A Multi-Distributional, Conceptual Framework For the Study of Perceived Risk

John W. Vann, University of Missouri-Columbia
ABSTRACT - This paper presents a multi-level framework within which perceived distributions of outcomes associated with alternatives may be viewed. Risk associated with the shape and location of these distributions is discussed along with the manner in which the risk perceived at subordinate levels in the hierarchy may be combined by the consumer to derive the risk perceived at superordinate levels. Two types of risk are differentiated: the risk associated with an alternative and the risk associated with a choice between alternatives. Bettman's concepts of inherent risk and handled risk are positioned within the framework. Opportunities for viewing information within the framework are presented.
[ to cite ]:
John W. Vann (1984) ,"A Multi-Distributional, Conceptual Framework For the Study of Perceived Risk", in NA - Advances in Consumer Research Volume 11, eds. Thomas C. Kinnear, Provo, UT : Association for Consumer Research, Pages: 442-446.

Advances in Consumer Research Volume 11, 1984      Pages 442-446

A MULTI-DISTRIBUTIONAL, CONCEPTUAL FRAMEWORK FOR THE STUDY OF PERCEIVED RISK

John W. Vann, University of Missouri-Columbia

ABSTRACT -

This paper presents a multi-level framework within which perceived distributions of outcomes associated with alternatives may be viewed. Risk associated with the shape and location of these distributions is discussed along with the manner in which the risk perceived at subordinate levels in the hierarchy may be combined by the consumer to derive the risk perceived at superordinate levels. Two types of risk are differentiated: the risk associated with an alternative and the risk associated with a choice between alternatives. Bettman's concepts of inherent risk and handled risk are positioned within the framework. Opportunities for viewing information within the framework are presented.

INTRODUCTION

Consumer behavior involves risk in the sense that any action of a consumer will produce consequences which he cannot anticipate with anything approximating certainty, and some of which at least are likely to be unpleasant (Bauer 1960. D. 389).

This characterization of risk and others which have been used in both theoretical and empirical investigations consider two aspects of the distribution of potential outcomes, on some dimension, that may accrue to a decision-maker following a choice. The first is the shape of the distribution (in this case expressed as uncertainty, but also variously expressed as probability, variance, semi-variance, semi-standard deviation, skewness, kurtosis, etc.). The second is the evaluation of the location of that distribution or a portion or portions of that distribution on the respective outcome dimensions (e.g., ". . . some of which at least are likely to be unpleasant"). This evaluation may be expressed in either objective or subjective (perceived) terms. (Evaluation may be included in some of the previously mentioned shape assessments, e.g., semi-variance, as well as through other measures such as expected value.)

This paper will be concerned with perceived as opposed to "objective" outcome distributions. In other words, it is concerned with the subjective frequency distributions of the potential outcome levels (on the relevant dimensions) which a consumer associates with an alternative or with a product class. It is presumed that these distributions may be based either on personal experience or on information received from external sources.

Perceived risk has been examined on many dimensions (e.g., performance risk, financial risk, physical risk, psychological risk, social risk, time risk, etc.). Presumably, consumers combine these dimensions in some way to arrive at an overall risk assessment (Humphreys and Kenderdine 1979, Zickmund and Scott 1973). For simplicity, the present discussion will be limited to functional performance outcome distributions. It is assumed that the principles suggested for performance would apply to other dimensions as well.

It is the contention of this paper that much of the confusion surrounding the study of perceived risk can be reduced by recognizing several perceived distributions on each risk dimension rather than just one. The proposed framework presumes that perceived outcome distributions are cognitively represented at multiple levels of abstraction within the consumer's semantic network. That is, that consumers will have perceived outcome distributions at subordinate category levels and at superordinate category levels. Additionally, it is assumed that the distributions at superordinate category levels are inferred from those at subordinate levels. The structural framework presented here will presume that consumers' perceptions of different aspects of a brand's performance are subordinate to their perceptions of the overall level of a brand's performance and that their brand perceptions are subordinate to their perceptions of product categories. Other, alternative ways of perceptually classifying reality are possible. (See Lynch and Srull (1981) for a discussion of grouping by brands versus grouping by attributes.) However, the principles discussed here should apply to these alternate classification schemata as well.

