Jim Stewart's paper proposes that there are multiple kinds of ad hocness, showing significant differences between two of those kinds (token and type ad hocness). 

Abstract
Defining the concept of ad hocness has proven notoriously difficult. I propose that one reason for this difficulty is the fact that there are multiple kinds of ad hocness that have been unwittingly lumped together. To this end, I argue that there are clear differences between at least two kinds of ad hocness: token ad hocness and type ad hocness.  

• Token ad hocness is seen in hypotheses that postulate the existence of a specific token for which there is no previous evidence (such as the first postulations about the yet-to-be-discovered planet Neptune).  
• Type ad hocness is seen in hypotheses that postulate the existence of an entirely new type of entity or phenomenon. (An example of type ad hocness is found in the neutrino hypothesis of Wolfgang Pauli.)  

The criteria for these kinds of ad hocness are quite different. Failing to distinguish between these has heightened the difficulty of defining ad hocness and our ability to separate ad hoc from non-ad-hoc hypotheses.