Message discrimination is a self-report measure of media exposure. In survey interviews, respondents are asked to recall information about a particular topic that they have encountered in various media in the recent past. For example, respondents might be asked, “What have you seen or heard on television about family planning in the last month?” Open questions like this are repeated for other media (e.g., newspapers, radio, magazines, books). Responses are recorded verbatim and coded into “messages.” The message discrimination measure is the sum of messages reported by a respondent for the topic of interest across all of the media.
The term was coined by Peter Clarke and F. Gerald Kline (1974) in an article introducing the measure and its use in research on media effects. The approach appears to have been presaged in the work of Edelstein (1974). His comparative research in the United States and Yugoslavia involved asking respondents to nominate important problems confronting their local area and the world. Respondents were then queried about the sources of information on these problems (media), what solutions had been proposed, and what actors were involved in addressing the problems. Message discrimination features this “respondent-centered and open-ended” approach, as Clarke and Kline termed it. Clarke was a colleague of Edelstein at the University of Washington prior to joining Kline on the faculty at the University of Michigan in the early 1970s. Clarke and Kline (1974) illustrated the message discrimination measurement approach with a small study that employed Edelstein’s “problem – solution” questioning design. In subsequent research, the recall of information about solutions to a respondent-nominated problem was replaced by recall of information about a topic supplied by investigators.
Motives for Using Message Discrimination Measures
The rationale offered by Clarke and Kline for the use of open recall of information as a measure of media exposure rests on two claims. First, they asserted that frequently used measures of media exposure, which focus on reported time expenditure or frequency of contact with a medium, are too crude to capture what might be called “meaningful” exposure to the channel. At most, such measures denote the potential for exposure to particular content. Because they are so gross, the standard “exposure” measures are unlikely to show much predictive power when used as independent variables in media effects research. Clarke and Kline argued that media effects had been generally underestimated in the literature because “time spent” or “frequency of use” measures were so commonly employed as predictors.
The second rationale for the message discrimination measure was normative. Clarke and Kline struck a decidedly egalitarian note when they declared that previous media research had relied too much on “researcher” definitions of media exposure and outcomes.
They contended that the extensiveness of respondents’ cognitions should define exposure and that respondents’ reported “information-holding” should replace elite “textbook knowledge” as the likely outcome of exposure. Thus, message discrimination may be seen as part of a more general reaction in mass communication scholarship in the early 1970s against the “limited media effects” viewpoint that had dominated in the previous two decades (see Klapper 1960). In sum, the contention was that genuine media effects had been overlooked in research to date because the independent and dependent variables had been erroneously defined.
Applications in Empirical Research
An example supporting the latter argument may be seen in Clarke and Fredin (1978). This study involved analysis of data from the 1974 National Election Study post-election interviews. The hypothesis examined was that message discrimination about the campaign and about respondent-nominated problems facing the country would better predict “information-holding” (in this case, the number of reasons offered for supporting or opposing senatorial candidates) than would estimated frequency of television news viewing or number of daily newspapers read. Message discrimination measures for television and newspapers did predict the number of reasons offered by respondents better than did the number of newspapers read or the frequency of television news viewing, controlling for education and political interest. A finding for which the authors had no ready explanation was the fact that the number of messages discriminated from television was a negative predictor of information-holding. But the principle that exposure measured through message discrimination should predict knowledge better than exposure measures based on frequency of use was buttressed.
Two field experiments prominently featured message discrimination in their assessments of media use. The first study, conducted in Flint, Michigan, and Toledo, Ohio, focused on adolescent learning about such topics as family planning and drug abuse. The second experiment was conducted in several Minnesota cities and concerned heart disease prevention. Example publications from these studies are Kline et al. (1974) and Jacobs et al. (1986). There is some irony in the observation that the “respondent-centered” message discrimination measure received its major application in the evaluation of “researcher-driven” information campaigns.
Finnegan et al. (1989) examine aspects of the measurement technique in detail. The authors note that there was considerable operational variability in the approach across studies. While some investigations focused the message discrimination sequence with respondent-nominated issues, the field experiments queried respondents about topics of mass media campaigns (e.g., family planning, heart disease). Kline et al. (1974) asked respondents what they had seen, heard, or read about the topic of interest, repeating the question for each of several media. Jacobs et al. (1986) asked a more general question: “What specifically have you read, seen or heard about heart attack and stroke during the past few months from any of these sources?” (Respondents were handed a card listing various media.) The reference period for recall also varied across studies, with some delimiting a particular time frame (“the past two weeks” or “past month”), while others asked about more vague periods of time (“past few months” or “lately”). Thus, while the several studies that have employed message discrimination shared the free recall approach – allowing respondents to report on media exposure “in their own words” – there is a case to be made that message discrimination is not one measurement technique, but several.