THEORETICAL FRAMEWORK

Four perceived distributions will be discussed. First, outcomes may have multiple aspects. For example, if we were to think of the performance dimension for a particular meal at a restaurant, there could be perceived outcome distributions for that meal's sweetness, saltiness, portion size, temperature, etc. (Ahtola 1975, Sarel 1982). The second distribution is presumed to be a distribution of a person's perceptions of the performance (quality) of the meal, as a whole, across multiple experiences. Third would be the distribution of the perceptions of summary judgement regarding the meal (summarized across occasions). This distribution would include the perceptions of summary statements made by others as well as the person's own summary judgement. The fourth distribution is the perceived distribution of that meal as served by all restaurants within some relevant set (summarized for each restaurant across other's and own summary judgements/across occasions) .

The distributions illustrated in this example may be applied to any product or service category. At a more general level they are: first, the perceived outcome distributions for each aspect of an option's (i.e., brand's) performance; second, the perceived distribution of overall performance of the option (brand) across occasions (as experienced by a person); third, the perceived distribution of summary (across occasions) statements, made by others and self, of the option's (brand's) overall performance; and, fourth, the perceived distribution of overall performance of individual options (brands) within a product/service class, i.e., point estimates that the person would make for each brand, if requested (summarized across persons and occasions).

Each of these distributions will be discussed in greater detail below. Several aspects of perceived risk that have been discussed in the literature will be considered within the context of the emerging framework -- including Bettman's concepts of inherent and handled risk (1973). A distinction will also be made between the risk associated with one alternative and the risk associated with a choice.

Perceived Distribution of Performance on one Aspect of a Dimension Across Occasions

The first distribution to be considered is the distribution of a consumer's perceptions of outcomes accrued from one alternative on one aspect of performance. These outcomes are assumed to have occurred across multiple episodes in which the alternative has been experienced by the consumer. (Parallel distributions representing the experiences of others could also be perceived by the consumer.)

There is no reason to expect a direct mapping from the actual performance to the perceived performance for two reasons. First, this perceptual process would be subject to all of the normal biases which accompany any perceptual process (Bruner 1957). For example, the perceptions would be affected by expectations derived from prior experiences, by social influence, physiological state, emotional state, etc. Second, there is evidence that values on the subjective scale onto which the perceptions are mapped do not represent a linear transformation of values on the actual (objective) scale (Kahneman and Tversky 1979).

Consumers would be expected to have some uncertainty regarding their perceptions (Ahtola 1971, Sarel 1982). This would be related to their confidence in their perceived ability to judge that aspect of performance (e.g., "I just can't judge a wine's dryness"). This confidence in ability to judge has been incorporated into the Fishbein extended behavioral intentions model at the "attribute" level by Bennett and Harrell (1975, also see Pras and Summers (1978) for an attribute-level application).

Risk inferred for an alternative on one aspect of performance should reflect the shape and/or the location of the perceived distribution for this alternative on this aspect of performance.

Perceived Distribution of Perceived Performance on One Dimension Across Occasions

The next distribution would represent overall judgements of the performance (or physical harm, etc.) associated with an alternative across multiple episodes where each judgement is based on some combination of the aspects of performance which are relevant to the consumer (Bettman 1979). This combination process should also reflect perceptual biases. For example, the perception of performance which is extremely good or extremely bad relative to other occurrences may be distorted toward more frequent or "representative" levels (Hogarth 1980, p. 15). As with the previous one, this distribution would be associated with two sorts of confidence -- confidence in the ability to judge and confidence in the brand. (Confidence in the ability to judge a brand was examined by Barach (1969).) The confidence in the brand seems to be what Bauer had in mind in suggesting that risk was the inverse of confidence (1960).

