Probing the Locus of Causation in the Miscomprehension of Remedial Advertising Statements

Jacob Jacoby, New York University
Margaret C. Nelson, State University of New York-Albany
Wayne D. Hoyer, The University of Texas at Austin
Hal G. Gueutal, Rensselaer Polytechnic Institute
ABSTRACT - Application of a recently developed MDS algorithm (which is unique in that it permits empirical testing of its key assumptions) to a set of remedial advertising statements proposed by the FTC suggests that consumer miscomprehension of these messages may be due to decoding failures on the part of the recipients rather than encoding failures on the part of the source.
[ to cite ]:
Jacob Jacoby, Margaret C. Nelson, Wayne D. Hoyer, and Hal G. Gueutal (1984) ,"Probing the Locus of Causation in the Miscomprehension of Remedial Advertising Statements", in NA - Advances in Consumer Research Volume 11, eds. Thomas C. Kinnear, Provo, UT : Association for Consumer Research, Pages: 379-384.

Advances in Consumer Research Volume 11, 1984      Pages 379-384


Jacob Jacoby, New York University

Margaret C. Nelson, State University of New York-Albany

Wayne D. Hoyer, The University of Texas at Austin

Hal G. Gueutal, Rensselaer Polytechnic Institute


Application of a recently developed MDS algorithm (which is unique in that it permits empirical testing of its key assumptions) to a set of remedial advertising statements proposed by the FTC suggests that consumer miscomprehension of these messages may be due to decoding failures on the part of the recipients rather than encoding failures on the part of the source.


Except for one prior investigation, research on remedial advertising (i.e., affirmative disclosure and corrective)to date has focused on the impact that such advertising statements have on consumer beliefs, attitudes, and/or purchase intentions. Yet, according to prevalent models of communication and advertising impact (e.g. Engel, Blackwell, and Kollat 1978; Lavidge and Steiner, 1961; McGuire, 1976), a logically prior question is whether such advertising is accurately comprehended. If remedial statements are not accurately understood, then their impact on such subsequent states as beliefs, attitudes, and intentions becomes academic.

Using remedial statement, developed by the Federal Trade Commission for insertion in the advertising for two popular branded products, a prior investigation found that consumers more often extracted confused or incorrect meanings than they did the meaning intended by the FTC (Jacoby, Nelson, and Hoyer 1982). However, demonstrating that some level of miscomprehension exists is insufficient when remedial advertising has been mandated. Given that a problem of comprehension has been identified, it becomes important to also identify the likely causes of that problem so that ameliorative actions can be taken. The present investigation addresses this issue, i.e. the locus of such miscomprehension.

Assuming that the channel (or medium) does not distort the communication, such miscomprehension can be due primarily to either encoding problems introduced by the source in constructing the message, or to decoding problems introduced by the receiver when interpreting the message. Resolving the problem would require a different approach if the problem were due to source encoding, than if the source had done a good job of encoding the message and the miscomprehension lay with the receiver. Diagnostic research aimed at suggesting the probable locus of such miscomprehension will help ensure that efforts to ameliorate the problem are not misguided.

In other words, stated somewhat differently, three basic questions must be asked of any remedial statement being considered for inclusion in advertising: (1) Does miscomprehension of that remedial statement occur at sufficiently high levels so as to give reason for concern? (2) If so, can the locus of causation be identified, i.e., is the problem likely one of improper encoding on the part of the source or improper decoding on the part of the receivers? (3) Given that the locus of causation can be identified, what actions can be taken to rectify the problem? The earlier investigation (Jacoby, Nelson and Hoyer, 1982) addressed the first of these questions. The present investigation illustrates an approach which has relevance for addressing the second. Though by itself, it cannot identify the locus of causation with certainty, it can point to the existence or absence of a certain amount of fuzziness in source encoding.

Determining miscomprehension necessarily means assessment at the point of the receiver. Given findings which indicate that substantial numbers of respondents miscomprehended a given remedial advertising statement, how can one tell whether such miscomprehension is likely due to source encoding or receiver decoding? Here the widely accepted model underlying most psycholinguistic theory and research (see Lindsay and Norman, 1977, Spiro, Bruce and Brewer, 1980) proves useful.

A Psycholinguistic Model of the Communication of Meaning

When a source deliberately engages in communication, why does s/he do so? What does s/he hope to accomplish? At its most fundamental level, the answer is that the source has some thought in mind which s/he wishes the other par.y (i.e., the "receiver") to come to understand. For the moment, let us call this internal mental state of the source his/her "meaning."

