H i ‘ 1 A i ’ i 7—- ‘,,! W in W THE EE "E 833 0? W MiE "3 “MW TEEN On: A1E¢E2EANEJ SIEP. :E 7”.» H8 EBA. ATTHMD AME: ”EH ’RE‘: 1HC'N EMU Thesis for the Degree. 01‘ M. A. iiiiftlviifii‘ii‘i S‘i'A'f'E Uiil‘v’ER‘Si‘i‘Y ED‘A’ARE} LEE RAZi Nsirii’ .3 Q gr“: I ‘. II .1 ‘J . a llllllllllllfllllllllilllll 3 1293 limlimunlmlulyzwl E L IBRAR Y .‘iia‘.§:iqan Stat University ..__ :3 ABSTRACT THE EFFECTS OF MESSAGE QUALITY DETERIORATION ON MESSAGE AND SOURCE EVALUATION, ATTITUDE, AND INFORMATION GAIN by Edward Lee Razinsky The purpose of this study was to investigate the effects of deterioration in message quality on evaluations of the message and its source, on learning, and on persuasion. Message quality was defined as following or violating normative rules of Spelling, grammar and punctuation. Message quality deterioration was defined as increasing numbers of violations of the accepted rules. Dependent measures were: 1) Evaluation of Message: logic, information quality, clarity, overall quality, and evaluation based on style; 2) Evaluation of the message source: competence and trustworthiness; 3) Information gain; and, u) Persuasive impact of the message. Four hypotheses were tested: As frequency of incorrect uses of the language code increases, (1) other aSpects of the message are evaluated less favorably; (2) perceptions of the message source become more derogatory; (3) amount of information learned decreases; and, (u) attitude change decreases. Message quality was manipulated by systematically introducing increasing numbers of errors into a basic, SOC-word message. Four message versions resulted: (l) No-Brror version-~the original message; (2) Moderate-Error version--with a total of 12 errors: 5 Spelling errors, Edward Lee Razinsky u punctuation and 3 grammatical; (3) High-Error version--with a total of 2a errors: 10, 8, and 6 in the previous categories; and, (u) Extreme- Error versions--with a total of #8 errors: 20, 16, and 12. Two experiments were performed, with 66 subjects in the first and 55 in the second; all 83 were undergraduates at Michigan State University. All subjects were asked to read one version of the experimental message and evaluate the message and its source. A separate test booklet tapped attitude toward the topic and also provided a lO-item information quiz. In Experiment II, subjects were asked to find and mark errors perceived. These results were reanalyzed: based on the frequency of errors actually perceived, the pool of subjects was divided into quartiles. Data from EXperiment I lent only partial support for the hypotheses. Results of Experiment II and the Reanalysis indicated that message and source evaluation and information gain tend to decrease with increasing perception of errors. There were no differences in attitude toward the topic on the basis of message errors. All four message versions persuaded equally well. THE EFFECTS OF MESSAGE QUALITY DETERIORATION ON MESSAGE AND SOURCE EVALUATION, ATTITUDE, AND INFORMATION GAIN By Edward Lee Razinsky A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of MASTER OF ARTS Department of Communication 1967 Accepted by the faculty of the Department of Communication, College of Communication Arts, Michigan State University, in partial fulfillment of the requirements 5% Director ofvThesis for the Master of Arts degree. / « , / [/- Committee: 29//<;,:;?:;;4éééi/¢fii// (L ACKNOWLEDGMENTS Primary thanks must go to my committee members whose patience and skill helped me to avoid many pitfalls, extricate myself from those not avoided, and learn from the whole eXperience: Dr. Bradley Greenberg, Chairman; Dr. Verling Troldahl; and Dr. Hideya Kumata. It might, but should not, go without saying that the orientation, training and philosophy of the Department of Communication provides impetus for any research undertaken by its students. We will be eternally thankful for exposure to this way of knowing the world. A statement by a former employer must also be acknowledged here: "You've been studying communications for over a year now. What have you learned that will help me publish better magazines?" While solving practical, magaEine-publishing problems is not an overall goal for me, I would hOpe to reserve some generalizability for his problem also. Or, in other words, I hOpe he doesn't find my work completely irrelevant to the practical business of mass communications. ii Table of Contents CHAPTER Page I Q RATIONAI—‘B O O O O O O O I O O O O O O C O O O O C O O O 1 Background and Prior Research Hypotheses II 0 METHOD 0 O O O O O O O O O O O I O O O O O O O O O O 0 10 Independent Variables Dependent Variables Subjects Message Design Procedure III 0 RESULTS 0 O O O O O I O O O O O O O O O O O O O O O O O 16 Overall Analysis, Perceptions of Message, Perceptions of Source, Attitude Toward Topic, Information Gain Initial Experiment I Experiment II Reanalysis Hypotheses Supported Iv. DISCUSSION 0 O O O C O O O O C O O O O O O O O O O O O 25 Conclusions Factors Influencing the Findings Implications BIBLIOGRAPHY O O O C O O O O O O O O O O O O O O O O O O O O O O O 3” iii List of Tables TABLE 1. Possible Evaluation of Source, Assertion and Concept After Reading the Experimental Message 0 O O O C I C O O O O O O O I O O O O O O 2. Initial Experiment: Means of Message Quality Effects 0 O O O O O O O I O O O O O O 0 O 3. Experiment II: Means of Message Quality Effects. u. Reanalysis of Experiment II--Subjects Divided into Quartiles According to Errors Perceived: Means of Message Quality Effects . . . . . . . . 