Copy Testing, Thought Elicitation and Attitude Theory: Two's Company, Three's a Crowd

ABSTRACT - A number of thought elicitation coding schemes have been proposed by academics and advertising researchers to study the effects of advertising copy. Few of them appear to have as strong a conceptual foundation as they might. An alternative three-phase coding scheme is proposed that is based on several attitude theories and a study to test the coding scheme and several thought elicitation task instructions is described.

Citation:

Peter R. Dickson and Paul L. Sauer (1987) ,"Copy Testing, Thought Elicitation and Attitude Theory: Two's Company, Three's a Crowd", in NA - Advances in Consumer Research Volume 14, eds. Melanie Wallendorf and Paul Anderson, Provo, UT : Association for Consumer Research, Pages: 177-181.

Advances in Consumer Research Volume 14, 1987      Pages 177-181

COPY TESTING, THOUGHT ELICITATION AND ATTITUDE THEORY: TWO'S COMPANY, THREE'S A CROWD

Peter R. Dickson, The Ohio State University

Paul L. Sauer, State University of New York at Buffalo

ABSTRACT -

A number of thought elicitation coding schemes have been proposed by academics and advertising researchers to study the effects of advertising copy. Few of them appear to have as strong a conceptual foundation as they might. An alternative three-phase coding scheme is proposed that is based on several attitude theories and a study to test the coding scheme and several thought elicitation task instructions is described.

INTRODUCTION

Over the last decade interest in specific aspects of attitude theory peaked and now seems to be in decline amongst marketing academics and researchers. It has partly been replaced by an interest in memory processes, often involving the use of thought elicitation measurement techniques. Thought elicitation is a rather fancy name for asking the subjects to think out-loud or to retrospectively write down all the thoughts that came to mind or that are still coming to mind related to a topic or a stimulus object. They are sometimes also called verbal protocols or cognitive responses (Ericsson and Simon 1984).

Just exactly what thought elicitation measures seems to depend very much on the nature of the task and instruction set. It is a way (perhaps the only way) of attempting to capture a stream of mental consciousness whose structure and valences can be analyzed. Inferences about memory structure, cognitive schema, beliefs. attitudes and values can be made from such structural categorization.

Surprisingly, there appears to have been little effort directed at developing thought elicitation coding schemes based on basic attitude theory. The reasons for such a lack of effort are probably numerous. Attitude theorists may not believe that unstructured, introspective verbal responses can be used to measure attitudes. Most attitude models require the use of semantic differential scales that are analytically more tractable. The major issues have been choice of such scales to represent "means' and "ends" constructs and the combination of such operational measures for diagnostic and predictive purposes. It may also be simply a lack of temporal intersection between a measurement fad (thought elicitation) and a theoretical fad (attitude modelling) that used a very different approach to measuring attitudes. An intriguing related question is whether measurement method myopia has limited the development of attitude theory? If so, it is an example of how theory can be constrained by measurement rather than vice-versa (Anderson 1985).

Meanwhile, advertisers and their agencies continue to wrestle with the vexing question of how to better test advertising copy. Over $200 million is spent annually on copy testing. Lipstein and Neelankavil (1984) state that simple copy-point playback is the main measure used to test copy effectiveness. Reactions to the product and commercial are also often measured. Misunderstanding of messages is claimed to approach 20% in copy-point playback. This helps to explain advertisers' focus on simple message transmission tests of the effectiveness of rough executions and finished commercials. Applied advertising researchers do appear do implicitly accept that measurement should be based on some theory or model of persuasion. Like their academic colleagues, however, there is not a lot of published evidence that when they use thought elicitations, their coding of the responses is based on well accepted attitude or information processing theory. This paper explores whether more theory based thought elicitation coding schemes can be developed for advertising copy testing and how they might be tested against current coding schemes.

What if thought elicitation cannot be used to operationalize attitude constructs such as beliefs, values, feelings and intentions and their relationships? Does this reflect poorly on thought elicitations or the attitude theory? Most scholars would probably question the method rather than the theory, for an intuitive concern with thought elicitations is that they are a restricted or filtered dump of thoughts and mental schema. Consequently thought elicitations do not reflect natural structure. We discuss how this concern might be examined by manipulating the task instructions to encourage thought elicitations that are streams of consciousness (like James Joyce's writing style!) rather than a list of collected and composed, staccato statements.

