Recessiveness (and, to some extent,
dominance) of
genes was first noted by
Gregor Mendel in his
pea plant experiments. It's a gross oversimplification, but entirely understandable given the lack of knowledge about the
inherent basis for genetic traits and
biochemistry. In most cases, a
recessive gene will encode the
absence of a
protein (well, more usually it'll be the reduction in
functionality of the gene rather than its complete absence - imagine an
enzyme with a malformed
active site, for instance). If (and this is a fairly big if) the level of the activity of the enzyme is not
critical (so only a relatively small amount of
functional enzyme needs to be present compared to the amount usually present) then carrying one copy that fails to encode a functional enzyme will still leave you with 50% of the normal level and everything will still work. With two copies, the level will drop further, the
pathway will no longer work properly and a new
phenotype will be expressed. This is classical recessiveness.
On the other hand, what if the concentration of the enzyme is important? As above, assume that our "recessive" is completely non-functional. With no copies of the recessive, we will have 100% of the normal level of enzymatic activity. With two copies of the recessive, we will have 0% of the normal level of enzymatic activity. With one copy of the recessive, we will have 50% of the normal level. This may be sufficient to have a phenotypic effect, which may well be halfway between the other two phenotypes. This is usually called incomplete dominance, but depends more on the characteristics of the recessive than anything else.
For an even more convoluted (but still observable) example, imagine a locus at which we have three alleles. A is normal. a is a typical recessive, reducing activity of the enzyme encoded at the locus by 50%. a' is a completely non-functional mutation. AA will be "normal", Aa will be "normal" (the activity level is 75% of normal), aa will be the "recessive phenotype" (the activity level is 75% of normal). However, Aa' will also be the "recessive phenotype". a' is dominant to A which is in turn dominant to a, but a' and aa both give the same phenotype.
gene interaction confuses things even further. Tubulin, the protein which polymerises into microtubules, is made up of two subunits (alpha and beta) encoded for by different genes. We'll call them A and B. A recessive mutation forms malformed subunits that are capable of binding to the other subunit but not of polymerising. AABb will have 50% of normal tubulin levels - 50% of the beta subunit will be malformed and therefore 50% of the tubulin formed will be unable to polymerise. AaBB will be the same. AAbb will have no functional tubulin and the organism will die. aaBB is the same. So far, we appear to have two recessive lethal mutations. However, AaBb will only have 25% of the normal levels of functional tubulin (simple probability) and will also be lethal.
The idea of a recessive gene is only really useful when genes are treated like black boxes. Nowadays we have sufficient understanding of the processes involved that different alleles can be catagorised on the basis of their biochemical effects. Trying to map these onto "recessive" and "domainant" just confuses people when things start behaving in ways that they don't expect. Even something thought of as a classical recessive may turn out to be significantly more complicated. Can't we stop trying to oversimplify things?