5. Comparison of Errors Introduced and Errors Actually Perceived in the Experiment II; 'Errors Perceived" Represents the Mean Values for Quartiles . . . . . . . . . . . . . . . . . . 6. Errors Introduced Compared to Errors Actually Perceived in Experiment II . . . . . . . . . . . iv List of Appendices Appendix: Test Booklet Page Title Experiment I: Cover Sheet . . . . . . . . . . . . . . . . . . . . 36 Experiment II: Cover Sheet . . . . . . . . . . . . . . . . . . . . 37 All Administrations: No Error-Message . . . . . . . . . . . . . . 38 All Administrations: Low-Error Message . . . . . . . . . . . . . HO All Administrations: Moderate-Error Message . . . . . . . . . . . M2 All Administrations: Extreme Error Message . . . . . . . . . . . nu Experiment I: Message Perception Scalesl . . . . . . . . . . . . #6 Experiment II: Message Perception Scales . . . . . . . . . . . . . #7 All Administrations: Attitude Statements . . . . . . . . . . . . #8 Experiment I: Information Quiz . . . . . . . . . . . . . . . . . 51 Experiment II: Information Quiz . . . . . . . . . . . . . . . . . S2 Source perceptions scales appear at the end of the attitude state- ments in the Initial Experiment, and at the end of the message perception items in Experiment II. CHAPTER I RATIONALE Background and Prior Research Over time and across cultures, language behavior has evolved with systematic structure. (9) An illogical sequence of words or unfamiliar word might be noticed and be found unacceptable even without formal knowledge of rules. Further, receivers might be expected to make the message producer aware of his linguistic indiscretions, or alter his evaluation of the message and its producer, or both. For present purposes, all forms of formal and informal, written and unwritten rules imposed on written message production will be considered as defining conditions for "message quality," i.e. an explicit or implicit "This is the right or wrong way to encode messages." Work has been done to demonstrate that "meaning" for correct structure is acquired even before formal training in the grammar of the native tongue. Roger Brown (u) asked pre-school youngsters (3-5 years old) to identify appropriate pictures with plural or singular nouns or verbs. The children correctly designated, for example, action with the "ing" form of a nonsense word ("The man is Nissing.") They knew that "The man has three Lattgr but has "Some Latt." It may be assumed that some concept of correct usage of the code is acquired in the process of normal language acquisition, with or without the Speaker's knowledge of formal rules. But it is suggested that the effects of deviatiom3from normative rules have not been extensively explored. One might conceive of a continuum of acceptability which includes deviations ranging from completely acceptable to totally unacCeptable-- gradually deteriorating message quality. At the more acceptable end, there may be an occasional extra Space between words or a 'widow'--a single word as the last line of a paragraph. Toward the middle might be the incorrect but understandable Pennsylvania Dutch form, e.g. "Throw Momma from the train, a kiss," or the sometimes ponderous translations of operating instructions which accompany German cameras. At more extreme positions on the continuum, there could be misspelled words with just one letter out of place, or words in which the "i" follows the "e" when it should precede; further, reach examples of inapprOpriate lexical items, misused parts of speech; and finally reach the point where the words so violate acceptable usage that the native Speaker would consider them a random sequence. Relative placement of deviations along the fictional continuum has been Speculative. But it does suggest that there are gradations of acceptability and perhaps differential fidelity accompanying successively more deviant forms. At some points, variations may not be perceived; at others, perceived but not influential in the receiver's evaluation of the message or its producers. Previous research dealing generally with message quality might be divided into that which treats language or codes either within, or apart from, the context of social interactions. Experiments dealing with structure of language per se, in the realm of the linguist or psycho— linguist, have worked with such variables as the recall of content words vs. functional words, psychological evidence for grammatical structure, semantic generalizations, and association strength. (13) The distinction can be made between this context-free research and that which treats language as a connecting link in a social dyad, i.e. as communication. The present study is in this second category. Research in this second area has