CONVERGENCE IN EVOLUTION 261 
It is limited in its acquisitions by what has been 
already established, just as an architect in making 
an addition to a house is limited by the existing 
style. Flying birds differ greatly in detail, but 
within a relatively narrow range. No mutation 
can be considered, so to speak, which risks a loss 
of the secret of flight. Along this line of thought 
we may get another gleam of light on convergence. 
Other solutions than that adopted are possible to a 
theoretical organism, but not to this or that par- 
ticular organism which has to run in the trammels 
of a hereditary organization. Why should all the 
ordinary mammals of our acquaintance be restricted 
to four legs? The dachshund often looks as if it 
could do with another pair about half-way along, 
and so does a sow—also a somewhat artificial 
creature. The proposition of an extra pair cannot 
be considered, however, for higher vertebrates are 
hereditarily tied down to a maximum of two pairs of 
limb-buds, just as they probably are to a maximum 
of twelve cranial nerves, where an increase would be 
more readily effected. It is interesting, however, 
to force the question back and ask why some of the 
lower vertebrates, notably fishes, might not have 
tried the experiment of an additional pair of limbs. 
In all probability the answer is simple: that paired 
limbs were established at the level of fishes, and 
that their primary function was in connection with 
balancing, not with locomotion, the locomotor 
organ in almost all fishes being the so-called tail. 
Now, for balancing purposes the choice is between 
