The magnetic shielding effect of a refuelling pellet is considered by first briefly reviewing the existing balloon model. The limitation of the model is pointed out and discussed. Since solid deuterium is an insulator and the ablated plasma is expected to be cold and dense, it is felt that the existence of a field-free balloon is questionable; rather some field will be trapped by the pellet. On the assumption that the flow of the ablated plasma is governed by a flexible magnetic nozzle, it is shown that the reduction of the ablation rate of the pellet is not due to the exclusion of the field by a balloon, but to the reduction of the ablated plasma pressure through the trapping of the field inside the nozzle. The model also indicates that the penetration of a mm-size pellet into the centre of a low-β reactor might be possible, provided an injection speed around 104 m centerdots−1 can be attained and no appreciable deceleration process is present.