What permits some systems to evolve and adapt more effectively than
others?  Gell-Mann (1990) has stressed the importance of
``compression'' for adaptive complex systems.  Information about the
environment is not simply recorded as a look-up table, but is rather
compressed in a \emph{theory} or \emph{schema}. Several conjectures
are proposed: (I) organisms generalize for compression; (II)
compression occurs more easily on a ``smooth'', as opposed to a
``rugged'', adaptive landscape; and (III) constraints from compression
make it likely that natural languages evolve towards smooth adaptive
landscapes.  We have been examining the role of such compression for
learning and evolution of formal languages by artificial agents.  Our
system does seem to conform generally to these expectations, but the
tradeoffs between compression and the errors that sometimes accompany
it need careful consideration.