Unfortunately, the thought which exists in the mind of a source cannot be directly transposed into the mind of a receiver. For the source to communicate a thought (i.e., evoke the intended meaning in the mind of the receiver), he must convert it into some externally denotable form such as spoken words, written words, or some visible gesture. The receiver must then decode this overt expression and extract meaning from it--hopefully, the same meaning intended by the source.

In terms of the vocabulary that has evolved (cf. Lindsay and Norman, 1977, p. 470-471), meaning structure is the label used to designate the thoughts that exist in the minds of individuals. These meaning structures are never visible. Surface structure is the label used to designate the externally visible expressions of these thoughts. That is, to communicate his/her thought (">leaning Structure 1"), the source must use some surface structure in an attempt to evoke Meaning Structure in the mind or the receiver. Should the source succeed in evoking precisely this one thought and only this thought, then we have correct, faithful, and entirely accurate communication. Should s/he fail to evoke Meaning Structure 1 but succeed in evoking some other meaning in the mind of the receiver (say Meaning Structure 2 or 3 or 4), when we have miscommunication. Should s/he succeed in evoking Meaning Structure 1 but also evoke one or more other meaning structures, then we have ambiguous, confusing communication perhaps a combination of accurate and inaccurate communication). [For purposes of simplification, we have written as if the communication of Meaning Structure 1 is an all-or-none phenomenon; i.e., the receiver, he either does or does not comprehend Meaning Structure 1. Although we employ this simplistic fiction throughout, the reader should be alerted to the fact that miscommunication, in the form of "misinterpretation and misunderstanding, will usually vary from subtle nuances of meaning to complete distortion of the intended message" (Harris and Monaco 1978, p. 2).]

We turn now to a specific case -- in which remedial advertising is currently being considered for two popular products -- to illustrate how these concepts may be used to probe the likely locus of causation for miscomprehension.


The case involves two familiar analgesic products manufactured by the Bristol-Myers Company -- Bufferin and Excedrin. The Federal Trade Commission contends that an integral part of Bufferin advertising has been the claim that Bufferin is a faster pain reliever and gentler to the stomach than plain aspirin. Similarly, it is asserted that Excedrin advertising contains the claim that the product is a more effective pain reliever than is plain aspirin. The FTC contends that a significant proportion of American consumers now hold these beliefs as a result of exposure to television advertising.

The problem, according to the FTC staff is that the degree of empirical proof available to support the claims is not as conclusive as the FTC staff believes it should be. Hence, the FTC staff would like consumers to know that, while a certain amount of proof is available to show that these claims may be true, the degree of proof is insufficient according to some authorities.

The rejoiner made by Bristol-Myers is that, while no "irrefutable and incontrovertible proof positive" exists to support these claims, the proof that is available does meet commonly accepted standards for such evidence.

With the expectation that the FTC Administrative Law Judge would rule in its favor, the FTC Complaint Counsel sent a letter to the Judge. The letter provided two sets of remedial statements, one for each product, and proposed that one of the statements in each set be mandated for inclusion in advertising for the respective products.

The three "equally acceptable" statements proposed by the FTC Complaint Counsel for use in remedial Excedrin advertisements are asterisked in Table 2. A similar set was proposed for remedial Bufferin ads. Further in a letter to Montgomery K. Hyun, Administrative Law Judge, [Fisherow, W. Benjamin. Complaint Counsel's statement concerning corrective advertising for Excedrin and Bufferin. Letter dated January 31, 1979, and sent to Montgomery K. Hyun, Administrative Law Judge. In the matter of: Bristol-Myers Company, Ted Bates and Company, Inc., and Young and Rubicam, Inc. Docket No. 8917, Federal Trade Commission.] it was stated that these phrases were equally acceptable.

It can be seen that the remedial statements proposed by the FTC are basically three different surface structures designed to convey a common meaning structure, i.e., they are different "equally acceptable" ways of saying the same thing. Indeed, there would be no logical reason for trying to convey different meanings. However, the presence of such alternative surface structures which provides an opportunity to explore whether miscomprehension is likely due to encoding or decoding problems.


The earlier investigation (Jacoby, Nelson and Hoyer, 1982) found that the three surface structures proposed by the FTC more often led receivers to extract confused or incorrect meanings than it did the intended (or accurate) meaning. By itself, this finding tells us nothing regarding probable the locus of causation for miscomprehension. It could be in the first link between the intended meaning and the surface structures, or in the second link between the surface structures and the meanings extracted. If one could reduce the possibility that either of these links contained the probable locus of the problem, then efforts to ameliorate miscomprehension could be more productively focused on the remaining link.

