Here we compare the applicability, the information provided and the cost-benefit of three sampling methods usually used in the study of rocky benthic assemblages. For comparative purposes, sampling was performed seasonally and along a depth gradient (0-50 m) in the Cabrera Archipelago (western Mediterranean). The destructive scraping (collection) method was the least cost-effective but provided the best qualitative and quantitative information. The in situ visual method was the most time-effective but provided low levels of taxonomic resolution and its accuracy decreased with depth due to the increasing difficulty of recognizing species in situ due to nitrogen narcosis, reduced light and cold. The photoquadrat method showed intermediate values of cost-effectiveness and information but was not suitable for multilayered assemblages, as it only accounted for the overstory. A canonical correspondence analysis showed that depth was highlighted as the main environmental gradient (16.0% of variance) by the three methods. However, differences due to the sampling method (7.9% of variance) were greater than differences due to temporal variability (5.8% of variance), suggesting that the three methods are valid but their selection has to be carefully assessed in relation to the targeted assemblages and the specific goals of each study.