Emotion-Eliciting Advertising: Effects on Long Term Memory and Judgment

Marian Friestad, University of Wisconsin-Madison
Esther Thorson, University of Wisconsin-Madison
ABSTRACT - Six to eight weeks after a single viewing of emotional and neutral advertising messages, subjects were phoned and given memory sad judgment questions about the messages and brands. The original viewing had involved two encoding instructions, one that encouraged episodic processing; the other, semantic processing of the messages. Results showed stronger memory and more positive judgments associated with emotional messages. Memory was stronger with episodic processing, while judgment was more positive with semantic processing. Emotion and encoding instructions showed some interaction. The findings are discussed in terms of the viewing conditions under which emotional messages will enhance memory and judgment.
[ to cite ]:
Marian Friestad and Esther Thorson (1986) ,"Emotion-Eliciting Advertising: Effects on Long Term Memory and Judgment", in NA - Advances in Consumer Research Volume 13, eds. Richard J. Lutz, Provo, UT : Association for Consumer Research, Pages: 111-116.

Advances in Consumer Research Volume 13, 1986      Pages 111-116

EMOTION-ELICITING ADVERTISING: EFFECTS ON LONG TERM MEMORY AND JUDGMENT

Marian Friestad, University of Wisconsin-Madison

Esther Thorson, University of Wisconsin-Madison

ABSTRACT -

Six to eight weeks after a single viewing of emotional and neutral advertising messages, subjects were phoned and given memory sad judgment questions about the messages and brands. The original viewing had involved two encoding instructions, one that encouraged episodic processing; the other, semantic processing of the messages. Results showed stronger memory and more positive judgments associated with emotional messages. Memory was stronger with episodic processing, while judgment was more positive with semantic processing. Emotion and encoding instructions showed some interaction. The findings are discussed in terms of the viewing conditions under which emotional messages will enhance memory and judgment.

INTRODUCTION

The study reported here is part of a research program that explores the role of emotion in memory and judgment about advertising messages. One objective of the study was to test memory and attitudinal differences for emotional and neutral messages after an extensive delay (six to eight weeks). The second objective was an examination of the effects of encoding instructions designed to initiate either episodic or semantic processing of the messages. a e third objective was to see if emotion operates differently in episodic and semantic memory.

Emotion's Role in the Processing of Commercials

Attempts to understand and define human emotion have a long history (e.g., Darwin 1872/1965; James 1890/1950; Canon 1914; Arnold 1968). In this paper, emotion will be treated as a process that occurs over time, varies in intensity (Simon 1982) and valence (Russell 1980), and exhibits a variety of concomitant indices: conscious appraisals of internal states and physiological changes in heart rate, blood pressure, and adrenalin levels. Emotion's ability to redirect attention and produce arousal distinguishes it from the concept of mood (Simon 1982). Finally, emotion is assumed to be a powerful influence on memory and judgment for at least three important reasons: emotion's developmental privacy (Handler 1984), its evolutionary importance (Plutchik 1980), and the general pervasiveness of its impact on human functioning (Lazarus 1980).

A review of emotion in the marketing and advertising literature is not feasible here (see Kreschel, 1984, for a good introduction). It should be noted, however, that ia marketing and consumer behavior, emotion has typically been defined in four ways. First, emotion has been treated as a type of appeal that is opposite from appeals emphasizing factual attributes of products (Preston 1968; McEwen and Leavitt 1976). Second, emotion has been defined as how such viewers like a commercial (Mitchell and Olson 1981; MacKenzie and Lutz 1982; Moore and Hutchinson 1983). Third, emotion has been defined as mood state during the viewing of commercials (Srull 1983). And fourth, emotion has been identified in the structure of cognitive responses that people make about commercials (Wright 1973; Leavitt 1968; Stout and Leckenby 1984).

