We examined effects of wetting and then progressive drying on nitrogen (N) mineralization rates and microbial community composition, biomass and activity of soils from spinifex (Triodia R. Br.) grasslands of the semi-arid Pilbara region of northern Australia. We compared soils under and between spinifex hummocks and also examined impacts of fire history on soils over a 28 d laboratory incubation. Soil water potentials were initially adjusted to -100 kPa and monitored as soils dried. We estimated N mineralization by measuring changes in amounts of nitrate (NO3--N) and ammonium (NH+--N) over time and with change in soil water potential. Microbial activity was assessed by amounts of CO2 respired. Phospholipid fatty acid (PLFA) analyses were used to characterize shifts in microbial community composition during soil drying. Net N mineralized under hummocks was twice that of open spaces between hummocks and mineralization rates followed first-order kinetics. An initial N mineralization flush following re-wetting accounted for more than 90% of the total amount of N mineralized during the incubation. Initial microbial biomass under hummocks was twice that of open areas between hummocks, but after 28 d microbial biomass was < 2 mu g(-1) ninhydrin N regardless of position. Respiration of CO2 from soils under hummocks was more than double that of soils from between hummocks. N mineralization, microbial biomass and microbial activity were negligible once soils had dried to -1000 kPa. Microbial community composition was also significantly different between 0 and 28 d of the incubation but was not influenced by burning treatment or position. Regression analysis showed that soil water potential. microbial biomass N, NO3--N, % C and delta N-15 all explained significant proportions of the variance in microbial community composition when modelled individually. However, sequential multiple regression analysis determined only microbial biomass was significant in explaining variance of microbial community compositions. Nitrogen mineralization rates and microbial biomass did not differ between burned and unburned sites suggesting that any effects of fire are mostly short-lived. We conclude that the highly labile nature of much of soil organic N in these semi-arid grasslands provides a ready substrate for N mineralization. However, process rates are likely to be primarily limited by the amount of substrate available as well as water availability and less so by substrate quality or microbial community composition. (c) 2007 Elsevier Ltd. All rights reserved.