[en] The electromagnetic form factors are crucial for our understanding of the inner structure of the proton. Recently it has become feasible to measure them by the use of polarisation transfer techniques in addition to the traditional Rosenbluth separation method. Thereby emerged an incompatibility of the results obtained by these two different experimental methods. It is commonly assumed that the discrepance is induced by higher order corrections to the cross section, especially through two-photon exchange processes. Unfortunately these processes cannot be calculated in a model independent manner because off-shell photon nucleon vertices arise. Effective chiral lagrangians contain already local two-photon couplings and therefore seem exceptionally well suited to study the anomaly contribution to the two-photon exchange. These couplings give two-photon exchange contributions that can be understood as the coupling of the nucleon to pions, decaying into two virtual photons. A particular contribution emerges from the chiral anomaly of QCD, that describes the two-photon decay of the neutral pion. The most important goal of this work is the calculation of the contribution arising from the anomaly to the elastic electron-proton scattering. The results are expected to be widely model independent since the anomaly directly reflects a QCD property. Based on the Skyrme model the protons are realized as soliton solutions in effective chiral theories. The next to leading order contribution to the cross section is given by the interference between the one- and two-photon exchange. The latter contains an ultraviolet divergence, which is renormalized by a local effective counterterm. This counterterm contributes to the width of the neutral pion decay which determines the finite part of the counterterm coefficient. The affect of the anomaly to the Rosenbluth separation of the electromagnetic form factors as well as the discrepance regarding the polarization measurements is extensively discussed within the Skyrme model. In addition, the correction as expected from the effective two-photon coupling of the non-linear σ model is estimated. (orig.)