Detailed paleoseismic investigation of the Alpine Fault, South Island, New Zealand, has been undertaken at locations which bracket the central and north sections of the fault, between the Hokitika and Ahaura River. A total of seven trenches and pits have been excavated at four localities along approximately 75 kilometres of the fault. From these excavations a total of 16 radiocarbon dates provide age constraints on the timing of the most recent two earthquakes. This trenching demonstrates that the most recent rupture occurred after 1660 AD, and most probably around 1700 - 1750 AD. There is consistent evidence for this event in the trenches in the central section of the fault. The surface rupture has extended into the north section of the fault as far as the Haupiri River area, which is 25 km northeast of the Alpine Fault junction with the Hope Fault. An earlier event at around 1600 AD can be recognised throughout the study area, and this is the most recent event in the trench locations north of the Haupiri River. An updated record of landslide and aggradation terrace ages is consistent with two earthquakes over this period, but this does not significantly refine the estimates of their timing. However, the analysis of indigenous forest age in Westland and Buller reveals two periods of synchronous regional forest damage at 1625 ± 15 AD and 1715 ± 15 AD. I infer that these two episodes of forest damage correspond to the two earthquakes revealed in the trenches for this same time period. Analysis of growth rings in trees which are old enough to have survived these earthquakes indicates that the most recent event occurred in 1717 AD. The growth ring anomalies also indicate a northeast earthquake limit near the Haupiri River. The most recent 1717 AD event appears to have been a synchronous rupture for a distance of over 375 km, from Milford Sound in the south Westland section of the fault, northeast to the Haupiri River. Based on the forest disturbance record, the earlier earthquake at 1625 ± 15 AD had a rupture length of at least 250 km, but further work is required to determine the southwest and northeast limits of this event. A range of methods is used here to estimate the probability of the next earthquake occurring on the central section of the Alpine Fault and all the calculated probabilities are relatively high. The most robust method, that of Nishenko and Buland 1987, suggests a conditional fifty-year probability in the order of 65 ± 15%. A sensitivity analysis indicates that the conditional probabilities of rupture are not significantly affected by assumptions regards the exact timing of the last earthquake, or even the number of most recent earthquakes, and conditional fifty-year probabilities of rupture remain at around 50% or higher. Based on the previous earthquake events, the next Alpine Fault earthquake is likely to have a Moment Magnitude of 8 ± 0.25, and will have a widely felt regional impact. Very strong ground shaking will occur in the epicentral area of the Southern Alps and central Westland. For most of the central South Island the ground shaking is likely to be stronger than that experienced in any other historical earthquake. Landslides and liquefaction will cause the greatest immediate damage to the natural environment, and in the longer-term increased sediment loads will cause aggradation, channel avulsion, and flooding in the numerous rivers which drain the epicentral region. There will also be substantial and widespread damage to the built environment, in some cases at a considerable distance from the epicentre. Because of the rugged nature of the topography of the central South Island, and the expected regional extent of the earthquake shaking, one of the greatest problems during the post earthquake recovery phase will be difficulty in communication and access.