The Story, the Frame, and the Choice

ABSTRACT - Much of the information processing literature has focused on how individuals handle information and make choices in particular task settings. Dawes (1975) has suggested that 'the performance observed may well be task specific, and that the behavior exhibited depends on the task structure. This research investigates how individuals come to a specific task (whether to purchase property insurance or not), how that initial response affects the reference points selected, and how both relate to the resultant Attitudes and intentions. The results indicate that "subjects' initial Affect and Cognitive Story do relate significantly with the reference points selected and with the specific attitudes and intentions relating to the property insurance task.

Citation:

James W. Gentry, Joshua L. Wiener, and Melissa Burnett (1987) ,"The Story, the Frame, and the Choice", in NA - Advances in Consumer Research Volume 14, eds. Melanie Wallendorf and Paul Anderson, Provo, UT : Association for Consumer Research, Pages: 198-202.

Advances in Consumer Research Volume 14, 1987      Pages 198-202

THE STORY, THE FRAME, AND THE CHOICE

James W. Gentry, Oklahoma State University

Joshua L. Wiener, Oklahoma State University

Melissa Burnett, Oklahoma State University

ABSTRACT -

Much of the information processing literature has focused on how individuals handle information and make choices in particular task settings. Dawes (1975) has suggested that 'the performance observed may well be task specific, and that the behavior exhibited depends on the task structure. This research investigates how individuals come to a specific task (whether to purchase property insurance or not), how that initial response affects the reference points selected, and how both relate to the resultant Attitudes and intentions. The results indicate that "subjects' initial Affect and Cognitive Story do relate significantly with the reference points selected and with the specific attitudes and intentions relating to the property insurance task.

INTRODUCTION

There are a multitude of approaches that can be employed to explain and predict how individuals sake choices under uncertainty. In recent years, the traditional expected utility approach has been supplemented by prospect theory. Prospect theory, as developed by Kahneman and Tversky ,1979), proposes that the task environment directly influences how the decision is framed (i.e., how the individual perceives the various alternatives, the consequences derived from selecting an alternative, and the likelihood of particular consequences occurring). As an individual edits a problem, she/he codes the potential outcomes as being either above or below a particular reference point. Potential outcomes are framed as a gain or a loss which directly influences choice. Individuals are theorized to be risk takers in the domain of losses and risk avoiders in the domain of gains.

Numerous experiments have found that, in highly structured task environments, prospect theory can predict choices. See Kahneman and Tversky (1984) for a literature review. The task environments are so structured that, in the words of Kahneman and Tversky (1979), "Either the original formulation of prospects leaves no room for editing or the edited prospects can be specified without ambiguity." Thus, it is assumed that the subject's decision frame is forced by the task she/he is given by the experimenter.

Prospect theory has proven less potent when used to explain how choices are made in unstructured situations. In a recent set of experiments, Lichtenstein, Wagenaar, and Keren (1985) investigated this issue by using story problems that had common deep structures and different surface structures. A story problem's surface structure is the story as presented, and its deep structure is the representation of the problem used by the respondent or as specified by psychological theory (Lichtenstein, Wagenaar, and Keren 1985). They investigated the relationship between surface structure and choice by developing eleven variations of a problem that, according to subjectively expected utility theory, had a common deep structure. They found that changing aspects of the surface structure (e.g., whether the individuals at risk were island residents or hostage children) greatly changed the proportion of respondents selecting the risk averse alternative. After discussing prospect theory's editing operations, Lichtenstein, Wagenaar, and Keren (1985) conclude that these operations do not provide a basis for understanding why their surface structure variations produced such different choices.

