We present  formal models for some of the known algorithmic design paradigms. 
Ideally, a formal model of a paradigm should:
(1). Include all or nearly all of the known algorithms intuitively classified 
as examples of the technique;
(2). Capture the intrinsic characteristics of the paradigm;
(3). Be useful in lower bounding the limits of the technique through proofs 
of negative results for all algorithms in the formal model.

Building on \cite{BNR} we give submodels for greedy algorithms and dynamic 
programming.