Unfortunately, none of these definitions capture three critical attributes of emotion: the fact that it resides in individuals rather than in commercial message structure, that it is intense enough to elicit "feelings experienced," and that it fluctuates over tine. Under the present approach, an "emotional message" refers to a vehicle that creates over-time flow of feelings that people report as emotional experience. It is not the type of information in the message, the mood of the viewer, or how much the viewer can articulate liking the ad.

The paper turns now to an episodic-semantic model of consumer memory that incorporates this conceptualization and measurement of emotion.

Episodic and Semantic Memory and Consumers

Endel Tulving (1972) first suggested that a distinction should be made between memory that stores information about specific events experienced by a person, and memory that stores general knowledge about the world in a logical and categorized way. The first kind of memory he termed episodic, and the second, semantic. All incoming experience is stored initially as episodic information. In this system, associations between events are generally in terms of their contiguity in time. Semantic information must be derived by performing mental operations on information in episodes. When episodes are categorized, judged, evaluated, generalized, compared or in some other way acted upon by mental operations that information then becomes part of semantic store.

When people watch commercials, there is episodic processing of the audio and video events in the ads. There is also simultaneous processing of events in the viewing context (e.g., other people in the room, a doorbell ringing) and internal events such as physiological states (hunger, pain), thoughts, and feelings. Episodic traces of commercials contain all of this information. Furthermore, the traces vary in strength as a function of various parameters such as the intensity of the experience, subjective importance and meaningfulness of the experience, and the individual's ability to understand the message. One of the most important determinants of the strength of the episodic trace is emotion. In fact, there is extensive evidence that episodes experienced while people are emotionally aroused leave much stronger episodic traces than those experienced without arousal (Meltzer 1930, Postman 1947: Pillemer 1984).

While marketers like to think of consumers carefully operating on advertising episodes to enter product-related information into semantic memory, such operations may not occur, or they may occur only at some time after the episode is experienced. There are factors that increase the probability such operations will occur. These include relevance of episodes to the individual's goal states, the meaningfulness or importance of the message, and whether the episode arouses emotion. It seems reasonable to assume that under ordinary low-involvement viewing of television (Krugman 1965), primarily episodic memory traces are created. On the other hand, if subjects are instructed to make judgments as they vies, both episodic and semantic traces should be created.

As can be seen, then, it is important to consider two kinds of memory structure that can result in response to advertising messages. It can also be seen that emotion is relevant to the structure of both memories, as well as to the likelihood that mental operations will create semantic information from episodic traces. These relations are represented in a flow diagram of memory processes shown in Figure 1.

This motel suggests that emotion experienced while viewing a commercial will strengthen the episodic memory trace for that message. The model does not assume, as do many involvement theorists (Mitchell and Olson 1981; Batra and Ray 1983; Lutz, Mackenzie, and Belch 1983; Moore and Hutchinson 1983; Petty, Cacioppo, and Schumann 1983; Park and Young 1984), that low-involvement processing is based on emotional processing and high involvement is cognition-based; but rather that an emotional response may be present in both high and low involvement situations. The model also differs from the distinction (Gartner, Mitchell, and Russo 1985; between brand and non-brand processing. Certainly brand processing would be largely semantic, but non-brand processing (e.g., evaluating an ad) could also be semantic. The approach closest to the present one is the distinction between retrieval and computational processing (Srull, 1983). Retrieval instructions would.produce episodic processing. Computational instructions would demand semantic memory.

FIGURE 1

EPISODIC AND SEMANTIC MEMORY PROCESSING OF TELEVISION COMMERCIALS

The paper turns now to a study that explores the predictive use of the episodic-semantic model for messages that differ in their ability to generate emotional responses in viewers.

Experimental Design and Hypotheses

The present design is a subcomponent of a larger study that manipulated subjects' encoding and retrieval processes (episodic/semantic) and type of message (emotional/neutral). objects were exposed to a series of television program segments and promotional messages. In the episodic processing condition, the viewers were simply told to watch all of the material carefully. Subjects' processing of the stimuli was, therefore, unstructured by any imposed task. Subjects in the semantic processing condition were instructed to watch all of the messages and evaluate them in terms of their ability to "Influence your thoughts, opinions, and beliefs." These instructions imposed a semantic processing task on the subjects' encoding of the stimulus material.

