The Madeiran archipelago consists of Madeira itself, Porto Santo and t
he Deserta islands. On Madeira, the forest and the coastal floral asso
ciations are so different that their faunas are effectively isolated a
nd have undergone largely independent development. There are different
faunal associations on the eastern peninsula and in the SE coastal re
gion, which may have been separated from each other in the past. On Po
rto Santo, western and eastern hills have different faunas. Most obser
vations on the fauna are compatible with evolution by allopatric speci
ation, consequent upon isolation on different islands or mountains, as
opposed to parapatric or sympatric processes following disruptive sel
ection. Some cases where the taxonomy is difficult to unravel may, how
ever, indicate parapatric speciation; examples belong to the genera Di
scula and Heterostoma (Helicidae) and Amphorella (Ferussaciidae). Most
evidence relating to species composition in communities is compatible
with a balance of random immigration and extinction, rather than sele
ctive interaction, allowing clusters of similar sympatric species to a
ccumulate. However, this impression may indicate that test procedures
are insufficiently sensitive to detect interactions, and detailed ecol
ogical studies are required. Questions about speciation and distributi
on would be clarified if dates of divergence were established. (C) 199
6 The Linnean Society of London