A systematic discussion of the structure of fermion mass matrices in terms of quantum numbers is presented. Small ratios between fermion masses and small mixing angles are related to a fine structure of scales around the unification scale. We argue that in higher dimensional models all small fermion masses should be explained from symmetry considerations since no free small Yukawa couplings are available. This leads to a scanning procedure selecting higher dimensional models consistent with realistic fermion mass patterns. We present a six-dimensional model admitting “compactifications” with only SU(3) c ×SU(2) L×U(1) γ gauge symmetry, a vanishing cosmological constant and three generations of quarks and leptons. The field equations have solutions with a gauge hierarchy for weak symmetry breaking for a large range of model parameters without the need of fine tuning. The weak scale M W is a free integration constant and the mechanism determining its order of magnitude is not yet identified. These solutions have a good chance to be classically stable. For one particular solution the largest fermion mass is the top quark mass which is of the same order as M W. At the next level the fermion masses m b, m τ and m c are supressed by a small ratio of symmetry breaking scales γ. For the mixing between the second and third generation one finds θ 23≈ m b/ m t≈ γ. The relation m b( M) = m τ ( M)(1+O γ)) is predicted. Corrections of order γ 2 quark and the muon with the relation m s (M) = 1 3 m μ(M) . This reproduces the qualitative order of magnitude m s/ m b ≈ m c/ m t. Unfortunately this particular solution fails by predicting maximal Cabibbo mixing and θ 13 ≈ γ. The model can be interpreted as a subgroup analysis for E 8 × E 8 superstrings. We also give a systematic discussion of higher dimensional scalar fields in non-trivial representations of the gauge group. We describe the higher dimensional Higgs effect which can lead to a stabilization of the ground state.