Risk may also be modeled within this distribution using information regarding the shape and/or the location of the distribution. However, this risk would be the risk associated with one particular brand (e.g., Peter and Ryan 1976), and not the risk associated with the choice between brands. This distinction will be maintained throughout the paper.

Perceived Distribution of Summary Judgements on nne Dimension (Summarized Across Occasions)

The third distribution consists of the perceptions of statements representing the summary judgements of others and the consumer's own summary judgement, where each summary judgement is derived from the distribution of overall evaluations across occasions. (There is no reason that the consumer would not also have a corresponding distribution of summary judgements across others and self at the "attribute" level. For example, someone could make the summary statement: "That cola always seems very carbonated.")

It is likely that attribution processes would be involved in the weighting of performance/occasions to form summary judgements. Extreme evaluations (outliers) may be discounted by attributing them to special circumstances. On the other hand, they may be more easily recalled and thereby weighted more heavily (i.e., the availability heuristic -- "A person is said to employ the availability heuristic whenever he estimates frequency or probability by the ease with which instances or associations could be brought to mind" (Tversky and Kahneman 1973, p. 208).) Recent episodes may be more heavily weighted because they are more available or because they are seen as being more representative of the current state of affairs.

Knowledge of others' summary statements would probably interact with the consumer's confidence in his/her own ability to judge performance in this product category to influence his/her own summary statement (Festinger 1950, 1954; Asch 1951). If the consumer had no perceptions regarding others' evaluations of the brand then the only point on this third distribution would be that representing the consumer's own summary judgement. Alternatively, if the consumer had no experience with the brand then his/her own summary judgement would not be represented and the only information that s/he might have would be the summary judgements of others.

This distribution may also be modeled to depict the risk associated with one brand. The primary difference between this distribution and the previous one is that this one explicitly considers additional information regarding the summary judgements of others that has been gained through social interaction.

Perceived Distribution of Brands' Performances

The consumer may combine information across individuals' summary judgements from the previous distribution to arrive at an overall summary judgement (i.e., a point estimate) for the performance (or other relevant dimension) of a brand. The distribution of these overall summary judgements (summarized across persons/aspects-of-the-dimension/dimensions/occasions) across brands within a product class comprises the fourth distribution of interest. (Note that for some manufacturers, there would be another distribution which would represent the perceived point estimates for the performance of multiple products covered by a family brand name.)

Consumers would be expected to differentially weight different persons' summary judgements in arriving at the overall summary judgement for each brand. Another's perceived similarity in standards (tastes) should affect the weight attached to that person's summary judgement as should his/ her credibility due to particular expertise (perceived ability to judge) with respect to that particular product category (e.g., 'Ne usually like the same foods, so if she says it's good . . ."; or "She really knows her wines, so. . .").

Summary statements which are widely divergent from the consumer's own may be discounted. Attributions will likely be made regarding the standards of the persons making these divergent statements as well as the number and quality of the performance/occasions which they must have experienced (e.g., "He doesn't like any Mexican food;" or "He must have just hit them on a bad day").

The inference process involved in arriving at the summaries for the brands should reflect Kelley's covariance approach to attribution. Person-brand-occasions may be represented as cells in Kelley's Cube (see Mizerski, Golden, and Kernan 1979, p. 126). The three dimensions of the cube are brands (entities), persons, and occasions (time) respectively. The overall evaluations for each brand on various occasions by one person would represent one "tube" through Kelley's Cube (i.e., the experiences of one person for one brand on various occasions). The consumer's confidence in the inferred summary estimate for each brand should increase with the consistency of the performance levels across occasions and with the consensus in reported evaluations across persons (Kelley 1967, p. 197).