Lack of Adoption of Message Discrimination Measures in Research
As Finnegan et al. (1989) note, the message discrimination approach has not supplanted the time-based or frequency-of-use media exposure measures in reaction to which Clarke and Kline offered their alternative. Apart from research in which they, their colleagues, and their students were involved in the 1970s and 1980s, message discrimination has not played a role in the literature. This situation obtains despite the undeniable weaknesses of the standard exposure measures and some empirical support for the proposition that message discrimination better predicts information gain. It is worth speculating on why this is the case.
One reason may be ambiguity in the measurement of the concept. Message discrimination is supposed to measure media exposure, but it does so by asking respondents to recall information that they have received via the media. This means that exposure is confounded with information gain, which is usually treated as a dependent variable in media effects research. The confounding is intentional – Clarke and Kline argue that exposure is topic-specific and that it has not occurred without evidence that the putative audience member has “discriminated some symbols” relevant to the topic. But how are these “symbols” to be distinguished from knowledge about the topic that is the outcome of exposure? Clarke and Kline assert that message discrimination is distinct from “information-holding,” the dependent variable in their model of media effects. As Salmon (1986) observed, however, both message discrimination and information-holding rely on long-term recall of information. There is nothing in the measurement procedure as applied to date to ensure that message discrimination picks up information that is antecedent to knowledge, or even to ensure that the “symbols” recorded are actually the result of contact with media. Unlike Southwell et al. (2002), who assessed exposure to a health information campaign through recognition of campaign messages, researchers who have employed message discrimination have not verified that recalled messages originated in the media. While in some instantiations of the message discrimination technique, respondents have been asked to recall information that they encountered in different media – as noted above – the “messages” reported were not checked against a compendium of media content so as to validate the reports.
Another, maybe more powerful reason for the lack of widespread adoption of the message discrimination measure is that it is more laborious and expensive to execute than the time-allocation or frequency-of-use measures. Like any open questioning approach, message discrimination involves questioning, probing, and careful recording of responses. Interviewers must be extensively trained and the interviews they conduct will take longer than ones that rely on closed questions. Interviewer variance is apt to be greater for this sort of measure than for closed question approaches. Once obtained, the responses must be unitized (“messages” must be identified) and placed in categories by coders. Inter-coder reliability is an additional concern. The message discrimination approach can be burdensome to respondents too, despite the claim that it allows them to “speak in their own words.” Responding to open questions often requires more difficult comprehension, retrieval of relevant information, organization of thoughts and articulation than is involved in responding to closed questions. Greater familiarity and ease of use probably gave an advantage to the traditional media exposure questions that Clarke and Kline sought to replace with message discrimination.
Advances in survey technology and computational linguistics could make application of message discrimination measurement more tractable for media effects researchers. Computer-assisted interviewing, voice recognition, and automated content analysis software could reduce the burden and expense of collecting and summarizing reports of information gleaned from media sources. Validation of these reports and clear differentiation of them from outcome measures of media effects is required for the message discrimination approach to achieve more widespread adoption.
References:
- Clarke, P., & Fredin, E. (1978). Newspapers, television and political reasoning. Public Opinion Quarterly, 42(2), 143 –160.
- Clarke, P., & Kline, F. G. (1974). Media effects reconsidered: Some new strategies for communication research. Communication Research, 1(2), 224 –240.
- Edelstein, A. S. (1974). The uses of communication in decision-making. New York: Praeger.
- Finnegan, J. R., Viswanath, K., Hannan, P. J., Weisbrod, R., & Jacobs, D. R. (1989). Message discrimination: A study of its use in a campaign research project. Communication Research, 16(6), 770 –792.
- Jacobs, D. R., Luepker, R. V., Mittelmark, M. B., et al. (1986). Community-wide prevention strategies: Evaluation design of the Minnesota heart health program. Journal of Chronic Disease, 39(8), 775 –788.
- Klapper, J. (1960). The effects of mass communication. New York: Free Press.
- Kline, F. G., Miller, P. V., & Morrison, A. J. (1974). Adolescents and family planning information: An exploration of audience needs and media effects. In J. Blumler & E. Katz (eds.), The uses of mass communication. Beverly Hills, CA: Sage, 113 –136.
- Salmon, C. T. (1986). Message discrimination and the information environment. Communication Research, 13(3), 363 –372.
- Southwell, B. G., Barmada, C. H., Hornik, R. C., & Maklan, D. M. (2002). Can we measure encoded exposure? Validation evidence from a national campaign. Journal of Health Communication, 7(5), 445 – 453.
Back to Communication and Social Change.