Immediately after viewing, subjects were tested on their recall of the commercials, products, and brands, as well as their attitudes about the messages. These data are still being analyzed and not included in this paper. After a delay of approximately two months, the subjects were telephoned, reminded of their participation in the study, and again tested on memory for and judgments about the commercials (primarily an episodic task).

Hypotheses

Both the theoretical importance of emotion in strengthening episodic and semantic memory traces, and previous empirical findings that emotional commercials are more strongly remembered under short term memory conditions (Batra 1985; Friestad and Thorson 1985) led to the expectation that for long term recall:

H1: Commercials that produce strong emotional responses during viewing will be more likely to be recalled, and recalled in more detail than commercials that produce little or no emotional response during viewing.

Expectations about the effects of encoding instructions were more exploratory. It could be that episodic encoders, although they had also developed semantic traces during the immediate testing conditions, retained strong episodic traces and therefore when the long term memory task (primarily episodic in nature) was given, would perform better. This notion leads to the expectation that:

H2a: Memory performance for viewers who encode episodically will be better than memory performance for those subjects who encode the messages semantically.

On the other hand, since all commercial information must first be recorded episodically, the semantic encoders would, during viewing, lay down structures in both memory systems (double trace strength) and hence, remain at an advantage even for the long-term episodic task. This alternative conception leads to the expectation that:

H2b: Memory performance for viewers who encode semantically will be better than memory performance for those subjects who encode the messages episodically.

It should be noted that more complex recall conditions had been used in the immediate recall condition, but these differing recall conditions were evenly distributed between the two encoding groups.

A second set of exploratory hypotheses concerned whether there exist independent or interactive relationships between message type and encoding instructions. Because emotional experience may be stored in both episodic and semantic memory, equally strong operations of emotion in both encoding conditions (i.e., no interaction) seemed a reasonable possibility. However, a theoretical case for expecting greater impact of emotional messages in either episodic or semantic encoding could be made. For example, it might be that because episodic encoders presumably do not lay down semantic information during viewing, emotion creates greater strength differences between emotional and neutral message traces. This notion would lead to the expectation that:

H3a: Emotion will show greater recall enhancement effects for episodic encoders.

But it might also be that because semantic encoders must share processing capacity between episodic and semantic trace production, their episodic traces are generally weaker, and enjoy greater benefit of the presence of emotion than do those of episodic encoders. Hence:

H3b: Emotion will show greater recall enhancement effects for semantic encoders.

Finally, although the model of episodic and semantic memory shown in Figure 1 does not have theoretical representations of message liking and judgments of influence, there is empirical support in short-term memory tasks that emotional commercials receive more positive responses than do neutral ones (Stout and Leckenby 1985; Batra 1985; Friestad and Thorson 1985) therefore:

H4: Emotional messages will be better liked and produce more positive effects on judgments of influence than neutral messages.

In addition, there is evidence consistent with the idea that semantic processing involves making product and commercial evaluations (Srull 1984). In the present study, the semantic encoders would be more likely to have these responses available, even long after the experience. The episodic encoders would have to retrieve the relevant episodic trace and then make judgments. Given the weakened state of the trace, more neutral Judgments would be expected. Bence:

H5: Semantic encoding would lead to more positive message liking and perceived message influence for both emotional and neutral messages.

METHOD

Emotional and Neutral Commercials

Target messages were selected to create two levels of emotional response in subJects: strong emotional, and neutral. The messages were chosen on the basis of previous research (Friestad and Thorson 1985) in which subjects turned a dial up from a neutral point as they experienced positive feelings, and down as they experienced negative feelings. The resultant dial-reading patterns were averaged over each 30-second message. The five emotional messages had produced the largest mean divergence (X - 28.1) from the neutral dial reading, and the five neutral messages had produced the least mean divergence (I - 15.4).

