Orange roughy, which spawn in an aggregation around a seamount off north-eastem Tasmania, were surveyed acoustically with vessel-mounted and deep-tow transducers from 1990 to 1993. Orange roughy dominated at 700-900 m (95% of individuals) and were less numerous both shallower and deeper. Echo integration estimates from the vessel-mounted transducer were consistently lower than those from the deep-tow transducer by a factor of ~1.8, owing to surface bubble attenuation, vessel movement and acoustic beam thresholding among other factors. The absolute estimate from the deep-towed transducer of the prefishery orange roughy biomass (98200 t ; coefficient of variation, 6.7%) compared well with estimates from stock reduction analysis (95000 to 110000 t) and an egg survey (96900 t; CV, 47%). The acoustic surveys had lower sampling error (CV, 7-15%) than that of the egg survey and required less survey time. The present results support the use of initial absolute acoustic biomass estimates to set a management strategy for the fishery. Stock reduction analysis based on the relative acoustic estimates and catch over the four years indicated that orange roughy target strength was -50 dB � 3 dB, which is consistent with results from models of target strength and from measurements in situ and on dead specimens.