In one respect the coding of these characters differs fundamentally from that of the lexical characters. In the case of the latter we restricted ourselves to a single well-attested dialect of each language in order to assure some degree of uniformity in the data. For the characters coded here that would be self-defeating, because (a) there is ample evidence that these characters are far likelier to reflect the true tree—so that we want to recover as much evidence as possible for each character in each language—and (b) we know a great deal about the prehistory of some details for some languages. Thus we have not excluded Old Latin, G3thic Avestan, morphological relics, and so on in the coding of the phonological and morphological characters, whereas we did exclude archaic and dialectal forms, as well as relic words, from our lexical database. The number of morphological characters is much smaller than we would have preferred , for the following reasons. Many potential characters exhibit the same state for all the languages, or all the languages but one; they are compatible with every possible tree and are thus uninformative, and including them would not help to find the true tree. Still others exhibit distinctive innovations for each well-established subgroup, so that including them simply confirms the existence of the known subgroups without contributing to the reconstruction of their interrelationships; we have included a few exceptionally interesting ones, and further examples can be found in the standard handbooks. Finally, many potential morphological characters exhibit clear parallel development; the loss of various grammatical categories (such as the dual) is an obvious example, and other examples can be found in the development of the verb system and of pronouns. Several characters that were originally included in our database were eventually removed for that and other reasons (see the discussion at the end), and new evidence shows that one which is still listed below—M11— should be removed as well (see the discussion under that character; we continue to include it here because it was used in Ringe, Warnow, and Taylor 2002). A few uninformative characters have been included in spite of the considerations just outlined because of their interest to specialists in the field. Except as noted, our morphological characters are not directed. helpful discussion of particular morphological characters and for several references.