First, we review conventional thought elicitation coding techniques. We then offer two alternative theory based coding techniques based on Fishbein and Ajzen (1975) and Rosenberg (1956) and discuss how involvement and centrality should affect verbal protocols. An experimental study that will compare three different thought elicitation coding schemes across four different task instruction sets is then described.

Thought Elicitation Coding Techniques

Thought elicitation coding schemes have categorized thoughts under various labels such as counterarguments, defensive reactions, favorable thoughts, neutral thoughts and recipient generated thoughts. For example, Wright (1973) utilized a simple coding scheme consisting primarily of counterargument, source argument and source derogation. He showed that this coding scheme could be used to measure the mediating effect of responses to ad intention on both formation of attitude toward a product and intention to purchase the product. He also found by using subjects' categorization of their thoughts as to whether they were advertisement-originated, recipient-modified or recipient generated (e.g., past product experience), that only recipient-modified or generated thoughts had a significant effect on attitude and intention.

Cacioppo, Harkins and Petty (1981) suggest three dimensions that they claim to be orthogonal and capture the essence of previous classifications of cognitive responses to persuasion. The first dimension is Polarity, the degree to which the statement is in favor or opposed to the advocacy. The second dimension is the Origin of the thought. It can be direct copy-point play back, a modification of or reaction to the advertising copy, or a statement pertinent to the issue (e.g., recall of a past product usage experience or evaluation) but not directly related to any parts of the advertising copy. A classification problem with the first two dimensions is how does a judge code the polarity of a thought which may have a valence but the thought is unrelated to the advocacy? The third dimension, Target, defines the focus of attention of the thought. It can be the product, or a particular feature or component of the ad such as a comment on the medium, the picture, text, spokesperson, layout, print style, or even a comment about the audience the ad appears to be directed toward. The authors suggest that other dimensions such as saliency and emotionality might also be used. The structure of the verbal protocols can also be studied (see Zajonc 1960) by measuring the number of distinct thought topics (differentiation), number of themes within a thought topic (complexity), the logical dependency between thoughts (unity), and the integrative dominance of a particular thought over other thoughts (organization).

Ostrom (1969) used a theory based coding scheme not to categorize individual thoughts but 168 verbal statements of "attitude toward church" whose source was not disclosed. Expert and naive judges classified the statements using the tripartite (cognitive, affective, conative) classification of attitudes (Allport 1935; Rosenberg and Hoveland 1960; McGuire 1969). Perceptual responses and beliefs are classified as cognitive. Feelings, sympathetic nervous responses and verbal statements of affect are classified as affective. Behavioral intention statements are classified as conative. It was found that highly reliable classification results can be obtained using even relatively untrained judges. If the tripartite classification scheme can be used to classify an assemblage of attitude statements (that are then used to create conventional attitude measurement scales) there does not appear to be any reason why the classification scheme cannot be used to reliably categorize a single individual's verbal protocols or elicited thoughts.

Other verbal protocol coding schemes have been used in consumer information processing research but they have been designed to study memory or cognitive processes rather than persuasion. Bettman and Park (1980) distinguished between brand and attribute processing as one dimension and between personal relevance and message source relevance on the other dimension. Their scheme is not particularly relevant as it was developed to study internal and external search patterns and linear compensatory choice processes rather than identifying key phrases in ad copy that trigger attitude formation and change. They do, however, provide encouraging evidence that much more can be extracted from verbal protocols than simple meaning or valence.

Alternative Theory Based Coding Schemes

In addition to a tripartite coding scheme, two other coding schemes can be derived from attitude theory. Fishbein and Ajzen (1975) conceptualize attitude as "the amount of affect for or against some object," A simple frequency account of pro and counter arguments has been found to be as satisfactory a measure of the persuasive impact of a message as any technique to date (Cacioppo, Harkins and Petty 1981). This may be because this measure, in large part, operationalizes the above definition of attitude. The question is whether it can be improved on using more of Fishbein and Ajzen's conceptual framework? Their constructs build on the tripartite model except they call cognition, belief; affect, attitude, and conation, intention. Using an information-processing approach they claim that attitude toward an object is based on salient beliefs about that object. Attitude is produced from a set of beliefs (expectancies and values), and intention in turn from attitude.

