Non-conformable arguments while predicting with glmmadmb_model

I am attempting to do an observed versus predicted assessment of a negative
binomial glmm.

My first step was to simply use the predict( ) function with my glmmadmb
object on the dataset used to generate the model:

glmm_dataset$prediction<-predict(glmmadmb_object, glmm_dataset,
type="response")

But the error message results:
Error in X %*% beta : non-conformable arguments

What has baffled me is that when I use the same glmmadmb_object to predict
onto newdata, in the same way above, it runs the predictions fine, with no
error message and credible results.

Does anybody have ideas about what is going on here?