treated message quality in various ways. For example, readability_has been manipulated as a function of word and sentence length and number of personal references. (13) Readability has been measured in terms of comprehension (recall) or number of paragraphs read by an audience (10). Sensationalism in written material has been indexed in terms of style of writing and topic of the message. (18) Structure, in terms of organization of arguments, has been manipulated to effect comprehension of controversial news Stories; ability to answer questions about content. (S) Redundancy_in written material has been measured by Taylor's "Cloze" procedure: the more redundant, the higher the Cloze score and the higher the presumed comprehension of the copy. (19) But in these areas, it is not clear which message treatment is the .norm, or most acceptable version, and which is the deviate, or unacceptable version. Copy with a Flesch Reading Ease score of 20 might be completely unacceptable for a popular digest, but eSpecially acceptable for a scientific journal. A news story judged as sensational might be considered the standard of excellence for a tabloid or the classic example of poor writing for staff members of the Christian Science Monitor. In short, the research described seems to treat variations g§_ structures and strategies as much as deviations from correct or acceptable form. They demonstrate dimensions of message quality, but not deviations from general norms of acceptable code usage. More direct tests of the effects of message deterioration were made recently by Sencer (written messages) and Miller and Hewgill (oral presentations). Sencer (15) introduced increasing numbers of grammatical errors into a lOOO-word message and measured subjects' evaluation of the quality of the writing, attitude toward the message, comprehension, and interaction between increasing errors and verbal aptitude. Sencer found that as number of grammatical errors increased, in- creasing proportions of subjects indicated that the message was "badly written," attitudes toward the message (attitude toward the topic was not a variable) became less favorable, and subjects with higher verbal aptitude were more cognizant of increasing error frequency. He found no difference in comprehension (as measured by "Cloze" procedure) among message treatments. Miller and Hewgill (12) manipulated verbal nonfluencies: the "vocalized pause" (or "Uhhh"), and the "repetition" or stammer, where a first syllable was uttered and then repeated before the message continued. They established five conditions for each type of nonfluencey: 0, 25, 50, 75, and 100 nonfluencies-in the identical lOOO-word, seven minute taped Speech. With increasing nonfluencies, of each type, the audience made less favorable evaluations of the Speaker. The present study expanded on the previous research: it directly tested message quality deterioration, increased the number of error types and measured attitude change. The basic question proposed is this: What are the effects of following or violating the implicit or explicit rules of message quality on the effectiveness of a particular message? Hypotheses A In this study, the variable of message quality was conceptualized in terms of prescribed use of the code, and operationalized as increasing numbers of inaccuracies in these categories: spelling, grammar and punctuation. Effect was measured in terms of four dependent variables: 1) evaluation of the message; 2) evaluation of the source of the message; 3) the persuasive impact of the message; and, u) the amount of information .learned from the message. Pour hypotheses were tested: Hypothesis 1: As number of errors increases, related components of the message will be evaluated less favorably. Work with response generalization suggests that the perception of errors will structure reSponses to related message components, Specifically: logic of arguments, clarity of presentation, style of writing, information quality and overall evaluation. Impression formation research by Asch (1) indicated that the response to certain key words in a personal description, e.g. "hot-cold," generalized to many personality traits of an individual. For example, the "cold" individual was also perceived as being ungenerous, shrewd, etc. Greenberg (7) showed that attitude change induced for one concept will generalize to related concepts. And Sencer's (l6) experiment with grammatical errors indicated that as number of errors increased, subjects were more likely to evaluate the message generally as "badly written." Hypothesis 2: As the number of message inaccuracies increases, perceptions of the message source will become less favorable. Again, the findings in response generalization and impression formation suggest that the perception of errors in the message will alter the receivers' evaluation of the source, Specifically on the competence and trustworthiness aspects of credibility. In a more direct test of this hypothesis, Miller and Hewgill (12) found that as the number of nonfluencies increased, ratings of the source became less favorable. Dynamism ratings decreased