The shape and/or location of this distribution should be reflected in the perceived risk associated with the product class. This risk is different from the risk associated with one alternative. This is the perceived risk associated with the choice between/among alternatives. It should reflect the likelihood of choosing the "wrong" brand, as well as the consequences of such a choice. The likelihood of a wrong choice is reflective of the shape of the distribution. A wrong choice would occur if, on this occasion, a rejected brand would have performed better than the chosen one. Viewed in this way, making a wrong choice involves the simultaneous occurrence of two events (which can occur separately), i.e., the rejection of a satisfactory brand and the purchase of an unsatisfactory one. (Popielarz (1967) considered that there was a separate risk associated with each.) Thus, a wrong choice could result in foregone opportunities (i.e., opportunity loss (Zickmund and Scott 1973)) associated with the rejected alternative or in negative outcomes incurred with the chosen one. In other words, both the positive and negative ends of the brand- level distributions could be considered in evaluating the risk associated with the choice between brands. Pras and Summers suggested that (at the attribute level) the positive end of an outcome distribution would be relevant for risk takers and the negative end would be relevant for risk voiders (1978). It seems more reasonable to imagine risk takers as trading-off the possibility of high gain against the possibility of a loss, rather than only considering the positive end of the distribution. Models of risk associated with the choice between/among brands which do not consider the positive as well as the negative ends of the distribution cannot represent such trade-offs.

The likelihood of choosing the "wrong" brand is related to the degree of overlap of the separate brand distributions (i.e., the distributions of summary statements of others and self). Increases in the dispersion across brands should decrease the overlap of their corresponding distributions, and, consequently, the perceived risk associated with choosing between brands. The opposite should be true or the dispersion within brands. As the dispersion of he perceived distributions of performance (or other outcome dimension) for each of the brands under consideration increases, the likelihood of different distributions overlapping increases and the risk associated with choosing between/among them should also increase. (Festinger discusses this idea of "cognitive overlap" in relation to cognitive dissonance -- hypothesizing that the greater the overlap, the less the dissonance following the choice 1957. a. 41).)

The risk associated with a choice should also reflect the location of the brands in contention. If all are good, hen the risk associated with the choice should be low. t is to this final distribution that Bettman's concept of inherent risk seems most relevant

RECONCILING BETTMAN'S DEFINITIONS OF RISK WITH THE FRAMEWORK

Inherent Risk

Bettman's definitions of inherent and handled risk are essentially empirical. Inherent risk is the level of risk which consumers would tell you they perceived in a product Lass if they were confronted with making a choice in an imaginary store in which they could not tell which brand was which (Bettman 1973). ". . . inherent risk deals with the riskiness a consumer feels if no information is assumed" (p. 184). Inherent risk would be reflected in e perceived distribution of brands within the product Lass (i.e., the fourth distribution). The paradox is that this distribution can only be known if the consumer has information regarding the brands' performance. When consumers truly buy "blind," as when buying within a product category for which they have no information, they will also not know the shape or location of the fourth distribution. Inherent risk can only be estimated by people for whom it no longer exists.

Inherent Risk, Variance, and the Risk Associated with a Choice

Examination of the many ways in which the shape of outcome distributions has been incorporated into risk models is beyond the scope of this paper. However, the definition of inherent risk as being associated with "buying blind," when compared against the more common situation in which consumers would be presumed to buy what they thought was the "best" alternative, results in some counterintuitive notions regarding the relationship between the variance of the fourth distribution and the perceived risk associated with the choice

Thus, the risk inherent in a brand choice situation within a product class will depend upon the degree to which a buyer believes he can construct a reasonable decision rule for making a brand choice. . . . Further, the goodness of the buyer's decision rule is hypothesized to depend upon his perceived distribution of quality over the brands of the product class. . . . the higher the perceived variation in quality for the product class, the less the likelihood of constructing an effective brand choice decision rule. Variation is measured by the variance of the perceived quality distribution (Bettman 1973, pp. 184-185).

It is true that for consumers buying "blind" an increase in the variance of the fourth distribution would increase the risk associated with the choice of a brand. However, within the framework presented here, in which the consumer has some information, an increase in the variance across brands within the product class actually should decrease the risk even using Bettman's conception of inherent risk as the ". . . innate degree of conflict the product class is able to arouse" (1973, p. 184) -- since the overlap of the brands' performance distributions should be reduced. There would be less likelihood that any of the rejected brands would have performed better than the chosen one on that occasion and there should be less conflict before the choice.