Fischhoff (1983) first pointed out that prospect theory's tests had been limited to highly structured basis in which the subject's frame was not measured. To overcome this problem he presented subjects with problems that could be framed in multiple ways, measured their reference points, and recorded their choices. He found no systematic association between a subject's frame and his or her choice. Puto (1985) has suggested that, if generally true, Fischhoff's (1983) findings imply that "the domain of prospect theory may be limited to highly structured gambles" and reference points might be unimportant for explaining choice. He also suggests that Fischhoff's (1983) findings may be attributable to poor reference point measures. Consequently, Puto (1985) conducted a study in which he sought to manipulate the frames used by subjects by varying the task description, measuring their reference points, and then recording their choices. The results of Puto's (1985) study suggest that the reference point is related to choice: however, it leaves open the question of how useful prospect theory is for explaining unstructured choices.

We argue that prospect theory's difficulty with unstructured task environments is partially attributable to the inadequacy of its editing phase. The editing phase consists of six basic operations. As stmfltarized by Fischhoff (1983) and Lichtenstein, Wagenaar, and Keren (1985), they are: (a) coding (describing each outcome in terms of a reference point), (b) segregation (isolating the riskless from the risky components of a prospect), (c) cancellation (discarding the components that are shared by all prospects), (d) combination (adding probabilities that are associated with identical outcomes), (e) simplifications (rounding probabilities or outcomes), and (f) detection of dominance (eliminating dominated prospects).

One weakness is that it is a pure encoding approach to the issue of how surface structures are transformed into deep structures. Prospect theory explicitly predicts how a specific set of probabilities and outcomes will be transformed into a decision frame, i.e., how the manner in which the objective data are presented influences the subject's framing of the decision.

We maintain that an individual's framing of a decision is influenced by the information she/he accesses from memory. The information accessed is influenced by the surface story. Both our criticism of prospect theory's editing phase and our argument that accessed information influences framing have been made by previous researchers, including Tversky and Kahneman (1981). In their words, "a diversity of factors determine the reference outcome in everyday life. The reference outcome is usually a state to which one has adapted; it is sometimes set by social norms and expectations; it sometimes corresponds to a level of aspiration, which may or may not be realistic" (Tversky and Kahneman 1981, p. 456).

Our contribution is that we will investigate the role played by two types of stored beliefs. Specific actions may be viewed as intrinsically good or bad, and these feelings (Affect) may influence how the decision is framed.

The second type of stored belief is a simple cognitive description (Story) of the choice situation. Although the exact nature of this stored description, whether a script, schema, or some far richer representation, is unknown (Alba and Hasher 1983), there is little controversy over the fact that one's response to a choice situation can be influenced by one's recalled beliefs about the situation. These recalled beliefs can focus the subject's attention upon particular dimensions of the decision, or even suggest (contain) a reference point.

In addition to influencing how a decision is framed, affect and the simple cognitive story may have two other consequences. First, either or both may directly influence choice. Second, either or both may mediate the consequences of surface task factors. If memory has a powerful influence upon the reference point, then the choice of a reference point may not be influenced by how the experimenter presents the task.

The role of these memory factors will be investigated in the context of an insurance purchase decision. Reasons for utilizing insurance are two-fold. First, an insurance purchase decision can be framed in multiple ways. Certain potential framings predict that insurance will be purchased, while other potential framings predict that it will not be purchased.

Slovic, Fischhoff, and Lichtenstein (1982) suggest the most straightforward decision frame is the decision frame which uses the no gain/no loss state as the reference point. Using this reference point, the insurance premium is viewed as a small certain loss while the option of not insuring is viewed as a risk involving a large uncertain loss at worst and no gain at best. Since risk seeking is common in the domain of losses (Hershey and Schoemaker 1980; Payne, Laughhunn and Crus 1980; Slovic, Fischhoff, and Lichtenstein 1982), individuals using this frame would be expected to prefer the risk of a large uncertain loss over the cost of a small certain loss; i.e., they would not be expected to purchase insurance.

Another reference point suggested by Slovic, Fischhoff, and Lichtenstein (1982) is the loss of the premium. A possible rationale for the existence of this reference-point would be that buying insurance is socially desirable. Using this reference point, the decision is framed as though the premium were paid and the small certain loss taken. Individuals using this form would view the amount of the premium as being in the domain of gains, and the difference between the premium and the total loss incurred as the possible loss. Because the value function is steeper for losses than gains, the risk of a large uncertain loss would be viewed negatively. Thus, it would be more attractive to stay at the reference point of paying the insurance premium.

