Co-occurrence and correlation of pairs of species is an important element in ecological data analysis. When using the co-occurrence indices, it is, however, important to understand that it matters both how the correlation is computed and how the base set is selected. If the base set is selected improperly the observed correlation can be due to some relatively trivial reason, such as both species existing on the same continent. We show how to apply spatial, temporal, and taxonomic criteria to select a proper base set. Similarity indices such as the Jaccard index sidestep this problem by ignoring locations in which neither of the species exist, but as a result, these indices are not suitable indicators of existence or non-existence of correlations.