If elicitated thoughts tap into attitude structure and mental scheme, can they reveal not only salient beliefs, attitudes and intentions that result from exposure but also evidence of the connections between these constructs? For example, a stream of thoughts (consciousness) that first describes a belief about a brand attribute or performance, then an expression of liking for the brand because of the attribute and finally an expression of intent to buy the brand, captures the salient belief, attitude and intention. Just as importantly it demonstrates the relation between the belief, the attitude and the intention that is theorized in Figure 1.1 in Fishbein and Ajzen (1975, pg. 15). It should be noted that the subject is not being asked to report the effect that his or her beliefs had on attitude and intention. This is fortunate for Nisbett and Wilson (1977) have convincingly demonstrated that subjects are very poor at such introspective skills. But just as motives and behavior are ascertainable from analysis (Freud 1924) and beliefs and attitudes can be derived from both verbal protocols and prompted measurement scales, perhaps the connectedness between beliefs, feelings, attitudes and intentions can also be tracked using verbal protocols. The contiguity of thoughts, the grammar and the evident logical flow provide us with direct measures of the schematic relationship between these constructs.

An alternative means-end model of attitude (Rosenberg 1956; McGuire 1969) would suggest coding the thoughts that arise as the result of exposure to advertising copy into statements that describe brand features, and statements that describe the benefits that result from the use of the brand. When an elicited thought contains an advertised brand feature statement connected to a resulting benefit connected to an expression of affect toward such a benefit then this would appear to be; (1) clear support for the means-end (instrumentality-value) attitude theory, and (2) clear evidence that a particular part of the advertising copy has had a powerful persuasive impact.

Caveats and Compromise

Some may consider that we are far too ambitious in our attempt to extract so much information out of thought elicitation. This may be true. Certainly, such a measure will not capture sensory derived feelings that cannot be adequately articulated. How can anyone adequately describe, in words, how they feel when they;see the Lincoln Memorial in Washington, smell the sap from the Christmas tree, listen to Beethoven's Fifth, or sip a gin and tonic as they watch the sun set over a South Pacific ocean? Poets make a living out of almost succeeding, which only serves to prove how rare is such verbal skill. So there will be errors of omission and commission. The question is whether these errors will be any greater or perhaps even less than measuring attitude and testing different attitude models by using batteries of measurement scales and correlational analysis. None of the techniques seem to adequately capture the impact of "feeling" ads (Mitchell 1985).

At least the saliency and potency of beliefs can be directly observed in verbalized thoughts. A severe limitation of conventional attitude measurement is that it can prompt subjects to consider beliefs that otherwise never would have factored into their attitudes. It also relies heavily on correlational relationships to establish the potency of beliefs on attitude. There is also no escaping the fact that asking subjects to weight the importance or value of beliefs really stretches our introspective abilities. In short, we consider that if we can elicit information rich thoughts (and that is a big if) we will not be leaning on the data any more than conventional attitude researchers.

A further point that can be made in support of using thought elicitation to explore deep attitude structure is that prominent attitude researchers have already recommended using dimensions of such cognitive responses, other than content analysis, to examine attitude. They have suggested studying the range, hierarchy and interrelationships of different thoughts. Our suggested classification scheme seeks to identify logical chains of thought that flow from belief to affect or even belief to intentions. It is no less objective than the above structural dimensions of an individual's verbal protocol.

In a recent review paper Mitchell (1985) has raised some interesting attitude influence issues that should be able to be studied using measures of verbal response to advertising copy. He points out that thought elicitation cannot be expected to measure the effect of low involvement processing of an advertisement because, by definition, low involvement seems to mean minimal thought about the ad, during or after exposure. However, what thoughts that do occur should be directed at message elements and the source (spokesperson) rather than the message according to Gardner, Mitchell and Russo (1978) and Petty and Cacioppo (1981). This would further suggest that thoughts recorded after exposure to an ad might be appropriately categorized into beliefs and attitudes about the ad and beliefs and attitudes about the product (see Batra and Ray 1983). This would dichotomize Cacioppo, Harkins and Petty's (1983) "target" dimension. The prediction is that the increased impact of attitude to the ad on final product attitude should show up in the verbal protocols of the subjects in low involvement conditions.