Kahneman and Tversky (1979) have suggested that an individual could use zero assets as their reference point. Use of this reference point would lead to the framing of all potential outcomes in the domain of gains. Consequently, the risk averse course of action of purchasing insurance would be selected. Kahneman and Tversky (1979) argue that this particular framing is the one assumed by expected utility theory. There is some empirical support for the contention that some people use the expected utility frame. In Wiener, Gentry, and Miller's (1985) study, one frame the individuals could use corresponded to the frame posited by expected utility theory. Individuals who used this frame were more likely to express a willingness to purchase flood insurance than those who did not use this frame.

Fischhoff (1983) has suggested that decisions are framed so that the worst case is the reference point. Applied to the insurance decision, this implies that the consequences of an uninsured event, such as a flood, would be the reference point. Use of this framing would place all alternatives in the domain of gains, and should lead to an increased likelihood of purchasing insurance.

Besides providing a variety of possible reference points, a second reason for investigating insurance is that it is a familiar decision that numerous researchers have observed can be influenced by social norms (Hershey, Kunreuther, and Schoemaker 1982; Kahneman and Tversky 1979; Slovic, Fischhoff, and Lichtenstein 1982). Two possible paths by which these social norms influence the decision have been proposed. Kahneman and Tversky (1981, 1984) argue that the norms will influence how the decision is framed. Hershey, Kunreuther, and Schoemaker (1982) suggest that the social norms will dominate the framing process. Consequentially, the purchase decision may not be influenced by the reference point. At present, all that is known is that subjects are much more likely to pay for protection when the term "insurance premium" is used rather than the term "sure loss" (Hershey and Schoemaker 1980; Schoemaker and Kunreuther 1979; Slovic, Fischhoff, and Lichtenstein 1982; Slovic, Fischhoff, Lichtenstein, Corrigan, and Combs 1977). Hence insurance provides a means for investigating whether, in an unstructured setting, the reference point influences the individual's decision.

METHODOLOGY

A two-stage data collection process was employed. During the first stage the beliefs concerning insurance, stored in the subject's memory, were measured. In the second stage subjects were shown an advertisement, and given the (role play) opportunity to purchase a theft insurance policy. At this time their perceptions of the advertisement, reference point, and choice were measured. Below the key elements of the study, and their measurement, are discussed.

Affect. Subjects gave their affective reaction to the term "insurance" using seven-point scales with the labels "Good/Bad," "Like/Dislike," and "Positive/Negative."

Story. Two possible semantic descriptions of insurance have been suggested by previous researchers. Slovic et al. (1977) have suggested that individuals resist buying insurance because they view it as a sure loss. Kahneman and Tversky (1984) have suggested that people willingly buy insurance because they perceive it to be protection. Wiener, Gentry, and Miller (1985) fount that people who perceived it as protection were more likely to indicate a willingness to buy insurance than those who perceived it as a sure loss.

Three approaches were used to measure the story used by the subjects. First, the subjects were shown statements about insurance: 1) Buying insurance is a gamble involving losses; the insurance premium is a certain loss that is bet against the small possibility of incurring a large loss; or 2) Buying insurance protects my assets; I pay a small fee--premium--to make certain that I will be able to keep what I have). They were asked to make at least three statements in support of each position and at least three that are not in support. Second, the respondents were asked to indicate their level of agreement with the two general perspectives on seven-point "Strongly Agree/Strongly Disagree" scales. Last, they were asked to choose the perspective which best represents their view of the insurance purchase. It was expected that respondents that viewed insurance in one way (a gamble, for example) would find it much easier to write supportive statements for "insurance as a gamble," write contradictory statements for "insurance as protection," express a higher level of agreement with the "insurance as a gamble" statement, and choose that as their initial representation of insurance.

