We are developing a theory of probabilistic language learning in the c
ontext of robotic instruction in elementary assembly actions. We descr
ibe the process of machine learning in terms of the various events tha
t happen on a given trial, including the crucial association of words
with internal representations of their meaning. Of central importance
in learning is the generalization from utterances to grammatical forms
. Our system derives a comprehension grammar for a superset of a natur
al language from pairs of verbal stimuli like Go to the screw! and cor
responding internal representations of coerced actions. For the deriva
tion of a grammar no knowledge of the language to be learned is assume
d but only knowledge of an internal language. We present grammars for
English, Chinese, and German generated from a finite sample of about 5
00 commands that are roughly equivalent across the three languages. Al
l of the three grammars, which are context-free in form, accept an inf
inite set of commands in the given language.