An extreme case would be the situation in which the variation in quality for the product class increases until there is no overlap between the distributions of the brands under consideration (either across others' and own summary statements (distribution three) or across the individual's overall evaluations for performance/occasions (distribution two) -- whichever is relevant). There should be little risk associated with a choice between such alternatives. This situation illustrates the concept of stochastic dominance which is used in the finance literature.

Probability distribution A is said to stochastically dominate probability distribution B i' the cumulative probability of achieving any rate of return up to some specified level for distribution A is less than or equal to that same cumulative probability for asset B, and, at least at one point, the less-than inequality holds (Francis and Archer 1979, p. 391).

Handled Risk

It is unclear exactly how Bettman's concept of handled risk may be incorporated into this framework.

Handled risk is the amount of conflict the product class is able to arouse when the buyer chooses a brand from a product class in his usual buying situation. That is, handled risk to a first approximation represents the end results of the action of information and risk reduction processes on inherent risk. . . . For example, a consumer may feel there is a great deal of risk associated with the product class aspirin. However she has a favorite brand which she buys with confidence. In this case, inherent risk is high, but handled risk may be low for aspirin (Bettman 1973, p. 184, emphasis added).

The problem arises from the different ways in which the consumer may approach the buying decision. If the consume considers all of the alternatives on every purchase occasion, then handled risk could reflect the shape and location of the fourth distribution (i.e., the risk associated with a choice). However, if this purchase has been routinized, then handled risk would be a reflection of the second or third distributions. That is, handled risk would then be the risk associated with one brand rather than the risk associated with a choice.

Bettman's choice of the term, "handled risk," connotes the idea of risk that can be "handled" or "coped with" by the consumer. His definition states that handled risk reflect the ". . . action of information and risk reduction processes on inherent risk" (1973, p. 184, emphasis added). He makes it explicit that handled risk applies to persons who are buyers within the product class and who have under taken whatever risk reduction strategies were necessary to enable them to cope with the risk associated with their purchase.

On the one hand, then, we have the concept of inherent risk which is irrelevant to those who have information regarding the product class and unknown to those for whom it is relevant, and on the other hand we have the concept of handled risk (whether it be reflected in the product-class or the brand distribution) which only applies to those who have elected to take the risk associated with the purchase of a brand from the product class. Neither concept applied to those who have information regarding the distribution o performance of a brand or brands within a product class an have decided to not buy any brand because there is too much risk associated with the choice (i.e., they cannot "handle or cope with the risk which they perceive).

The present framework avoids such problems because it simply presents the perceived outcome distributions which consumers have regarding the performance (or other outcome dimension) of a brand or brands. These distributions are presumed to reflect all perceptual distortions which the information may have suffered including (but nor requiring any which may have served to reduce risk to tolerable levels. This could be important to managers of brands within a product class for which the consumers "know" where the brands are on the fourth distribution, yet they refuse to purchase any of the brands because the perceived risk i too high. Information regarding the source of the high perceived risk would allow the managers to attempt to reduce it (through advertising or through product quality control, or warranty (Bearden and Shimp 1982) modifications).

All of the distributions presented thus far provide an opportunity to resolve the manner of incorporating information regarding the particular distribution's shape and location into a model of perceived risk. In addition, however, the entire framework also provides an opportunity to model the consumer's information search activities intended to reduce risk.

VIEWING INFORMATION SEARCH WITHIN THE FRAMEWORK

Information gathering has been presented as one consumer strategy for reducing perceived risk (Locander and Herman 1979, Zickmund and Scott 1973). Within the present framework, the information gathered would give the consumer sample points on one or more of the distributions which have been presented. For example, product trial by a consumer would be seen as the gathering of data points on either the aspects-of-a-dimension distributions or the overall evaluation scales for given dimensions. These points along with points gathered in repeated trials (if they exist) would comprise the perceived outcome distribution for that brand for that consumer.