Reference. There are many potential reference points that can be used for tho insurance decision. Reference points c n be formulated either (1) in terms of assets or (2) in tenDs of gains, losses, and neutral outcomes. Prospect theory research suggests that the latter formulation is more common; hence it is the one we have used. This formulation implies that there are four potential consequences of the property insurance purchase decision: (1) owning insurance/no theft; (2) owning/theft; (3) not owning insurance/theft; and (4) not owning insurance/no theft.

The first two consequences have the same monetary consequence, i.e., a premium is paid. Hence, as Lichtenstein, Slovic, and Fischhoff (1983) have suggested there is a sound b sis for considering "owning insurance" to be a potential reference point. On the other hand, as Loomes and Sugden (1982) have argued, an individual may care about more than the payoff she/he receives. She/he may also care about the payoff she/he would have received if she/he had made a different choice. Prom this perspective, the two consequences are obviously very different. In light of these two reasonable (albeit inconsistent) viewpoints, two sets of measurements were made of the decision frame. In the first, the two consequences corresponded to two potential frames. In the second, a single potential frame "owning insurance" was used. The remaining two consequences directly correspond to the "worst case" frame suggested by Fischhoff (1983) and the "straightforward" frame suggested by Slovic, Lichtenstein, and Fischhoff (1983). A reference point can be measured in one of two ways: (1) by selecting one of the reference points, as was done by Fischhoff (1983) and Puto (1985) and (2) by indicating their level of agreement to statements containing the reference points as was done by Wiener, Gentry and Miller (1985).

In the first set of measures subjects were asked to select the reference point that they: thought of first, paid most attention to, and compared other consequences to. In the second set they were asked their level of agreement (seven-point Likert Scale) with the assertion that they thought of a specific reference point first, and then asked to select the consequence that was most important to them when making their decision.

Multiple questions (first, attention, compare, and important) were used to explore the relationship between the reference point, and other (more familiar) points of interest. Previous research has established that the initial value (consequence) considered can serve as an anchor. If a value (consequence) serves as an anchor then other values (consequences) are being viewed relative to the initial value. If alternatives are being compared to a particular consequence then this consequence will both be attracting attention (i.e. mental processing effort will be focused on this consequence) and be important.

Ads. As noted above, previous research has concentrated upon investigating how the communicator's presentation of a problem influences framing and choice. In our study advertisements were used as the communicator-generated source of influence. Three advertisements were used, with one-third of the subjects seeing each one. The advertisements were presented in print form each, stressing one of the possible reference points (the loss of all one's possessions, the no loss/no gain state, and the loss of the premium) suggested by Slovic, Fischhoff and Lichtenstein (1982).

The advertisements were professionally created and produced to minimize the possibility of the results being biased by the quality of the ads. All ads contained illustrations, approximately the same length of copy, same size and type, and similar layouts. Evaluated by eight semantic differential style questions, the advertisements were checked to insure that they created the frame they sought to induce.

Attitudes and Purchase Intentions. Individuals were initially asked to assess the likelihood that they would buy the advertised property insurance. This was followed by seven Likert questions focusing on their purchase intentions and attitudes towards the policy. Finally, they were forced to decide whether they would or would not bus the policy.

HYPOTHESES

The expected associations are summarized in Figure 1. It is assumed that even though affect and cognitive story are related, they have independent effects. Affect will encompass the positive associations generated by many factors, of which protection is only one. It is assumed that in our design the advertisements are of equal persuasive power, hence they influence choice solely through their influence on the reference point. The expected associations are discussed in more detail below.

H1: Liking insurance is associated with buying insurance.

One's general Affect toward insurance is related to the choice decision made in specific situations. To some extent, the choice decision will depend upon the task structure as represented by the decision theoretic details (consequences and probabilities), but there also will be utility (disutility) that is derived from buying and/or owning insurance. Social norms may play a dominant role in the individual's willingness to buy insurance (Hershey, Kunreuther, and Schoemaker 1980), and those norms may determine the nature of the individual's affective response to insurance.

H2: People who like insurance will think of it as protection.

