A boiling water reactor SVEA-96+ fresh fuel lattice has been used as the basis for a benchmark study of the void reactivity coefficient at assembly level in the full voidage range. Results have been obtained using the deterministic codes CASMO-4, HELIOS, PHOENIX, BOXER and the probabilistic code MCNP4C, combined for almost all cases with different cross section libraries. A statistical analysis of the results obtained showed that the void reactivity coefficient tends to become less negative beyond 80% void and that the discrepancies between codes tend to increase from less than 15% at voidages lower than 40% to more than 25% at voidages higher than 70%. The void reactivity coefficient results and the corresponding differences between codes were isotopically decomposed to interpret discrepancies. The isotopic decomposition shows that the minimum observed in the void reactivity coefficient between 80% and 90% void is largely due to the decrease in the relative importance of the 157Gd(n, ) rate with increasing voidage, and that the fundamental discrepancies between codes or libraries are mainly governed by the different predictions of the 238U(n, ) variation with voidage. [All rights reserved Elsevier].