as a function of in- creases in the repetition-type nonfluency only. As suggested by Osgood and Tannenbaum (In), both attitude toward source and attitude toward message topic are alterable when the two are linked in a communication situation. Thus, linking an unknown source to a poor message should yield less favorable source perceptions. Hypothesis 3: As message quality deteriorates, the amount of in- formation learned from the message will decrease. It is suggested that manipulation of language factors in violation of accepted practice is analogous to introducing "noise" into the communication system. Noise interferes with the acquisition of information. Therefore, as noise level increases, information gain decreases. (15) Hypothesis 4: Attitude change varies inversely with message quality deterioration, such that as number of message errors increases, attitude change will decrease. This hypothesis can be supported from a number of theoretical positions. The bulk of the research dealing with source credibility indicates that attitude change and credibility are positively related. (8) Since it was hypothesized that ratings of source credibility would become less favorable as message errors increased, it is reasonable to eXpect that persuasive impact would decrease as source credibility ratings de- creased. The principle of congruity also suggests support for the attitude change hypothesis offered here. (1”) The experimental situation can be described this way: A highly valued concept (toothbrushing) is attacked by a source who is less highly valued than the concept (source is a college freshman; receivers are college juniors or seniOrs). In terms of the congruity model, this is an incongruous situation. In his application of the congruity model to oral communication, Bettinghaus (3) has suggested that evaluation of the source can occur in stages: an initial evaluation before presentation and an evaluation based on presentation. So while the writer of the message might have been perceived as neutral just after an introduction, he would be rated negatively after being credited with an error—laden message. It would be this second evaluation which would serve as the context for source- assertion-concept cognitions, e.g. S+A~C+ in the No-Error Treatment just prior to readjustment toward congruity. Based on this inference, it is expected that increasing numbers of message errors will tend to decrease the source's ratings during the course of the message. This prediction is Hypothesis 2: less favorable credibility rating with increasing message quality deterioration. Given this situation, just after reading the message the receivers might have S-A-C structures Similar to those in Table 1. Sources ranging from neutral or slightly positive (in the No-Error treatment) to highly negative (in the Extreme-Error treatment) made negative assertions (constant in all message versions) about a highly positive concept (toothbrushing). Congruity theory suggests that the subject will attempt to reduce or eliminate his incongruity. With an initially neutral source and a fairly sound argument, the source might be slightly positive after the presentation, and attitude change would be expected. But as the source is, in effect, discredited by increasing message errors, incongruity could be more easily reduced by lowering the evaluation of the source and not changing attitude toward the concept. Table 1. Possible Evaluations of Source, Assertion and Concept After Reading the Experimental Message Extent of Predicted Treatments Source Assertion Concept Incongruity Attitude Chang: No-Error + - +++ Most Most Moderate-Error - - +++ Slight Minimal High-Error -- - +++ Minimal Slight Extreme-Error --- - +++ Congruous Little or None This study provides two tests of these hypotheses-~Experiment I, which allowed free perception of errors; and.Experiment II, which forced attention to errors in an attempt to test maximum message quality deterioration effects. CHAPTER II METHOD Independent Variables: Message quality was manipulated by feur systematic introductions of increasing numbers of errors into the basic message. The error types--Spelling, punctuation and grammarb-were selected arbitrarily since differential effect of error classes was not under investigation. However, it seemed reasonable that errors in these categories were representative of the class of errors of concern here. The errors were constructed to appear to have been made naturally. For example, the change of "constgnt" to "const_e_nt"; "tmdgsirable" to "undisirable". Punctuation errors took the form of deletion of necessary marks (period at the end of a sentence) or inclusion of superfluous marks (comma in the middle of a phrase). Grammatical errors included "...Studies indicate§_that... toothpastes and powders has been found..." Here is a sample passage from the Extreme-Error message. Punctuation and grammar errors are indicated by parentheses; spelling errors by under- lining: "Many pe0p1e brush their teeth more or less automaticaly_ after each meal without realizing that of late, medical reports have been calling this procedure into question. Recent medical and biological studies (indicates) that the beneficial effects of constant tooth brushing have been (exaggerating). Furthermore() it has(,) been demonstrated(,) that a number of bad (effect) can result from brushing teeth so often. Constant gum iritation can result in infection and even mouth cancer." 