Because of the complexity of flashed commercials it is difficult, if not impossible, to control all sources of variation. A number of steps were taken to minimize this problem. First, a variety of production techniques and executional styles were represented in both message types. Second, while many advertising and consumer behavior studies have used only one or two commercials to represent a particular category, in this study, a total of five messages in each message type were used. The specific messages in each category were as follows: Neutral: a) a home economics teacher and students, in a classroom setting, comparing a smoked sausage product to other cuts of meat; b) an off-camera spokesman describing an air freshener in strip form; c) a small child drawing on a chalkboard and asking for understanding rather than pity for disabled people; d) a slow-paced close-up of a light beer product with voice-over; and c) an on-screen woman presenting a new combination tomato sauce/tomato paste product. Emotional: a) children watching a Punch & Judy puppet show with a male voice-over talking about child abuse; b) a family reunion scene promoting a soft drink product; c) a series of outdoor scenes presented in a slow-paced manner with musical accompaniment sponsored by a bank; d) a young couple in a romantic setting with a male voice-over stating the woman had been killed by a trunk driver; and d) a series of family scenes with voice-over and Jingle promoting house paint. It should also be noted that the commercials and public service announcements used in the study are similar to those used by other researchers examining emotion in advertising (Batra 1985; Aaker, Stayman, and Hagerty 1985).

Finally, because familiarity has been associated with increased memorability, it was important to have the two message types as comparable as possible on this dimension. Information from a separate group of subjects (also college students) was used to establish ratings of the "familiarity" of the brands or organizations sponsoring each of the messages. On a ten-point scale, (10-high) mean familiarity of emotional messages was 4.59, and mean familiarity for neutrals was 5.18 (n.s.t.).

Programming and Buffer Commercials

Subjects viewed the ten target commercials embedded in an environment of program segments and buffer commercials. This procedure was used to control for privacy and recency effects, as well as to provide a more naturalistic viewing context. The program segments used were five- to six- minute "human interest" stories similar to what might appear on "PM Magazine." Each subject saw four buffer commercials, a program segment, five target commercials, a second program segment, five target commercials, a third program segment, and four buffer commercials. The order of the program segments and buffer commercials remained constant across the three counterbalanced orders of target commercials .

Subjects

The subjects in this study were undergraduate students who had participated in a laboratory experiment for course credit during the previous semester. Ninety-four of the original 170 participants were telephoned and interviewed for this study. No reward was given for participation in the phone interview. The length of delay between the initial laboratory session in which the advertisements were viewed, and the telephone contact ranged from 48 to 71 days (mean delay - 61 days).

Procedures

In the initial experimental session, subjects were told the purpose of the study was to "aid in the understanding of how viewers watch TV," and that they would be watching both programming and commercials. There were two encoding conditions: episodic and semantic. Instructions for the episodic encoding conditions were simply to watch the programs and commercial messages carefully, thereby encouraging an unstructured processing situation. The instructions for the semantic encoding conditions were to think about the programs and commercial messages in terms of their effectiveness in shaping thoughts, opinions, and beliefs. These instructions imposed an evaluative task on the subjects' processing of the messages. After viewing, subjects were asked to retrieve everything they could about the viewed commercials, and then answer additional questions about each of the target and buffer commercials. The laboratory study took an average of 75 minutes.

Sis to eight weeks later, interviewers telephoned subjects and identified themselves as calling on behalf of the Wisconsin Department of Journalism and Mass Communication. Subjects were reminded of their participation in the original study, and asked to answer some additional questions. If they agreed to participate, they were given a free recall test, product recognition questions, and finally asked for some judgments.

Dependent Measures

There were five kinds of dependent measures: free recall, additional free recall, recognition, liking for the messages, and judgment of how influential the messages were.

