Here are a few references that used the Wizard of Oz approach:

Oviatt S. Predicting Spoken Disfluencies During Human-Computer
Interaction. 
     Computer Speech and Language 9(1):19-35, January 1995.  

Issacs E, Wulfman CE, Rohn JA, Lane CD, Fagan LM. Graphical Access to 
     Medical Expert System: IV. Experiments to Determine the Role of 
     Spoken Input, Methods Inform Med, 32(1):18-32, 1993.

You can check out http://www.cs.umbc.edu/~mikeg/annbib.html 
for related papers by these authors. I believe that Oviatt at the
Oregon Graduate Institute and Issacs at Stanford both use this
method a lot.

Diaper, D (1989) The Wizard's Apprentice: A Program to Help Analyse 
Natural Language Dialogues. In: People and Computers V, Sutcliffe, a. and 
Macaulay, L., Eds, Cambridge Univ Press, pp 231-243