An individual's affective response is expected to be consistent with the cognitive description the subject associates with the action. The description should support the feeling, i.e., if the action is viewed as being positive then the cognitive story should contain positive associations. This hypothesis, that positive affective judgments should be associated with the favorableness of the information accessed from memory, is consistent with existing memory based accounts of the judgment process (Kisielius and Sternthal 1986; Petty and Cacioppo 1981).

H3: Liking insurance is associated with using the worst case reference point.

Previous researchers (Kahneman and Tversky 1983; Puto 1985; and Thaler 1985) have suggested that affective responses and framing should be linked. We expect that the decision will be framed so that the act's coding (in terms of the reference point) and the affective reaction are consistent. A simple consistency requirement is that the more favorably an act is perceived, the more likely it is to be perceived as a gain, and the less likely it is to be perceived as a loss.

H4: Thinking of insurance as protection is associated with using the worst case reference point.

H5: It is expected that the cognitive story will be associated with both choice and the reference point.

Specifically, those who view insurance as protection should be more likely to indicate a willingness to buy insurance. This is expected because protection is an insurance benefit that is independent of the actual monetary consequences. As Wiener, Gentry, and Miller (1985) have argued the gamble versus protection semantic framing is parallel to the no gain/no loss versus worst case reference point framing. We expect to find these associations.

H6: Using the worst case reference point is associated with buying insurance.

It is expected that the subject's choice of a reference point will be associated with his choice. Subjects selecting the no gain/no loss reference point should be far less likely to purchase insurance than those selecting the worst case reference point. Prospect theory predicts that individuals selecting either owning insurance consequence should be more likely to buy insurance than those selecting the no gain/no loss reference point. The alternative view, suggested by Loomes and Sugden (1982), implies that individuals who use theft/own insurance as the reference point will be far more likely to buy insurance than those using the no theft/own insurance reference point.

RESULTS

This section will discuss the relationship between the components of the model presented in the previous section. First, though, we will discuss the descriptive statistics since they provide insight into how consumers view the insurance purchase decision.

Descriptive Statistics

Table 1 presents a summary of the responses on the Affect, Story, Frame, and Decision variables. The means for the three Affect variables (6.1, 5.6, and 5.4, with 4.0 being the neutral point) indicate that the term "insurance" elicited a generally positive response from the respondents. The story was measured in three ways: a count of the supportive and the contradictory statements given for the two alternative insurance stories ("protection" or a "gamble"), the level of agreement with the two stories, and the choice with which they agreed most. The protection story elicited more cognitive responses than did the gamble story, indicating that the respondents may have been more familiar with the logic of the protection story. There were more statements made in opposition to both stories than were made in support. There was little difference in the overall mean agreements with the two stories, although the respondents agreed slightly more with the protection story (X = 3.1) than with the gamble story (X = 2.9). However, when the respondents were forced to choose one story, 71% chose the protection story.

The respondents rated the three ads similarly, as there were no significant differences in the mean ratings across groups. Moreover, all three ads were rated positively (means greater than the neutral point) on all dimensions except for the informative/uninformative one. The ads were not significantly related to any other measures.

Three different measures were used to determine which reference point people use when contemplating the purchase of property insurance. All three measures indicate that the respondents focus on the possible negative consequences. Given that more than 80% of the respondents focus on the possible property loss rather than the security associated with owning insurance or the opportunity cost associated with the insurance premium, prospect theory would predict that the decision would be to purchase insurance (as discussed earlier). One other important finding is that the choice of reference point can vary substantially, depending upon the wording of the question (first thing thought of, situation receiving the most attention, or the situation used as a basis of comparison). For example, 83% selected a reference involving the loss of property when asked about the situation receiving the most attention, while only 66% selected a similar reference point when asked which situation they thought of first.

The intention and attitude measures also indicate that the respondents were receptive to the notion of property insurance. When asked if they would buy insurance if they had to make a decision immediately, 54% said that they would.