If the three remedial statements proposed by the FTC were truly "equally acceptable" and substitutable, we would likely expect that a larger set of statements communicating various related meanings would cluster together when plotted in multi-dimensional space together with confirming this expectation would reduce the possibility that miscomprehension lies at the encoding stage, since their "substitutability" would be supported. However, failure to confirm this expectation would suggest that at least some (perhaps even most) of the confusion could be traced back to the encoding stage. In turn this would imply that resolution of the problem might be accomplished by first attempting a revision in message encoding (i.e., language).



In keeping with the underlying rationale, it was reasoned that an educationally up-scale sample would suit the objectives of this exploratory investigation. Remember, the purpose was to determine how similar in meaning the test statements were to each other ('ink 1) not to determine whether they would be miscomprehended (link 2). Whereas this latter question is best addressed by using representative samples, the former (i.e., link 1) question is best examined using subjects whom, it can be assumed, are less likely to experience confusion with these statements. The subjects were 133 upper-level undergraduates and master's level students enrolled in a consumer behavior course during the spring of 1980. Their participation was entirelY voluntarY.

Instrument Procedure

Because of their greater simPlicity and demonstrated ability to be better comprehended (see Jacoby, Nelson, and Hoyer, 1982), attention in this study focused on the three Excedrin statements rather than on their Bufferin counterparts. Seven other relevant statements were developed and added to the three FTC generated statements, thus forming a set of 10 test statements (see Table 1). Two of these interpretations (Statements 6 and 9) were deliberately structured to also communicate the meaning that the authors believed the FTC intended to convey, while another interpretation (Statement 10) was structured to communicate the opposite meaning. [Statements which have nothing to do with the issue (e.g., "The cat was brown") could have also been constructed. Under such circumstances, demonstrating that the three FTC statements clustered together when placed within a set of such statements would provide little insight.]

Each of the ten statements was paired once with every other statement on the list, yielding 45 pairs of statements. The instrument consisted of a 45-page paired comparison questionnaire, with each page containing one pair of statements and a 9-point similarity scale labeled "extremely dissimilar" and "extremely similar" at the ends. Each page asked the subJect to judge the similarity in meaning of the statement pair appearing on that page. Ten of the 1.1962 x10 56 possible orderings of these 45 pairs were selected at random to create ten different forms of the instrument. Instructions on how to use the scale included six examples and practice items and encouraged questions from participants who did not understand the procedure. All subjects participated during a single group session at which the forms were randomly distributed. After all subjects had finished, full explanation of the study was provided.

Scaling Algorithm

The similarity data were analyzed via the Horan Scaling Model (1969), as implemented in the computer algorithm COSPA (Schonemann, James, and Carter 1978). [The authors are extremely grateful to Dr. Peter Schonemann for discussing application of these programs to the present problem and reviewing the data print-outs and manuscript prior to submission.] This algorithm, as all multidimensional scaling (MDS) techniques, seeks to represent graphically the perceptual space in which the stimulus objects are viewed. However, unlike any of the MDS algorithms now being used for marketing applications, the Horan model does possess a unique advantage. Typically, the researcher employing MDS algorithms has no way to assess whether, in fact, his/her data are appropriate for use with the scaling model (i.e., whether the data meet the assumptions of the model). Thus the conclusions drawn from such research are tenuous. The Horan model, on the other hand, provides two test indices (derived in Schonemann, James, and Carter 1979) which allow the user to determine whether the data meet the two key assumptions of the model. The first of these, the common space index, assesses the degree to which subjects perceive the stimulus set in a common m-dimensional space. The second, the diagonality index, determines whether the user may assume that this solution is orientationally invariant.




In its current form, the COSPA algorithm can scale data from a maximum of 100 subjects simultaneously. Accordingly, the 133 subjects in the present study were randomly divided into two groups, termed Replications 1 and 2 (67 and 66 subjects, respectively). This enabled the scalings from the separate subsamples to be compared and permitted an assessment of the stability of the solutions across the entire sample. For each replication, scaling solutions were derived in one, two and three dimensions.

The first step in determining the optimal solution involves assessment of the Horan model's test indices. For both replications, the common space and diagonality constraints were satisfied (p < .001). Thus, the assumption that subjects viewed the statements along a common set of dimensions and the assumption of rotational invariance in the solutions appear justified.