10 11 An attempt was made to have the errors appear as a violation of the message code stemming from ignorance. Simply adding random letters to a word to construct a miSSpelling might seem unbelievable. Or, the inversion of an "i" (and an "e" might be attributed to typographical inability rather than the source's violation of rules for acceptable use of the code. Overall, sentences were not restructured by the introduction of errors, nor were words deleted; all message treatments were of identical length and wording. No more than one error was introduced for a given word, i.e. no one word was both miSSpelled and grammatically incorrect. As far as possible, both number and type of error were introduced uniformly throughout the message. Four message treatments resulted: l) No-Error version--The unaltered message. 2) Moderate-Error version-«with a total of 12 errors: 5 Spelling errors, 4 punctuation and 3 grammatical. 3) High-Error version--with a total of 2a errors: 10, 8 and 6 in the previous categories. u) Extreme-Error version-with a total of ”8 errors: 20, 16, and 12; approximately one error per 10 words. This was the same error rate used by Sencer (17) in his high error version. Errors were introduced ordinally, i.e. the "Moderate-Error" message contained a basic set of 12 errors; the "High-Error" version was constructed by adding l2 additional errors to the basic set; the "Extreme-Error" treatment contained the 21+ errors from the less error-laden messages, plus 2!: new ones. l2 Dependent Variables: Pour dependent variables were examined in this study: 1) Generalization between message quality deterioration and related message aspects. Subjects rated the logic, amount and quality of in- formation, and clearness of the writing on 5-point scales from "not good at all" to "very good". Subjects also evaluated the overall message. Five message perception items were used: I. In general, how would you rate the logic of the arguments presented? 2. In general, how would you rate the (amount and) quality of information in the essay? 3. In general, how would you rate the clearness with which the essay is written? u. What would you give as a letter grade for the overall quality of the theme? 5. If the essay had no "simple" errors in it (for example, Spelling errors), and you were basing a grade only on the content and style of writing, what would you give as a letter grade for it? 2) Perceived credibility of source as a function of message quality was measured with two seven-point semantic differential scales. Subjects rated the source's competence and trustworthiness, two independent factors in source credibility. 3) Information gain was measured with a lO-item information quiz based on factual material in the essay. For example, "What does the essay say is the most undesirable effect of toothbrushing?" No attempt was made to construct "trick" questions. For the most part, the questions paraphrased the statement in the essay which included the correct answer. Subjects l3 1 received one point for each correct answer. a) Subjects' attitude toward toothbrushing was measured with four 15-point agree-disagree statements. The four toothbrushing items were included among eight other opinion statements relating to health practices: regular physical checkup, chest X-Ray, use of penicillin, etc. A score of 15 on each indicated maximum agreement with the opinion statement (all positively stated). Such a score (max = 60) would mean minimal agreement with the position advocated by the experimental message; that the message had no effect on initial attitudes. The four attitude statements were (Numbered as in test booklet): 1. Everyone should brush his teeth after every meal if at all possible. u. There are almost no disadvantages to regular and frequent toothbrushing. 7. Frequent toothbrushing is a very healthy practice. 10. The best way to prevent tooth decay is to brush one's teeth frequently. Subjects: In the initial experiment (February, 1966) subjects were 66 undergraduates in four sections of a junior-level Speech class at Michigan State University. For the Experiment II (May, 1966) subjects were 57 undergraduates in one lecture section of a general business communications class at Michigan State University. One question in the initial experiment ("What was the title of the essay?") was replaced with "What was the writer's definition of frequent toothbrushing?" in the Experiment II. It was found that the original item did not differentiate among subjects in various treatment groups; almost no subject could answer the item correctly. 