We discuss contrasting approaches to cladistic character definition and thus to cladistic data matrix compilation. The conventional approach considers character states as alternate forms of the “same thing” (the character). A review of the challenges to this convention is presented, and their implications evaluated. We argue that the recognition of structures which are alternate forms is a vital stage of primary homology assessment and is equivalent to the conceptualization of a transformational homology. Such a view complies with the demand that characters are independent and that character states are hierarchically related. We identify one justifiable solution to the inapplicable data coding problem (coding for organisms which have red tails, blue tails or no tails), and show that alternative approaches to character definition support spurious solutions which deny the relation of structures which are “the same but different”. We propose that the term character can be defined, in a cladistic context, as the descriptive label referring to a transformational homology evidenced by the similarity criterion.