Following the previous sections, the set of arguments can be partitioned in two different ways:
![\includegraphics[scale=0.4]{/home/lagasq/recherche/argumentation/eval-accep/JAIR-final/cmna4bis-angl.eps}](img494.png)
![\includegraphics[scale=0.4]{/home/lagasq/recherche/argumentation/eval-accep/JAIR-final/cmna5ter-angl.eps}](img495.png)
A very natural and interesting question is: is it possible to find a
semantics
and a gradual valuation
such that the associated
partitions have some compatibilities?
The following examples show that the class of the well-defended arguments does not correspond to the class of the cleanly-accepted arguments (in some cases, some uni-accepted arguments are even not well-defended).