R help - Aug 2008 - Covariance structure determination when lmer has false convergence.

I have fit a model with a more complex covariance structure, but the fit reports
a false convergence.  I have read from past posts that this can be an indication
of over-specification.  I went ahead and fit a model with a simpler covariance
structure.  It doesn't seem like I can compare the two likelihoods or the
AIC or BIC to compare the two model since the one model had false convergence. 
Is this correct?  If so, what would be a good approach to comparing the two
models to validate the smaller model?
 
Thanks for the time.
 
Todd Remund
	[[alternative HTML version deleted]]