A study is undertaken using anatomical measurements of specimens attributed to six species of Geospiza, the ground finches from the Galápagos archipelago. In a demonstration of method, a probabilistic approach associated with “sigma taxonomy” is adopted to assess the probability that pairs of specimens are or are not conspecific. We use a definition of a species based on morphometric analyses of the kind previously undertaken on extant vertebrate taxa (including mammals, birds and reptiles), using pairwise comparisons of anatomical measurements in regression analyses of the form y = mx + c from which the log-transformed standard error of the m-coefficient is calculated (“log sem”). The latter statistic is a reflection of variability in morphology. There is a high probability that at a species level, specimens attributed to G. magnirostris are different from those attributed to G. fulginosa, G. difficilis or G. scandens. Results of this study, using probabilistic sigma taxonomy, confirm the refutation of a single species hypothesis. In addition, we apply the log sem method to demonstrate that in case of comparisons between G. fortis and G. scandens (which are known to hybridise), there is a high probability that they are not different at a species level.