To investigate the biases associated with manta tow surveys of Acanthaster planci, counts obtained by manta-towed observers were compared with counts made on SCUBA swims under a limited range of conditions. Five 10 m wide strip transects on different parts of two reefs and with different densities of A. planci were surveyed. On average, 22.7% of starfish counted on SCUBA searches were counted on manta tows over the same transect (SD=12.0%, n=69). This proportion is termed sightability. As sightability decreases with increasing transect width, we estimate that, on average, less than 5% of the A. planci present are counted on routine manta tows which are conducted over transect of undefined width. Multiple regression analysis was used to determine which of 33 variables explained most of the variation in sightability in 10 m wide transects. The most influential variables were the proportion of cryptic starfish and an index of the degree of reef complexity. A regression equation designed to improve estimates of the abundance of A. planci on routine manta tows was developed. However, as it explained only 39% of the variation in sightability, this equation is of limited value in stabilising the negative bias associated with manta tow counts. In view of the variability of this bias, the manta tow technique is not suitable for estimating absolute densities of A. planci. Manta tow surveys are appropriate for identifying gross relative differences between densities of A. planci, and thus for determining broad-scale patterns of abundance. As such, they are a cost-effective method of estimating the geographical extent of A. planci outbreaks.