That I know of there is no theoretical limit on element size relative to the thickness, either large or small. However there are discretization considerations:
 - the elements should be small enough to capture the   
   distribution of the thing you are examining.
 - however, there is no point in going too small.
In general the rule that I have used is that an element need be no smaller than 2xthickness. It will depend on what you are examining though.

TERRY

I really dont know. Should the aspect ratio with reosect to thickness matter at all. i belive the formulation should take care of thickness wise variation.

The rule of thumb aspect ratio for thin shell theory has to do with the actual part being modeled not the elements.  You can decrease the size of the elements for refinement and, if they are thin shell formulated elements, the analyis wiill converge to thin shell theory.

caviac,

Governing equations are physically accurate *globally*, and for shell theory arguments two directions should much greater than the 3rd direction this is correct, e.g 10 times like you stated (note that many fea software solvers have thick element formulations which are quite accurate).

                  {F} = {K}{D}     GLOBAL MATRICES

Thees equations can also be written for individual elements, or "locally"

                   {f} = {k}{d}    LOCAL, ELEMENTS

But the key is the elements equations / matrices are then assembled into teh global matrices, and it is the global matrices that are solved / inverted. The element equations are only physically reasonable when assembled in into the GLOBAL equations. Governing equations applied only on an element and not assembleed into a global matrix, must also follow shell theory assumptions for them to be accurate.

Hope this Helps,

George