The recent installation of a full metal, ITER-like, first wall provided the opportunity to study the impact of the plasma-facing materials on plasma initiation or breakdown. This study for the first time presents a full experimental characterisation of tokamak breakdown at JET, using all discharges since 2008, covering both operations with a main chamber carbon and a beryllium ITER-like main chamber wall. It was found that the avalanche phase was unaffected by the change in wall material. However, changes in out-gassing by the wall and lower carbon levels resulted in better controlled density and significantly lower radiation during the burn-through phase with the ITER-like wall. Breakdown failures, that usually developed with a carbon wall during the burn-through phase (especially after disruptions) were absent with the ITER-like wall. These observations match with the results obtained from a new model of plasma burn-through that includes plasma–surface interactions (Kim et al 2012 Nucl. Fusion52 103016). This shows that chemical sputtering of carbon is the determining factor for the impurity content, and hence also radiation, during the burn-through phase for operations with a carbon wall. As seen experimentally, with a beryllium main wall, the plasma surface effects predicted by the model do not raise the radiation levels much above those expected for pure deuterium plasmas. With the ITER-like wall, operation with higher pre-fill pressures, and thus higher breakdown densities, was possible, which helped maintaining the density after breakdown.