While Wittgenstein shows us there is a wide family of uses for the word "read," not all of which fit easily into this picture of reading as mental mechanism, he also argues that even in the most seemingly clear-cut cases, we are wrong to identify an internal mechanism. In game sixty-seven, he imagines a case of humans as "reading machines" that are simply trained to take written words from a page and pronounce them aloud. Untrained humans might occasionally guess a word by accident, but the trainer insists that they are not yet reading. Wittgenstein points out that a student who guesses one word correctly, then another, and slowly passes from a state of what the trainer would call "not reading" to one that the trainer would call "reading," does not make the transition from "not reading" to "reading" at any identifiable point.

This argument is not simply meant to show that we cannot know at what point the student stopped guessing and started reading, it is meant to show that no such point exists. This language game is based entirely on the behavior of the people reading. The game does not make any reference to their internal states. The trainer bases his judgment of whether or not someone is reading entirely on that person's response to written signs. "Read," in this language game, is defined by the trainer's observation of the students' behavior.

Suppose the word "table" was the first of a hundred that a student correctly pronounces in a row. When the student says "table," the trainer says the student is not reading, but one hundred words later, the trainer concludes that the student is now reading. We do not know whether or not the student read the word "table." We gather that the student is reading from his general behavior, because he correctly pronounces so many words in a row that we no longer call it an accident. When we say he is reading, we simply mean his behavior is of a certain general pattern. We are not pointing to an internal mechanism that might have been switched on when he looked at the word "table."

We should not mistake Wittgenstein as a behaviorist, however. Behaviorism is a psychological movement, popular in the middle of the twentieth century, which insists that all we can rightly claim about a person's psychology is based on that person's behavior. We cannot make any definite claims about people's internal structure or mechanisms, but can only observe the outward manifestations of their inner life. Wittgenstein differs significantly from this view in that he denies the necessity of an inner life to explain outward behavior. He is not simply saying that we can only observe the outward behavior of someone reading and cannot make any reasonable conjectures as to what sort of internal mechanism is at work. He is also saying that we have no good reason to suppose any kind of internal mechanism exists.

In game sixty-nine, Wittgenstein offers a possible general definition of reading, saying that reading is derived from the rule provided by the alphabet. This seems to be generally true of all instances of reading, which seems to refute Wittgenstein's insistence that there is no general definition for reading. However, as Wittgenstein has already shown, "following a rule" manifests itself in a wide variety of ways. Identifying reading as a form of rule-following gets us no closer to a general definition.

Wittgenstein concludes part I of the Brown Book by saying, "Our method is purely descriptive; the descriptions we give are not hints of explanations." He has given us numerous different examples of what "reading" is, how "can" is used, what "being able to go on" consists of, and so on. The point of these different examples is not to hint that there is some underlying commonality that all of these words share. The point is to show that there is a family of different uses for these expressions, and while some expressions share certain features with others, there is essential feature that make them what they are. In looking for internal mechanisms and other such secrets, we are barking up the wrong tree.