Detailed Analyses

Our analysis focuses upon three issues. The first is whether owning insurance is a meaningful reference point. The second is whether there is empirical support for the pairwise associations between variables that have been advanced above. The third is whether there is empirical support for the model depicted by Figure 1.

FIGURE 1

EXPECTED RELATIONSHIPS IN THE INSURANCE PURCHASE DECISION

Is Insurance a Reference Point? If owning insurance is a useful reference point, then there should be little difference between respondents who selected owning insurance/theft and those who selected owning insurance/no theft as the first (or most attention receiving, or point of comparison) consequence.

This issue was investigated by chi square analysis. In each chi-square test, subjects selecting the insurance/no theft consequence were compared to those selecting the insurance/theft consequence. In one set they were compared to their stated willingness to buy the advertised insurance policy, and in the other to their selection of the protection or the sure loss cognitive story. The findings strongly suggest that the two reference points are very different. Individuals who first thought of the theft/insurance consequence were more likely to think of insurance as protection (p < .05) and somewhat more likely to say they would buy insurance (p < .2). Those who used theft/insurance as the point of comparison were more likely (p < .1) to view insurance as protection. Those who reported paying most attention to theft/insurance were also more likely (p < .05) to say they would buy insurance. In all other cases the results were similar albeit not statistically significant.

There are a number of possible reasons why the two insurance reference points differ. One is that people think of the alternative they would have received if they had made a different choice. A second reason is that people are responding to an event (i.e. theft). In other words, if a person thinks of the situation where a theft occurs then s/he will: (1) think of insurance as protection and (2) want insurance.

Pairwise Findings. The hypotheses advanced above are viewed initially in terms of their disaggregated pairwise relationships. The results are presented in disaggregated form because neither the story nor reference point measures are well suited for being transformed into multi item single measures. This weakness stems from our use of so many nominal measures (i.e. select a story or select a reference point), and the mix of specific issues addressed (e.g. which of four consequences attracted the most attention, and which of three consequences were most important).

The strongest results of the pairwise analysis are listed below. Subjects selecting the protection cognitive story both held a more favorable attitude towards insurance and were likely to express a willingness to buy insurance. However, story choice and choice of a reference point were unrelated. An individual's Affect towards insurance was generally associated with both choice of a reference point (if you like insurance you select the worst-case reference point) and expressed intentions. A subject's level of agreement with a reference point (or his selection of it) was associated with purchase intentions in the manner predicted by prospect theory. The aforementioned results do not hold for all measures, for example agreement with a cognitive story is not significantly correlated with Affect, rather these results constitute common patterns of findings. We believe that the sensitivity of the results to how measures were worded is itself an interesting, albeit disquieting, finding. Tables detailing measures and findings will be furnished upon request.

The Model

Path analysis was used to analyze the model depicted by Figure 1. An initial analysis found that the Story (i.e., gamble vs. protection view of insurance) is not significantly related to either the reference point or buying intentions when Affect is included in the analysis. A more detailed analysis revealed that the pairwise story/reference point and story/intent to buy relationships are spurious, i.e. they are attributable to Affect. Consequently the model was reanalyzed without the cognitive story. The estimated standardized beta coefficients (p < .01) are depicted in table one. It is of interest that only 13% of affect's influence upon choice operates through the reference point. It is of equal interest that only 18% of the total association between the reference point and choice is spurious, i.e. attributable to Affect.

TABLE 1

RESULTS OF THE PATH ANALYSIS

CONCLUSIONS

This study investigated the relative importance of non-task constructs (Affect, Story) and task-determined constructs (Advertisements, Reference Points) on Choice. It was found that Affect relates directly to Choice and indirectly to Choice through the Reference Point. Attempts to manipulate the Reference Point selected through print advertisements were not successful.

Major problems encountered in this study were the need to structure the task so that the reference points and its measures could be understood by the subjects. We used a variety of such measures, with mixed results.

Future research should continue to develop better measures of relevant reference point usage in applied settings. Also, future research should continue to investigate the role that non-task variables (Affect, Story) play on the task's dependent variable (Choice).