Free recall was measured by asking respondents to remember all they could about messages presented during the initial test study. > is question produced six recall measures: Number of messages recalled, total information in the recall protocols, and numbers of executional, product category, claim, and brand names contained in the protocols. After respondents were able to recall no more messages, they were given an additional recall prompt: were there any other images, thoughts, or events you can recall from the messages you saw?"

Both free and additional recall responses were transcribed word for word, and later coded into the four categories of executional elements, product category, claims, and brand name. Three coders performed these tasks, and their intercoder reliabilities were: executional elements (.92), product category (.98), claims (.98), and brand name (.98).

Next, respondents were given product categories and asked whether they remembered an at for each one (Recognition). Subjects were asked to rate on seven-point scales how much they liked each message (Liking) and how much it had influenced them (Influence). Finally, because syndicated delayed memory scores are often a combination of free recall and recognition (e.g., Bruzzone) or cued recall (e.g., Burke), the number of messages generated by a combination of the two methods (Recall Plus Recognition) was used.

RESULTS

Each of the dependent measures was subjected to a 2 (Message Type) x 2 (Encoding Condition) analysis of variance (anova) with message type a within-subJects variable, and encoding a between-subjects variable. Occasional differences in number of subjects per analysis are attributable to missing data.

Free Recall

The anova for the number of messages recalled shoved a significant main effect of message type (F(1,90)=35.29, p <.0000). Encoding had no effect and there was no interaction. As can be in Figure 2, Hypothesis 1 was supported, in that emotional messages were more likely to be recalled then neutral messages.

FIGURE 2

MEAN NUM8ER OF MESSAGES RECALLED

Free recall of executional elements, produce category, and brand name mentions showed the same pattern of results, with emotional messages producing significantly higher response levels than neutral messages. As in number of messages recalled, episodic coding was always higher than semantic coding, but not significantly. Only free recall of claim (no significant differences) failed to show this pattern, presumably because claim scores were 80 low.

It is interesting to note that although the means for brand nave recall were quite low, there was a marginally significant main effect for encoding condition (F(1,90) 3.9, p < .08), with the episodic mean 5 s .36) higher than the semantic (X - .10). This main effect provides marginal support for Hypothesis H2a. The levels of brand recall in this study ranged from .10 to .46, and are very similar to levels of delayed recall (28 days) reported by Craig, Sternthal, & Leavitt (1976) in their study of long-term repetition effects.

Total Amount of Information in Free Recall

A sum of the number of executional elements, product category references, product claims, and brand names recalled was calculated for the two message types. Again, in support of Hypothesis 1, there was a significant main effect for message type (P(l,90) - 4.72, z < .03). The wean amount of information recalled for emotional messages, collapsed across encoding conditions, was 3.17. The mean for neutral messages was 1.60.

Additional Free Recall

After the respondents had spontaneously recalled as many commercials as possible, they were prompted with a request for any other images, thoughts or events that came to mind. These responses were coded in the same way as the initial free recall responses. For this variable, there was a main effect for message type (F(1,91)"15.61, v < .001). The mean number of additional emotional messages recalled was .42; the mean for neutral messages was .10.

A sum of the four different types of information (execution, product category, claim, and brand) was also analyzed for these additional messages. In this case there was a main effect for encoding (F(1,91)-4.49, z < .03), with episodic subjects recalling more information (t - 2.0) than semantic subjects (X - .33).

Although encoding effects were generally weak in the memory data, there was a consistent direction in the differences. Episodic encoding always produced more recall than semantic encoding. a is result, together with the two instances of marginal or significant differences, provides preliminary support for H2a, namely that episodic encoding will produce stronger memory than semantic encoding .

Recognition

The number of commercials not recalled but correctly recognized was next analyzed. In this anova, there were no significant effects.

Liking and Influence

Figure 3 shows the results of respondents' rating of how much they liked each message. Here, message type produced a significant main effect (F(1,86)-47.42, p < .0000), with emotional messages better liked. There was no interaction, but encoding had a marginally significant effect (F(1,86)-3.02, p < .09), with semantic encoding producing more positive liking than episodic.