Involvement, Centrality and Thought Elicitation

What should verbal protocols reveal when involvement is high and the copy has an impact on central attitudes. Self-schema theory (Markus 1980) tells us that individuals create and use knowledge structures built around self experience to facilitate processing of incoming relevant information and subsequent retrieval of such information during a recall task. Rogers, Kuiper and Kirker (1976) note that, "the self appears to function as a superordinate schema that is deeply involved in the processing interpretation and memory of personal information" (p. 677). They found that self-referencing induced superior incidental recall compared to purely semantic rating tasks. If, when exposed to an advertisement, an individual relates the message to himself or herself then the involvement of the person should be high and the impact should be greater on cental attitudes (Krugman 1964; Olson 1984).

McConville and Leavitt (1968) fount that meaningfulness, as measured by the number of "related references" made by subjects in response to questions asked after exposure to an advertisement, can be used to predict recall of message content. Advertisements which enable subjects to make more connections between products and product-claims facilitate recall. Leavitt, Waddell and Wells (1970) contended that Krugman's involvement measure assessed reaction to the ad rather than to the product in the commercial. They devised a Personal Product Response (PPR) protocol coding scheme to assess product connections. Rules of coding judged the use of the first person singular or plural and personal product experience thoughts to be personal product links. Batra and Ray (1983) also attempted to do this to a limited extent.

A View Response Profile (VPR) based on coding thoughts has found that personal relevance of the product message is a major factor in explaining the effective impact of ads (Schlinger 1979; Olson 1985; Olson, Schlinger and Young 1982; Olson 1984). Other similar techniques have also indicated the importance of personal relevance (Moldovan 1985). Relevance measures the connectiveness of the consumer to the use or benefit described by the ad. It may relate to past, current, or future use and involves associations contained within the self-schema of relevant experiences. The VPR was derived from the work of Wells (1964) and Leavitt (1970) who developed sets of at response scales that revealed the influence of advertising "meaningfulness" and "personal relevance". Such scales indicate the personal relevance of an at but they to not show which portions of an at are more relevant and telling in their impact. Establishing such connections is very important to copy writers who spent hours seeking and crafting phrases and images that will spark such associations. Verbal protocols are more likely to reveal the "hot" copy or features in an ad.

It would appear to be worthwhile to overlay a personal involvement coding scheme on top of any scheme that identifies beliefs, affect, intentions and their connectedness. Cacioppo, Harkins and Petty (1981) claim that the greater the topic importance the more likely it is that the polarity of the cognitive responses will predict attitude change, as conventionally measured. In other words, evidence of personal involvement in the elicited thoughts should power up the explanatory effect of the thoughts on attitude change. But this may only occur if the source of the thought is from the advertisement. Thoughts about previous personal purchase experience may reinforce rather than change previously held attitudes, making them more central and resistant to change. Consequently any personal involvement coding scheme should also be crossed with source categorization, to identify the involvement impact of specific copy claims separately from the subject's general usage involvement with the product category.

If the elicited thoughts: (1) do not reveal involvement or attitude centrality through the use of the first person in the thoughts and other cues, or (2) such personal connections are observed not to be related to attitude formation or change, then either the theories of involvement and attitude centrality are not supported or the measure does not capture the dominant features of mental schema that produce attitude.

A Test of Coding Schemes and Instruction Sets

In our proposed study we intend to examine whether we can: (1) elicit richer thoughts by changing the task instructions, and (2) evaluate a phased means-end coding scheme against the more conventional coding scheme proposed by Wright (1973) and the scheme indirectly proposed by Ostrom (1969). We will also block subjects by their self-reported involvement and prior product familiarity.

As previously mentioned, task instructions have been shown to affect cognitive responses. Instructions to "gather your thoughts" and then write them down on a sheet, which often has a prenumbered space for each thought, discourages spontaneity and continuity, and encourages filtering, construction and disjuncture between the thoughts. We propose four between subject instruction set manipulations to explore the reactivity of subjects and whether richer data (that provides much more information) can be elicited by the use of different thought elicitation instructions.