REFERENCES

Alba, Joseph and Lynn Hasher (1983), "Is Memory Schematic?" Psychological Bulletin, 93 (No. 2), 203-23.

Dawes, Robyn M. (1975), "The Mind, The Model and The Task" in Cognitive Theory, Vol. 1, Frank Restle et al., eds., Hillsdale, NJ: Lawrence Erlbaum, 119-139.

Fischhoff, Baruch (1983), "Predicting Frames," Journal of Experimental Psychology: Learning, Memory, and Cognition, 9 (No. 1), 103-16.

Hershey, John C., Howard C. Kunreuther, and Paul J. h. Schoemaker (1982), "Sources of Bias in Assessment Procedures for Utility Functions," Management Science, 28 (No. 8, August), 936-54.

Hershey, John C. and Paul J. H. Schoemaker (1986), "Risk Taking and Problem Context in the Domain of Losses: An Expected Utility Analysis," Journal of Risk and Insurance, 47 (No. 1, March), 111-132.

Kahneman, Daniel and Amos Tversky (1979), "Prospect Theory: An Analysis of Decision Under Risk," Econometrica, 47 (No. 2, March), 263-291.

Kahneman, Daniel and Amos Tversky (1984), "Choices, Values, and Frames," American Psychologist, 39 (No. 4), 341-350.

Kisielius, Jolita and Brian Sternthal (1986), "Examining the Vividness Controversy: An Availability-Valence Interpretation," Journal of Consumer Research 12 (No. 4, March), 418-31.

Lichtenstein, Sarah, Willem A. Wagenaar, and Gideon Keren (1985), "Islanders and Hostages: Deep and Surface Structures of Story Problems," Decision Research Report 85-7, Eugene, OR.

Loomes, Graham and Robert Sugden (1982), "Regret Theory: An Alternative Theory of Rational Choice Under Uncertainty," The Economic Journal 92 (December), 805-24.

Payne, John W. K Dan J. Laughhunn, and Roy Crum (1980 "Translation of Gambles and Aspiration Level Effects in Risky Choice Behavior," Management Science, 26 (No. 10, October). 1039-60.

Payne, John, R. Crum, and D. J. Laughhunn (1982), "Translation of Gambles and Aspiration Level Effects in Risky Choice Behavior," Management Science, 39, 361-350.

Petty, Richard and John Cacioppo (1981), Attitudes and Persuasion: Classic and Contemporary Approaches, Dubuque, IA: William C. Brown.

Punj, Girish N. and David W. Stewart (1983), "An Interactive Framework of Consumer Decision Making," 10 (No. 2, September), 181-196.

Puto, Christopher P. (1985), "The Framing of Buying Decisions," Journal of Consumer Research, forthcoming.

Schoemaker, Paul J. B. and Howard Kunreuther (1979), "An Experimental Study of Insurance Decisions," Journal of Risk and Insurance, 46, 603-618.

Slovic, Paul, Baruch Fischhoff, and Sarah Lichtenstein (1982), "Response Mode, Framing, and Information Processing Effects in Risk Assessment," in Robin M. Hogarth, ed., Question Framing and Response Consistency San Francisco: Jossey-Bass Inc., Publishers, 21-36.

Slovic, Paul, Baruch Fischhoff, Sarah Lichtenstein, Bernard Corrigan, and Barbara Combs (1977), "Preference for Insuring Against Probable Small Losses: Insurance Implications," The Journal of Risk and Insurance, 44 (No. 2, June), 237-58.

Thaler, Richard, "Mental Accounting and Consumer Choice," Marketing Science 4 (Summer), 199-214.

Tversky, Amos and Daniel Kahneman (1981), "The Framing of Decisions and the Psychology of Choice," Science, 211, 453-458 .

Wiener, Joshua L., James W. Gentry, and Ronald K. Miller (1986), "The Framing of the Insurance Purchase Decision," in Richard J. Lutz, ed., Advances in Consumer Research, Vol. 13, Provo, UT: Association for Consumer Research, 251-256.

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