Harvey et al.: Improving the statistical power of length estimates of reef fish 
77 
publications and that measurement error is minimized to 
allow realistic interpretation of the data and comparisons 
of assemblages or individual species. Because of these er- 
rors and biases, it is likely that many studies lack the sta- 
tistical power to detect small, but biologically important, 
changes in fish length within reef-fish communities. 
Increasing numbers of volunteers are assisting with 
sampling for monitoring programs, particularly in third 
world countries (Halusky et ah, 1994; Mumby et al., 1995; 
Darwall and Dulvy, 1996). Darwall and Dulvy (1996) not- 
ed that the advantages of using volunteers in surveys in- 
clude not only greater manpower, enabling large spatial 
surveys, but also financial savings and increased public 
awareness of environmental issues through participation. 
To overcome the problem of subjectivity in visual esti- 
mates and to enhance the accuracy and precision of length 
estimates, and ultimately the power of a monitoring pro- 
gram, an underwater stereo-video system could be used. 
The use of an underwater stereo-video system provides an 
opportunity for volunteers to be involved in monitoring 
programs without compromising the quality of the data. 
In assessing the statistical precision of visual length es- 
timates, we have focused solely on the extra variability 
in the estimates introduced by measurement error. This 
error can also introduce a bias in the estimates. For ex- 
ample, the experienced scientific divers used by Harvey et 
al. (2000) had a mean relative error of -8.6% (SE=1.9%), 
suggesting that they would on average underestimate the 
true length by between 5% and 12%. In the type of study 
for which we have predicted the power to detect a differ- 
ence in mean length, this bias will make little, if any, con- 
tribution to the loss of precision because the true lengths 
