Replies to Sally Jackson, Daniel O'Keefe, and Scott Jacobs' article (same issue), maintaining that randomness requirements can not be relaxed for generalizing from message samples, since systematic samples are not truly random. (MS)

Argues that generalizing to message populations by treating messages as a random variable is inappropriate for complex messages, and proposes meta-analytic techniques for investigating biases in message samples and other methodological factors that can limit generalizability of communication research. (MS)