Analysis of experiments with electron cyclotron resonance heating (ECRH) requires a good knowledge of the ECRH power profile. This profile is reconstructed by analysis of the transient process after on-axis ECRH switching on in special experiments with suppressed sawtooth oscillations in the T-10 tokamak. The calculations show that the absorbed ECRH power, , determined by the change in time derivative of the electron temperature at the region of ECRH power input, and the absorbed ECRH power, , determined by the magnetic measurements, are several times different. Depending on the plasma density and plasma current, their relation, , changes from 0.2 to 0.4. Analysis of different explanations for this effect shows that adequate description of the transient process demands introduction of a ballistic jump in the total heat flux just after on-axis ECRH switching on. The effective heat diffusivity increases up to values of 10–15 m2 s−1 in the first 100–200 µs and decreases down to values of 1.5–2.0 m2 s−1 during the following 1–2 ms. Note that such a non-monotone dependence of the effective heat diffusivity cannot be described by the modern critical gradient models. It seems that plasma reacts directly to the deposited power but not to the corresponding consequences (the increase in temperature or gradients). Different physical mechanisms could be proposed for this process (partial destruction of magnetic surfaces, fast transition of information through the turbulent cell connections), but each of them needs further confirmation.