Estimating length frequency distributions of large reef fish underwater

We describe the training of divers to recognise and remove bias in estimating lengths of fish underwater. Divers were asked to allocate objects, from a population (N=50) with a known length frequency distribution, to ten 100 mm size classes. Observed and expected distributions were then compared and the divers informed of their errors. Training continued until divers consistently produced length frequency distributions that were not significantly different from the expected distribution (=0.8) by the one sample Kolmogorov-Smirnov (K-S) test. Divers were trained in five trials, but after six months they had lost all their ability and had to be retrained. Three trained divers observing the same population of the large reef fish Plectropomus leopardus (Serranidae) produced length frequency distributions that were not significantly different (P>0.1) on 67% of occasions. Data collected by divers can be used to detect small but real differences in length frequency distributions of populations when analysed using the two sample K-S test. We suggest a means of determining within site variation in length frequency relative to between site variation.