On Misinterpreting the `Phylogenetic Correction'

The last 10 years have seen important advances in methodology for taking phylogeny into account when analysing a comparative dataset. This commentary is about a class of interpretive procedures associated with these new statistical methods. We will call the interpretive logic 'phylogenetic correction' (PC for short), because this phrase summarizes the approach. Our essential message, however, is that a PC procedure is not in fact a 'correction', an adjustment to remove errors. Rather, it is a conceptual decision to give priority to one interpretation over another. Accordingly, it is an error to believe that PC is a methodology that must routinely be applied in all comparative analyses. The present Forum was proposed by the Editor during the review process for a paper about comparative ecology of seed mass (Leishman et al. 1995; see p. 517). Accordingly, examples will be drawn from the literature on seed mass, but the points made would be true of many species attributes. By 'comparative dataset', we mean a table with present-day species as rows, and attributes as columns. Some columns describe phylogenetic relations among the species, and others describe present-day attributes affecting the ecology of each species. Ecological attributes may include both those recording the behaviour of species in the field, such as capacity to establish in small canopy gaps (Kelly & Purvis 1993), and those that would be manifested even in glasshouse or garden, such as seeds per carpel (Hodgson & Mackey 1986). The analyses we are concerned with examine between-species variation in a focal ecological trait (here, seed mass), by investigating its correlation with other columns in the dataset.