Antibodies against primary structure

[Matthew Collins]

What is the minimum number of amino acid residues that an antibody can 
detect?  We are trying to model the loss of immunological epitopes in 
fossil proteins as a function of chain scission, the rate of which is 
being monitored as the release of free amino acids.  We are using 
polyclonal antisera, but assume that the most persistent epitopes are to 
primary structure and these are the ones that are capable of detecting 
samples which are > 1 million years old.   

To get the model to fit our data requires a minimum length of intact 
sequence that an antibody can detect of between three and four residues 
but this is much shorter than the 6-7 reside minimum I was taught as an 
undergraduate, implying that our model is wrong.

Does anyone know the answer (preferably with a citable source)?