In earlier studies, it has been shown that hydrocyclones have a higher removal efficiency of porly developed fibres than screens. Our hypothesis is that primary stage pulp already contains large amounts of fibres, good for paper making. Thus, it ought to be possible to reduce the energy consumption by hydrocyclone fractionation of primary stage pulp, at a freeness around 300 ml, and separate refining of these un-developed fibres. A study has therefore been made in full-scale production in the Holmen Paper Braviken mill, aiming at investigating the possibility to increase the energy efficiency by the use of fractionating hydrocyclones. Two different process configurations were evaluated. In the reference run, the total refining energy was 2430 kWh/ton to a freeness level of 58 ml and about 20% of the energy input was in the reject refiner. The total reject rate was 30%, i.e. the sum of the reject mass flows from the screens and hydrocyclones was 30 percent of the total production. In the trials with increased hydrocyclone reject rate, no secondary refiners were used. The total refining energy was 2030 kWh/ton to a freeness level of 61 ml and about 30% of the energy input was in the reject refiner. The total reject rate was 120 %, i.e. the amount of pulp to the reject refiner was 1.2 times the total production. The pumping energy for hydrocyclones was 70 kWh/ADT. All pulp properties in the increased fractionation trials were at the same levels as in the reference run. The shive content in the pulp from the trials with increased fractionation was even somewhat lower. Thus, it seems possible to increase the total refining energy efficiency by the use of fractionating hydrocyclones. These trials indicate a reduction in energy consumption around 15 %. Tensile index and fibril coefficient was measured on fibre fractions sampled as injects, accepts and rejects from screens and hydrocyclones. These data showed clearly that hydrocyclones have higher separation efficiency of insufficiently refined fibres than that of screens, i.e. according to the degree of fibre surface development or fibre bonding potential. One conclusion of the study is that a more energy efficient TMP process can be designed, with only single stage refining in the main-line, if fibre fractionation is used in an efficient way. In such a process, the primary refining is made to a relatively high freeness, screen reject rate should be kept low and the cyclone reject rate maximised. No optimisation of reject refining was made in these trials.