FIGURE 3

MEAN LIKING FOR ALL RECALLED AND RECOGNIZED MESSAGES

The anova on perceived influence shoved a similar pattern. Encoding did not produce a significant effect, but semantic encoding was more positive than episodic. Message type was again significant (F(1,86)-29.03, p < .0000). There was no interaction.

The liking and influence results show clear evidence for H4, that emotional messages are responded to more positively than neutral messages. There is weaker evidence for H5, that semantic encoding leads to more positive delayed responses than episodic coding.

Recall Plus Recognition

In this measure, messages free-recalled and those recognized were combined. There was a significant main effect for message type (F(1,92)-35.53, z < .001), no effect of encoding, and a marginally significant interaction (F(1,92)-2.89, p < .09). Follow-up two-tailed t-tests shoved a significant (p < .05) difference only for neutral messages. As can be seen in Figure 4, contrary to Hypotheses 3a and 3b, emotion did not differentially affect episodic and semantic encoders. Instead, encoding condition affected memory only for neutral ads which shoved higher recall under episodic coding.

FIGURE 4

MEAN RECALL PLUS RECOGNITION

DISCUSSION

The most striking aspect of these results is the positive long-term effect of emotional messages on memory. For every dependent variable except free recall of claims, where memory exhibited a floor effect, the emotional messages showed stronger memory than the neutral messages. This finding is in contrast to the popular conception in advertising that emotional ads "don't recall well" (Zielske, 1982), but consistent with recent scientific tests (Choi and Thorson 1983; Thorson and Friestad 1984; Batra 1985).

A second important aspect is the directional support of memory differences as a function of encoding instructions. While the main effect of encoding did not reach statistical significance as often as did message type, the consistency in the direction of the differences make the encoding results worthy of consideration. The fact that episodic encoding consistently showed better memory than semantic encoding lends support to the notion that encoding instructions did lead to differential memory processes. The alteration of this pattern in the judgment tasks adds further credence to the idea that consumers may differentially store information in memory as a function of differing processing objectives (Srull, Lichtenstein, and Rothbart 1985).

Finally, there are the findings of interaction between message type and encoding. The only dependent measure that indicated an interaction was the combination measure, Recall Plus Recognition. Here, encoding mate no significant difference for emotional messages. However, neutral messages benefitted from episodic encoding. Because the encoding effect was weak, interpreting the interaction patterns should await analysis of the data on immediate performance. Nevertheless, some speculation on the delayed performance results are in order. Given that most of the individual recall measures (number of messages recalled, product category, brand name mentions, executional, and claim information) show parallel effects of message type and encoding, the data support the idea that memory trace enhancement from the presence of emotion plays a similar role in both the episodic and the semantic systems. The interaction for Recall Plus Recognition under this interpretation may indicate that when total memory performance is indexed, the enhancing effect of emotion is strong enough to overwhelm encoding effects, and that the advantage of episodic coding can only be observed in neutral messages.

One of the inherent frustrations in the type of research presented here is that the most interesting variables are also those that carry with them significant methodological problems. The limitations of the data reported in this paper are directly tied to: 1) the problems associated with using complex stimuli (commercials), and 2) the difficulty of effectively manipulating the thought processes of individuals (encoding instructions). For example, it is not possible to "know" in a completely objective sense if emotion was the critical difference between the two sets of messages. Nevertheless, the steps taken to control for some of the more likely confounding variables (e.g., familiarity), the use of multiple messages, and the variation in products and executional styles help to address this problem.

Further, the difficulty of determining processing strategies of individualist is a problem for all researchers interested in cognition. In the data presented here, it cannot be unequivocally claimed that the manipulation was successful. With more effective control over strategies and without the immediate testing condition, the differences as a function of encoding condition might have been stronger. It is likely that the short-term performance data will help to clarify this issue.

