The goal of the present experiment was to test the predictions of Central Bottleneck and Central Capacity Sharing models. According to the Central Bottleneck model, dual task interference, as observed in the PRP paradigm, is caused by an all-or-none bottleneck in information processing. The Central Capacity Sharing model postulates that dual task interference is caused by a capacity limited process that can allocate capacity in a graded fashion. The Central Bottleneck model predicts no change in RT1 with decreasing SOA, whereas the Central Capacity Sharing model predicts that RT1 will increase with decreasing SOA and that the slope of the RT1 SOA effect will depend upon the difficulty of task 2. Subjects were required to perform a tone pitch judgement and shape-matching task in rapid succession. Task order was randomized and the SOA between the first and second stimulus varied from 50 to 1250 ms. Results from this experiment favour the Central Capacity Sharing model. The results were then run through simulations of both the Central Bottleneck and Central Capacity Sharing models. Results from the simulations also favoured the Central Capacity Sharing model. As the difficulty of the second task increased, more capacity was allocated to it, confirming the prediction that as task 2 difficulty increases, the RT1 SOA slope increases. The proportion of capacity allocated to the first task varied from.78 to.91 indicating that capacity can be allocated in a graded fashion.