Inferences from the sample points to more complete distributions could b,e made through various attributions. Consumers may infer that a few sample points are highly representative of the underlying distribution. For example, even with sophisticated sampling techniques, scientists appear to view small samples as being representative of the larger distributions from which they are drawn (i.e., the "law of small numbers" (Tversky and Kahneman 1971)). Meyer (1981) suggested that the less information that was known to a consumer regarding the performance of a brand, the more strongly that consumer would rely on the average for the product class as an estimate for that brand's performance.

Word-of-mouth and the reading of advertisements or other written material about a brand may also be represented within the present framework. This would be seen as the gathering of points on the distribution of summary statements (distribution three). For example, a person attending a conference in a large city may have no information regarding the various restaurants in that city. One common source in situations such as this is taxi drivers. The taxi driver is presumed to give his/her summary judgement(s) regarding the recommended restaurant(s). Of course, if a restaurant is recommended, that information also suggests the restaurant's relative location on the fourth distribution. In fact, the consumer may make attributions that the taxi driver's judgement is itself a summary across others' summary statements (garnered from previous customers) and is therefore a point on the taxi driver's fourth distribution. It is presumed that information points gathered in this process would be weighted through the attribution processes discussed earlier in the paper.

Some sources, such as Consumer Reports, provide information regarding the location of brands on the fourth distribution. In fact, these points often specifically represent judgements which have been summarized across occasions/ aspects-of-dimensions/dimensions/individuals.

Whether information search reduces or increases perceived risk should depend upon the emergent perceived distribution(s). Information search would, therefore, be better positioned as a strategy of risk assessment rather than of risk reduction.

SUMMARY AND CONCLUSIONS

This paper makes explicit the separate distributions which comprise the overall framework within which consumers may assess the risk which they perceive as being associated with brands and with choices between/among brands. These separate distributions also provide researchers with a framework within which to examine the effects of shape and location on perceived risk for the respective distributions. In particular, it seems that it should be fruitful to distinguish between the risk associated with one brand versus the risk associated with the choice between brands. It seems likely that the dispersion (however it may be modeled) of the distribution within a brand may act in the opposite way from the dispersion of the distribution across brands in affecting perceived risk. Increases in the dispersion of outcomes for one brand should increase the risk associated with that brand and the risk associated with a choice for which that brand is in contention, while increases in the dispersion of outcomes across brands should decrease the perceived risk associated with a choice among those brands.

Other research opportunities which may be cast within this framework deal with information search and integration. At which level(s) (i.e., within which distribution(s)) do consumers tend to search to assess risk? How do they combine information at the aspects level to arrive at an overall evaluation for a dimension on one occasion? -- across occasions? How is dimensional information combined to arrive at an overall evaluation at the brand level on single occasions and across occasions? How is information from multiple sources combined to estimate risk? What sorts of attributions are made? What are other influences on differential weighting of information from different sources?

Perhaps this framework can contribute to answering some of these questions and can contribute to reducing the confusion which exists within the perceived risk literature.

REFERENCES

Ahtola, Olli T. (1975), "The Vector Model of Preferences: An Alternative to the Fishbein Model," Journal of Marketing Research, 12, (February), 52-9.

Asch, Solomon E. (1951), "Effects of Group Pressure upon the Modification and Distortion of Judgements," in Groups, Leadership, and Men, ed. H. Guetzkow, Pittsburgh Carnegie Press.

Barach, Jeffrey A. (1969), "Advertising Effectiveness and Risk in the Consumer Decision Process," Journal of Marketing Research, 6, (August), 314-20.

Bauer, Raymond A. (1960), "Consumer Behavior as Risk Taking," Dynamic Marketing for a Changing World, Proceedings, 43rd Annual Conference, American Marketing Association, 389-98.

Bearden, William O. and Shimp, Terence, (1982), "The Use of Extrinsic Cues to Facilitate Product Adoption," Journal of Marketing Research, 19, (May), 229-39.

