When manipulating objects, humans rely on their sense of touch to perceive subtle movements and micro slippages. This synergy between sensations and motion permits them to manipulate an impressive range of objects of different sizes, shapes, and surface properties. This incredible dexterity relies on fast and unconscious adjustments of the grip force by placing a 20% safety margin before slip that holds an object strong enough to avoid a catastrophic fall yet gentle enough not to damage it. In addition to being accurate, this regulation is swift: only a hundred milliseconds after first making contact, grip forces are already adjusted by taking into account the actual frictional strength of the contact. This astonishing performance is owed to the sense of touch, which informs on the physical properties of the surrounding world and contact state. Within the fingertip, thousands of mechanoreceptors convert the complex mechanical interaction into action potentials. However, how the brain copes with large amounts of data to infer the state of the contact is still debated. This thesis covers how the cutaneous tactile afferent made it possible for a swift and precise regulation of the grip. Firstly, I show that humans can assess friction without slippage, suggesting that the radial stretch of the skin can provide enough information to regulate grip at the contact initialization. Secondly, I show that the perceptual system uses a compact code to estimate the safety margin from the skin deformation during an incipient slip, suggesting a mechanism to explain the rapid reactions. Finally, I expose a new model based on contact mechanics to quantify the sensitivity of the mechanoreceptors to the patterns of skin deformation highlighted in the first two chapters. This model also correlates the spatial and temporal detection threshold to detect a moving stimulus, suggesting a persistence of touch that bridges discrete sensations into a continuous stimulus. Taken together, these results reveal how the perception of friction is encoded in the spatio-temporal deformation of the skin. The findings are useful for designing bio-inspired tactile sensors for robotics or prosthetics and for improving haptic human-machine interactions.