Recording from neuronal populations is a major theme of present day Neuroscience that requires the collaboration of scientists with a wide range of expertise from solid state Physicist (designing and building the recording electrodes) to Mathematicians like interacting particle systems experts to model (and make sense of) the experimental outcome. In any such scientific endeavor the experts at any one stage tend to take for granted what the experts at the other stages do or provide. Such an attitude is definitely "efficient" on a short term basis but can lead to problems on a longer term. The goal of this talk is therefore to present to mathematically oriented audience how data are practically obtained. This implies a discussion of: i) the recording techniques with their strength and weaknesses; ii) the (non neutral) processing applied to the data; iii) the different brain regions from which the recordings are obtained--do not forget that we are dealing with an intrinsically biological problem and that one cannot totally ignore the differences between the retina and the cerebellum!