Bennett, Peter D. and Harrell, Gilbert D.(1975), "The Role of Confidence in Understanding and Predicting Buyer's Attitudes and Purchase Intentions," Journal of Consumer Research, 2, (September), 110-17.

Bettman, James R. (1973), "Perceived Risk and Its Components: A Model and Empirical Test," Journal of Marketing Research, 10, (May), 184-90.

Bettman, James R. (1979), An Information Processing Theory of Consumer Choice, Reading: Addison-Wesley.

Bruner, Jerome S., "On Perceptual Readiness," Psychological Review, 64, No. 2, 123-52.

Festinger, Leon (1950), "Informal Social Communications," Psychological Review, 57, 271-80.

Festinger, Leon (1954), "A Theory of Social Comparison Processes," Human Relations, 7, 117-40.

Festinger, Leon (1957), A Theory of Cognitive Dissonance, Stanford: Stanford University Press.

Francis, Jack C. and Archer, Stephen H. (1979), Portfolio Analysis , 2nd Ed., Englewood Cliffs: Prentice-Hall.

Hogarth, Robin M. (1980), Judgement and Choice: The Psychology of Decision, Chichester: John Wiley & Sons.

Humphreys, Marie Adele and Kenderdine, James M. (1979), "Perceived Risk and Consumer Decision Making: An Alternative View of Uncertainty," in Educator's Conference Proceedings, eds. Neil Beckwith, Michael Houston, Robert Mittlestaedt, Kent B. Monroe, Scott Ward, Chicago: American Marketing Association.

Kahneman, Daniel and Tversky, Amos (1979), "Prospect Theory: An Analysis of Decision Under Risk," Econometrica, 47, (March), 263-91.

Kelley, Harold H. (1967), "Attribution Theory in Social Psychology," in Nebraska Symposium on Motivation, 1967, ed. David Levine. Lincoln: University of Nebraska Press.

Locander, William B. and Herman, Peter W. (1979), "The Effect of Self-Confidence and Anxiety on Information Seeking in Consumer Risk Reduction," Journal of Marketing Research, 16, (May), 268-74.

Lynch, John G., Jr. and Srull, Thomas R. (1982), "Memory and Attentional Factors in Consumer Choice: Concepts and Research Methods," Journal of Consumer Research, 9 (June), 18-37.

Mizerski, Richard W., Golden, Linda L., and Kernan, Jerome B. (1979), "The Attribution Process in Consumer Decision Making," Journal of Consumer Research, 6 (September), 123-40

Peter, J. Paul and Ryan, Michael J.(1976), "An Investigation of Perceived 'Risk at the Brand Level," Journal of Marketing Research, 13, (May), 184-8.

Popielarz, Donald T. (1967), "An Exploration of Perceived Risk and Willingness to Try New Products," Journal of Marketing Research, 4, (November), 368-72.

Pras, Bernard and Summers, John O. (1978), "Perceived Risk and Composition Models for Multiattribute Decisions," Journal of Marketing Research, 15, (August), 429-37.

Sarel, Dan (1982), "Models of Behavioral Choice Under Risk--A Review," in Proceedings of the Annual Meeting of the Southern Marketing Association, eds. John H. Summer, Blaise T. Bergiel, and Carol B. Andrews, Carbondale: Southern Marketing Association.

Tversky, Amos and Kahneman, Daniel (1971), "Belief in the Law of Small Numbers," Psychological Bulletin, 76, (August), 105-110.

Tversky, Amos and Kahneman, Daniel (1973), "Availability: A Heuristic for Judging Frequency and Probability," Cognitive Psychology, 5, (September), 207-232.

Zickmund, William G. and Scott, Jerome E. (1973), "A Multivariate Analysis of Perceived Risk, Self-Confidence, and Informational Sources," in Advances in Consumer Research, Vol. 1, eds. Scott Ward and Peter Wright, Urbana: Association for Consumer Research.

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