementary motor area (pre-SMA) as the two main inhibitory cortical nodes with a pivotal role in response suppression (Aron et al. JNS 2007; Chambers et al. 2009; Chikazoe 2010 MORE REF). Indeed, there is evidence that rcIFG is critically involved in stopping behaviour (Rubia, Smith, Brammer, & Taylor, 2003) contributing more to response inhibition rather than to monitoring performance or adjusting behaviour (Chevrier, Noseworthy, & Schachar, 2007). Substantial evidence indicates that pre-SMA may also contribute to successful implementation of stopping (Nachev, Wydell, O’Neill, Husain, & Kennard, 2007; Chen et al. 2009; Aron et al. 2012; Obeso et al. 2013) as well as to other forms of action control such as altering a motor plan or switching from one action to another (Curtis, Sun, Miller, & D’Esposito, 2005; Li, Huang, Constable, & Sinha, 2006; Swick et al., 2011). 
However, none of these areas is directly connected to the primary motor cortex (M1) (REF). It has been proposed that the rcIFG or pre-SMA suppression of motor output is permitted primarily by their projection to the subtalamic nucleus (STN) (see Hamani et al., 2003; REF Sara). When STN is activated, the internal segment of the globus pallidus becomes activated and motor output is generally suppressed (Verbruggen REF Sara). Therefore, as far as we are interested in a direct inhibitory control over the primary motor cortex, accumulating evidence seems to suggest that in the caudal areas of the premotor complex there is another inhibitory cortical node that should be added to the list introduced above, namely the dorsal premotor cortex (PMCd) (Filevich et al. 2012 MORE REF). Single units in monkeys PMCd have been shown to exhibit a significant modulation either when an action should be suppressed with respect to when it should be executed (Kalaska & Crammond 1995) or when a motor response should be rapidly stopped (Mirabella et al. 2011). In humans, cortico-cortical connectivity between left PMCd and the contralateral M1, assessed by dual-coil TMS, was found to be inhibited when prepared actions should be suppressed (Koch et al. 2006). A previous study of our group has provided further support to the hypothesis of a putative role of PMCd in action inhibition, by finding a prominent action-related inhibitory connectivity between PMCd and the ipsilateral M1 while the subject had to withhold a movement until the appearance of a go signal (Parmigiani et al. submitted). This hypothesis seems also to be consistent with an electrical stimulation study finding negative motor responses following PMCd stimulation (Mikuni et al. 2006) as well as with lesion studies in nonhuman primates and humans: monkeys revealed increased frequency of impulsive and uncontrolled reaching movements following PMCd ablation (Moll & Kuypers, 1977) and loss of the ability to withhold movements after temporary inactivation of PMCd (Sawaguchi et al., 1996). In humans, stimulation of the premotor cortex causes ``negative motor responses'' characterized by the inability to perform certain voluntary movements or to sustain a voluntary muscle contraction (REF Sara). Epileptic discharges in the premotor cortex generate negative motor phenomena characterized by a loss in postural tone (REF Sara epileptic negative myoclonus). Furthermore, humans with focal lesions, especially in the left superior portion of BA6 (encompassing putative PMCd), have been demonstrated to succumb to an increased number of false alarms, thus revealing a clear deficit in inhibiting responses to a no-go stimulus (Picton et al., 2007).
More ref and data?
In the present study, we focussed on the functional properties of PMCd in stopping behaviour using single pulse transcranial magnetic stimulation (TMS). Although the abovementioned studies speak all for a role of PMCd in action inhibition, the clear-cut evidence for the idea that PMCd may exert a distinctive inhibitory function in stopping behaviour is still missing. The present study aimed at starting to fill this gap. As described above, differently from rcIFG and pre-SMA, PMCd is densely connected to M1 (Muakkassa and Strick 1979; Tokuno and Tanji 1993; Kiefer et al. 1998; Hatanaka et al. 2001; add ref??). Our conjecture is therefore that PMCd might be a good candidate site for exerting direct cortical motor inhibition, thus mediating, at the behavioural level, the ability to stop an ongoing action. To test this conjecture, we stimulated PMCd and measured the performance of healthy volunteers during one of the most established task to test stopping behaviour in the laboratory (Noorani, 2017), the Stop-signal task (SST). This task relies on bottom-up sensorimotor associations and allows probing how fast and accurate individuals might be in stopping a planned action triggered by a go-signal when a stop-signal is presented after a certain delay (Verbruggen and Logan 2008).
 