There will be two test ads in the initial study. The order of presentation will be controlled and factored into the analysis to study task learning effects. After exposure to each test advertisement, subjects will be asked to either think out-loud or write down their thoughts. Exposure time and response time will be limited to a maximum (but not minimum) number of minutes. In this study subjects will know before hand that the intent is to obtain their reactions to an advertisement. We make no attempt in this study to disguise the cask at the outset for several reasons. First, most of what we know about cognitive responses has come from persuasion studies that did not disguise the intent of the study. Subjects expected to have to react to the material. Second, if we cannot observe and record interesting theory based dimensions in the thoughts of subjects who expected to have their opinions measured then there is little hope of using the technique in situations with surprised and naive subjects. We accept that we are testing the method under favorable conditions. Third, much of the applied industry advertising copy testing is undertaken in situations where the subject is well aware of the purpose of the exercise.

Instruction Set Manipulations

The first instruction set condition will involve a minima' prompt to write down or verbalize all thoughts about the advertisement and the topic. The majority of our study will involve subjects writing down thoughts in a group session but we intend to also obtain the oral thoughts of subjects (that will be taped and then transcribed) in individual testing sessions. We expect that we will observe more connections when subjects think out loud compared to when they write down more "composed" thoughts. The second instruction set condition will involve a much more detailed prompt. It will start with the minimal prompt but then elaborate in the following manner:

"On the following pages please write town all your thoughts about the advertisement and the advertised product. This should include everything you thought about while reading or examining the advertisement, including information about the product, its brand name, its features, the claims made about the product, the benefits that would result or would not result if you or other people used the product, the benefits and disadvantages of the various features of the product, and your feelings about trying, buying and using the product. If certain features of the at or phrases in the ad impressed you, and now come to mint, then tell us also about them. Don't feel you have to compose your thoughts. Just write town your thoughts as they come to mint. Never mind about the spelling or the grammar for we are much more interested in you telling us all your thoughts and how they flow together."

Having read this instruction set several times, subjects will be asked to turn the page and will not be allowed to refer back to the instructions. This is to prevent diligent subjects from systematically providing all of the requested information, even in the correct order! The third condition will involve providing the subjects with the task instruction set to read several times before exposure to the test ads. In the final condition, subjects prior to the exposure to the test ads will also be given a demonstration ad and then three examples of especially composed verbal protocols. One will be typical of the usual verbal protocol obtained under the minimal instruction set. It will be rather terse and short. At the other extreme will be a long and rich verbal protocol that contains beliefs, feelings and intentions connected together using the first person. The third protocol will fall between the other two in terms of the richness, openness and flow of the responses. It is anticipated that the prior illustrated explanation of what will be sought may relax some subjects suffering from evaluation apprehension. may shape the way that others attend and react to the ad but will generally encourage subjects to be much freer and open in their thoughts than they otherwise might have been. We are fully aware of the demand characteristics of this manipulation and in fact it was created to observe the effect of providing these "leading" protocol examples ranging from high to low in their "structural" richness. All reactive copy testing can be criticized for its demand characteristics and the recommended method of studying and hence controlling such effects is to deliberately manipulate the factor. Contrast tests will be run between conditions 1 and 2, 2 and 3,and 3 and 4 to explore the incremental effects of explaining what information is being sought and providing demonstration examples of several verbal protocols. Providing the example ad and verbal protocol responses after exposure to the first test ad would interfere with the recall of information and reactions to the first test at and for this reason was not considered to be a viable manipulation

The Tested Coding Schemes

Three schemes will be used to cote each of the subjects' responses. The first "affect scheme" will use the conventional pro and counter argument approach advocated by Wright (1973) but will include coding thoughts by their source and whether they are directed at the product or features of the at (see Batra and Ray 1983). The second "tripartite scheme" will use the approach developed by Ostrom (1969) but will also cote the thoughts by their source and whether they are directed at the product or the ad. The third will test a new coding scheme that we have developed (see Table 1). It is different in several ways from the above two schemes. First, it will involve three sets of judges each making separate but phased sweeps of the subject's thoughts. We believe that this will make the coding task much easier and also increase inter-judge reliability.

