Messages as replications: Toward a message‐centered design strategy

defense of single-message designs from our suggestion that messages be replicated within experiments. We respond to each of these controversies. First, we examine their claim that controlled single-message designs completely eliminate all confounding of manipulated variables with other possible inßuences on the dependent variable and show it to rest on manifestly implausible assumptions. Second, we show why researchers should plan for the possibility of nonuniform treatment effects across mes.sages, and so use multiple-message designs; contrary to Hunter et al. s suggestion the available empirical evidence shows that treatment effects can and do vary across messages. Third, we discuss the advantages of having message replications both within and between studies, as opposed to Hunter et al.s suggestion that such replication occur only between studies; multiple-message designs provide greater reliability in estimation of treatment effects, equivalent power for detection oj variability in treatment effects, and easier identification of moderator variables. Other issues raised by Hunter, Hamilton, and Allen (nested vs. crossed designs, the desirability of experimental manipulation of messages, the benefits of meta-analysis) are in fact not controversial. MUCH communication research aims at the development and justification of general propositions about messages, but certain diflScult methodological problems! must be confronted if these aims are to be achieved. Jackson and Jacobs (1983)1 formulated these problems as having to do with generalizability and considered some abstract strategies for coping with them. The principal recommendation made by Jackson and Jacobs was that multiple messages be incorporated as replications in experiments aimed at general conclusions about messages and that messages be treated as an explicit random factor in statistical analysis. These suggestions and their rationale have been the subject of ongoing discussion (e.g., Bradac, 1983;