In general, the long-term delay in this study created a difficult test of the episodic-semantic encoding distinction and the effects of viewer emotion during commercials. After only a single exposure and a long delay, emotional messages nevertheless hat a decided advantage over neutral messages. The effects of the episodic-semantic encoding distinction are only suggestive as yet, but they point toward a model of processing that has potential significance to students of consumer behavior.

REFERENCES

Aaker, David A., Douglas M. Stayman, and Michael R. Hagerty (1985), "Warmth in Advertising: Measurement, Impact and Sequence Effects," University of California at Berkeley Working Paper.

Arnold, Magda et. (1968), The Nature of Emotion, Baltimore: Penguin Books.

Batra, Rajeev (1985), "How Emotional Ads Work: Theory and Evidence," in Proceedings of the American Academy of Advertising, et. Nancy Stephens, Tempe, AZ: Arizona State University.

Batra, Rajeev and Michael L. Ray (1983), "Advertising Situations: The Implications of Differential Involvement and Accompanying Affect Responses," in Information Processing Research in Advertising, et. Richard Harris, Hillsdale. NJ: Erlbaum.

Canon, Walter 8. (1914), Recent Studies of Bodily Effects of Fear, Rage and Pain. Journal of Philosophy. Psychology and Scientific Methods, 11, 162-165.

Choi, Young and Esther Thorson (1983), "Memory for Factual, Emotional, and Balanced Ads Under Two Instructional Sets," in Proceedings of the 1983 Conference of the American Academy of Advertising, ed. Alan D. Fletcher, Knoxville: University of Tennessee.

Craig, C. Samuel, Brian Sternthal and Clark Leavitt (1976), "Advertising Wearout: An Information Processing View," Journal of Marketing Research, 17(May), 173-186.

Darwin, Charles (1872-1965), The Expression of the Emotions in Man and Animals, Chicago: University of Chicago Press.

Friestad, Marian and Esther Thorson (1985), "The Role of Emotion in Memory for Television Commercials," paper presented at the annual meeting of the International Communication Association. Honolulu, HA.

Gartner, Meryl P., Andrew A. Mitchell, and J. Edward Russo (1985), "Low Involvement Strategies for Processing Advertisements," Journal of Advertising, 14(2), 4-12.

James, William (1890/1950), "The Emotions," in The Principles of Psychology, Volume II, New York: Dover.

Kreshel, Peggy J. (1984), "Emotion and Advertising: The "Grande Idea" of the 80's." Advertising Working Paper No. 17, College of Communications, University of Illinois.

Krugman, Herbert E. (1965), "The Impact of TV Advertising: Learning Without Involvement," Public Opinion Quarterly, 29, 349-356.

Lachman, Roy, Janet L. Lachman, and Earl C. Butterfield (1979), Cognitive Psychology and Information Processing, Hillsdale,-NJ: Erlbaum.

Lazarus, Richard S., Allen D. Kanner, and Susan Folkman (1980). "Emotions: A Cognitive Phenomenological Analysis," in Emotion: Theory, Research and Experience, eds. Robert Plutchik and Henry Kellerman, New York: Academic Press.

Leavitt, Clark (1968), "Response Structure: A Determinant of Recall," Journal of Advertising Research, 8, 3-6.

Lutz, Richard J., Scott B. MacKenzie, and George E. Belch (1983), "Attitude Toward the Ad as a Mediator of Advertising Effectiveness: Determinants and Consequences," in Advances in Consumer Research Vol X, eds. Robert P. Bagozzi and Alice M. Tybout Ann Arbor: Association for Consumer Research.

MacKenzie, Scott B. and Richard Lutz (1982), "Monitoring Advertising Effectiveness: A Structural Equation Analysis of the Mediating Role of Attitude Toward the At." Working Paper #117 UCLA, Center for Marketing Studies, Los Angeles.

McEwen, William J. and Clark Leavitt, (1976), "A Way to Describe TV Commercials." Journal of Advertising Research, 16, 35-39.

Mandler, George (1984), Mind and Body, Nev York: W.W. Norton.

