The essay is concerned with Norman Malcolm's book "Dreaming" (1959). Malcolm's conceptual analysis is strongly influenced by Wittgenstein, but it has its starting point in the skeptical problem of distinguishing between dream and reality. Malcolm claims that what determines when dreams occur during sleep is nothing but "grammar", and that such a grammar does not tell us the precise time when a sleeper dreams. It follows that a naturalistic account of dreaming, which tries to correlate dreams to eye movements or to neurochemical processes, is not possible; what scientists actually do is to stipulate a new meaning of the word "dreaming". The essay accepts Putnam's criticism of Malcolm's verificationism but rejects the use of the theory of "natural kinds" in this context. It also suggests that Putnam's criticisms were unneeded, since Malcolm simply misdescribed the ordinary concept of dreaming, according to which, in fact, we do have chance to determine the temporal occurrence of dreams during sleep; this is enough to make the empirical correlation supposed by scientists possible.