Phenomenological analysis of two-photon exchange effects in proton form factor measurements

We perform a model-independent phenomenological analysis of experimental data on proton form factors ratio. We find that only one of the two photon exchange amplitudes, which we call $\delta G_M$, is responsible for the discrepancy between Rosenbluth and polarization methods. The linearity of Rosenbluth plots implies that $\delta G_M$ is approximately linear function of $\epsilon$. The slope of $\delta G_M$ is extracted from experimental data and is shown to be consistent with theoretical calculations.