Minds, Brains, and Programs (1980) [pdf]

He makes an important distinction between being able to compute and actually understanding what one is doing. If a computer can do it, a person can then simulate the computer that did it, but that person wouldn't need to understand a word of Chinese to do any of the computations (since they are just instructions that need following).

Searching "Chinese Room" will bring up a ton of resources about this, and Wikipedia has a thorough entry as well.

https://en.wikipedia.org/wiki/Chinese_room

This is a classic article introducing the Chinese room argument:

https://en.wikipedia.org/wiki/Chinese_room

http://plato.stanford.edu/entries/chinese-room/

Searle pushes the implication of "Strong AI" that intelligence is independent of any particular type hardware (such as the human brain) if we have the right algorithm and implementation...e.g. that a mechanism of strings, beer cans and windmills can have the same sorts of beliefs that humans have given the right sort of "program" relative to mechanisms of strings, beer cans, and windmills.

Searle describes a mechanism, the Chinese Room, with external behaivor that we are inclined to say amounts to human thought. Yet the core operations are dependent on a human in a role analogous to a CPU, and when we look at that human, we are inclined to say that the human does not have the "right kind of thoughts" to justify our belief that the entire mechanism amounts to human thought.

In context it is a rebuttal of radical materialism (e.g. John McCarthy's argument that thermostats have beliefs) as a solution to the mind-body problem.