This paper presents the development of XKey, a tool for generating taxonomical identification keys by means of decision tree construction. The tool is based on an XML standard for the representation of general taxonomical information, which makes it ideal for different fields of application. The article analyses the problem by examining the adaptation of machine learning techniques to the sphere of biology so as to incorporate the viewpoints of biologists and computer science experts. It also analyses the effect of using various division criteria on a set of real data: the Gymnosperm plant groups present in the Iberian peninsula.