As a part of a Brain-Machine Interface, we define a model for learning and forecasting muscular activity, given sparse cortical activity in the form of action potential signals (spike trains). Whereas very impressive results such as [1] exist where a reaching task is successively performed from the sole interpretation of cortical signals, we focus our efforts in formalizing how neural impulses can be transcribed into a flexion of the index finger. We have a collection of experiments in which a trained monkey (macaca nemestrina) performs a precision grip. Its neuronal activity is partially recorded as the monkey clasps two levers between its index finger and thumb. In these experiments, 33 corticomotoneuronal (CM) cells from the hand area of the motor cortex (area 4) were recorded with glass-insulated platinum-iridium micro-electrodes, refer to [2] for more details about retrieving and filtering the data in our particular experiments. The main objective of this work is to treat the data in a way that allows us to provide an effective input/output functional. The underlying model parameters being interpreted with respect to the physiological aspects, though the model itself is not a bio-physical one. The method used here is based on a system of first degree linear equations involving the firing rate of the recorded neurons, two sets of thresholds associated to them, and the variation of the global neuronal activity. The learning formula is validated over a training set and tested over an estimation set.