The second step in deriving the optimal solution involves determining the number of dimensions which should be retained. The scalings of these data seem to indicate that a two-dimensional solution is most appropriate. This conclusion is based on several factors. First, little improvement in explained variance was obtained by proceeding from a two- to a three-dimensional solution. Based on the common space index (cf. Schonemann, James and Carter 1979), the unidimensional solution accounted for 35.9% and 34.6% of the variance in Replications 1 and 2, respectively. The variance accounted for in the two dimensional solution was 61.2% and 60.1% in Replications 1 and 2 respectively. In moving from the two-dimensional to the three-dimensional solution, the explained variance increased only 6% to 7% (to 68.0: and 66.9%). Since the observed increase between the unidimensional and two-dimensional solutions was of substantially greater magnitude than that between the two- and three-dimensional solutions, the variance represented by the addition of the third dimension is likely mostly error variance.

A second criterion for maintaining that the two-dimensional solution is optimal is the consideration of interpretability. In both replications, the unidimensional solution is interpretable and appears to represent a continuum of "degree of proof for Excedrin's greater effectiveness" (see Table 1). In the two-dimensional solution, this initial dimension clearly splits into one termed "certainty" (shown on the vertical axes of Figures 1 and 2) and another term "favorableness" (shown on the horizontal axes).

As can be seen in Figures 1 and 2 all of the inconclusively worded statements (1, 2, 3, 6, 8, and 9) fall at one end of the certainty dimension, while all of the conclusively phrased statements (4, 5, 7, and 10) fall at the other. Given that the FTC's purported objective is not to inform consumers that Excedrin is not more effective than aspirin, but only to raise doubts (i.e., increase uncertainty in their minds regarding the evidence), it follows that a remedial statement should fall near the "uncertainty" end of Dimension 1 in order to be considered a candidate for implementation. Interestingly, all the "inconclusive" statements cluster near the middle of the effectiveness continuum, while Statements 7 and 4 ("Excedrin is not a more effective pain reliever than aspirin" and "Excedrin is a less effective pain reliever than aspirin") form the "ineffective" end and Statements 5 and 10 ("Excedrin is a more effective pain reliever than aspirin" and "it has been consistently and conclusively proven that Excedrin is a more effective pain reliever than aspirin") form the effective end. When the data are mapped into three dimensions, these two dimensions are clearly recovered but the new dimension is, in the opinion of the authors, not interpretable.





Additional evidence for the two-dimensional solution is provided by the convergence of the solutions across replications. The unidimensional and two-dimensional solutions correlated almost perfectly across replications. The correlation between Replications 1 and 2 was r = .99 (p < .001) for the unidimensional solution. In the two-dimensional solution, dimension 1 from Replication 1 and dimension 1 from Replication 2 correlate r = 1.00 (p s .001). Dimension 2 from Replication 1 and dimension 2 from Replication 2 also correlate r = 1.00 (p < .001). Indeed, inspection of Figures 1 and 2 reveals that they are nearly identical graphic depictions. However, when the data were mapped into three dimensions, the third dimension did not converge nearly as highly across replications. The correlation for dimension 1 across replications was only r = 57 (p < .05).

In sum, based on the two Horan model test indices, percent of variance accounted for, interpretability, and convergence of the replicated solutions, it appears that a two-dimensional solution is most suitable for describing these data. These two dimensions have been labeled "certainty' and "favorableness." It also appears that the three FTC statements, along with three others developed by the authors to communicate the same essential meaning, did cluster together meaningfully in multi-dimensional space.


The objective of this investigation was to illustrate a procedure for probing where the problem might reside when a remedial advertising statement -- indeed, any advertising statement -- is miscomprehended. This procedure was illustrated using a set of three remedial advertising statements devised by the FTC for inclusion in the advertising for two popular branded products. The findings revealed that these test statements did indeed cluster together in multi-dimensional space and were meaningfully interpretable in terms of a priori expectations, much as the FTC would have liked . Had this finding not materialized, it would have suggested that the statements were not reflecting a single meaning structure and that some (or perhaps all) of these surface structures were inadequate for communicating the intended meaning. Under these circumstances, there would have been ample basis for believing that the miscomprehension found in the earlier investigation was due to problems occurring in the encoding link between intended meaning structure and surface structure, rather than in the subsequent decoding link between surface structure and extracted meaning structure. This, in turn, would have implied the need to revise the proposed statements. Further, while the statements in the present study did cluster meaningfully and as one would predict, one cannot automatically expect that all future remedial statements to do likewise. Thus, the present study illustrates an approach for the re-testing remedial statements and this technique may be useful in identifying problem statements in future cases.