Meltzer, B. (1930), "The Present Status of Experimental Studies on the Relationship of Feeling to Memory," Psychological Review, 37, 124-137.

Mitchell, Andrew A. and Jerry C. Olson (1981), "Are Product Attribute Beliefs the Only Mediator of Advertising Effects on Brand Attitudes?" Journal of Marketing Research, 18, August, 318-332.

Moore, Danny L. and J. Wesley Hutchinson (1983), "The Effects of Ad Affect on Advertising Effectiveness," in Advances in Consumer Research Vol X, eds. Richard P. Bagozzi, and Alice M. Tybout, Ann Arbor: Association for Consumer Research.

Park, C. Whan and S. Mark Young (1984), "The Effects of Involvement and Executional Factors of a Television Commercial on Brand Attitude Formation," Marketing Science Institute Report No. 84-100.

Petty, Richard E., John T. Cacioppo, and David Schumann (1983), "Central and Peripheral Routes to Advertising Effectiveness: The Moderating Role of Involvement," Journal of Consumer Research, 10 (September), 135-146.

Pillemer, Davit B. (1984), "Flashbulb Memories of the Assassination Attempt on President Reagan," Cognition, 16, 63-80.

Plutchik, Robert (1980), "A General Psycho-evolutionary Theory of Emotion," in Emotion: Theory, Research and Experience, eds. Robert Plutchik and Henry Kellerman, New York: Academic Press.

Postman, Leo (1947), The History and Present Status of the Law of Effect." Psychological Bulletin, 44, 489-550.

Preston, Ivan L. (1968), "Relationships Among Emotional, Intellectual and Rational Appeals in Advertising," Speech Monographs, 35, 504-511.

Russell, James A. (1980), "A Circumplex Model of Affect." Journal of Personality and Social Psychology, 39, 1161-1178.

Simon, Herbert (1982), "Comments," in Affect and Cognition, eds. Margaret Clark sat Susan Fiske, Hillsdale, NJ: Lawrence Erlbaum.

Srull, Thomas t. (1984), "The Effects of Subjective Affective States on Memory and Judgement," in Advances in Consumer Research, Vol. XI, ed. Thomas C. Kinnear, Provo, UT: Association for Consumer Research.

Srull, Thomas R. (1983), "Affect and Memory: The Impact of Affective Reactions in Advertising on the Representation of Product Information in Memory," Advances in Consumer Research, 10, 520-525.

Srull, Thomas R., Meryl Lichtenstein, and Myron Rothbart (1985), "Associative Storage and Retrieval Processes in Person Memory, Journal of Experimental Psychology: Learning, Memory, sat Cognition, 11(2), 316-345.

Stout, Patricia and John D. Leckenby (1984). "The Rediscovery of Emotional Response in Copy Research," in American Academy of Advertising Proceedings, ed. Donald R. Glower, Lincoln, Nebraska: School of Journalism, University of Nebraska-Lincoln.

Stout, Patricia and John Leckenby (1985), "The Pendulum Swings--A Return to Emotion in Copy Research," in American Academy of Advertising Proceedings, ed. Nancy Stephens, Tempe AZ: Arizona State University.

Thorson, Esther (1984), -Episodic and Semantic Memory: Implications for the Bole of Emotion in Advertising," paper presented at the Advertising Division of the Association for Education in Journalism and Mass Communication, Gainesville, Florida.

Thorson, Esther and Marian Friestad (1984), "Emotion and the Recall of Television Commercials," in Proceedings of the American Psychological Association, Consumer Science Division, ed. David Stewart, Toronto, Canada.

Tulving, Endel (1972), -Episodic and Semantic Memory," in Organization of Memory, eds. Endel Tulving and W. Donaldson. NY: Academic Press, 381-403.

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

Zielske, Hubert A. (1982), "Does Day-After Recall Penalize 'Feeling' Ads?- Journal of Advertising Research, 22(1), 19-23.

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