Benthic community analysis: Are we interpreting our non-metric multidimensional analyses correctly?

Ecologists commonly summarize community and environmental data through the use of multivariate statistical methods.  A common issue with ordination approaches is determining how many axes need to be interpreted in order to provide a useful summary, and various tests and tools have been suggested to provide guidance.  Community ecologists use of non-metric multidimensional scaling (NMDS) has increased considerably in recent years given the availability of different tools to determine the “significance” of solutions.  NMDS is arguably now the multivariate method most commonly used by ecologists.  However, there has been little consideration of the performance characteristics of this approach and, as a consequence, when and how we should use it appropriately.  I examine how well these guidance tools perform in selecting the correct number of dimensions in NMDS using a set of simple ecological communities that are characteristic of various environmental relationships.