1” Message: The basic message was a persuasive message deve10ped by McGuire in his studies of resistance to attitude change. It was a 500- word attack on the merits of toothbrushing. (11) The message attacks a "cultural truism"--a positively evaluated concept in the American culture which is seldom subject to criticism. For example, "...brushing teeth so frequently tends to push back the gums and expose the non-enameled parts... often cause our gums to bleed...can produce mouth cancer..." Desigg: An after-only design was used, with the No-Error treatment group serving as the control or baseline for effect of the unadulterated message. Since the differential effects of relative numbers of errors was of primary interest, pretest measures were not conducted. It was assumed that individual differences among subjects, including ability to perceive errors, initial attitude toward toothbrushing, and familiarity with the arguments presented, would be effectively controlled through random assignment of subjects to treatments. However, previous findings on evaluation of the concept and effectiveness of the message were used as additional comparison points. (11) Procedure: In both administrations, subjects were randomly assigned to treatments. The course instructors introduced the author as a staff member of the University Office of Institutional Research. The purpose of the study was given as "...an attempt to examine and hopefully improve upon the manner in which student themes, essays and term papers are handled by classroom instructors." Subsequent discussion with the subjects indicated that the guise was effective. 15‘ The experimenter distributed a booklet to each class member con- taining one of the four versions of the experimental message. Distribution was random in each of the course sections. The cover sheets for both the Experiment I and Experiment II described the same ploy. Subjects were told that many classes in the University were being asked to read and evaluate a variety of themes actually submitted in freshman courses, and that booklets contained themes mechanically reproduced in the form submitted. The fictional freshmen had been given the assignment of writing a MOO-600 word theme on any every- day topic, after doing some library research for background information. The themes had already been graded by instructors; this study was designed to evaluate the adequacy of current grading procedures. In the initial eXperiment, subjects were asked to read the theme carefully, and underline the main points made by the writer. In the second experiment, subjects were asked to indicate any errors that were noticed by circling misspelled words, putting an X over incorrect or missing punctuation and underlining errorsin grammar. The cover sheet was followed by the experimental message, the five message evaluation items and source credibility scales. When subjects had completed this portion, they were told that a second part of the study would be distributed and were asked not to refer back to Part I. Part II consisted of 12 opinion statements (including the four toothbrushing items); and Part III was the lO-item information quiz. After the subjects completed this section, the test booklets were collected, and the true purpose of the exPeriment was explained. CHAPTER III RESULTS Analyses of data will be presented in four sections: (1) Experiment I; (2) Experiment II; (3) Reanalysis; and, (u) Support for hypotheses across all analyses. The overall analysis of data was a Friedman two-way analysis of variance (17). This treated the nine dependent variables as replications, ranking the mean treatment scores for each variable, then analyzing the overall treatment differences in terms of their rank score sums. Single classification analysis of variance was computed for each .variable to test significance of difference, overall, among treatment groups. Experiment I Overall Analysis: The data (Table 2) indicated that across the set of dependent measures three of the nine F-values were significant. However, ranks of the means across treatment groups were in the predicted direction, yielding significant results on the Friedman analysis of variance (p(.02) . Perceptions of Message: Deterioration in message quality led to significantly different evaluations on logic of content, overall evaluation, and evaluation based on writing style (p<<}05). It Should be recalled that the content of the arguments was identical across treatments; only frequencies of errors were manipulated. 16 l7 Table 2. Experiment I: Means of Message Quality Effects Measures Treatments No 12 2a H8 Errors Errors Errors Errors 2* I. Perceptions of Message: a. LOgic of Content 3.7 3.8 3.3 3.2 .05 (2)** (l) (3) (A) b. Information 3.0 2.9 3.1 2.7 n.s. Quality (2) (3) (l) (u) c. Writing Clarity 3.8 3.3 3.1 2.8 n.s. (l) (2) (3) (A) d. Overall 3.3 3.H 3.3 2.u .05 Evaluation (2.5) (l) (2.5) (4) 3. Style Evaluation 3.” 3.6 3.5 2.8 .05 (3) (l) (2) (4) II. Perceptions of Source: a. Competence u.6 4.4 u.3 3.5 n.s. (l) (2) (3) (4) b. Trustworthiness 4.9 u.1 4.9 3.8 n.s. (1.5) (3) (1.5) (A) III. Attitude Toward Topic: 38.7 no.9 39.5 37.9 n.s. (2) (u) (3) (1) IV. Information Gain: 5.6 s.u 5.6 H.6 n.s. (1.5) (3) (1.5) (4) n: 16 16 16 18 Sum of Treatment Ranks: (16.5) (20) (20.5) (33) Predicted Ranks: 9 18 27 36 (Analysis of Variance of Ranks: X2R = lo,u, df = 3, p.