Experiments were performed on JET where high-density plasmas with an internal transport barrier (ITB) were created by means of combined use of lower hybrid current drive (LHCD) and pellet injection before the barrier formation. Attempts were also made to use pellets to fuel the plasma and to sustain the density during the ITB phase. It was found that shallow pellets ablating in the region r/a ⩾ 0.8 and far from the foot of the barrier did not destroy the ITB, whereas deeper pellets penetrating up to 0.6 ⩽ r/a ⩽ 0.7 affected the barrier and led to its disappearance. Modelling of these experimental scenarios has been performed with transport and fluid turbulence codes. The codes used in the analysis were: JETTO, a 1.5 dimensional transport code, TRB, a global electrostatic fluid turbulence code and CUTIE, a global electromagnetic fluid turbulence code. The results show that for the shallow pellet case all codes reproduce the general features of the experiment, whereas for the deep pellet case, there are differences in the degree of agreement between the different codes and the experiment. Runs performed varying the pellet penetration depth indicate that not only the pellet penetration, but also the barrier strength plays a key role in the dynamics of the pellet-ITB interaction.