BackgroundThis study compares the efficiency of identifying the plants in an area of semi-arid Northeast Brazil by methods that a) access the local knowledge used in ethnobotanical studies using semi-structured interviews conducted within the entire community, an inventory interview conducted with two participants using the previously collected vegetation inventory, and a participatory workshop presenting exsiccates and photographs to 32 people and b) inventory the vegetation (phytosociology) in locations with different histories of disturbance using rectangular plots and quadrant points.MethodsThe proportion of species identified using each method was then compared with Cochran’s Q test. We calculated the use value (UV) of each species using semi-structured interviews; this quantitative index was correlated against values of the vegetation’s structural importance obtained from the sample plot method and point-centered quarter method applied in two areas with different historical usage. The analysis sought to correlate the relative importance of plants to the local community (use value - UV) with the ecological importance of the plants in the vegetation structure (importance value - IV; relative density - RD) by using different sampling methods to analyze the two areas.ResultsWith regard to the methods used for accessing the local knowledge, a difference was observed among the ethnobotanical methods of surveying species (Q = 13.37, df = 2, p = 0.0013): 44 species were identified in the inventory interview, 38 in the participatory workshop and 33 in the semi-structured interviews with the community. There was either no correlation between the UV, relative density (RD) and importance value (IV) of some species, or this correlation was negative.ConclusionIt was concluded that the inventory interview was the most efficient method for recording species and their uses, as it allowed more plants to be identified in their original environment. To optimize researchers’ time in future studies, the use of the point-centered quarter method rather than the sample plot method is recommended.