<;o2) *Based on values of F; critical value of F 2.76 (.05), df = 60 **These parenthetical numbers represent the ranking of the means across the four treatments for each dependent measure. high (most favorable) to low (least favorable) for all variables except the Attitude Toward Topic measure. more the attitude corresponded to the direction of the message. The means are ranked from For that, the lower the score, the 18 All message perception items, including clarity of writing and in- formation quality, showed least favorable ratings in the Extreme-Error condition. Perceptions of Source: Both items in the source evaluation-- competence and trustworthiness--showed highest rankings for the No- Error treatment, with the Extreme-Error treatment ranked as fourth. Attitude Toward Topic: Results were contrary to predictions, both in terms of statistically significant differences and direction of ranks. Since McGuire got pre-attack mean attitude scores greater than 14 on a 15-point scale using these cultural truisms (11) the four versions were apparently equally effective in attacking the topic of the message. Information gain: Differences in information scores across the four message treatments were non-significant, although they were in the predicted direction. Experiment II In this experiment, subjects were asked to mark all errors per- ceived. This was mainly intended as a device to force attention to errors, thereby maximizing their effects, if any. This also allowed a measure of errors actually_operating, as distinct from errors experimentally in- troduced. Overall Analysis: Results of the Friedman AOV (Table 3) indicated even stronger trends in the same direction (p<<}001) than those found in Experiment I. Predicted rankings would have the No-Error treatment with a rank of l on each dependent measure, and the Extreme-Error treatment with 19 (Analysis of Variance of Ranks: 2 XR 21.0, df = 3, p <:.001) Table 3. EXperiment II: Means of Message Quality Effects Measures Treatments No 12 24 48 Errors Errors Errors Errors Hp* I. Perceptions of Message: a. Logic of Content 3.9 3.2 3.0 2.9 n.s. (l)** (2) (3) (4) b. Information 3.8 3.3 2.6 2.3 .05 Quality (1) (2) (3) (4) c. Writing Clarity 4.5 3.1 2.8 2.4 .001 (l) (2) (3) (4) d. Overall 4.1 3.0 2.6 2.2 .001 Evaluation (1) (2) (3) (4) e. Style Evaluation 4.3 3.3 2.9 3.1 .01 = (l) (2) (4) (3) II. Perceptions of Source: a. Competence 5.5 4.1 3.7 3.3 .01 (l) (2) (3) (4) b. Trustworthiness 5.1 4.3 4.1 4.1 n.s. (1) (2) (3.5) (3.5) III. Attitude Toward TOpic: 38.0 42.8 37.7 44.4 n.s. (2) (3) (l) (4) IV. Information Gain: 6.3 5.2 5.3 4.6 n.s. (l) (3) (2) (4) n: 8 l4 l6 17 Sum of Treatment Ranks (10) (20) (25.5) (34.5) Predicted Ranks 9 18 27 37 *Based on values of F; critical value of F = 2.84 (.05), 4.31 (.01), 6.60 (.001); df = 40 **These parenthetical numbers represent the ranking of the means across the four treatments for each dependent measure. high (most favorable) to low (least favorable) for all variables except the Attitude Toward Topic measure. more the attitude correSponded to the direction of the message. The means are ranked from For that, the lower the score, the 20 a rank of 4, yielding a sum of ranks of 9, 18, 27, and 36 across the four error treatments. These predictions compared favorably with those obtained: 10, 20, 25.5, and 34.5 Perceptions of Message: F-values for four of the five measures were significant. The item which asked for evaluation of "amount and c1uality" of information in Experiment I, dealt with quality alone here; the F was significant (p<<:05). F values were Significant for three other measures: Writing Clarity (p<<;001); Overall Evaluation (p (3001); and, Style Evaluation (p<<:01). Evaluations based on perceived logic were not significantly different across treatment groups. A perfectly predicted rank ordering was observed across message treatments in all items except evaluation based on style. Perceptions of Source: The competence item significantly differentiated among treatment groups on ranking of means as predicted, F-values (p< B C D F 5. If the essay had no 'simple' errors in it (for example, Spelling errors), and you were basing a grade only on the content and style of writing, what would you give as a letter grade for it? A B C D F 6. Now, please tell us what you think of the writer of this theme? How competent do you think the writer is? competent : : : : : : incompetent very very How trustworthy or responsible do you think the writer is? trustworthy : : : : : : untrustworthy very very 48 ALL ADMINISTRATIONS: ATTITUDE STATEMENTS Part II. Sometimes, instructors and teachers are concerned over the possibility that their own attitudes may affect the manner in which they grade student papers. To determine whether grading is affected by such opinions, you are asked to indicate your own personal opinions toward the subject matter of the theme you have just read. In this section, please indicate your own opinions, at this moment, toward each item, and not necessarily what the essay migh¥_have said. EaCh statement below will be accompanied by a scale. You