Tunnelling experiments have played an essential role in the development and verification of theories for superconductivity. They are usually interpreted within the so-called semiconductor model, in which the tunnelling current is given by the convolution of the density of states of the tunnelling electrodes. With a break junction technique, we have measured the current-voltage characteristics I(V) in the high Tc superconductor Bi2Sr2Can-1CunO2(n+2) at different temperatures and doping levels. The derivative dI/dV shows the well-known peak-dip-hump structure. These results, including their dependence on temperature and doping, are similar to that of other tunnelling experiments like scanning tunnelling microscopy, grain boundary junction or mesa structure experiments. In mesa structures it is known that self-heating of the structure is a severe problem. At a bath temperature of 4.2 K, temperatures as high as Tc were measured in the mesa stack already when a voltage comparable to the dip-position was applied. The typical current-voltage characteristics can be described with a heating model, in which all the non-linearities are explained by only the temperature dependence of the junction resistance. It seems that the peak-dip-hump structure is an artefact of Joule heating. This heating model can in principle also explain the peak-dip-hump structure of our break junction measurements. Many features like the temperature or doping dependence of the I(V) characteristics are described naturally. Furthermore we realised that the so-called Kohlrausch relation gives an estimate of the junction temperature in function of the voltage that is consistent with the heating model. The typical features of our spectra are independent of the contact area between the broken parts: this observation has often been used as a proof against heating. The Kohlrausch relation shows that this argument does not hold. For low bath temperatures the relation predicts that a voltage U = 3.6 kBT/ε has to be applied to reach the contact temperature T. This similarity to the BCS relation 2Δ = 3.52 kBTc makes clear that care has to be taken when an experimental feature appears at twice the BCS gap energy Δ: it might be caused by heating. Due to problems of thermal dilatation of the sample holder, the validity of the heating model could only be tested qualitatively for our break junction technique. However, data of grain boundary structures allowed a more precise verification. These results are in very good agreement with the model predictions: the heating model is a valid alternative to the semiconductor model. This work opens many questions concerning the validity of the semiconductor model. We argue that further investigation of heating effects in tunnelling experiments is important.