We had previously identified (Parmigiani et al., 2015; submitted) a small region in the left PMCd, along the superior frontal sulcus, exerting a robust short-latency effect on the ipsilateral portion of M1 involved in controlling and executing mouth movements. Indeed, we found both an inhibitory effect of PMCd over M1, when the task required withholding a preloaded action, and a reversal from inhibition to facilitation when the right time to release the action was coming (Parmigiani et al., submitted). In contributing in a cardinal behaviour as the precise motor control, a candidate brain region could be tested not only in his involvement in initiating movements with appropriate timing, but also in stopping them. Therefore, in the present study, we targeted with real and sham TMS the same region in PMCd as in Parmigiani et al. (2015; submitted) and used mouth actions during the SST. Participants had either to fully lift a stick with their lips (no-stop trials) or to suddenly stop their lifting movement (stop-trials), when a stop-signal appeared after a very short personalized fixed delay. Participants’ performance accuracy was measured as a function of stimulation, real vs. sham.
 
 
Introdurre SMA
Indeed, the premotor complex is canonically parcelled in the rostro-caudal dimension into rostral, prefronto-dependent area, and caudal, parieto-dependent areas (REF). Since only the caudal areas communicate directly with M1, if we want to look a direct premotor-motor inhibitory control, we need to investigate whether also the caudal premotor regions are involved in action inhibition
 
In order to clarify more precisely the spatial specificity of our site, we carried a second experiment in which the adjacent medial premotor cortex was stimulated, while a different group of participants performed the same SST. The supplementary motor area (SMA) is a neighbouring area of PMCd, located more medially and directly connected to M1, which could be a good candidate for motor inhibition as well. Indeed, SMA invasive stimulation (REF) and data from epileptic patients indicate that the output of cortical stimulation or the epileptic activity from SMA may be of negative (inhibitory) nature. For instance, Meletti and colleagues suggested that the involvement of frontomesial areas in their epileptic patient could account for the negative motor events through a disruption of primary motor cortex output (Meletti, Tinuper, Bisulli, & Santucci, 2000).
 
n epilepsy and/or tumour surgery, resection of the SMA leads to transitory motor and speech deficits called the supplementary motor area syndrome (Krainik et al. 2001, 2003, 2004). This SMA syndrome is characterized by disorders of initiation, with variable severity ranging from complete suppression of motor and speech production to reduced spontaneous motor and speech output (Krainik et al. 2004). We decided therefore to test SMA with the double purpose of clarifying the spatial specificity of our findings in PMCd and of investigating the possible behavioural effects of SMA stimulation of our SST. On one side, PMCd and SMA are neighbours areas and it could be argued that the observed effect could be due to a diffusion of the stimulus. On the other side, it could be of interest to investigate whether there are also at the behavioural level effects compatible with the stimulus-induced silent periods and the cortical negative myoclonus.
Short conclusion: we know from the literature that action stopping / inhibition requires the integrity of the IFG and the pre-SMA regions. In the present data we show that at a more distal, hierarchically lower level, the pathway leading to voluntary inhibition of action passes from the PMCd, which probably exerts a direct inhibitory control over M1. We report evidences showing that the caudal PMCd can be an important inhibitory cortical node. In fact, the portion of PMCd we stimulated seems to be necessary for performing the action stopping behaviour required in the SST. 
Our results are in line with the hypothesis that Go and stop processes could relate to the relative activation of different neural pathways. Rely on different inhibitory nodes