Two closely related studies were recently published in the "Special Reports" section of SM: the first by McCroskey and Dunham¹ was an admittedly speculative and a posteriori analysis of what appeared to be confounded data; the second by Holtzman² attempted to gain confirmation for the McCroskey-Dunham speculations by means of a tighter, a priori design. The purpose of this paper is to consider some assumptions therein and to suggest a solution to one of the problems raised by these researchers.

McCroskey and Dunham were interested in two basic questions: (1) What is the initial ethos level of an unseen, unknown, tape-recorded speaker in an experimental setting? (2) What effect does the experimenter's ethos have on an unseen, unknown, tape-recorded speaker's initial ethos? (The authors are unclear as to whether introductions to the speaker were used; an introduced speaker can be unseen, unknown, and unheard.) On the first question the researchers decided that an unseen, unknown, tape-recorded speaker in an experimental setting is not a neutral source. Why? Because data gathered could not be predicted from homeostatic models of attitude change if the unseen, unknown, tape-recorded speaker were assumed to have a neutral level of ethos.

There seem to be several flaws in this line of reasoning. First, I know of no study in which a homeostatic model has perfectly predicted attitude change. Berlo and Gulley³ claimed to have positive support for the congruity model by predicting accurately the attitude change for only two of three subjects which, as Anderson and Clevenger observed, is only "somewhat better than chance alone."⁴ At this point we cannot be sure whether the fault is with the homeostatic model or the ethos model.

Second, the authors failed to measure their unseen, unknown, tape-recorded speaker's initial ethos level. Why did they fail to do so? Because, they said, "it seemed doubtful that subjects could, or would, give meaningful responses to a scale concerning their opinions of an unseen, unknown, as yet unheard speaker."⁵ There is, however, direct evidence which contradicts this assumption: in at least one study researchers measured the effect of the neutral introduction of an unseen, anonymous, tape-recorded speaker; the measurements were taken immediately after the introduction. The mean response of the subjects hearing the neutral introduction was precisely on the middle of a seven-point scale: the point defined verbally as a neutral source. In fact, through pilot tests, the researchers found that an unseen and anonymous speaker was the most likely source to be perceived by subjects as neutral.⁶ In short, readers of the McCroskey-Dunham report should not infer that such an experimental speaker is inevitably something other than a neutral-ethos source.

On the second question McCroskey and Dunham concluded that an experimenter's ethos has an effect on the ethos of an unseen, unknown, tape-recorded speaker. Why? Because two groups of subjects (admittedly drawn from different populations) did not respond identically to measurements of the experimental speaker's ethos. These differences were ascribed to differential effects of experimenter's ethos. Because of the acknowledged weakness of this design, Holtzman was encouraged to confirm the McCroskey-Dunham hypothesis. In one treatment "casually dressed graduate students" were the sponsors; in the other the subjects' speech instructors were the sponsors. Differences found were said to confirm the "sponsorship effect," i.e., that different levels of experimenters' ethos can confound data gathered on other variables.

It is finding which suggests the theoretical reasons for doubting the universality of the first of the McCroskey-Dunham speculations. For example, in any case where an experimenter (or instructor) introduces a variable by verbal directions or information, his ethos becomes an additional or possibly confounding variable. If his ethos is "high" (positive or favorable), the instructions will be believed; hence, the treatment will "take" as intended. A positive "sponsorship effect" would thus be desirable, if not mandatory, in many experimental settings.

It is difficult to conceive of a case in which one would desire a negative "sponsorship effect." As a corollary, it is difficult to conceive of a case in which one would, for purposes of acquiring pedagogical data or data on formal situations, desire "conditions, tests and stimulus materials . . . presented . . . in classrooms other than speech classrooms, by casually dressed graduate students and without explanations"--as was the case in the Holtzman study.

What is the solution to the problem of negative or other unwanted "sponsorship effects"? McCroskey and Dunham suggest that the level of ethos be measured in any study of persuasive communication.⁷ Holtzman has a similar suggestion: "All experimental designs should account for ethos effects (including effects of perceived sponsorship."⁸ McCroskey and Dunham assume that in circumstances such as their experiments "only indirect estimates can be made of the subjects' initial image of the speaker."⁹ We can infer from Holtzman's a priori design that he agrees that such measurements should be indirect estimates coming after exposure to the speech.

The dangers in this procedure should be apparent. First, we have already seen the possibility for error in making estimates from homeostatic models. Second we know that ethos is a dynamic rather than static variable. Frequently the speaker's ethos at the conclusion of a speech is significantly different from what it was at the beginning. (In fact, the homeostatic models embraced by McCroskey and Dunham would posit such changes in most speaking situations.) Finally, it can be shown that a speaker's ethos changes from moment to moment--thus making indirect estimates of initial ethos levels a risky enterprise. Therefore, the solution to the problem is not to be found in the McCroskey-Dunham and Holtzman reports.

One simple and workable solution would be to measure directly the initial ethos of the speaker--before he speaks. this can and has been done. This can and has been done. This procedure yields not only a direct measurement of the initial ethos level but, in addition, indicates whether the verbal introductions and manipulations of ethos were perceived by the subjects as intended by the experimenter--a direct measurement of the effects of the experimenter's ethos.

It is conceivable that a modification of this technique could clarify some of the ambiguity McCroskey found in reviewing the experimental studies of the importance of evidence in persuasion. It is not possible that a speaker can affect this own ethos by quoting "authoritative testimony"? The quoted authority's ethos would then become a confounding variable. I am not confident that all researchers in the past have defined "authoritative" from the point of view of the subjects', as opposed to the experimenter's, point of view. College sophomores may at times have definitions of authoritative which are inconsistent with those of each other and/or those of the experimenter. Therefore, in such studies would it not be wise to make a direct measurement of this "secondary" or "absorbed" ethos?