As a by-product, the findings of the present investigation also yield insight on how to improve the remedial advertising statements developed by the FTC. While no test is available for calculating whether the distance between two points in multi-dimensional space is "significantly greater" than the distance between two other points, meaningful observations can be based on relative distances. In this regard, it can be noted that Statement 1 ("Excedrin has not been proven to be a more effective pain reliever than aspirin") lies farther toward the "ineffective" end of the favorableness dimension than either Statement 2 or 3. On this basis, it might be concluded that Statement 1 tends more than 2 or 3 toward having a "punitive" meaning. A punitive content would violate the STC's guidelines as to what remedial messages should, and should not accomplish. Interestingly, Statements 6 and 9 (which also convey the meaning desired by the FTC) fell closest of all to the neutral point of the favorable-unfavorable continuum, while still clustering among the "inconclusive" statements on the certainty dimension. Hence, either of the statements developed by the authors for use in the present study would seem to be preferable to the three devised by the FTC

The fact that remedial advertising statements could be developed and mandated without considering whether these messages would be adequately understood (ant, if not, why not and what could be done about it) suggests that there may be other important issues which remain unrecognized in the development and implementation of rem dial advertising. That is, the comprehension process is but one component of a larger system which needs to be applied to the development, assessment, and implementation of remedial advertising statements. One perspective as to the major components and interrelationships of such a system has recently been developed (Hoyer, Jacoby, and Nelson, 1982). Hopefully such broader perspectives will assist public policy makers in recognizing that effective remedial advertising requires more than simply developing statements which one thinks will do the job.


Engel, J.F., Blackwell, R.D. and Kollat, D.T., (1978), Consumer Behavior (3rd ed.), New York: Dryden Press.

Harris R.J. and Monaco, G.E. (1978), "The Psychology of Pragmatic Implication: Information Processing Between the Lines," Journal of Experimental Psychology: General, 197, 1-22.

Horan, C.B. (1969), "Multidimensional Scaling: Combining Observations When Individuals Have Different Perceptual Structures," Psychometrika, 34, 139-165.

Hoyer, W.D., Jacoby, J., and Nelson, M.C. (1982) "A Normative model for the Development, Evaluation, and Implementation of Remedial Advertising Statements," In D.R. Corrigan, F.B. Kraft, and R.H. Ross (eds.), 1982 Proceedings: Southwestern Marketing Association, 307-311.

Jacoby, J., Hoyer W.D., and Sheluga, D.A. (138(?) The Miscomprehension of Televised Communication, New York: American Association of Advertising Agencies.

Jacoby, J., Nelson, M.C., and Hoyer, W.D. (1982) "Corrective Advertising and Affirmative Disclosure Statements: Their Potential for Confusing and Misleading the Consumer," Journal of Marketing, 46(1), 61-72.

Lavidge, R.J. and Steiner, G.A. (1961) "A Model for Predictive Measurement of Advertising Effectiveness," Journal of Marketing, 25, 59-62.

Lindsay, P.E. and Norman, D.A. (1977) Human Information Processing: An Introduction to Psychology. New York: Academic Press.

Lingoes, J.C. (1971) "Some Boundary Conditions for a Monotone Analysis of Symmetric Matrices," Psychometrika, 36, 195-203.

McGuire, W.J. (1976), "Some Internal Psychological Factors in Influencing Consumer Choice," Journal of Consumer Re-search, 2, 302- 319.

Schonemann, P.H., James, W.L., and Carter, F.S. (1978), "COSPA, Common Space Analysis -- A Program for Fitting and Testing Horan's model Subjective Metrics Motel," Journal of Marketing Research, 15, 268- 270.

Schonemann, P.H., James, W.L., and Carter, F.S. (1979), "Statistical Inference in Multidimensional Scaling: A Method for Fitting and Testing Horan's Model," In J.C. Lingoes, E.E. Roskam, and I. Borg (eds.), Geometric Representations of Relational Data, Ann Arbor: Mathesis Press, 791-826.

Spiro, R.J., Bruce, B.C. and Brewer, W.F. (eds.) (1980) Theoretical Issues in Reading Comprehension. Hillsdale, Sew- Jersey: Lawrence Erlbaum Associates.