We conducted dedicated aerial surveys to determine the distribution and abundance of dugongs in Hervey Bay and the Great Sandy Strait in August 1988, November 1992 and December 1993. The methods followed those of Marsh and Sinclair (1989a, 1989b) and Marsh and Saalfeld (1989). Sightings were recorded in a strip 200 m wide on each side of the aircraft, from an altitude of 137 m. Two isolated, independent observers were used on each side of the aircraft so factors could be derived to correct for the dugongs visible, but missed by observers [based on a mark-recapture analysis of sightings (perception-bias correction factor: Marsh and Saalfeld 1989). Results were standardised to correct for dugongs not at the surface at the time the plane passed over (availability-bias correction factor; Marsh and Sinclair 1989b). The same parallel, east-west-oriented transects were flown on each survey, except that the number of transects in Block 1 (Great Sandy Strait) was doubled for 1992 and 1993. This increased the survey intensity in this block from 9.6% in 1988 to 17% and 16.4% in 1992 and 1993, respectively. Surveys were conducted only under good weather conditions (Beaufort sea state ≤ 3), and we avoided flying during periods of severe glare (early morning, late afternoon and midday). The 1992 and 1993 surveys of the Great Sandy Strait were timed to coincide with high tide over most of the area. As the transects were of variable length, the ratio method was used to estimate the density, population size and associated standard errors for each block. The standard errors were adjusted to incorporate the error associated with each correction factor, as outlined in Marsh and Sinclair (1989a). The significance of the differences between the dedicated dugong surveys of Hervey Bay conducted in 1988, 1992 and 1993 was tested by ANOVA both with and without the modal Beaufort sea state for each transect as the covariate. Blocks and times were treated as fixed factors and transect as a random factor nested within block. Input data for all analyses were corrected densities km-2, based on mean group sizes and the estimates of the correction factors for perception and availability bias, each line contributing one density per survey based on the combined corrected counts of both tandem observer teams. The densities were transformed [log10 (x+l)] for analysis, to equalise the error variances.