Our present understanding of the dependence of non linear optical (NLO) properties of charge transfer compounds on their dielectric environment is generally summarized as a set of universal response functions versus the ground state ionicity. Experimentally, these behaviors are recovered by piece-wise assembling measurements performed in solvents of increasing polarities on series of chromophores with varying acceptor and donor groups and/or conjugation paths. In this work, we will take advantage of the recent success in synthesizing pyridinium phenoxides with or without tert-butyl oxygen protection groups as well as with or without steric methyl groups to modify the twisting of the central diaryl bond. Thus, we are in a position to sweep over a wide range of zwitterionic character with essentially the same chromophore. This opens up a great opportunity to investigate the relationships between geometrical structure and NLO properties, to examine the validity of current formalisms, and to test numerical simulations at the semi empirical and density functional levels. In this work, we have carried out an experimental and theoretical study combining UV visible, IR, Raman, and Hyper-Rayleigh spectroscopies to extract information concerning the geometry, the electronic structure, and the NLO response of our compounds. In particular, we show that the steric effect is sufficient to push the chromophore to the full zwitterionic limit. More generally, the approach we followed here shows great potential in probing chromophore-environment interactions.