[R-sig-ME] fitted values for logit models using glmer

[Malcolm Fairbrother]

Dear all,

I have been trying to determine what I get with "fitted(model)" when  
the model in question is a logit, fitted using glmer. After some  
investigation, my current impression is that I get predicted  
probabilities. However, I've found at least three suggestions to the  
contrary scattered around the web:

http://psy-ed.wikidot.com/glm
http://clic.cimec.unitn.it/methods/R/lectures/regression3-ho.pdf  
(second last page)
http://tolstoy.newcastle.edu.au/R/e2/help/06/11/6209.html

Can anyone please tell me whether I'm right or they're right? I'm  
quite happy to be wrong, but I'd like to know for sure. Was there  
perhaps a change made to the package in this regard at some point?

Many thanks,

Malcolm Fairbrother

School of Geographical Sciences
University of Bristol