The concept of homology arose from classical studies of comparative mo
rphology, and took on a new significance with the advent of evolutiona
ry theory. It is currently undergoing another metamorphosis: many deve
lopmental geneticists now define homology as shared patterns of gene e
xpression. However, this new usage conflates definition with criteria,
and fails to recognize that meaningful assignments of homology must s
pecify a biological level. We argue that although developmental geneti
c data can help identify homologous structures, they are neither neces
sary nor sufficient, and do not in any case justify a new definition o
f homology.