The observable computing actions of a Turing machine，including moving and writing，can be categorized as the behavior of the Turing machine．In this paper，it is shown that the behavior of a Turing machine is learnable．That is，there exists a learning procedure which，given the behavior of a Turing machine，will yield a new Turing machine with the same behavior in a finite number of steps．According to the same mechanism，the behavior of an LR parser is also learnable．