The selection effects that govern the observations of Visual Binary Stars are investigated, in order to obtain a realistic statistical distribution of the mass-ratio q = M sec /M prim . To this end a numerical simulation programme has been developed, which 'generates' binary stars and 'looks' at them to determine whether an observer on Earth would be able to detect them. The simulations show that for mass-ratios q > 0.35, observations are expected to reveal the real q-distribution, while for massratios q < 0.35 selection effects begin to play a major part. It is found that the observed mass-ratio distribution for main-sequence systems, derived from the Index Catalogue of Visual Binary Stars (ids), can be explained by a distribution of secondary masses according to the Initial Mass Function (imf), i.e. φ(M) ∝ M −2.7 . From the Fourth Catalogue of Orbits of Visual Binary Stars (ovb) authors find a q-distribution that peaks strongly for q-values close to q = 1. It is shown that this mass-ratio distribution may be the result of a sampling selection effect. Due to this sampling selection effect, the ovb is a considerably more biased sample of the binary population in our Galaxy than the ids. Numerical simulations of biased sampling show that the q-distribution, found from the ovb, is not incompatible with the distribution of secondary masses according to the imf (for q > 0.35), found from the ids.Because of the selection effects, it is difficult to establish the real q-distribution for q < 0.35. If the real q-distribution departs from φ(q) ∝ q −2.7 for q ≤ 0.35, about 36% of all stars are in visual binaries (i.e. if the q-distribution is assumed to be flat for 0 < q < 0.35); if the distribution flattens for q ≤ 0.25, about 60% of the stars must be primaries of visual binaries.