[R-sig-ME] citation for significance testing in mixed models

[Elizabeth Crone]

Hello,

I cannot find a good (published) citation for the statement that significance tests from mixed models (e.g., the t-statistics produced by lmer) are approximate and should be interpreted as such.  It would make me particularly happy if I could attribute the statement to D. Bates.

I often need to explain the idea to colleagues (ecologists) and reviewers of articles in ecological journals.  It would be nice to be able to argue from authority, as well as from first principles.

Humble apologies if I am missing something obvious....

Thanks!
Elizabeth