The first judge will be asked to cote the thoughts following the "tripartite scheme". The second judge will then be asked to take the output of the first judge and code the extent to which each thought identified by the first judge is personalized (i.e., was it expressed in the first person, tit it refer to personal consequences, events or outcomes). The third judge will measure the extent to which the thoughts, if any, flowed in means-ends chains. For example, a thought that connects a feature to the advertised brand is evidence that the feature is salient and that the subject believes the brand possesses such a feature. A brand-feature thought connected to a benefit outcome would constitute evidence that (a) the feature is salient and the benefit is salient, and (b) direct evidence of feature instrumentality and hence direct evidence of an instrumentality-value based cognitive schema and attitude structure. If the benefit is connected to the self then this is suggestive of an internalization or centrality of the I-V structure to the self and clear evidence of personal relevance. Such a thought chain that concluded with some statement of purchase interest or intent would be considered to be evidence that the advertisement has had a powerful impact on persuasion. Each of these possible types of combinations of thoughts will be searched for and coded by the third judge. Our scheme enables the measurement of the impact of advertising copy on: (1) product knowledge, (2) beliefs, (3) affect toward the product and the ad, 44) intentions, (5) personal relevance, and (6) causal means-end thought chains that explain the origin of any expressed affect and intentions.

Analysis

After the subjects have completed the thought elicitation exercise they will be asked to complete a battery of standard measures of beliefs, values, attitudes and behavior intentions toward the target brand, a set of measures of attitude toward the ad and a set of product familiarity and involvement measures. It is also hoped that arrangements can be made to obtain a direct behavioral response measure by offering coupons offering major price discounts off the advertised test products in lieu of payment for participating. Mitchel (1985) has stated that we need i many different response measures in advertising studies and that correlational methods need to be used to examine the relationship between these variables. It is our objective to determine whether we can obtain higher correlations between various measures derived from our proposed verbal coding scheme and the later measures of attitude, intention and behavior, compared to the correlations between measures that we can derive from the other two coding schemes and the later measures. We are also interested in observing whether the task instruction manipulations influence such relationshiPs.

CONCLUSION

Aside from testing the effects of task instructions and the convergent validity of multiple measures, we are most interested in seeing whether we can directly observe, in the structure of subject's thoughts, what might be predicted by well established attitude theory. If we can't and yet the conventional measures and attitude modelling suggest such structure then there are at least two rather unpalatable competing explanations. Either thought elicitation cannot capture deep attitude structure (ant if this is so then how deep do any verbal protocols probe in capturing any structure?), or the thoughts do reflect substance and process and are a challenge to attitude theory that has been tested using mainly correlational relationships between combinations of batteries of measures. One possibility is that relationships between beliefs, values, attitudes and intentions operate at a non-verbal level. The problem with this is that it is difficult to argue that we have verbal access to our beliefs, values, attitudes and intentions but such verbal reports tell us nothing about when and how strongly they are connected.

The above research will attempt to develop a new thought elicitation taxonomy to capture relationships between well established constructs. Other advances are likely to be made by scholars interested in focusing on the nonverbal imaging and elaboration of subjects. Such activity may simply be under-reported because current thought elicitation task instructions do not request subjects to report on such topic specific and tangental activity. It may be very easy for subjects to report and categorize non verbal imaging but very difficult for them to describe its content in detail.

TABLE 1

THE THREE-PHASE CODING SCHEME

REFERENCES

Allport, G.W. (1935), "Attitudes," in C. Murchison (ed.) Handbook of Social Psychology, (Vol. 2), Worchester, Mass.: Clark University Press.

Anderson, Paul F. (1985), "On Method in Consumer and Buyer Behavior Research: A Critical Relativist Perspective." Working Paper, Harvard Business School, Boston, MA.

Batra, Rajeev and Michael L. Ray (1983), "Operationalizing Involvement as Depth and Quality of Cognitive Response," in Richard P. Bagozzi and Alice M. Tybout (eds.) Advances in Consumer Research, Vol X, Ann Arbor, Michigan: Association for Consumer Research, 309-313.