Perhaps the only disadvantage to this procedure would be a possible interaction effect between the initial ethos measurement and the experimental speech. This, however, is a matter for research.

1. James C. McCroskey and Robert E. Dunham, "Ethos: A Confounding Element in Communication Research," SM, XXXIII (November, 1966), 456-463.

2. Paul D. Holtzman, "Confirmation of Ethos as a Confounding Element in Communication Research," ibid, XXXIII (November, 1966), 464-466.

3. David Berlo and Halbert Gulley, "Some Determinants of the Effect of Oral Communication in Producing Attitude Change and Learning," ibid., XXIV (March, 1957), 10-20.

4. Kenneth Andersen and Theodore Clevenger, Jr., "A Summary of Experimental Research in Ethos," ibid., XXX (June, 1963), 68.

5. McCroskey and Dunham, p. 457.

6. Phillip K. Tompkins and Larry A. Samovar, "An Experimental Study of the Effects of Credibility on the Comprehension of Content," SM, XXXI (June, 1964), 120-123.

7. McCroskey and Dunham, p. 463.

8. Holtzman, p. 466.

9. McCroskey and Dunham, p. 457.

"The solution to the problem," Professor Tompkins writes, "is not to be found in the McCroskey-Dunham and Holtzman reports." We agree. We offered no solution--only several sets of evidence that there is a problem in persuasion research: that of accounting "for ethos effects (including effects of perceived sponsorship)." Further, we agree that Tompkins and Samovar seem to have admirably accounted for these effects in their cited study.

Wherein are the issues joined? Professor Tompkins takes exception to the line of reasoning offered by McCroskey and Dunham to arrive at the conclusion that an "unseen, unknown, as yet unheard speaker" does not necessarily have "neutral" ethos. The reasoning model may be subject to the criticisms raised. This is a needless argument, however, for it is self-evident that such a speaker has no ethos until he is somehow introduced--neither "neutral" nor "high" nor "low." Ethos is a variable derived by a listener from what is "known" about a speaker. Only after the experimental subject has learned something about a speaker does that speaker have ethos at variable levels along several dimensions.

This statement concerning the absence of ethos is not contradicted by the evidence in the Tompkins-Samovar report in which neutral ethos was engineered. There, initial ethos was determined (accounted for) by a direct measurement of the effects of introductions interacting with the ethos of the introducer (apparent sponsor). We judge from Professor Tompkin's comments that "a positive 'sponsorship effect'" was (and should be) assumed. Thus we are told, it seems, that the ethos effects of an introduction should not be assumed but measured. So far, we agree. But we seem also to be told that the ethos effects of the introducer need be merely assumed. If that assumption be wrong along any dimension, we contend an uncontrolled contaminant can produce confusing--if not invalid--experimental results.¹ This is our only important caveat.

We do not wish to be understood as suggesting, as solution, "that such measurements should be indirect estimates coming after exposure to the speech." On the contrary, our initial presentations grew out of McCroskey's criticism of the great bulk of studies in which ethos-effects of introductions were estimated or assumed or ignored, especially when ethos was not a manipulated variable.²

That the ethos of an experimenter "rubs off" on the source of an experimental message is well established in the literature on the social psychology of psychological experiments.³ That is the problem in research on persuasion as well. We think that this "ethos-by-association" was suggested by the McCroskey-Dunham post hoc analyses and demonstrated in the experiments by Holtzman's students. We therefore continue to urge, not that experiments emulate our research designs for exposing the problem, but that "to enhance the probability of contribution to a unified theory of persuasion, all experimental designs should account for ethos-effects (including effects of perceived sponsorship)."

Professor Tompkins' critique does not mention certain other findings in our reports which are also crucial for the researcher. These show that ethos tends to interact with other variables in persuasion. Thus we suggest that for dependable results those who study other variables in persuasion must also gather data on initial ethos.

We recommend, where research designs use recorded messages in studying persuasion, that tested introductions of sources be used to establish initial ethos--as Tompkins suggests. However, we would add the further precaution that the ethos-effects of the introductions be pretested, not with the subjects in the experiment proper, but with comparable subjects drawn from the same population. This would avoid the ethos sensitization to which Tompkins refers, which might occur if measurements of ethos are taken between introduction and recorded message. (It may be tautological to warn that the "apparent sponsor" must be standardized in the comparative procedures.) Data on ethos should, then, be reported and taken into account in drawing conclusions about the effects of other variables.

1. In one study the low credibilities of introduced, "highly skilled criminals" were assumed. When readers did not respond as expected, the experimenters were forced to conjecture that for that subject criminals had high credibilities on an authoritative dimension. See William J. McGuire and Susan Millman, "Anticipatory Belief Lowering Following Forewarning of a Persuasive Attack," Journal of Personality and Social Psychology, II (October, 1965), 475.

2. James C. McCroskey, "Experimental Studies of the Effects of Ethos and Evidence in Persuasive Communication," unpublished dissertation (The Pennsylvania State University, 1966). "[N]one of the researchers who has studied evidence reported any measurement of the audience's initial image of the communicator. Neither was any pretest measurement of the source's ethos reported. The unreported 'guestimates' were probably that ethos was unimportant because it was 'controlled' or that the ethos of the communicators was 'neutral.' We must reject both of these possible assumptions."

3. In addition to the studies cited by McCroskey and Dunham (pp. 456-457) see Milton J. Rosenberg, "When Dissonance Fails: On Eliminating Evaluation Apprehension from Attitude Measurement." Journal of Personality and Social Psychology, I (January, 1965) 28-42.

Click Here To Go Back To PERIODICALS