In this study, we introduced a novel zoomed 3D pseudo-continuous ASL (pCASL) technique at 7T with high spatial resolution (1-mm isotropic) and sensitivity to characterize layer-dependent resting and task activation induced perfusion activity in the human motor cortex. For the first time, multi-delay pCASL was applied to measure variations of arterial transit time (ATT) and resting state CBF across cortical layers, illustrating the dynamics of labeled blood flowing from pial arteries, arterioles to downstream microvasculature in the middle layers of cerebral cortex. Zoomed pCASL at the optimal post-labeling delay was then applied on the motor cortex for detecting and quantifying the layer-dependent activity of M1 during finger tapping and finger brushing. Our results showed the two motor tasks, which consisted of unilateral finger tapping (FT, sequential, frequency of ∼2 Hz) and finger brushing (FB, ∼2Hz) of the dominant hand. The two motor tasks elicited different laminar-dependent activation patterns. Strong and medium CBF increases evoked by the FT and FB tasks can be observed along M1 respectively. FT induced CBF changes show clearly two peaks in deep and superficial layers respectively, consistent with the hypothesis that FT engages neural activity of both somatosensory and premotor input in the superficial layers and motor output in the deep layers. FB-induced CBF increase was much smaller averaged cross cortical layers compared to that of the FT task. The perfusion response mainly peaked in superficial layers, consistent with the hypothesis that FB primarily engages somatosensory input and minimal motor output. These results demonstrate the high spatial specificity of ASL, capable of resolving and further quantifying layer-dependent input and output activity in human M1.