Bettman, James R. and C.W. Park (1980), "Implications of a Constructive View of Choice for Analysis of Protocol Data: A Coding Scheme for Elements of Choice Processes," in Jerry C. Olson (ed.) Advance; in Consumer Research, Vol. VII, Ann Arbor, Michigan, Association for Consumer Research.

Cacioppo, John T., Stephen G. Harkins, and Richard E. Petty (1981?,"The Nature of Attitudes and Cognitive Responses and Their Relationship to Behavior," in Petty, Richard E., Thomas M. Ostrom and Timothy C. Brock (eds.) Cognitive Responses in Persuasion, Hillsdale, New Jersey: Lawrence Erlbaum.

Ericsson, K. Anders and Herbert A. Simon (1984), Protocol Analysis: Verbal Reports as Data, Cambridge, MA: The MIT Press .

Fishbein,M. and I. Ajzen (1975), Belief, Attitude, Intention and Behavior: An Introduction to Theory and Research, Reading, Mass.: Addison-Wesley.

Freud, Sigmund (1924), Collected Papers, London: The Hogarth Press Ltd.

Krugman, H.E. (1965), "The Impact of Television Learning: Learning Without Involvement," Public Opinion Quarterly, 26, 622-634.

Leavitt, Clark (1970), "A Multidimensional Set of Rating Scales for Television Commercials," Journal of Applied Psychology, 54.5, 427.9.

Leavitt, Clark and Maureen M. McConville (1968), "Predicting Product Related Recall From Verbal Response," Proceedings 76th Annual Convention, American Psychological Association, 677-8.

Leavitt, Clark, Charles Waddell, and William Wells (1970), "Improving Day After Recall Techniques," Journal of Advertising Research, 10.3, 13-17.

Lipstein, Benjamin and James P. Neelankavil (1984), "Television Advertising Copy Research: A Critical Review of the State of the Art," Journal of Advertising Research, 24.2, April-May, 19-25.

Markus, Hazel (1980), "The Self in Thought and Memory," The Self in Social Psychology, D.M. Wegner and R.R. Vallacher. eds.. New York: Oxford Press, pp. 102-130.

McGuire, W.J. (1969), "The Nature of Attitudes and Attitude Change," in G. Lindzey and E. Aronson (eds.) The Handbook of Social Psychology, (2nd ed., Vol. 3) Reading, Mass.: Addison-Wesley.

Mitchell, Andrew A. (1985), "Theoretical and Methodological Issues in Developing an Individual Level Model of Advertising Effects," in K. Sentis and J.C. Olson (eds.) Advertising and Consumer Behavior, Praeger Publishing Co.

Olson, David W. (1984), "Validation of Copy Testing Measures Based on In-Market Performance: An Analysis of New Product Ads," in Methodological Advances in Marketing Research in Theory and Practice, Proceedings of EMAC/ ESOMAR Symposium, Copenhagen, 49-65.

Ostrom, Thomas M. (1969), "The Relationship Between the Affective, Behavioral and Cognitive Components of Attitude," Journal of Experimental Social Psychology, 5, 12-30.

Nisbett, R.E. and T.D. Wilson (1977), "Telling More Than We Can Know: Verbal reports on Mental Processes, Psychological Review , 84, 231-259.

Rogers, T.B., N.A. Kuiper and W.S. Kirker (1976), "Self-Reference and the Encoding of Personal Information," Journal of Personality and Social Psychology, 35:9, 677-688.

Rosenberg, M.J. (1956), "Cognitive Structure and Attitudinal Affect," Journal of Abnormal and Social Psychology, 53, 367-372.

Rosenberg, M.J. and C. Hovland (1960), "Cognitive, Affective and Behavioral Components of Attitudes," in M.J. Rosenberg, C.I. Hovland, W.J. McGuire, R.P. Abelson, and J.W. Brehm (eds.) Attitude Organization and Change, New Haven, Conn.: Yale University Press.

Schlinger, Mary Jane (1979), "A Profile of Responses to Commercials," Journal of Advertising Research, 19:2, pp. 37-48.

Wells, William D. (1964), "EQ, Son of EQ, and the Reaction Profile," Journal of Marketing, 28, 45-52.

Wright, Peter L. (1973), "The Cognitive Processes Mediating Acceptance of Advertising," Journal of Marketing Research, 10, 53-62.

----------------------------------------