We compare the effectiveness of two hypotheses, Natural Flavour Conservation (NFC) and Minimal Flavour Violation (MFV), in suppressing the strength of flavour-changing neutral-currents (FCNCs) in models with more than one-Higgs doublet. We show that the MFV hypothesis, in its general formulation, is more stable in suppressing FCNCs than the hypothesis of NFC alone when quantum corrections are taken into account. The phenomenological implications of the two scenarios are discussed analysing meson-antimeson mixing observables and the rare decays B s,d → µ + µ − . We demonstrate that, introducing flavour-blind CP phases, two-Higgs doublet models respecting the MFV hypothesis can accommodate a large CP-violating phase in B s mixing, as hinted by CDF and D0 data and, without extra free parameters, soften significantly in a correlated manner the observed anomaly in the relation between ε K and S ψK S . of renormalizable couplings contributing at the tree-level to FCNC processes, in multi-Higgs models, goes under the name of Natural Flavour Conservation (NFC) hypothesis.The idea of NFC has been with us for more than 30 years. During the last decade another concept for the suppression of FCNC processes has become very popular: the hypothesis of Minimal Flavour Violation (MFV) [4,5], whose origin, in specific new-physics (NP) models, can be traced back to [6,7]. The question then arises how NFC (and GIM) are related to MFV, and vice versa. Motivated by a series of recent studies about the strengths of FCNCs in multi-Higgs doublet models [8][9][10][11][12], in this paper we present a detailed analysis of the relation between the NFC and MFV hypotheses. As we will show, while the two hypotheses are somehow equivalent at the tree-level, important differences arise when quantum corrections are included. Beyond the tree level, or beyond the implementation of these two hypotheses in their simplest version, some FCNCs are naturally generated in both cases. In this more general framework, the MFV hypothesis in its general formulation [5] turns out to be more stable in suppressing FCNCs than the hypothesis of NFC alone. This analysis will also give us the opportunity to compare the various formulations of MFV present in the literature and to clarify which of the multi-Higgs models proposed in [8][9][10] are consistent with the MFV principle, and thus are naturally protected against too large FCNCs.The phenomenological tests of these different concepts which can be obtained on the basis of meson-antimeson mixing observables, such as the CP-violating (CPV) observable ε K , the mass differences ∆M d,s , and the CP asymmetries S ψK S and S ψφ are also analysed. Beside being perfectly consistent with present data even for light Higgs boson masses, two-Higgs doublet models respecting the MFV hypothesis could even accommodate a large CP-violating phase in B s mixing, as hinted by CDF [13] and D0 [14,15]. However, as pointed out first in [16], this can happen only introducing flavour-blind phases, i.e. decoupling the breaking of the flavo...