Seven coral reef survey methods were compared in an experimental plot of 100 m² on a Caribbean shelf reef off southwest Puerto Rico. This area was mapped in detail by means of underwater photography and in situ drawings, in order to provide an objective standard against which to test the results obtained by the different survey methods. Minimal survey area was determined, and minimal sampling time deduced for one of the most laborious methods. Two 45-minute periods (i.e. 1.6 X minimal time) were chosen arbitrarily as the standard time alotted to each method. Three survey criteria were chosen: number of species observed, relative coverage (in %) and population densities (in colonies m²). The results of the latter two methods were compared to real values (as obtained from the map) and between each other by the Friedman two-way analysis of variance by ranks. It is concluded that point-intercept methods, whether linear (POLI) or planar (POSU), should be discarded. A randompoint method, the point-centered quarter method (POCQ) scores only moderately well. The same must be said for the line transect (LITR), so far the most popular method in reef surveys. The three remaining methods perform quite well for estimating the dominant species. The first one consists of in situ drawn maps of quadrats (ISMP). The second one is a photographic record of reef sections (PHRC); although giving good results, it is considered unpractical because of the equipment and the facilities involved, and the amount of time needed for working out the field data. Finally, counting of individual colonies and estimating relative coverage per species within 1 m² quadrats (ICCE) stands out as the most practical, versatile and reliable method.