iv INTRODUCTION. 
If the first great difficulty of classification be to find out in each case whether a 
given character denotes affinity or only analogy, the second great difficulty is to 
ascertain the relative value of the different characters. Unfortunately no rule can 
be laid down; but there are certain laws which govern the operation of classification 
by diagnosis. The characters of each group must belong to every species in that 
group, otherwise it is obvious that artificial groups could be made to any extent. 
On the other hand, half the species in an outside group may be excluded by one 
character, and the other half by another character. For example, in diagnosing the 
Pico-Passeres the absence of the ambiens muscle excludes part of the Columbz, and 
the absence of the accessory femoro-caudal muscle excludes the rest. In diagnosing 
the Columbe the presence of basipterygoid processes excludes all the Pico-Passeres 
except the Trogones, and the schizorhinal character of the nasals excludes all the 
Pico-Passeres except a few Passeres. This is perfectly legitimate, but the converse 
would not be legitimate. To include amongst the characters of the Columb 
ambiens or accessory femoro-caudal muscles present, or to include amongst the 
characters of the Pico-Passeres nasals not schizorhinal unless accompanied by the 
absence of basipterygoid processes, though perfectly true, is not admissible, as it 
would introduce a practice that would make the combination of widely separated 
groups into an unnatural group perfectly easy. 
By strictly obeying this law the number of possible classifications becomes greatly 
restricted, and the danger of forming unnatural groups is greatly diminished. 
The breaking-down of a character in one group is no reason why it should not be 
employed in groups where it does not break down. For example, the fact that 
some of the Tubinares possess basipterygoids whilst others are without those 
processes is no bar to the presence of basipterygoid processes in the Limicole being 
used as a character to exclude the Gaviz. 
The student who attempts to diagnose a group will probably begin by being very 
saucy in the selection of his characters, and will no doubt arrogantly reject many 
which are in his opinion too trivial. He will most likely end in gratefully accepting 
an apparently slight character if it will only run through all the species in the group 
he is trying to diagnose. He will discover that Nature has decided for him which 
characters are important and which are trivial, and in which groups they are im- 
portant and in which groups they are trivial. It may almost be accepted as an axiom 
that every character breaks down somewhere. The range of variation in Birds is 
so small that there are scarcely any characters which are not passing through the 
process of being acquired or lost, as the case may be, in one of the thirty or forty 
suborders into which birds may be naturally divided. 
Fossil birds throw little or no light on the relative age of the various characters. 