are asked to indicate the extent of your agreement or disagreement with the statement. You may find yourself agreeing strongly with some of the statements, and disagreeing just as strongly with others of the statements. Some of the statements deal directly with the subject matter of the theme. Others deal with related topics. The following is an example of how the scale is to be used: At the present time in the U.S., life expectancy is greater for people living in rural areas than for those in urban areas. Definitely Mildly Neutral Mildly Definitely Disagree Disagree Agree Agree Notice that the scale has five categories ("Definitely Disagree," etc.) and that each of these categories has three divisions. You are asked to indicate your agreement with the statement by marking an "X" in whichever of these divisions best shows your opinion. For example, suppose that you agree completely with the above statement, without reservation. In this case, you would put your "X“ over in the place to the extreme right, as Shown below: Definitely Mildly . Neutral Mildly Definitely Disagree Disagree Agree Agree If you strongly agree with the statement, with some slight reservations, you may want to put the "X” down toward the left end of the "Definitely Agree" category, as follows: Definitely Mildly Neutral Mildly Definitely Disagree Disagree Agree Agree If, on the other hand, you completely disagree with the statement, with- out reservation, you should mark an "X" in the Space at the extreme left, as follows: Definitely Mildly Neutral Mildly Definitely Disagree Disagree Agree Agree We want your personal opinion on each statement, as you think at this moment. Please react to the statements on the next page. 49 1. Everyone should brush his teeth after every meal if at all possible. Definitely Mildly Neutral Mildly Definitely Disagree Disagree Agree Agree 2. Everyone should see his doctor at least once a year. Definitely Mildly Neutral Mildly Definitely Agree Agree Disagree Disagree 3. If everyone were to get a complete physical checkup once every year more good than harm would result. Definitely Mildly Neutral Mildly Definitely Agree Agree Disagree Disagree 4. There are almost no disadvantages to regular and frequent tooth-brushing. Definitely Mildly Neutral Mildly Definitely Agree Agree Disagree Disagree 5. Even though one may not have any reason for suspecting TB, it is a good idea to have regular chest X-ray examinations. Definite 1y Mildly Neutral Mildly Definitely Agree Agree Disagree Disagree 6. One of the greatest single advances in the history of medical science was the discovery of penicillin. Definitely Mildly Neutral Mildly Definitely Disagree Disagree Agree Agree 7. Frequent tooth-brushing is a very healthy practice. Definitely Mildly Neutral Mildly Defifiitely Agree Agree Disagree Disagree 8. People Should be urged to have a complete medical checkup as often as once a year. Definitely Mildly Neutral Mildly Definitely Agree Agree Disagree Disagree 52 EXPERIMENT 2: INFORMATION QUIZ Part III. Finally, we would like to see how well you, as the reader, can remember the points made in the theme. We are not interested in whether you agreed or disagreed with the theme, but can you recall what the writer had to say. Please answer the following questions as well as you can. Do not look back to the essay for the answers. 1. The essay suggests two possible bad effects of frequent tooth-brushing. What are they? 2. Instead of frequent tooth-brushing, what does the essay suggest as an alternative? 3. What does the essay say is the most undesirable effect of tooth-brushing? 4. Why do tooth-pastes and tooth-powders contain harsh abrasives, according to this essay? 5. What two things does frequent tooth-brushing do to the enamel that opens the way fer decay bacteria? 6. What was the writer’s definition of 'frequent tooth-brushing?' 7. Here are two true-false items. Circle the correct answer. True False a. The writer refers to medical studies to support his viewpoint. True False b. Tooth enamel covers the exPosed and unexposed portions of the tooth. If you have any general reactions to thiS kind of study, please use this Space: 50 9. All things considered, getting an annual chest X-ray for detecting TB is a very wise practice. Definitely Mildly Neutral Mildly Definitely Disagree Disagree Agree Agree 10. The best way to prevent tooth decay is to brush one's teeth frequently. Definitely Mildly Neutral Mildly Definitely Agree Agree Disagree Disagree ll. Penicillin may well be considered a "wonder drug," because of the many advantages to its use. Definitely Mildly Neutral Mildly Definitely Disagree Disagree Agree Agree 12. We should all have medical checkups, not only when we feel ill, but also at frequent intervals even when we feel well. Definitely Mildly Neutral Mildly Definitely Disagree Disagree Agree Agree Now, please tell us what you think of the writer of this theme. How competent do you think he(she) is? competent : : : : : : incompetent very very How trustworthy do you think he is? trustworthy : : : : : : untrustworthy very very HICHIGRN STRTE UNIV. LIBRRRIES 31293101470783