Discussing the relationship between the astrometric binaries and the family of double stars is somewhat like dealing with the ‘philosophy’ of this type of binaries. Available evidence points toward the assumption that we encounter here primarily a specific technique of observation rather than a specific type of objects. This technique addresses itself to moderately wide pairs among nearby stars, with a in the ‘grey area’ around 10 AU, objects whose detection by other methods of observation presents difficult problems. In this context astrometric binaries may be considered an extention of the class of visual binaries, as illustrated by systems where the perturbing unseen component has been rendered visible by more powerful instruments, as in the case of Sirius B, predicted by Bessel, Procyon B, and Ross 614.